AN INTEGRATIVE APPROACH TO UNDERSTANDING PI(3)P SIGNALING AND
AUTOPHAGY
By
Katie Renee Martin

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Cell and Molecular Biology
2011

ABSTRACT
AN INTEGRATIVE APPROACH TO UNDERSTANDING PI(3)P SIGNALING AND
AUTOPHAGY
By
Katie Renee Martin
Phosphatidylinositol-3-phosphate (PI(3)P) is an intracellular signaling lipid which
recruits lipid-binding proteins, tethering them to subcellular compartments where they function
(Gaullier, Simonsen et al. 1998). PI(3)P is required for proper endocytosis and autophagy, two
membrane trafficking processes pivotal to cellular homeostasis, and as such, its production,
maintenance, and turnover must be tightly controlled (Funderburk, Wang et al. 2010). PI(3)P is
produced by the class III member of the mammalian phosphoinositide-3-kinase (PI3K) family,
Vps34 (vacuolar protein sorting 34) (Funderburk, Wang et al. 2010). Vps34-PI(3)P signaling is
antagonized by PI(3)P lipid phosphatases, such as MTMR3 and MTMR14 (hJumpy), which
function in autophagy (Vergne, Roberts et al. 2009; Taguchi-Atarashi, Hamasaki et al. 2010).
Given the complexity of Vps34 function and the number of proteins essential for its function, we
hypothesized that additional protein phosphatases exist to regulate Vps34-PI(3)P signaling.
To investigate this hypothesis, we performed a cell-based RNA interference screen to
identify phosphatases whose loss of function alters cellular PI(3)P . We found that reduced
expression of a receptor-like protein tyrosine phosphatase, PTPsigma, increased the abundance
of PI(3)P-positive vesicles in cells. Intriguingly, the vesicles in these cells mimicked those
observed in autophagic cells. Using a variety of cell biology approaches, we confirmed that
PI(3)P signaling and autophagy are elevated in the absence of PTPsigma.
As a role for PTPsigma in PI(3)P signaling was unprecedented, much work was required,
and still remains, to fully characterize its function in this process. We uncovered that PTPsigma

resides on PI(3)P-positive vesicles during both basal and induced autophagy. Further, its
internalization from the cell surface and presence on these vesicles is likely controlled through
defined proteolytic processing. Finally, loss of PTPsigma in cells appears to promote Vps34
activity as measured <italic>in vitro</italic> and PTPsigma is capable of interacting with both
Vps34 and at least one of its binding partners, Rubicon. We propose a working model where
PTPsigma downregulates PI(3)P signaling through control of a phosphotyrosine substrate, yet to
be identified, within or closely related to an endocytic Rubicon-containing Vps34 complex.
These results are summarized in Chapter 2.
Although still elusive, the function of PTPsigma in PI(3)P signaling and autophagy may
prove to have important consequences for cell fate. Autophagy is essential for cell survival
during stress and accordingly, we would predict PTPsigma suppression to enhance autophagymediated survival (Meijer and Codogno 2004). The ability to increase autophagy could
potentially provide therapeutic benefit in the treatment of several diseases, notably those
involving neurodegeneration (Hara, Nakamura et al. 2006; Komatsu, Waguri et al. 2006). With
this in mind, we utilized an <italic>in silico</italic> screening approach, outlined in Chapter 3,
to identify small molecule inhibitors of PTPsigma. If selective, these compounds could prove to
be useful molecular probes in the study of autophagy.
Finally, we aimed to capture a comprehensive understanding of autophagy dynamics
through mathematical modeling. To this end, we utilized kinetic live-cell microscopy to develop
an accurate and predictive model of autophagy vesicle dynamics, detailed in Chapter 4. This
model, and its framework for a more large-scale autophagy network, can be used to generate
novel hypotheses and be implemented as a tool to test experimentally-derived hypotheses, such
as those presented for PTPsigma.

ACKNOWLEDGEMENTS

First, I thank my mentor, Dr. Jeff MacKeigan, for his guidance, encouragement, and
support during the course of my graduate studies. He has believed in me and my abilities
unconditionally, even during times when I was unsure of myself. The confidence he has instilled
in me and the enthusiasm for science he has fostered will remain with me through my future
endeavors.
Second, I want to thank my lab-mates for making my time in graduate school enjoyable
on a daily basis. Each person that has passed through the lab, whether for a few weeks or years,
has impacted me personally and had a positive influence on my project. I extend an especially
heartfelt thank-you to my fellow graduate students in the lab – the future Drs. Natalie Niemi,
Laura Westrate, Jon Karnes, Megan Goodall, and Dani Burgenske. Together, we have shared
both the frustrations and joys of lab life and helped each day pass a bit easier with laughter. I will
take with me fond memories of our tissue culture room pow-wows, lunches, cardboard cut-out
heists, carbohydrate buffer, iPod Fridays, retreats, British accents, midnights in the lab, and
many other experiences. I also thank the rest of my lab-mates for their support and camaraderie Dr. Brendan Looyenga (my bench-mate for nearly five years), Dr. Nate Lanning, Dr. Vanessa
Fogg, Audra Kauffman, and Amy Nelson. I also express my gratitude for student interns who
have helped with my project over the years – Joe Church, Ryan Davis, Michael Shaheen, James
Hogan, and Anna Plantinga. And finally, I express a very special thank you to my limp-a-long,
Dr. Jamie Kopper. We were inseparable from our first days in graduate school, overcame the
difficulties of classes and prelims together, and will continue to be friends throughout the rest of
life’s journeys.

iv

I am also extremely grateful to the members of my guidance committee – Dr. Walt
Esselman, Dr. Christina Chan, Dr. John LaPres, and Dr. Cindy Miranti. They have each provided
critical insights into my project and supported its vision from the very beginning. Their kind,
accommodating, and encouraging nature has made my graduate experience a positive one.
My project was entirely dependent on the expertise and effort of collaborators including
Michel L. Tremblay, Yong Xu, H. Eric Xu, Richard Posner, William Hlavacek, Dipak Barua,
Edward Stites, Nikolai Sinitsyn, Srabanti Chaudhury, Nathalie Meurice, Joachim Petit, and Pooja
Narang. I thank them for their willingness to contribute to my project and for teaching me the
value of collaboration.
Last and most importantly, I thank my family for their unwavering support. My mom,
dad, grandparents, and siblings have accepted my career pursuits and offered encouragement
every step of the way. The work ethic and character of each of my family members motivates
and inspires me to do positive and constructive things with my life. Also, I would not be where I
am today without the love and support of my husband, Steve Martin. He has been there for me
every day, encouraged me when I needed it most, and serves as a constant reminder of what is
truly important in life.

v

TABLE OF CONTENTS

List of Tables………………………………………………………………………………..…..viii
List of Figures…………………………………………………………………………...…….....ix
CHAPTER 1. INTRODUCTION…………………………………………………………………1
Figures……………………………………………………………………………………19
CHAPTER 2. CHARACTERIZATION OF PTPSIGMA AS A NOVEL REGULATOR OF
Vps34-PI(3)P SIGNALING AND AUTOPHAGY……………………………………….……..23
SECTION 1. Identification of PTPsigma as a novel regulator of autophagy……....…....24
Abstract……………………………………………………………………….….25
Introduction…………………………………………………….………………...26
Results………………………………………………………………….………...28
Discussion………………………………………………………………………..34
Materials and Methods……………………………………………………….…..38
Tables.……………………………………………………………………………46
Figures……………………………………………………………………………55

SECTION 2. A potential role for PTPsigma as a Vps34 complex effector and the
discovery of autophagy-relevant processing events……………………………………..73
Abstract……………………………………………………………………….….74
Introduction……………………………………………………………………....76
Results……………………………………………………………………….…...79
Discussion………………………………………………………………….…….85
Materials and Methods……………………………………………………….…..88
Tables..……………………………………………………………………….…..91
Figures……………………………………………………………………………93

CHAPTER 3. IN SILICO-BASED IDENTIFICATION OF SMALL MOLECULE INHIBITORS
TARGETING PTPSIGMA……………………………………………………….…………….101
Introduction……………………………………………………………………..102
Results………………………………………………………………………..…106
Discussion……………………………………………...……………………….110
Materials and Methods……………………………………………………….…114
Figures………………………………………………………………………..…116
CHAPTER 4. MATHEMATICAL MODEL OF AUTOPHAGIC VESICLE DYNAMICS......130
Introduction…………………………………………………………………..…131
Results…………………………………………………………………………..133
Discussion………………………………………………………………………144
Materials and Methods……………….………………………………….…...…147
vi

Figures………………………………………………..…………………………151
CHAPTER 5. SUMMARY AND FUTURE DIRECTIONS…………………………………...167
Figures………………………………………………………………………..…182
REFERENCES…………………………………………………………………………………186

vii

LIST OF TABLES

Table 2.1. siRNA-mediated knockdown of human phosphatase genes alters cellular PI(3)P.......47
Table 2.2. Potentially phosphorylated tyrosine residues of the early autophagic machinery…....91

viii

LIST OF FIGURES

Figure 1.1. Overview of autophagy…………………………………………………..………….19
Figure 1.2. Structure and proteolytic processing of PTPsigma…………...………………......…21
Figure 2.1. Cell-based siRNA screen identifies PTPsigma as a modulator of PI(3)P……......…55
Figure 2.2. Loss of PTPsigma hyperactivates autophagy……………………………………..…57
Figure 2.3. Loss of PTPsigma increases autophagic vesicle abundance as measured by electron
microscopy……………………………………………………………………………………….59
Figure 2.4. Exogenous PTPsigma localizes to PI(3)P vesicles and rescues the siRNA
phenotype………………………………………………………………………………………...61
Figure 2.5. Localization of PTPsigma to vesicular structures does not require PI(3)P……….…63
Figure 2.6. Target genes are effectively knocked down by siRNA……………………………...65
Figure 2.7. PTPsigma knockdown increases the abundance of autophagic, but not endocytic,
vesicles…………………………………………………………………………………...………67
Figure 2.8. FL-PTPsigma colocalization with mRFP-LC3 and mock control for FL-PTPsigma
immunofluorescence…………………………………………………………………………..…69
Figure 2.9. PTPsigma dephosphorylates phosphotyrosine, but not PI(3)P, in vitro……………..71
Figure 2.10. PTPsigma potentially functions as a Vps34 effector…………………………...….93
Figure 2.11. Phosphotyrosine analyses following PTPsigma knockdown….………………...…95
Figure 2.12. In vitro kinase activity of functional Vps34 complexes …………………………..97
Figure 2.13. Proteolytic processing of PTPsigma………………………………………………..99
Figure 3.1 Workflow overview for PTPsigma inhibitor search……………………………...…116
Figure 3.2. In silico screen for compounds which dock into PTPsigma………………………..118
Figure 3.3. Primary in vitro screening of lead scaffolds filters for potency……………………120
Figure 3.4. In vitro screen of additional compounds with structural similarities to
the 4 leads………………………………………………………………………………………122
Figure 3.5. Optimization of in vitro screening conditions for selectivity analysis……………..124
ix

Figure 3.6. Relative inhibitions of PTPsigma and PTP1B by lead compounds………………..126
Figure 3.7. Methods for building specificity to small molecule inhibitors of PTPsigma…..….128
Figure 4.1. Overview of autophagy and key molecules involved in vesicle dynamics…….…..151
Figure 4.2. Experimental design for measuring GFP-LC3 vesicle dynamics………..…………153
Figure 4.3. Initial GFP-LC3 data collection……………………………………………………155
Figure 4.4. Model simulations after fitting of experimental data……………………………....157
Figure 4.5. Simulations take into account observed system noise…………………………...…159
Figure 4.6. Model prediction and test: Vps34 inhibition…………………………………...…..161
Figure 4.7. Model prediction and test: Atg9 depletion…………………………………………163
Figure 4.8. Model prediction: LC3 concentration and vesicle size………………………...…..165
Figure 5.1. Working model of PTPsigma function……………………………………………..182
Figure 5.2. Contact map of the mammalian autophagy network…………………….…………184

x

CHAPTER 1

Introduction

1

INTRODUCTION

Vps34 and phosphatidylinositol-3-phosphate (PI(3)P) signaling in Autophagy
Phosphoinositide-3-kinases (PI3Ks) comprise an important class of enzymes in cellular
signaling responsible for generating 3’-phosphorylated phosphoinositides, a group of diverse
lipid messengers. Mammalian PI3Ks are categorized into three classes - I, II, and III - based on
sequence and function. The class I PI3Ks are comprised of 110 kDa catalytic subunits and
regulatory subunits and function primarily in growth factor signaling through generation of
PI(3,4,5)P3. Class I PI3Ks are known to control cell survival, growth, and proliferation. As such,
the catalytic p110 subunits have been found to be amplified and activated in a number of
cancers, making these kinases the focus of intense research (Samuels, Wang et al. 2004;
Manning and Cantley 2007). The class II PI3Ks are homomeric enzymes that are quite large
multi-domain enzymes compared to class I PI3Ks. The class II PI3Ks enzyme function remains
to be elucidated, despite the defining features of a lipid kinase catalytic domain and a defined Cterminal motif (Domin, Pages et al. 1997).
The class III PI3K family is represented by a single enzyme in mammals, hVps34
(encoded by the PIK3C3 gene), which selectively generates PI(3)P on intracellular membranes.
This enzyme is considered the primordial PI3K as it is the only member found in yeast, where it
was initially identified (Herman and Emr 1990; Schu, Takegawa et al. 1993). Mutant Vps34
strains were characterized by defects in vacuolar protein sorting (Vps) and subsequently, it was
discovered that Vps34 harbors lipid kinase activity (Schu, Takegawa et al. 1993). Studies with
Vps34 yeast mutants and knockdown in mammalian systems have determined that Vps34catalyzed PI(3)P is critical for homotypic endosomal fusion, multivesicular body formation,
2

protein sorting, and receptor recycling (Herman and Emr 1990; Stack, Herman et al. 1993;
Simonsen, Lippe et al. 1998; Christoforidis, Miaczynska et al. 1999; Futter, Collinson et al.
2001). In support of these functions, a PI(3)P-specific lipid binding module, termed the FYVE
domain, has been characterized and found within several effector proteins with known endocytic
functions, for example, early endosomal autoantigen 1 (EEA1) (Gaullier, Simonsen et al. 1998).

Vps34 in Autophagy
Following establishment of Vps34 as a critical mediator of endocytosis, additional
evidence ignited interest in Vps34 as an integral component of an additional cellular process,
termed macroautophagy, which similarly involves dynamic membrane trafficking.
Macroautophagy (henceforth, autophagy) is a catabolic process in which portions of the cytosol,
including proteins and entire organelles, are encapsulated in double-membrane vesicles
(autophagosomes) and subsequently delivered to the lysosome for degradation (Klionsky 2007).
Sequestered material is broken down into basic biochemical building blocks, which are then
recycled and reused by the cell as an energetically favorable alternative to do novo synthesis.
Autophagy occurs constitutively in nearly all cells to maintain cellular homeostasis but is
dramatically activated in response to cellular stress, namely, nutrient starvation, where it can
function as a survival mechanism (Meijer and Codogno 2004).
Autophagy is executed in four stages: initiation, nucleation, maturation, and completion
(Figure 1.1). Nutrients and growth signaling activate mTORC1 (mammalian target of rapamycin
complex 1), the key controller of autophagy induction. While activating processes which
contribute to cell growth, proliferation, and survival during times of low stress and high nutrient
content, mTORC1 concurrently down regulates autophagy. This is accomplished through direct

3

inhibitory phosphorylation of an autophagy-initiating complex, ULK1-mAtg13-FIP200 (Ganley,
Lam du et al. 2009; Hosokawa, Hara et al. 2009; Jung, Jun et al. 2009). ULK1 (Unc-51-like
kinase 1; the mammalian homolog of Atg1) is a serine/threonine kinase required for autophagy.
Upon mTORC1 inhibition (i.e. during starvation), dephosphorylated ULK1 is liberated, and
functions to permits the nucleation of an isolation membrane, or phagophore (Ganley, Lam du et
al. 2009; Hosokawa, Hara et al. 2009; Jung, Jun et al. 2009). The synthesis of this cup-shaped
double-membrane structure is promoted in large part by Vps34 and its catalysis of PI(3)P. PI(3)P
decorates early autophagic membranes and serves as a subcellular tag, recruiting lipid-binding
effectors, such as DCFP1 (double FYVE domain–containing protein 1), and WIPI1 and WIPI2
(WD repeat domain phosphoinositide-interacting proteins 1 and 2) in mammals (ProikasCezanne, Ruckerbauer et al. 2007; Axe, Walker et al. 2008; Polson, de Lartigue et al.
2010).Vps34 activity is required for the downstream activation of Atg9 cycling, a process
whereby this transmembrane protein cycles from peripheral locations to the site of
autophagosome synthesis, putatively bringing lipids or membranes with it to build the vesicle
(Young, Chan et al. 2006; Webber and Tooze 2010).
Expansion of the phagophore and eventual closure into a mature autophagosome is
executed by two ubiquitin-like conjugation systems. The first conjugation involves covalent
binding of Atg12, a ubiquitin-like protein, to Atg5 and subsequent incorporation into a large
oligomer with Atg16L at the phagophore (Mizushima, Noda et al. 1998). The second system
involves the most well-known autophagy protein and classically used autophagosome-marker,
LC3 (Atg8 in yeast), which after processing by the Atg4 protease, becomes covalently attached
to phosphatidylethanolamine (PtdEns) on the autophagosome (Ichimura, Kirisako et al. 2000;
Kabeya, Mizushima et al. 2000; Kirisako, Ichimura et al. 2000). The location and allowance of

4

LC3 conjugation to the autophagosome is controlled by Atg5-Atg12-Atg16L, which functions as
an E3-like enzyme (Fujita, Itoh et al. 2008). The completion of this process is marked by the
fusion of the autophagosome directly with a lysosome (generating an autolysosome), or more
frequently, with an endosome destined for the lysosome (generating an amphisome), and
eventual degradation of sequestered cargo (Dunn 1990; Berg, Fengsrud et al. 1998; Klionsky
2007).
Vps34 was identified as part of the autophagic machinery in a study discovering that the
mechanism of action of the autophagy inhibitor, 3-methyladenine (3MA), was through selective
inhibition of class III PI3K activity (Petiot, Ogier-Denis et al. 2000). The 3MA-induced
autophagy blockade was rescued by feeding cells with synthetic PI(3)P, demonstrating the
importance of this lipid product in autophagy (Petiot, Ogier-Denis et al. 2000). In the years since
this observation, several key findings have validated a role for PI(3)P in autophagy: a) autophagy
is ablated in mutant Vps34 yeast strains and in cells from higher eukaryotes lacking Vps34
(Kihara, Noda et al. 2001; Juhasz, Hill et al. 2008), b) PI(3)P localizes to autophagic membranes
(Juhasz, Hill et al. 2008; Obara, Noda et al. 2008), and c) several autophagy proteins (notably,
Atg18) have been shown to bind PI(3)P (Proikas-Cezanne, Ruckerbauer et al. 2007; Axe, Walker
et al. 2008; Polson, de Lartigue et al. 2010). Further, defined complexes of Vps34 and its binding
partners have been shown to function in early autophagy (Sun, Fan et al. 2008; Matsunaga,
Saitoh et al. 2009; Zhong, Wang et al. 2009). While these advances have highlighted an absolute
requirement for PI(3)P in autophagy, little is known about the exact function PI(3)P serves in
autophagy nor its complex regulation.

5

Subcellular Complexes of Vps34
In sum, it is apparent that Vps34 serves several distinct roles in cellular functions. This
context specificity is controlled, at least in part, by the compartmentalization of Vps34 into
several distinct complexes. In yeast, Vps34 resides in two defined complexes: complex I which
regulates autophagy and complex II which functions in vacuolar protein sorting (Funderburk,
Wang et al. 2010). While both share a common core of Vps34, Vps15, and Atg6, the protein
sorting complex includes Vps38 and the autophagic complex includes Atg14 (Funderburk, Wang
et al. 2010).
In mammals, Vps34 similarly exists in two or more complexes but its association with
binding partners appears to be far more complex. The mammalian Vps34 core complex contains
Vps34, Vps15 (p150), and Beclin1 (coiled-coil myosin-like BCL2-interacting protein; Atg6
homolog). This core is recruited to nascent autophagic membranes via an interaction with
Atg14L (Barkor), the functional ortholog of yeast Atg14 (Sun, Fan et al. 2008; Matsunaga,
Saitoh et al. 2009; Zhong, Wang et al. 2009). Atg14L has a novel hydrophobic motif, termed
BATS (Barkor/Atg14L autophagosome targeting sequence), which directly binds to PI(3)Ppositive membranes and contributes to the tethering of this subcomplex to the phagophore (Fan,
Nassiri et al. 2011). Expression of Atg14L is stabilized by Beclin1, to which it directly binds,
and together, they positively influence Vps34 kinase activity (Zhong, Wang et al. 2009). The
origin of autophagosomal membranes is debated; however, growing evidence supports the
existence of “omegasomes”, cup-shaped platforms which support the assembly of
autophagosome machinery (Axe, Walker et al. 2008). Analysis by 3D-electron tomography has
shown that these omegasome structures are cradled by portions of rough endoplasmic reticulum
(ER) and contribute to the nucleation and dispersal of autophagosomes (Hayashi-Nishino, Fujita

6

et al. 2009). An elegant live-cell microscopy-based investigation of the PI(3)P-binding protein,
DFCP1, and its intriguing starvation-responsive ER localization led to this discovery (Axe,
Walker et al. 2008).
In addition to the Atg14L-containing complex, a mutually exclusive Vps34 complex
resides on endocytic compartments and is distinguished by the inclusion of UVRAG (UV
radiation resistance-associated gene protein; homolog to yeast Vps38), which competes with
Atg14L for binding to Beclin1 (Funderburk, Wang et al. 2010) . This complex, more abundant
than the Atg14L complex, has been reported to promote the maturation of both endosomes and
autophagosomes, although its exact contributions to these two processes is again debated (Liang,
Feng et al. 2006; Itakura, Kishi et al. 2008; Matsunaga, Saitoh et al. 2009; Zhong, Wang et al.
2009). Mechanistically, UVRAG functions in several ways. On the most fundamental level, it
promotes the catalytic activity of Vps34 (Liang, Feng et al. 2006). Secondly, UVRAG interacts
with the class C-VPS/HOPS complex (via the Vps16 subunit), a guanine nucleotide exchange
factor (GEF) for the late endosomal GTPase, Rab7 (Liang, Feng et al. 2006; Liang, Lee et al.
2008). This interaction stimulates activity of the complex, leading to GTP-loading of Rab7, and
enhanced endosomal maturation (Liang, Lee et al. 2008).
The Vps34-UVRAG complex can associate with an additional regulator, Rubicon (RUNdomain protein as Beclin1-interacting and cysteine-rich containing). Recently discovered in two
independent screens for Beclin1-interactors, Rubicon exists only in higher eukaryotes and
uniquely, functions as an inhibitor of endocytic and autophagic maturation (Matsunaga, Saitoh et
al. 2009; Zhong, Wang et al. 2009). Rubicon-mediated inhibition is achieved through two
mechanisms: the direct binding and suppression of Vps34 activity and the competitive
sequestration of UVRAG from C-VPS/HOPS. A direct interaction of Rubicon’s RUN domain

7

with the kinase domain of Vps34 was recently described and found to suppress PI(3)P
production (Sun, Zhang et al. 2011). In addition, Rubicon depletion enhances UVRAG binding
to C-VPS/HOPS, facilitates Rab7 activity, and increases endocytic maturation (Sun, Westphal et
al. 2010). Interestingly, a feed-forward loop was revealed whereby activated Rab7 sequesters
Rubicon from UVRAG, thus liberating UVRAG to continually engage C-VPS/HOPS, and
promote further loading of GTP to Rab7 (Sun, Westphal et al. 2010).
In addition to the control of Vps34 function conferred by the existence of unique
subcomplexes, it is also plausible that phosphatases, which antagonize Vps34-catalyzed PI(3)P
production, aid in the exquisite control of cellular PI(3)P and its specific cellular functions
(Zeng, Overmeyer et al. 2006). This latter concept is the focus of the research presented here.

Protein tyrosine phosphatase receptor-type sigma (PTPsigma/PTPRS)
It has been estimated that 30% of the human proteome is subject to phosphorylation
(Tautz, Pellecchia et al. 2006). Phosphorylation is a post-translational modification balanced by
the actions of kinases, which catalyze the addition of phosphate moieties to specific amino acids,
and phosphatases, which catalyze their removal (Tautz, Pellecchia et al. 2006). Tyrosine
phosphorylation accounts for less than 0.1% of the total phosphorylation in mammalian cells;
however, it represents a critical regulatory mechanism in signal transduction (Tautz, Pellecchia et
al. 2006). By actively controlling the level of phosphorylation in cells, protein tyrosine
phosphatases (PTPs) serve fundamental roles in signal transduction, making them vital to most
cellular programs.
Of the 107 human PTPs, 38 members comprise the classic PTP family and exhibit strict
specificity for tyrosine residues (Andersen, Mortensen et al. 2001). A subtype of this family
includes the dual-domain receptor-like PTPs which express at the cell surface and contain two
8

cytosolic active sites in tandem, termed D1 and D2 ((Andersen, Mortensen et al. 2001); Figure
1.2A)). The membrane-proximal D1 domains have robust catalytic activity, while the D2
domains are generally inactive. It has been postulated that the D2 domains may regulate the
activity, stability, or substrate-specificity of the D1 domains (Barr, Ugochukwu et al. 2009).
Three members of this subtype constitute the LAR family, including the ancestral LAR
(PTPRF), PTPdelta (PTPRD), and PTPsigma (PTPRS). These three phosphatases are
distinguished structurally, by their large ectodomains containing immunoglobulin and fibronectin
repeats, and functionally, by their prominent role in homeostasis and neuronal development
((Andersen, Mortensen et al. 2001); Figure 1.2A)).

PTPsigma in Neurons and Development
An original role for the LAR family in neuronal function came from the discovery that
expression of a Drosophila ortholog, DLAR, was required for the proper guidance of motor and
photoreceptor axons as well as formation of synapses in flies (Krueger, Van Vactor et al. 1996).
This finding instigated research and eventual confirmation of a similar function of vertebrate
family members in nervous system development.
While it is apparent that LAR, PTPdelta, and PTPsigma all play important roles in
neurons, they do so in both overlapping and distinct manners as evidenced in animal models
targeting each of these genes. Mild abnormalities were observed in a murine model essentially
lacking LAR expression (generated by gene trapping), including hippocampal and cholinergic
defects (Yeo, Yang et al. 1997; Van Lieshout, Van der Heijden et al. 2001). Loss of PTPsigma or
PTPdelta, however, resulted in more pronounced developmental defects. Over half of Ptprd

- /-

mice died within a few weeks of birth as a result of starvation, and those surviving to adulthood
9

exhibited growth retardation and defects in learning and memory (Uetani, Kato et al. 2000).
Sixty percent of Ptprs

-/-

mice died as neonates within two days of birth, and few survived longer

than three weeks (Elchebly, Wagner et al. 1999; Wallace, Batt et al. 1999). Major abnormalities
included neuroendocrine dysplasia, decreased brain size, structural irregularities of the central
nervous system, growth retardation, wasting, spasms, and abnormal limb flexion (Elchebly,
Wagner et al. 1999; Wallace, Batt et al. 1999). In addition to independent functions, PTPsigma
and PTPdelta likely fulfill at least some complementary roles because deletion of both caused
more pronounced developmental defects. All double Ptprs

-/-

/ Ptprd

-/-

mice died immediately at

birth, showing evidence of respiratory failure and paralysis (Uetani, Chagnon et al. 2006). This
severe phenotype most likely stems from failed motor neuron targeting late in development
(Uetani, Chagnon et al. 2006).
In support of its neuronal function, PTPsigma has been characterized as an important
mediator of neurite outgrowth and axon guidance. In many contexts, PTPsigma has been shown
to inhibit neurite outgrowth. First, Ptprs

-/-

mice showed enhanced nerve regeneration following

injury in several models including sciatic nerve crush, optical nerve crush, and facial nerve crush
(McLean, Batt et al. 2002; Thompson, Uetani et al. 2003; Sapieha, Duplan et al. 2005).
Similarly, dorsal root ganglion neurons from Ptprs

-/-

mice exhibited enhanced outgrowth, an

effect mediated by N-cadherin (Siu, Fladd et al. 2007). In agreement with this inhibitory role, it
was shown that overexpression of PTPsigma could suppress neurite outgrowth in primary
sensory neurons (Faux, Hawadle et al. 2007; Siu, Fladd et al. 2007). Most recently, it was
reported that loss of PTPsigma promotes nerve regeneration following spinal cord injury (SCI)
and that the interaction of the ectodomain of PTPsigma with chondroitin sulfate proteoglycans
10

(CSPGs) released at the site of injury, normally inhibits this process (Shen, Tenney et al. 2009;
Fry, Chagnon et al. 2010).
There has also been evidence to the contrary that PTPsigma may function as a positive
regulator in axon outgrowth. First, it was shown that PTPsigma promotes axon growth in retinal
ganglion cells of chicks and that this involved an interaction with heparin sulfate proteoglycans
(HSPGs) (Ledig, Haj et al. 1999; Aricescu, McKinnell et al. 2002; Rashid-Doubell, McKinnell et
al. 2002). In fact, a substrate of purified PTPsigma ectodomains could directly support the
growth of these retinal axons (Sajnani, Aricescu et al. 2005). A study combining crystallography
and colocalization data reconciled these findings, demonstrating that CSPGs and HSPGs
compete for occupation of a shared binding site in the ectodomain of PTPsigma to inhibit or
promote outgrowth, respectively (Coles, Shen et al. 2011).

PTPsigma – Cytosolic Functions and Proteolytic Processing
While the function of the ectodomain of PTPsigma and its interactions with ligands has
been well described in the control of axon growth, far less has been established concerning the
cellular mechanisms by which PTPsigma functions, and more elusive the identity of its cytosolic
substrates. N-Cadherin and neurotrophin receptors (TrkA, TrkB, TrkC) are the only substrates
demonstrated to be directly dephosphorylated by PTPsigma and also supported in vivo (Faux,
Hawadle et al. 2007; Siu, Fladd et al. 2007).
An important feature of regulation for PTPsigma and its relatives involves proteolytic
processing away from the cell surface (Figure 1.2B). PTPsigma is translated as a pro-protein and
undergoes Furin-mediated cleavage upstream of its transmembrane domain while processed
through the Golgi. PTPsigma is then destined to the cell surface where it exists as two non-

11

covalently associated subunits. The extracellular E-subunit is shed during a process called
ectodomain-shedding. For PTPsigma, this was shown to occur in cells in response to high cell
confluence, calcium influx, and treatment with the phorbol ester, TPA (12-Otetradecanoylphorbol-13-acetate) (Aicher, Lerch et al. 1997). Ectodomain shedding is mediated
by extracellular metalloproteases, and is triggered by internalization of a membrane-tethered Cterminal fragment (CTF). While the target of the internalized catalytic domains has not been
identified, in the case of PTPsigma, the internalized catalytic domains appear as punctate
cytosolic aggregates (Aicher, Lerch et al. 1997). A final proteolytic processing event, evidenced
for LAR, involves gamma-secretase-catalyzed intramembrane cleavage, leading to the
generation of a soluble intracellular domain (ICD) that is targeted inside the cell (Ruhe, Streit et
al. 2006; Haapasalo, Kim et al. 2007). PTPsigma contains similar cleavage residues to LAR,
making it therefore plausible that PTPsigma is also processed into an ICD (Figure 1.2).

Computational Models of Signal Transduction
Systems Biology – a Shifting Paradigm in Biology
Systems Biology is an interdisciplinary field founded on the principle that complex
physiological networks can be best understood through a holistic integration of numerous
scientific approaches (Kitano 2002; Kitano 2002; Kirschner 2005). When processes are analyzed
in their entirety, through the complementation of experimentation and computational methods,
emergent properties can be revealed which would not have been detected if individual
components were studied in isolation (Janes and Yaffe 2006). Integral to the success of this field
is the acquisition of large, detailed, high-quality data sets and the ability to analyze, interpret,
apply, and model them to reveal higher-order behaviors of systems (Kitano 2002).

12

The study of signal transduction pathways is of interest to the Systems Biology field.
Signaling pathways are comprised of numerous molecules (i.e. proteins, lipids, and metabolites)
which interact dynamically in a spatial and temporal manner to control cellular programs and
more broadly, network (e.g. immune system) and organismal behaviors. The large number of
molecules involved and the infinite interactions that are possible necessitate the use of
computational or mathematical models to comprehensively understand them (Hlavacek, Faeder
et al. 2006). To date, models have been generated to describe signaling of EGFR, immune
system receptors, and NF-κB, among others (Wiley, Shvartsman et al. 2003; Goldstein, Faeder et
al. 2004; Tay, Hughey et al. 2010).

Conventional Modeling
The most basic approach to modeling a signaling pathway is through conventional model
specification, that is, the description of molecules making up a system and the nature of their
relationships to one another written as linear chemical reactions. Differential equations
(generally ordinary; ODEs) are defined for each of these reaction schemes (Hlavacek, Faeder et
al. 2006). Standard computing methods can be used to diagram a pathway, define equations, and
calculate a model (Hlavacek, Faeder et al. 2006). These types of conventional models can be
used to describe simple processes when the number of molecules and reactions is manageable.
Simple models have been utilized to study cell signaling and are useful in identifying important
design principles behind cellular processes (Kholodenko, Demin et al. 1999; Kholodenko,
Hancock et al. 2010; Tay, Hughey et al. 2010).

13

Rule-Based Modeling
The need for a new type of a model manifests when considering a large number of
proteins, all with numerous binding sites, modifiable residues, and activities. The potential
number of interactions, modifications, and reaction consequences in this case is difficult to
comprehend and results in “combinatorial complexity” (Endy and Brent 2001; Blinov, Faeder et
al. 2006). A new paradigm in modeling, the use of rule-based models, was created to overcome
the inability of conventional mechanics to handle this complexity (Faeder, Blinov et al. 2005).
Here, rules are written to specify protein-protein interactions and the activities of molecular
species, or agents. A simple rule would be “The SH2 domain of Protein A binds tyrosine residue
100 of Protein B whenever tyrosine residue 100 is phosphorylated.” When written in welldefined machine-readable format, such as BioNetGen Language (BNGL), rules can be
interpreted and used to generate chemical reactions and new species based on initial starting
conditions (Blinov, Faeder et al. 2004; Hlavacek, Faeder et al. 2006). An example of a rule-based
model accounting for complexity overlooked in a conventional model was reported for EGFR
signaling by Blinov and colleagues (Blinov, Faeder et al. 2006).

Utility of Mathematical Models
Regardless of the methodology chosen for model specification, the aim should be to build
a model which accurately reproduces biological observations and subsequently, allows for the
testing of biologically-derived predictions and further, generates new, otherwise unpredictable,
ideas about the framework or function of a system (Hlavacek, Faeder et al. 2006). As human
diseases are often hallmarked by alterations in signaling, it is this predictive and testable power
that will provide utility in disease research and the development of therapeutics (Hopkins 2008).

14

Considerable benefit will be yielded by a model in which the physiological response to very
specific changes (i.e. genetic alterations or treatment with targeted-compounds) can be
accurately predicted (Hopkins 2008).

15

Rationale for this Study
This project originated as an effort to identify phosphatases controlling Vps34-PI(3)P
signaling, especially those which function in autophagy. Through a loss-of-function screen of all
human phosphatase genes, we identified PTPsigma as a potential inhibitor of this signaling axis.
Further, RNAi-mediated knockdown of PTPsigma was reported to confer chemoresistance to
HeLa cells in culture and PTPsigma expression was decreased significantly in a study of
metastatic prostate cancer tissue samples (MacKeigan, Murphy et al. 2005; Tomlins, Mehra et al.
2007). Although there was no precedence for PTPsigma to have a role in autophagy, the robust
phenotype in conjunction with potential disease relevance, warranted follow-up investigation.
We aimed to investigate the potential novel role for PTPsigma in PI(3)P signaling and autophagy
using an integrative approach combining reductionist cell biology and biochemical techniques
along with a systems-level mathematical modeling approach.

Specific Aim 1. To characterize the regulation of PI(3)P by PTPsigma.
Hypothesis: PTPsigma regulates cellular PI(3)P through direct dephosphorylation.
We discovered that RNAi-mediated knockdown of PTPsigma increased cellular PI(3)P
levels by an unknown mechanism. Structural modeling demonstrated that the membraneproximal catalytic domain of PTPsigma bears a unique conformation allowing
accommodation of PI(3)P, suggesting this lipid may serve as a direct substrate.
Accordingly, our aim was to characterize the control PTPsigma exerts on PI(3)P in cells
and further characterize the biochemical mechanism by which PTPsigma regulates this
lipid using in vitro approaches.

16

Specific Aim 2. To establish a role for PTPsigma in autophagy and chemoresistance.
Hypothesis: PTPsigma functions as a negative regulator of autophagy acting on
autophagic vesicles and consequently, loss of PTPsigma drives chemoresistance through
enhanced autophagy-mediated cell survival.
We found that upon loss of PTPsigma, cells accumulated abundant PI(3)P-positive
vesicles which appeared autophagic. It has been shown that PTPsigma knockdown also
confers chemoresistance to cancer cells in culture and its expression is lost in metastatic
prostate cancer (MacKeigan, Murphy et al. 2005; Tomlins, Mehra et al. 2007). In this
aim, we intended to a) establish the role of PTPsigma as a negative regulator of
autophagy, b) determine the processing events which target PTPsigma to PI(3)P-positive
autophagic membranes, and c) demonstrate that enhanced autophagy, as in the absence of
PTPsigma, confers a survival advantage to cells subjected to chemotherapy. We further
performed in silico-based identification of small molecule inhibitors of PTPsigma.

Specific Aim 3. To develop a data-driven mathematical model of autophagy.
Hypothesis: A data-driven model of autophagy will simulate and accurately predict
perturbations to key autophagic machinery (Vps34 inhibition and loss of Atg9).
Autophagy is a central regulator of cellular function and has emerging implications in
human disease. The elucidation of complex signaling processes, such as this, can be aided
through the generation of computational models. We aimed to create a model, driven
primarily by kinetic cell-based data, which would accurately simulate autophagic vesicle
dynamics and allow for the generation and testing of novel predictions. Constructed
initially as a simple model and described with conventional mechanics and stochastic

17

simulations, it will serve as the foundation for a comprehensive, rule-based model in the
future.

The results of the cell-based interrogation of PTPsigma and its regulation of PI(3)P and
autophagy are summarized in Chapter 2. The in silico-based identification of small molecule
inhibitors of PTPsigma is discussed in Chapter 3. Finally, the mathematical model constructed to
understand mammalian autophagy dynamics is outlined in Chapter 4. Chapter 5 concludes with
both an overreaching summary and discussion of future directions of these projects.

18

FIGURES

Figure 1.1. Overview of autophagy. Autophagy is executed in four stages: 1) mTORC1
controls autophagy initiation through inhibition of the ULK1/Atg13/FIP200 complex; 2) ULK1
activity permits nucleation of the double-membrane phagophore which is largely executed by the
Vps34 complex, PI(3)P-binding effectors (i.e. WIPI proteins), and the transmembrane protein,
Atg9; 3) membrane maturation into an enclosed autophagosome is accomplished by two
ubiquitin-like conjugation events involving LC3 and Atg5-Atg12-Atg16; 4) autophagy is
completed via degradation when the autophagosome fuses with a lysosome to form an
autolysosome or alternatively, when it first fuses with an endosome to form an amphisome
intermediate organelle that subsequently fuses with the lysosome. For interpretation of the
references to color in this and all other figures, the reader is referred to the electronic version of
this dissertation.

19

Figure 1.1 (cont'd)

UVRAG
Beclin1

mTORC1

Vps34

Vps15

Rubicon

1. Initiation
Atg13
ULK1 FIP200

PI3

proLC3

P

Atg4
LC3-I

PE

Vps34

LC3-II

endosome

PE

Beclin1

2. Nucleation

Atg7/3
Vps15

Atg14

Atg5

3P

Atg16

PI

phagophore

12

At
g

Atg9

At

I1

Atg4

g5

PI3

g
At

P

16

Atg12

WIP

LC3

-II

Atg23

LC3-I

Atg7/10
Atg12

autophagosome
3. Maturation

20

autolysosome
4. Degradation

Figure 1.2. Structure and proteolytic processing of PTPsigma. (A) PTPsigma consists of an
N-terminal ectodomain made of (3) immunoglobulin (Ig)-like repeats and fibronectin type IIIlike repeats (up to 8 in the longest isoform). A single transmembrane domain precedes two Cterminal cytosol-facing phosphatase domains, termed D1 and D2. (B) PTPsigma is generated as
a large pro-protein and following furin-cleavage within the Golgi, is exported to the cell surface
as two non-covalently linked subunits (E and P). Cleavage by extracellular metalloproteases
results in ectodomain shedding and generation of a membrane-tethered C-terminal fragment
(CTF) which can be internalized to the cell. Potential further processing by presenilin (PS) and
gamma secretase is predicted to generate a intracellular domain (ICD) which is liberated from
the membrane and may also be targeted inside the cell.

21

Figure 1.2 (cont'd)

A
D1

N
Ig-like (3)

B

FNIII (8)

D2

C

TM PTP domains

E-subunit

Ectodomain
Shedding
Metalloprotease
(ADAM/TACE)

extracellular
space

D1

D1

D2

D2

P-Subunit

C-terminal
Fragment
(CTF)

Furin

cytosol
γ-secretase/PS
(putative)

D1
D2
Intracellular
Domain (ICD)

D1
D2
Pro-Protein

22

CHAPTER 2

Characterization of PTPsigma as a novel regulator of Vps34-PI(3)P signaling and
autophagy

23

CHAPTER 2 SECTION I

Identification of PTPsigma as an autophagic phosphatase

Modified from
Martin KR, Xu Y, Looyenga BD, Davis RJ, Wu CL, Tremblay ML, Xu HE, Mackeigan JP
(2011). Identification of PTPsigma as an autophagic phosphatase. J Cell Sci. 2011 Mar 1;124(Pt
5):812-9.

24

ABSTRACT

Macroautophagy is a dynamic process whereby portions of the cytosol are encapsulated
in double-membrane vesicles and delivered to the lysosome for degradation.
Phosphatidylinositol-3-phosphate (PI(3)P) is concentrated on autophagic vesicles and recruits
effector proteins critical for this process. The production of PI(3)P by the class III
phosphatidylinositol 3-kinase (PI3K), Vps34, has been well established; however, protein
phosphatases which antagonize this early step in autophagy remain to be identified. To identify
such enzymes, we screened human phosphatase genes by RNA interference (RNAi) and found
that loss of PTPsigma, a dual-domain protein tyrosine phosphatase (PTP), increases cellular
PI(3)P. The abundant PI(3)P-positive vesicles conferred by PTPsigma loss strikingly
phenocopied those observed in amino acid-starved cells. Accordingly, we discovered that loss of
PTPsigma hyperactivates both constitutive and induced autophagy. Finally, we found that
PTPsigma localizes to PI(3)P-positive membranes in cells and this vesicular localization is
enhanced during autophagy. Our findings propose a novel role for PTPsigma and provide insight
into the regulation of autophagy. Mechanistic knowledge of this process is critical for
understanding and targeting therapies for several human diseases, including cancer and
Alzheimer’s disease, in which abnormal autophagy may be pathological.

25

INTRODUCTION

In addition to the well-characterized role of PI(3)P in endocytosis, recent evidence has
uncovered a critical requirement for this lipid in macroautophagy (autophagy) (Petiot, OgierDenis et al. 2000; Axe, Walker et al. 2008; Obara, Noda et al. 2008). Autophagy occurs
constitutively in nearly all cells to maintain homeostasis, but is further activated in response to
stress as a survival or adaptive mechanism. During autophagy, a double-membrane phagophore
forms and elongates around portions of cytosol, matures into an enclosed autophagosome, and
delivers its contents to the lysosome for degradation (Klionsky 2007). Basic biochemical
components (i.e. amino acids and fatty acids) are exported from the lysosome and recycled by
the cell, representing an energetically favorable alternative to de novo synthesis. In mammalian
systems, the lipid kinase Vps34 forms a complex with several proteins including Vps15, Beclin1,
Atg14L, UVRAG, and Bif1 to generate PI(3)P on autophagic vesicles (Itakura, Kishi et al. 2008;
Zhong, Wang et al. 2009). PI(3)P then recruits and tethers effector proteins, such as WIPI-1
(Atg18), which are required for proper membrane formation (Proikas-Cezanne, Waddell et al.
2004; Obara, Sekito et al. 2008). The critical requirement for PI(3)P in this process is evidenced
by the fact that autophagy is ablated in mutant Vps34 yeast strains and genetic Vps34 knockouts
in Drosophila (Kihara, Noda et al. 2001; Juhasz, Hill et al. 2008). Despite knowledge of PI(3)P
production, the antagonistic phosphatases which regulate PI(3)P during autophagy have
remained elusive. Two myotubularin-related phosphatases, MTMR3 and MTMR14 (hJumpy),
have recently been shown to dephosphorylate autophagic PI(3)P in various contexts (Vergne,
Roberts et al. 2009; Taguchi-Atarashi, Hamasaki et al. 2010). However, given the complexity of
autophagy execution and the number of proteins in the autophagy network, we predict that

26

additional protein phosphatases exist to regulate this process. Accordingly, we performed a highcontent cell-based RNAi screen using a fluorescent PI(3)P sensor to identify protein
phosphatases that function upstream of PI(3)P during autophagy.

27

RESULTS

RNAi screen identifies PTPsigma as a modulator of PI(3)P signaling
FYVE (Fab1, YOTB, Vac1, and EEA1) domains are cysteine-rich zinc-finger binding
motifs that specifically recognize and bind PI(3)P (Gaullier, Simonsen et al. 1998). An EGFP
molecule fused to two tandem FYVE domains, termed EGFP-2xFYVE, serves as an effective
cellular sensor of PI(3)P (Gillooly, Morrow et al. 2000). We analyzed U2OS cells stably
expressing this construct by fluorescent microscopy and found that PI(3)P predominantly
localized to punctate, often perinuclear, vesicles when cultured in complete growth media with
full nutrients (Figure 2.1A). RNAi-mediated knockdown of Vps34 reduced cellular PI(3)P
content and resulted in a diffuse cytosolic distribution of EGFP-2xFYVE (Figure 2.1B,F; Figure
2.6A). In contrast, a redistribution of EGFP-2xFYVE to small abundant autophagic vesicles
occurred when cells were deprived of amino acids to potently induce autophagy (Figure 2.1C).
To identify genes that down-regulate PI(3)P signaling, we utilized multiple siRNAs
targeting over 200 known and putative human phosphatases. The siRNAs were introduced into
U2OS-EGFP-2xFYVE cells, and the cells were subsequently monitored for PI(3)P signaling.
We identified several genes whose knockdown significantly increased cellular EGFP-2xFYVE
punctae abundance (Figure 2.1E, Table 2.1). Most notably, PI(3)P was substantially increased
following knockdown of the myotubularin family member, MTMR6 (Figure 2.6B,C), as well as
the dual-domain PTP, PTPRS (PTPsigma) (Figure 2.1D,E). While reduced expression of
MTMR6 was characterized by the appearance of enlarged perinuclear vesicles, the siRNAs
targeting PTPsigma caused a dramatic accumulation of abundant smaller vesicles throughout the
cytosol which phenocopied that observed during amino acid-starvation (Figure 2.1C,D).

28

Quantification of PI(3)P-positive vesicles revealed a 3.5-fold increase in abundance during
starvation-induced autophagy and a nearly 5-fold increase caused by PTPsigma knockdown
alone (Figure 2.1F). This phenotype was confirmed using four unique siRNA sequences
targeting PTPsigma (Figure 2.6D-K).
To validate a physiological increase in PI(3)P following knockdown of PTPsigma,
phospholipids were radiolabeled with

32

P-orthophosphate in vivo, extracted, and resolved by

thin layer chromatography. Indeed, PI(3)P levels were specifically elevated in the absence of
PTPsigma, while other lipid species remained unchanged (Figure 2.1D). In order to determine
the identity of the PI(3)P-positive vesicles formed by PTPsigma knockdown, we immunostained
cells with well-established markers of early endosomes (anti-EEA1 [early endosomal antigen 1])
and autophagic vesicles (AVs) (anti-LC3 [light chain 3]). We found that knockdown of
PTPsigma had no substantial effect on the presence of EEA1-positive endosomes (Figure 2.1H;
Figure 2.7A), but significantly increased the abundance of LC3-positive vesicles (Figure 2.1I).
From this, we hypothesized that PTPsigma functions during autophagy and focused our attention
on this enzyme as a candidate autophagic phosphatase.

Loss of PTPsigma hyperactivates constitutive and induced autophagy
The striking resemblance of PI(3)P-positive vesicles induced by PTPsigma knockdown to
AVs formed during amino acid-starvation led us to propose that autophagy is hyperactivated in
the absence of PTPsigma, despite the presence of nutrients. To test this, autophagy was analyzed
in U2OS cells by evaluating two ubiquitin-like proteins, Atg12 and LC3 (Atg8), which become
conjugated to AVs during autophagy. Following phagophore nucleation, Atg12 covalently binds
Atg5 and oligomerizes with Atg16L at the autophagic membrane (Klionsky 2007). To measure
29

vesicle abundance at this step, we immunostained cells for endogenous Atg12 and quantified
Atg12-positive punctae. We found that PTPsigma knockdown increased AV abundance 3- to 5fold from control when cells were cultured with rapamycin, a potent mTOR inhibitor and
autophagy inducer, or full nutrients, respectively (Figure 2.2A; Figure 2.7B).
The membrane-bound Atg5/12/16L complex permits lipidation of LC3, a classic marker
for AVs (Hanada, Noda et al. 2007). LC3 is unique among the autophagy proteins in that a
portion of it remains membrane-bound and is degraded in the lysosome along with vesicle cargo.
Therefore, lysosomal function can be inhibited [i.e. with bafilomycin A1 (Baf-A1) or
chloroquine treatment] and LC3 accumulation used as a measure of autophagic flux (Tanida,
Minematsu-Ikeguchi et al. 2005). We found that both PTPsigma knockdown and amino acidstarvation increased the abundance of LC3-II, the AV-lipidated form of LC3, when lysates were
probed with endogenous antibodies (Figure 2.2B). Similarly, we observed an increased number
of EGFP-LC3-positive punctae when PTPsigma expression was reduced under normal growth
conditions, and these structures accumulated substantially when cells were cultured with Baf-A1,
indicating their functionality (Figure 2.2C-F). PTPsigma knockdown caused an even greater
increase in EGFP-LC3 punctae from control when cells were treated with both Baf-A1 and
rapamycin (Figure 2.2G,H). Similar results were obtained when AVs were quantified from cells
immunostained for endogenous LC3 (Figure 2.7C).
To confirm hyperactive autophagy in the absence of PTPsigma independently of
fluorescent markers, we detected AVs by transmission electron microscopy (TEM).
Autophagosomes are hallmarked by unique double-membranes and by the presence of engulfed
cytosolic content- features which allow them to be detected by TEM. These vesicles fuse with
the lysosomes or with endocytic compartments destined for the lysosome, generating degradative

30

autolysosomal vesicles which can also be observed by TEM. While control cells contained very
few AVs, chloroquine treatment increased their abundance, most of which appeared to be
autolysosomal as expected (Figure 2.3A,B). Similarly, cells deprived of amino acids for one hour
harbored elevated numbers of AVs, as did cells transfected with PTPsigma siRNAs but cultured
in full nutrients (Figure 2.3C,D). These AVs generally appeared to be later stage degradative
structures, consistent with active flux through the pathway. To establish this phenotype
independent of RNAi, we examined autophagy during PTPsigma loss using wild-type (Ptprs
and PTPsigma knockout (Ptprs
generated Ptprs

–/–

+/+

)

–/–

) murine embryonic fibroblasts (MEFs). We have previously

mice by inserting a selectable neomycin resistance gene into the D1

phosphatase (catalytic) domain. From these mice, we generated MEFs that lack both Ptprs
transcript and protein, as measured by northern blot and western blot, respectively (Elchebly,
Wagner et al. 1999). TEM analysis showed that both Ptprs

+/+

and Ptprs

–/–

primary MEFs

contained a basal level of AVs; however, they were twice as abundant in Ptprs

–/–

MEFs (Figure

2.3E-G). Again, most structures appeared to be degradative late stage vesicles. Ptprs

–/–

MEFs

also contained a number of membrane whorls which may or may not be related to an autophagy
phenotype. Collectively, these results demonstrate that loss of PTPsigma, by RNAi and genetic
deletion, increases both constitutive and induced autophagy.

PTPsigma localizes to PI(3)P-positive vesicles and rescues the siRNA phenotype
Given the robust PI(3)P response elicited by PTPsigma knockdown, we hypothesized that
PTPsigma may regulate autophagy by functioning at the level of PI(3)P-positive vesicles.

31

PTPsigma is a bulky receptor-like PTP with an extracellular segment and two tandem cytosolic
phosphatase domains (termed D1 and D2). Complex processing events have been reported for
PTPsigma and related enzymes, including ectodomain shedding and internalization from the cell
surface (Aicher, Lerch et al. 1997; Ruhe, Streit et al. 2006). In order to determine the localization
of PTPsigma phosphatase domains, untagged full-length protein (FL-PTPsigma) was transiently
expressed in U2OS-EGFP-2xFYVE cells and detected by fluorescent microscopy using D1targeted monoclonal antibodies. We found that in addition to its presence at the plasma
membrane, PTPsigma localized to the perinuclear region and to numerous intracellular vesiclelike structures, many of which were PI(3)P-positive (Figure 2.4A). Strikingly, autophagy
induction by amino acid-starvation induced a redistribution of PTPsigma to smaller vesicles
which were abundant throughout the cytosol and were almost entirely PI(3)P-positive (Figure
2.4B,C). In support of the notion that this localization was autophagic, we discovered that
PTPsigma was capable of localizing to mRFP-LC3-positive punctae in the context of both basal
and induced autophagy as well (Figure 2.8B).
We further used exogenous PTPsigma expression in an RNAi rescue experiment to
demonstrate the specificity of the PTPRS siRNA-induced phenotype. The naturally-occurring
isoform of PTPsigma used in our studies lacks the fourth through seventh fibronectin domains
(present in the longest isoform): a region encompassing the sequence targeted by a potent siRNA
(siRNA-1; see Figure 2.6E,J,K) (Pulido, Serra-Pages et al. 1995). The accumulation of small,
abundant, non-perinuclear PI(3)P-positive vesicles induced by siRNA transfection was rescued
by exogenous expression of PTPsigma, an effect which was dose-dependent (Figure 2.4D,E).
This result suggests a target-specific effect of siRNA-mediated PTPsigma knockdown and a role
for this enzyme in PI(3)P signaling.

32

To verify that the PTPsigma-positive punctate structures were in fact vesicles functioning
in a lysosomal pathway, we monitored PTPsigma localization in Baf-A1-treated cells. Baf-A1
prevents maturing vesicles (e.g. endocytic and autophagic) from fusing with lysosomes and in
turn, they accumulate in a perinuclear region. We found that PTPsigma-positive vesicular
structures began to accumulate quickly (within 15 minutes) and densely populated the
perinuclear region within several hours (Figure 2.4F). This suggests that PTPsigma normally
localizes to vesicles destined for the lysosome.
Finally, to determine if PTPsigma functions upstream or downstream of PI(3)P at the
starvation-induced punctae, we analyzed its localization in cells depleted of the phospholipid.
Autophagy was induced by amino acid-starvation in cells treated with wortmannin, a potent and
irreversible inhibitor of Vps34 and other PI3K isoforms, or vehicle (DMSO). In vehicle-treated
cells, starvation induced the formation of abundant PI(3)P-positive vesicles which also contained
PTPsigma (Figure 2.5A). Conversely, wortmannin treatment essentially ablated the formation of
PI(3)P during starvation; however, PTPsigma was still recruited to the abundant punctate
structures (Figure 2.5B). This finding suggests that the localization of PTPsigma to intracellular
structures formed during autophagy occurs upstream, or independently, of PI(3)P.

33

DISCUSSION

Through use of a high-content cell-based RNAi screen, we have identified phosphatases
whose knockdown elevates cellular PI(3)P. Notably, RNAi-mediated knockdown of MTMR6
and several other phosphatases resulted in swollen and often perinuclear PI(3)P-positive vesicles.
Previous studies have shown similar phenotypes when endocytic PI(3)P is elevated, for example,
by constitutive activation of early endosomal Rab5, or knockdown of the PI5-kinase, PIKfyve
(Murray, Panaretou et al. 2002; Rutherford, Traer et al. 2006). Accordingly, the vesicles
observed following knockdown of these phosphatases are likely endosomal and these enzymes,
including MTMR6, may function in endocytic signaling. Of note, knockdown of both PTPN11
(SHP2) and PTPN13 (PTPL1) resulted in increased EGFP-2xFYVE punctae ( Table 2.1).
PTPN13, a phosphatase proposed to have both tumor suppressive and oncogenic functions, has
been implicated in several signal transduction pathways. Specifically, PTPN13 was shown to
inhibit PI3K/Akt signaling and thus, the PI(3)P phenotype elicited by knockdown could
potentially be explained by altered 3’-phosphoinositide metabolism (Dromard, Bompard et al.
2007; Abaan and Toretsky 2008). Mutations in SHP2 are associated with several human
diseases, most notably Noonan syndrome, LEOPARD syndrome, and juvenile myelomonocytic
leukemia (Araki, Mohi et al. 2004; Mohi, Williams et al. 2005; Kontaridis, Swanson et al. 2006;
Mohi and Neel 2007). Its activity has been linked to numerous signaling pathways, often
downstream of receptor tyrosine kinases, and the observed phenotype could be a consequence of
disruption of any number of substrates (Chan, Kalaitzidis et al. 2008).
Surprisingly, we did not identify MTMR3 or MTMR14 (hJumpy), the PI(3)Pphosphatases with reported roles in autophagy. The myotubularin phosphatases comprise a large,

34

highly conserved family of enzymes whose members have been shown to function as
heteromeric partners (Lorenzo, Urbe et al. 2006). As one example, MTMR3 and MTMR4, both
FYVE-domain containing phosphatases, have been demonstrated to interact and inhibit PI(3)P
(Lorenzo, Urbe et al. 2006). Accordingly, gene-by-gene loss of function analysis of this family
may not reveal phenotypes if compensation within the family occurs. Further, these enzymes
may serve cell- or context-specific functions not revealed in this study.
The most striking result from this study was the presence of abundant PI(3)P-positive
vesicles following PTPsigma knockdown which phenocopied that of an autophagic cell. We
confirmed hyperactive autophagy in the absence of PTPsigma through use of multiple autophagy
markers, as well as electron microscopy. Atg12 and LC3-positive autophagic vesicles were
substantially more abundant in the absence of PTPsigma when cells were cultured in full
nutrients (constitutive AVs) or treated with rapamycin (induced AVs). These autophagic vesicles
accumulated upon treatment with the lysosomal inhibitors, Baf-A1 and chloroquine,
demonstrating that they were functional and destined for lysosomal degradation. This phenotype
suggests PTPsigma regulates an early step in autophagy induction and its loss results in increased
autophagic vesicle generation. This is consistent with the fact that PI(3)P is generated on early
phagophores and is required for proper autophagic vesicle formation. A role for PTPsigma in
autophagy induction and specifically, PI(3)P regulation, is supported by our findings that
PTPsigma localizes to PI(3)P-positive vesicles which increase in number during autophagy.
It remains to be addressed how PTPsigma is targeted to autophagic vesicles. PTPsigma is
expressed at the cell surface in a two subunit complex comprised of a large ectodomain and a
membrane-spanning intracellular domain. Accordingly, it is implicated in cell-cell and cell-ECM
interactions, and it is a critical regulator of axon homeostasis and neuronal development (Aicher,

35

Lerch et al. 1997; Elchebly, Wagner et al. 1999; Wallace, Batt et al. 1999; Uetani, Chagnon et al.
2006). Relevant to our own work, ectodomain shedding and internalization of a membranebound carboxy-terminal fragment has been demonstrated previously (Aicher, Lerch et al. 1997).
Through immunofluorescent analysis of PTPsigma using D1 domain-specific antibodies, we
place intracellular PTPsigma on PI(3)P-positive autophagic vesicles. Autophagosomes
frequently fuse with endosomes during their maturation, forming hybrid organelles called
amphisomes, establishing the possibility that PTPsigma is internalized by endocytosis to arrive at
autophagic vesicles (Klionsky 2007). Further, the close relative of PTPsigma, LAR (PTPRF),
undergoes an additional proteolytic event whereby a soluble intracellular domain is formed and
targeted inside the cell, similar to the Notch receptor (Ruhe, Streit et al. 2006). PTPsigma
contains similar cleavage residues to LAR, making it therefore plausible that PTPsigma is
targeted from the plasma membrane to autophagic vesicles through a series of proteolytic events
in response to autophagic stimuli. Thus, this phosphatase may serve several unique functions
during various cellular conditions which are governed by its subcellular localization.
A critical finding presented here is the recruitment of PTPsigma to vesicular structures
during amino acid-starvation which occurs even in the absence of PI(3)P generation. This
finding, together with the hyperactivation of autophagy elicited by PTPsigma knockdown (as
measured by PI(3)P, Atg12, and LC3), suggests PTPsigma regulates autophagy at an early step
upstream of this lipid. In further support of this, we found that while almost all PTPsigmapositive vesicles are also positive for PI(3)P (EGFP-2xFYVE presence), fewer harbored LC3, a
marker which is incorporated into AVs at later maturation stages.
There are several potential mechanisms by which PTPsigma may function to regulate
autophagy. First, it is possible that PTPsigma could directly dephosphorylate PI(3)P following

36

recruitment to AVs. We did test the activity of recombinant PTPsigma in vitro, and while we
could not detect PI(3)P-phosphatase activity, it cannot be entirely excluded that PI(3)P does not
serve as a direct substrate in vivo (Figure 2.9). It is also possible that PTPsigma uses its robust
protein phosphatase activity to regulate the function of a PI(3)P-modifying enzyme, such as a
PI(3)P-phosphatase or a PI(4)- or PI(5)-kinase. Alternatively, PTPsigma could control the
activity of Vps34, which contains at least one phosphotyrosine site, or another component within
the larger Vps34 complex (Imami, Sugiyama et al. 2008). Finally, PTPsigma may contribute to
the regulation of autophagy at the earliest initiation step, which is executed by a complex of
autophagy proteins, namely ULK1 (Atg1) and Atg13. The functional formation of this complex,
which permits the generation of the PI(3)P-positive phagophore, was recently found to be tightly
regulated by phosphorylation events (Chang and Neufeld 2009; Ganley, Lam du et al. 2009;
Hosokawa, Hara et al. 2009; Jung, Jun et al. 2009). The aim of future work will be to determine
the precise mechanism by which PTPsigma functions to regulate autophagy.

37

MATERIALS AND METHODS

siRNA screen and validation
U2OS-EGFP-2xFYVE cells were seeded on 96-well plates (2,000 per well) in McCoy’s
5A medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS,
Invitrogen) at 37°C for 24 hours. Four siRNA molecules per phosphatase gene (phosphatase
siRNA library version 2.0, Qiagen, Valencia, CA) were transfected per well at a final
concentration of 25 nM using 0.2 μl HiPerfect transfection reagent (Qiagen) per well. After 48
hours, cells were fixed with 3.7% formaldehyde and nuclei were stained with Hoechst-33342
(Invitrogen). Cells were visualized at 40x on a Zeiss LSM 510 Meta confocal microscope
(Oberkochen, Germany) and EGFP-2xFYVE fluorescence was compared to that of control
siRNA-transfected cells within each plate. Triplicate wells from each gene were qualitatively
scored by two independent scorers on a scale from –100 (decreased EGFP-2xFYVE signal and
distribution) to +100 (increased) and mean scores were determined. Twenty-seven phosphatase
genes whose knockdown increased EGFP-2xFYVE fluorescence in the primary screen were used
in a secondary screen, where four siRNAs were individually transfected to eliminate off-target
hits. The primary score was multiplied by a binned secondary screen score (score of 1.0 for 3 or
4 of 4 siRNAs yielding a phenotype; 0.75 for 2 of 4 siRNAs; and 0 for 0 or 1 of 4 siRNAs).
Quantitative real-time PCR (qRT-PCR) assays with SYBR green dye (Roche, Basel,
Switzerland) and gene-specific primers confirmed that siRNAs effectively reduced mRNA
expression of target genes. For imaging, cells were cultured on number 1.5 coverglass,
transfections repeated as above, cells fixed, nuclei stained, and coverglass inverted into
microslides with mounting gel. A control siRNA transfected well was cultured for 3 hours in

38

amino acid-starvation media [Dulbecco’s phosphate-buffered saline (DPBS) with 10% FBS and
1 g/L D-glucose]. Cells were imaged using a 60x oil objective on a Nikon TE300 fluorescent
microscope (Tokyo, Japan). EGFP-2xFYVE-positive vesicles were quantified using image
analysis software.

Phospholipid labeling, extraction, and thin layer chromatography (TLC)
U2OS cells were seeded in McCoy’s 5A with 10% FBS at 200,000 cells per well of 6well tissue culture plates. After 24 hours, control or PTPRS siRNAs were transfected at a final
concentration of 25 nM using 2 μl HiPerfect transfection reagent per ml medium. Control siRNA
was All-star Negative Control (Qiagen) and PTPRS siRNAs were two unique sequences
(SI02759288, SI03056284, Qiagen). After 48 hours of transfection, the medium was replaced
with phosphate-free DMEM (Invitrogen) supplemented with 10% phosphate-free FBS for 30
min.

32

PO4 (0.25 mCi) was added per ml of medium for an additional 2 hours (Perkin Elmer,

Waltham, MA). Radiolabeling was quenched with ice-cold TCA (10% final concentration) and
cells incubated on ice for 1 hour. Cells were scraped, pelleted, and lipids extracted via an
acidified Bligh and Dyer method (Bird 1994). Lipids were lyophilized, resuspended in
chloroform:methanol (1:1), spotted on 20 cm x 20 cm silica gel TLC plates (Whatman,
Maidstone, UK), and resolved in a chamber using boric acid buffer (Walsh, Caldwell et al.
1991). A PI(3)P standard was generated by incubating synthetic phosphatidylinositol (diC16
PtdIns; Echelon, Salt Lake City, UT) with immunoprecipitated PI3K (using anti-p85, Cell
Signaling, Danvers, MA) and

32

P-ATP (Perkin Elmer). The TLC plate was exposed to film for

20 hours at –80°C and developed.

39

Fluorescent microscopy and western blot analyses of autophagy markers
U2OS cells were seeded at a density of 35,000 cells per well in McCoy’s 5A medium
with 10% FBS on number 1.5 coverglass in 24-well tissue culture plates (for fluorescent
imaging) or 150,000 cells per well on 6-well dishes (for western blot). After 24 hours, cells were
transfected with control or PTPRS siRNAs for 48 hours, as described above. Following, cells
were treated for 1-2 hours in amino acid-starvation media or with 50 nM rapamycin
(Calbiochem, San Diego, CA), 25 μM chloroquine (Sigma-Aldrich, St. Louis, MO), 100 nM
Baf-A1 (A.G. Scientific, San Diego, CA) or normal growth medium (full nutrients; McCoy’s 5A
with 10% FBS), as indicated. For western blots, cells were lysed [in 10 mM KPO4, 1 mM
EDTA, 10 mM MgCl2, 5 mM EGTA, 50 mM bis-glycerophosphate, 0.5% NP40, 0.1% Brij35,
0.1% sodium deoxycholate, 1 mM NaVO4, 5 mM NaF, 2 mM DTT, and complete protease
inhibitors (Sigma-Aldrich)] and 20 μg of total protein was resolved by SDS-PAGE. Proteins
were transferred to PVDF membranes and probed with primary antibodies (LC3B, Cell
Signaling Technologies; anti-α-tubulin, Sigma-Aldrich) for 16 hours at 4°C followed by
secondary antibodies (HRP-linked rabbit- or mouse- IgG, GE, Piscataway, NJ) for 1 hour at
room temperature. Proteins were detected by enhanced chemiluminescence. For EGFP-LC3
imaging, U2OS cells stably expressing ptfLC3 (Addgene plasmid 21074) (Kimura, Noda et al.
2007) were fixed with 3.7% formaldehyde and nuclei stained with Hoechst-33342 (2 μg/ml).
Coverglass were inverted onto microslides using mounting gel and cells imaged using a 100x oilimmersion objective on a Nikon Eclipse Ti fluorescence microscope. For immunofluorescence,
cells were fixed with 3.7% formaldehyde, permeabilized with 0.2% triton-X 100, and blocked
40

with 3% bovine serum albumin (BSA) in PBS. Antibodies (LC3B, Atg12, EEA1; Cell Signaling
Technologies) were added for 16 hours at 4°C followed by Alexa Fluor (AF)-488 conjugated
anti-rabbit IgG (Invitrogen) for 1 hour at room temperature. Nuclei were counterstained with
Hoechst-33342, coverglass inverted onto microslides using mounting gel, and cells imaged using
60x or 100x oil-immersion objectives on a Nikon TE300 fluorescence microscope (LC3, Atg12)
or a 63x water-immersion objective on a Zeiss LSM510 Meta confocal microscope (EEA1). For
quantification, punctae were counted using image analysis software after establishing an
intensity threshold.

Transmission electron microscopy (TEM)
U2OS cells in 10 cm dishes were transfected with control or PTPRS siRNAs for 48 hours
as described above. Cells (siRNA-transfected U2OS cells or primary MEFs) were briefly
trypsinized, pelleted, rinsed, and resuspended in 2% glutaraldehyde fixative (Sigma-Aldrich).
Cell pellets were embedded in 2% agarose, postfixed in osmium tetroxide, and dehydrated with
an acetone series. Samples were infiltrated and embedded in Poly/Bed 812 resin and polymerized
at 60°C for 24 hours. Ultrathin sections (70 nm) were generated with a Power Tome XL
(Boeckeler Instruments, Tucson, Arizona) and placed on copper grids. Cells were examined
using a JEOL 100CX Transmission Electron Microscope at 100 kV (Tokyo, Japan). Autophagic
2

structures were quantified from images encompassing approximately 8.5 μm of cell area each.
Electron microscopy services were performed by Alicia Pastor and the Michigan State
University Center for Advanced Microscopy (East Lansing, MI). MEFs were provided by Michel
L. Tremblay (McGill University, Goodman Cancer Center).

41

Exogenous PTPsigma expression and immunofluorescence
U2OS-EGFP-2xFYVE cells were seeded at a density of 20,000 cells per well in McCoy’s
5A medium with 10% FBS on number 1.5 coverglass in 24-well tissue culture dishes. Full-length
PTPsigma cDNA (BC104812) was inserted into pRK7 by EcoRI digestion and ligation to yield
FL-PTPsigma-pRK7 (FL-PTPsigma). DNA was transfected at 0.15 μg per well using 0.45 μl
FuGeneHD transfection reagent (Roche, Mannheim, Germany) in 50 μl Optimem and 450 μl
McCoy’s 5A with 10% FBS for 24 hours. For two hours, cells were cultured with full nutrient
media or starved of amino acids (Fig. 4A-C), or amino acid-starved while treated with DMSO or
100 nM wortmannin (Sigma-Aldrich) (Figure 2.5, Figure 2.8). Alternatively, cells were treated
with Baf-A1 (100 nM in full nutrient media) for 0, 15, 60, or 240 minutes (Figure 2.4E). Cells
were then fixed with 3.7% formaldehyde, permeabilized with 0.2% Triton-X 100, blocked in 3%
BSA, and stained with antibodies targeting the D1 domain of PTPsigma for 2 hours at room
temperature. AF-546-conjugated anti-mouse-IgG (Invitrogen) were incubated for an additional
hour at room temperature and nuclei stained with Hoechst-33342. Cells were imaged using oilimmersion objectives at 60x on a Nikon TE3000 or 100x on an Eclipse Ti fluorescent
microscope. For Fig. 4C, cells were treated as above and imaged using a 63x water-immersion
objective on a Zeiss LSM510 Meta microscope. Red (AF-546, FL-PTPsigma) and green (EGFP2xFYVE, PI(3)P) channels were captured with confocality through the Z-plane using 16
increments of 0.25 µm. Stacks through the indicated X and Y planes are shown at the border of
an image of the third Z-plane. For Figure 2.8B, U2OS cells stably expressing mRFP-LC3
(Addgene plasmid 21075) (Kimura, Noda et al. 2007) were seeded, transfected, treated (full
nutrient or amino acid starvation media for 2 hours), and stained as above. Images were captured
at 100x using an oil-immersion objective on an Eclipse Ti fluorescent microscope.

42

Rescue of siRNA phenotype
U2OS-EGFP-2xFYVE cells were seeded on number 1.5 coverglass in 24-well dishes at
20,000 cells per well in McCoy’s 5A medium with 10% FBS. 24 hours later, PTPRS siRNA-1
(CACGGCATCAGGCGTGCACAA; Qiagen) was transfected at a concentration of 25 nM using
1 μl oligofectamine per well per 500 μl McCoy’s with 10% FBS (Invitrogen). FL-PTPsigmapRK7 plasmid DNA was transfected 24 hours later at a concentration of 0.15 μg DNA per well
using 0.45 μl FuGeneHD transfection reagent in 50 μl Optimem and 450 μl McCoy’s 5A with
10% FBS for an additional 24 hours. Cells were fixed and immunostained as described above.
Cells were imaged using a 60x oil objective on a Nikon TE300 fluorescent microscope for
EGFP-2xFYVE phenotype (green) and FL-PTPsigma-pRK7 expression (red). FL-PTPsigmapRK7 expression levels were categorized as high or low. The presence or absence of a robust
PTPRS siRNA-induced EGFP-2xFYVE phenotype was determined (phenotype defined as the
presence of small, abundant, non-perinuclear EGFP-2xFYVE-positive vesicles; indicated with
white arrows in Figure 2.4D) for 30-40 cells each of low and high FL-PTPsigma-expressing cells
as well as cells transfected with PTPRS siRNA-1 but not FL-PTPsigma-pRK7.

In vitro phosphatase assays
GST-tagged recombinant PTPsigma containing all residues C-terminal to the
transmembrane domain (BC104812 cDNA; aa 883-1501) was generated in pGEXKG (Guan and
Dixon 1991). GST-tagged full-length MTMR6 (NM_004685.2) was generated in pGEXKG and
GST-tagged recombinant PTP1B was purchased (Upstate, Billerica, MA). Proteins were purified
from BL21 Escherichia coli after isopropyl β-D-1-thiogalactopyranoside (IPTG) induction and
used in phosphatase assays. For PI(3)P-phosphatase reactions, 1μg protein was suspended in 50
43

μl assay buffer (50 mM sodium acetate, 25 mM Tris-HCl, 10 mM DTT, pH 6.5) with 0, 25, 50,
or 200 μM diC8-PI(3)P and reactions carried out at 37°C for 25 min. For p-Tyr-phosphatase
assays, reactions were carried out as above using 0, 10, 25, or 100 μM p-Tyr peptide
(TSTEPQpYQPGENL; Upstate) at 37°C for 15 min. Released phosphates were detected with
malachite green (Upstate) and absorbance measured at 650 nm. Background levels from enzymeonly and substrate-only (p-Tyr or PI(3)P) reactions were subtracted and absorbance converted to
picomoles free phosphate released per minute using a standard curve.

44

ACKNOWLEDGEMENTS

We thank the Van Andel Institute Systems Biology lab for advice, analysis, and reagents.
This work was supported by the Department of Defense Prostate Cancer Research Program of
the Office of Congressionally Directed Medical Research Programs PC081089 to J.P.
MacKeigan. J.P. MacKeigan is also supported by Award Number R01CA138651 from the
National Cancer Institute.

45

TABLES

Table 2.1. siRNA-mediated knockdown of human phosphatase genes alters cellular PI(3)P.
U2OS-EGFP-2xFYVE cells were transfected with siRNAs targeting human phosphatase genes
for 48 h (4 siRNA sequences per gene per well; all four sequences displayed). Following
knockdown, EGFP-2xFYVE signal and distribution was visualized by confocal microscopy and
scored from -100 (decreased punctae from control cells) to +100 (increased punctae) and means
determined. Select genes were validated using multiple unique siRNA sequences and their
efficacy was incorporated into their scores (see Methods). Genes were ranked based on their
scores from 1 (most increased EGFP-2xFYVE) to 206 (most decreased).

46

Table 2.1 (cont'd)
Gene Symbol
ACP1
ACP2
ACP5
ACP6
ACPP
ACPT
ALPI
ALPL
ALPPL2
CDC25A
CDC25B
CDC25C
CDKN3
CTDP1
CTDSP2
CTDSPL
DKFZP566K0524
DOLPP1
DUSP10
DUSP11
DUSP12
DUSP13
DUSP14
DUSP15
DUSP16
DUSP18

Description
acid phosphatase 1, soluble
acid phosphatase 2, lysosomal
acid phosphatase 5, tartrate resistant
acid phosphatase 6, lysophosphatidic
acid phosphatase, prostate
acid phosphatase, testicular
alkaline phosphatase, intestinal
alkaline phosphatase, liver/bone/kidney
alkaline phosphatase, placental-like 2
cell division cycle 25A
cell division cycle 25B
cell division cycle 25C
cyclin dependent-kinase inhibitor 3 (CDK2-associated dual specificity
phosphatase)
CTD (carboxy-terminal domain, RNA polymerase II, polypeptide A)
phosphatase, subunit 1
CTD (carboxy-terminal domain, RNA polymerase II, polypeptide A)
small phosphatase 2
CTD (carboxy-terminal domain, RNA polymerase II, polypeptide A)
small phosphatase-like
DKFZP566K0524 protein
dolichyl pyrophosphate phosphatase 1
dual specificity phosphatase 10
dual specificity phosphatase 11 (RNA/RNP complex 1-interacting)
dual specificity phosphatase 12
dual specificity phosphatase 13
dual specificity phosphatase 14
dual specificity phosphatase-like 15
dual specificity phosphatase 16
dual specificity phosphatase 18

47

Score Rank
25
40
-17
176
25
41
-8
167
42
20
0
117
58
9
25
42
17
63
17
64
50
13
17
65
25

43

0

118

8

85

8

86

17
0
8
8
0
42
8
-42
0
33

66
119
87
88
120
21
89
198
121
29

Table 2.1 (cont'd)
DUSP19
DUSP2
DUSP21
DUSP22
DUSP23
DUSP24
DUSP3
DUSP6
DUSP7
DUSP8
DUT
ENPP1
ENPP2
ENPP3
FBP1
FBP2
FLJ32332
G6PC
G6PC2
G6PC3
ILKAP
IMPA1
INPP1
INPP4B
INPP5A
INPP5B
INPP5D
INPP5E
INPP5F
ITPA
LHPP

dual specificity phosphatase 19
dual specificity phosphatase 2
dual specificity phosphatase 21
dual specificity phosphatase 22
dual specificity phosphatase 23
dual specificity phosphatase 24 (putative)
dual specificity phosphatase 3 (vaccinia virus phosphatase VH1-related
dual specificity phosphatase 6
dual specificity phosphatase 7
dual specificity phosphatase 8
dUTP pyrophosphatase
ectonucleotide pyrophosphatase/phosphodiesterase 1
ectonucleotide pyrophosphatase/phosphodiesterase 2 (autotaxin)
ectonucleotide pyrophosphatase/phosphodiesterase 3
fructose-1,6-bisphosphatase 1
fructose-1,6-bisphosphatase 2
likely ortholog of mouse protein phosphatase 2C eta
glucose-6-phosphatase, catalytic (glycogen storage disease type I, von
glucose-6-phosphatase, catalytic, 2
glucose 6 phosphatase, catalytic, 3
integrin-linked kinase-associated serine/threonine phosphatase 2C
inositol(myo)-1(or 4)-monophosphatase 1
inositol polyphosphate-1-phosphatase
inositol polyphosphate-4-phosphatase, type II, 105kDa
inositol polyphosphate-5-phosphatase, 40kDa
inositol polyphosphate-5-phosphatase, 75kDa
inositol polyphosphate-5-phosphatase, 145kDa
inositol polyphosphate-5-phosphatase, 72 kDa
inositol polyphosphate-5-phosphatase F
inosine triphosphatase (nucleoside triphosphate pyrophosphatase)
phospholysine phosphohistidine inorganic pyrophosphate phosphatase

48

-8
33
0
0
0
-33
-33
25
33
42
42
0
-33
0
0
0
-17
0
50
0
0
-17
50
8
8
33
0
0
50
8
8

168
30
122
123
124
193
194
44
31
22
23
125
195
126
127
128
177
129
14
130
131
178
15
90
91
32
132
133
18
92
93

Table 2.1 (cont'd)
LOC387870

LOC389772

LOC391025

LOC400927
LOC442428

LOC474338
LPPR2
MINPP1
M-RIP
MTM1
MTMR2
MTMR3
MTMR4
MTMR6
MTMR8
MTMR9
PDP2
PDPR
PHOSPHO1
PHPT1
PIB5PA
PIP3AP
PLIP

similar to protein tyrosine phosphatase, receptor type, Q isoform 1
precursor; glomerular mesangial cell receptor protein-tyrosine
phosphatase; glomerular mesangial cell receptor protein-tyrosine
similar to Osteotesticular phosphatase; protein tyrosine phosphatase
receptor type V; protein tyrosine phosphatase receptor type W;
protein tyrosine phosphatase, receptor type, V
similar to protein tyrosine phosphatase, receptor type, U isoform 2
precursor; protein tyrosine phosphatase J; protein tyrosine
phosphatase receptor omicron; pi R-PTP-Psi
similar to TPTE and PTEN homologous inositol lipid phosphatase
isoform alpha; TPTE and PTEN homologous inositol lipid
similar to fructose-1,6-bisphosphatase 2; fructose-1,6-bisphosphatase
isozyme 2; D-fructose-1,6-bisphosphate 1-phosphohydrolase;
FBPase; muscle fructose-bisphosphatase; hexosediphosphatase
SUMO1 pseudogene 3
lipid phosphate phosphatase-related protein type 2
multiple inositol polyphosphate histidine phosphatase, 1
myosin phosphatase-Rho interacting protein
myotubularin 1
myotubularin related protein 2
myotubularin related protein 3
myotubularin related protein 4
myotubularin related protein 6
myotubularin related protein 8
myotubularin related protein 9
pyruvate dehydrogenase phosphatase isoenzyme 2
pyruvate dehydrogenase phosphatase regulatory subunit
phosphatase, orphan 1
phosphohistidine phosphatase 1
phosphatidylinositol (4,5) bisphosphate 5-phosphatase, A
phosphatidylinositol-3-phosphate associated protein
PTEN-like phosphatase

49

0

134

0

135

56

12

17

67

25

45

-8
0
-42
25
25
25
0
17
33
8
8
8
58
-8
17
25
0
17

169
136
199
46
47
48
137
68
33
94
95
96
10
170
69
49
138
70

Table 2.1 (cont'd)
PME-1
PNKP
PPAP2A
PPAP2B
PPAP2C
PPEF1
PPEF2
PPFIA1
PPFIA2
PPFIA3
PPFIA4
PPM1A
PPM1B
PPM1D
PPM1E
PPM1F
PPM1G
PPM1L
PPM2C
PPP1CA
PPP1CB
PPP1CC
PPP1R10
PPP1R11
PPP1R12A
PPP1R12B
PPP1R12C
PPP1R13B

protein phosphatase methylesterase-1
polynucleotide kinase 3'-phosphatase
phosphatidic acid phosphatase type 2A
phosphatidic acid phosphatase type 2B
phosphatidic acid phosphatase type 2C
protein phosphatase, EF hand calcium-binding domain 1
protein phosphatase, EF hand calcium-binding domain 2
protein tyrosine phosphatase, receptor type, f polypeptide (PTPRF),
interacting protein (liprin), alpha 1
protein tyrosine phosphatase, receptor type, f polypeptide (PTPRF),
interacting protein (liprin), alpha 2
protein tyrosine phosphatase, receptor type, f polypeptide (PTPRF),
interacting protein (liprin), alpha 3
protein tyrosine phosphatase, receptor type, f polypeptide (PTPRF),
interacting protein (liprin), alpha 4
protein phosphatase 1A (formerly 2C), magnesium-dependent, alpha is
protein phosphatase 1B (formerly 2C), magnesium-dependent, beta iso
protein phosphatase 1D magnesium-dependent, delta isoform
protein phosphatase 1E (PP2C domain containing)
protein phosphatase 1F (PP2C domain containing)
protein phosphatase 1G (formerly 2C), magnesium-dependent, gamma
protein phosphatase 1 (formerly 2C)-like
protein phosphatase 2C, magnesium-dependent, catalytic subunit
protein phosphatase 1, catalytic subunit, alpha isoform
protein phosphatase 1, catalytic subunit, beta isoform
protein phosphatase 1, catalytic subunit, gamma isoform
protein phosphatase 1, regulatory subunit 10
protein phosphatase 1, regulatory (inhibitor) subunit 11
protein phosphatase 1, regulatory (inhibitor) subunit 12A
protein phosphatase 1, regulatory (inhibitor) subunit 12B
protein phosphatase 1, regulatory (inhibitor) subunit 12C
protein phosphatase 1, regulatory (inhibitor) subunit 13B

50

33
17
8
8
8
17
8
0

34
71
97
98
99
72
100
139

0

140

25

50

25

51

-58
17
8
0
0
8
8
8
17
-17
17
-75
0
0
17
-8
50

204
73
101
141
142
102
103
104
74
179
75
205
143
144
76
171
16

Table 2.1 (cont'd)
PPP1R14A
PPP1R14C
PPP1R14D
PPP1R15A
PPP1R15B
PPP1R16B
PPP1R1A
PPP1R1B
PPP1R1C
PPP1R2
PPP1R2P9
PPP1R3A
PPP1R3B
PPP1R3C
PPP1R3D
PPP1R3F
PPP1R7
PPP1R8
PPP1R9A
PPP1R9B
PPP2CA
PPP2CB
PPP2R1A
PPP2R1B
PPP2R2A
PPP2R2B
PPP2R2C
PPP2R2D
PPP2R3A
PPP2R4

protein phosphatase 1, regulatory (inhibitor) subunit 14A
-17
protein phosphatase 1, regulatory (inhibitor) subunit 14C
0
protein phosphatase 1, regulatory (inhibitor) subunit 14D
-25
protein phosphatase 1, regulatory (inhibitor) subunit 15A
0
protein phosphatase 1, regulatory (inhibitor) subunit 15B
-25
protein phosphatase 1, regulatory (inhibitor) subunit 16B
-100
protein phosphatase 1, regulatory (inhibitor) subunit 1A
17
protein phosphatase 1, regulatory (inhibitor) subunit 1B (dopamine
33
and cAMP regulated phosphoprotein, DARPP-32)
protein phosphatase 1, regulatory (inhibitor) subunit 1C
67
protein phosphatase 1, regulatory (inhibitor) subunit 2
92
protein phosphatase 1, regulatory (inhibitor) subunit 2 pseudogene 9
17
protein phosphatase 1, regulatory (inhibitor) subunit 3A (glycogen and
0
sarcoplasmic reticulum binding subunit, skeletal muscle)
protein phosphatase 1, regulatory (inhibitor) subunit 3B
17
protein phosphatase 1, regulatory (inhibitor) subunit 3C
0
protein phosphatase 1, regulatory subunit 3D
-50
protein phosphatase 1, regulatory (inhibitor) subunit 3F
-17
protein phosphatase 1, regulatory subunit 7
-25
protein phosphatase 1, regulatory (inhibitor) subunit 8
0
protein phosphatase 1, regulatory (inhibitor) subunit 9A
-33
protein phosphatase 1, regulatory subunit 9B, spinophilin
0
protein phosphatase 2 (formerly 2A), catalytic subunit, alpha isoform
0
protein phosphatase 2 (formerly 2A), catalytic subunit, beta isoform
0
protein phosphatase 2 (formerly 2A), regulatory subunit A (PR 65), alph 25
protein phosphatase 2 (formerly 2A), regulatory subunit A (PR 65), bet 92
protein phosphatase 2 (formerly 2A), regulatory subunit B (PR 52), alph -8
protein phosphatase 2 (formerly 2A), regulatory subunit B (PR 52), bet 25
protein phosphatase 2 (formerly 2A), regulatory subunit B (PR 52), gam 33
protein phosphatase 2, regulatory subunit B, delta isoform
0
protein phosphatase 2 (formerly 2A), regulatory subunit B'', alpha
8
protein phosphatase 2A, regulatory subunit B' (PR 53)
33

51

180
145
188
146
189
206
77
35
7
2
78
147
79
148
200
181
190
149
196
150
151
152
52
3
172
53
36
153
105
37

Table 2.1 (cont'd)
PPP2R5A
PPP2R5B
PPP2R5C
PPP2R5E
PPP3CA
PPP3CB
PPP3CC
PPP3R1
PPP4C
PPP4R1
PPP5C
PPP6C
PR48
PSPH
PSTPIP1
PSTPIP2
PTEN
PTP4A1
PTP4A2
PTP4A3
PTPDC1
PTPLA
PTPLB
PTPN1
PTPN11
PTPN12
PTPN13

protein phosphatase 2, regulatory subunit B (B56), alpha isoform
protein phosphatase 2, regulatory subunit B (B56), beta isoform
protein phosphatase 2, regulatory subunit B (B56), gamma isoform
protein phosphatase 2, regulatory subunit B (B56), epsilon isoform
protein phosphatase 3 (formerly 2B), catalytic subunit, alpha isoform
(calcineurin A alpha)
protein phosphatase 3 (formerly 2B), catalytic subunit, beta isoform
(calcineurin A beta)
protein phosphatase 3 (formerly 2B), catalytic subunit, gamma
isoform (calcineurin A gamma)
protein phosphatase 3 (formerly 2B), regulatory subunit B, 19kDa,
alpha isoform (calcineurin B, type I)
protein phosphatase 4 (formerly X), catalytic subunit
protein phosphatase 4, regulatory subunit 1
protein phosphatase 5, catalytic subunit
protein phosphatase 6, catalytic subunit
protein phosphatase 2A 48 kDa regulatory subunit
phosphoserine phosphatase
proline-serine-threonine phosphatase interacting protein 1
proline-serine-threonine phosphatase interacting protein 2
phosphatase and tensin homolog (mutated in multiple advanced cance
protein tyrosine phosphatase type IVA, member 1
protein tyrosine phosphatase type IVA, member 2
protein tyrosine phosphatase type IVA, member 3
protein tyrosine phosphatase domain containing 1
protein tyrosine phosphatase-like (proline instead of catalytic arginine),
protein tyrosine phosphatase-like (proline instead of catalytic arginine),
protein tyrosine phosphatase, non-receptor type 1
protein tyrosine phosphatase, non-receptor type 11 (Noonan syndrome
protein tyrosine phosphatase, non-receptor type 12
protein tyrosine phosphatase, non-receptor type 13 (APO-1/CD95
(Fas)-associated phosphatase)

52

0
-50
17
-17
-25

154
201
80
182
191

-17

183

25

54

-17

184

-17
0
42
-8
-50
42
-8
8
-17
8
8
0
17
17
8
25
75
50
83

185
155
24
173
202
25
174
106
186
107
108
156
81
82
109
55
5
17
4

Table 2.1 (cont'd)
PTPN14
PTPN18
PTPN2
PTPN21
PTPN22
PTPN23
PTPN3
PTPN4
PTPN5
PTPN6
PTPN7
PTPN9
PTPNS1
PTPNS1L2
PTPRA
PTPRB
PTPRC
PTPRD
PTPRE
PTPRF
PTPRG
PTPRH
PTPRJ
PTPRK
PTPRM
PTPRN
PTPRN2
PTPRR
PTPRS
PTPRV
PTPRZ1
RNGTT

protein tyrosine phosphatase, non-receptor type 14
protein tyrosine phosphatase, non-receptor type 18 (brain-derived)
protein tyrosine phosphatase, non-receptor type 2
protein tyrosine phosphatase, non-receptor type 21
protein tyrosine phosphatase, non-receptor type 22 (lymphoid)
protein tyrosine phosphatase, non-receptor type 23
protein tyrosine phosphatase, non-receptor type 3
protein tyrosine phosphatase, non-receptor type 4 (megakaryocyte)
protein tyrosine phosphatase, non-receptor type 5 (striatum-enriched)
protein tyrosine phosphatase, non-receptor type 6
protein tyrosine phosphatase, non-receptor type 7
protein tyrosine phosphatase, non-receptor type 9
protein tyrosine phosphatase, non-receptor type substrate 1
protein tyrosine phosphatase, non-receptor type substrate 1-like 2
protein tyrosine phosphatase, receptor type, A
protein tyrosine phosphatase, receptor type, B
protein tyrosine phosphatase, receptor type, C
protein tyrosine phosphatase, receptor type, D
protein tyrosine phosphatase, receptor type, E
protein tyrosine phosphatase, receptor type, F
protein tyrosine phosphatase, receptor type, G
protein tyrosine phosphatase, receptor type, H
protein tyrosine phosphatase, receptor type, J
protein tyrosine phosphatase, receptor type, K
protein tyrosine phosphatase, receptor type, M
protein tyrosine phosphatase, receptor type, N
protein tyrosine phosphatase, receptor type, N polypeptide 2
protein tyrosine phosphatase, receptor type, R
protein tyrosine phosphatase, receptor type, S
protein tyrosine phosphatase, receptor type, V
protein tyrosine phosphatase, receptor-type, Z polypeptide 1
RNA guanylyltransferase and 5'-phosphatase

53

58
69
0
8
8
-25
8
67
0
25
-50
0
0
0
42
8
8
-17
25
0
25
0
33
25
8
8
25
17
100
50
42
33

11
6
157
110
111
192
112
8
158
56
203
159
160
161
26
113
114
187
57
162
58
163
38
59
115
116
60
83
1
19
27
39

Table 2.1 (cont'd)
SBF1
SGPP1
SKIP
SNAP23
SPAP1
TA-PP2C
TENC1
TPTE
TPTE2

SET binding factor 1
sphingosine-1-phosphate phosphatase 1
skeletal muscle and kidney enriched inositol phosphatase
synaptosomal-associated protein, 23kDa
SH2 domain containing phosphatase anchor protein 1
T-cell activation protein phosphatase 2C
tensin like C1 domain containing phosphatase
transmembrane phosphatase with tensin homology
transmembrane phosphoinositide 3-phosphatase and tensin homolog 2

54

0
0
-33
0
42
17
-8
25
25

164
165
197
166
28
84
175
61
62

FIGURES

Figure 2.1. Cell-based siRNA screen identifies PTPsigma as a modulator of PI(3)P. (A-D)
U2OS-EGFP-2xFYVE cells transfected with control siRNA (A), Vps34 siRNA (B), starved of
amino acids for 3 hours (C), or transfected with PTPRS siRNAs (D), were fixed and imaged at
60x magnification by fluorescent microscopy (green: PI(3)P, EGFP-2xFYVE; blue: nuclei).
Insets show 2x magnifications of small EGFP-2xFYVE vesicles. Bars, 10 µm. (E) Following
knockdown of phosphatases, EGFP-2xFYVE-positive punctae were scored from -100 (decreased
from control cells) to +100 (increased). Phosphatases are ranked and plotted by decreasing score
(left to right) with genes whose loss increased EGFP-2xFYVE fluorescence highlighted in green
and whose loss caused decreases highlighted in blue. PTPRS is identified. (F) Mean EGFP2xFYVE-positive punctate were quantified from cells under the conditions indicated using image
analysis software. Bars represent s.e.m., *p < 0.05. (G) Phospholipids were radiolabeled in vivo,
extracted, resolved by TLC, and visualized by autoradiography following transfection with
control or PTPRS siRNAs. A radiolabeled PI(3)P standard was resolved adjacent to extracted
lipids. (H,I) Endosomes were labeled by immunostaining with anti-EEA1 antibodies (H) and
autophagic vesicles were labeled with anti-LC3B antibodies (I) following transfection with
control or PTPRS siRNA (red: EEA1; green: LC3B; blue: nuclei). Insets show 2x magnifications
of LC3-positive vesicles. Bars, 10 µm.

55

Figure 2.1 (cont'd)

Control siRNA B

VPS34 siRNA

E

*

400

200
*

PI(3)P
PI(4)P
PI(5)P
PIP
2

VP

C
S3 on
AA 4 trol
-S siR
PT tar NA
PR va
S tion
si
R
N
A

0

H

PTPRS
siRNA
- +

300

100

G

56

Control
siRNA

Mean EGFP-2xFYVEpositive vesicles per cell

F

PTPRS siRNA

Standard

C

PTPRS
siRNA

AA-starvation
(Autophagy) D

100
75
50
25
0
-25
-50
-75
-100

Mean EGFP-2xFYVE score

A

PTPRS

1

50
100 150 200
Phosphatase genes
(ranked by score)

Early
Endosomes
(anti-EEA1)

I Autophagic

Vesicles
(anti-LC3B)

Figure 2.2. Loss of PTPsigma hyperactivates autophagy. (A) Atg12-positive punctae were
quantified from 60x images of cells transfected with control (black) or PTPRS siRNAs (white)
and treated for 1 hour with normal growth media (full nutrients) or 50 nM rapamycin. Values
represent relative Atg12-positive punctae per cell following normalization to control cells
cultured with full nutrients. Bars represent s.e.m., *** p < 0.001. (B) LC3-I and LC3-II were
analyzed by western blot using whole cell lysates from control siRNA-transfected cells, PTPRS
siRNA-transfected cells, or amino acid-starved cells. α-tubulin was probed as a loading control.
(C-H) EGFP-LC3-positive punctae were visualized in U2OS cells transfected with control
(C,E,G) or PTPRS (D,F,H) siRNAs following 2 hour treatment with normal growth media (full
nutrients; C,D), 100nM bafilomycin A1 (Baf-A1; E,F), or 50 nM rapamycin and 100 nM Baf-A1
(G,H) (green: EGFP-LC3; blue: nuclei). Insets are 2x magnifications of EGFP-LC3-positive
AVs. Bars, 10µm.

57

8

0
Full
RapaNutrients mycin

B
siRNA

LC3-I
LC3-II
Bafilomycin A1
Full Nutrients
Rapamycin

16

12

***

Full Nutrients

20

AA-Starvation

Relative ATG12-positive
punctae per cell (normalized
to control)

A

PTPRS

Control

Anti-LC3B
(endogenous)

Figure 2.2 (cont'd)

Control siRNA
PTPRS siRNA
***

4

C

α -tubulin

58

EGFP-LC3B
Control siRNA
PTPRS siRNA

D

E
F

G
H

Figure 2.3. Loss of PTPsigma increases autophagic vesicle abundance as measured by
electron microscopy. (A-D) Few autophagic vesicles (AVs) were found by transmission
electron microscopy (TEM) within control cells cultured in full nutrients (A), but were abundant
in chloroquine-treated (B), amino acid (AA)-starved (C), and PTPRS siRNA-transfected (D)
cells. Black arrows indicate autophagic vesicles (autophagosomal or autolysosomal). White scale
bars in (A-D) represent 1 µM. (E-G) Primary wild-type PTPRS (+/+, E) and knockout PTPRS (/-, F) MEFs were analyzed by TEM and AVs quantified (G). AVs, defined as autophagic
structures containing cytosolic components, generally degradative, were counted from 8 μm

2

sampling regions per cell. Values represent mean AVs per sampling area. Bars represent s.e.m.,
**p < 0.01. White scale bars in (E-F) represent 0.5 µM.

59

Figure 2.3 (cont'd)

A

Control B

Chloroquine

C

AA-Starvation D

PTPRS siRNA

Ptprs+/+

E

F

Mean AVs per
TEM sampling
area (8 μm2)

G

60

Ptprs-/-/-

4
3

**

2
1
0

+/+
-/PTPRS

Figure 2.4. Exogenous PTPsigma localizes to PI(3)P vesicles and rescues the siRNA
phenotype. (A,B) FL-PTPsigma was transiently expressed in U2OS-EGFP-2xFYVE cells and
PI(3)P and PTPsigma imaged by fluorescent microscopy following 2 hour incubation with full
nutrient media (A) or amino acid-starvation media (B) [green: PI(3)P; red: anti-PTPsigma (D1targeted antibodies); blue: nuclei]. Insets are 2x magnifications of boxed regions. Bars, 10 µm.
(C) U2OS-EGFP-2xFYVE cells transfected with FL-PTPsigma and amino acid-starved for 2
hours were imaged using D1-targeted PTPsigma antibodies. A Z-stack of 0.25 µm increments
was captured using sequential channel acquisition and confocal microscopy, with the third slice
displayed and Z-stacks through the X and Y planes shown at the border. Insets are 2x
magnifications of boxed regions. Bar, 10µm (green: PI(3)P, EGFP-2xFYVE; red: antiPTPsigma; yellow: colocalization). (D,E) U2OS-EGFP-2xFYVE cells were transfected with
PTPRS siRNA-1 for 48 hours, after which FL-PTPsigma (which lacks the sequence targeted by
siRNA-1) was introduced for an additional 24 hours. PI(3)P and PTPsigma were imaged as
previously described. The presence of siRNA-induced phenotype (abundant, non-perinuclear
EGFP-2xFYVE-positive vesicles; indicated by white arrows in (D)) was determined for cells
expressing no, low, or high levels of FL-PTPsigma (E). Bars, 10 µm. (F) FL-PTPsigma-positive
vesicular structures accumulate when lysosomal fusion is inhibited. U2OS cells expressing FLPTPsigma for 24 hours were treated with 100 nM Baf-A1 in full nutrient media for 0, 15, 60, or
240 minutes and FL-PTPsigma imaged using D1-targeted PTPsigma antibodies (red). Nuclei
were stained with Hoechst (blue). Bars, 10 µm.

61

Figure 2.4 (cont'd)

FL-PTPsigma:
anti-D1 domain

Overlay (with
Hoechst)

C

Full Nutrients

A

EGFP-2xFYVE

AAStarvation

B

60
20
0
FLPTPsigma
Expression

62

Bafilomycin A1

100

w
H
ig
h

E

k

Overlay

Lo

FLPTPsigma

F

M
oc

EGFP2xFYVE

siRNA-indued
Phenotype
(% Cells)

D

FL-PTPsigma
EGFP-2xFYVE
0 min

15 min

FL-PTPsigma
0 min
15 min

60 min

240 min

Figure 2.5. Localization of PTPsigma to vesicular structures does not require PI(3)P. (A-B)
U2OS cells expressing FL-PTPsigma were treated with vehicle (DMSO; A) or 100 nM
wortmannin (PI3K inhibitor; B) for 2 hours while cultured in amino acid-starvation media and
PI(3)P and PTPsigma imaged by fluorescent microscopy [green: PI(3)P, EGFP-2xFYVE; red:
anti-PTPsigma (D1-targeted antibodies); blue: nuclei]. Insets are 2x magnifications of boxed
regions. Bars, 10 µm.

63

Figure 2.5 (cont'd)

FL-PTPsigma:
anti-D1 domain

AA-Starvation
Wortmannin
DMSO

A

EGFP-2XFYVE

B

64

Overlay
(with Hoechst)

Figure 2.6. Target genes are effectively knocked down by siRNA. (A) Western blot analysis
of whole cell extracts following transfection with control or Vps34 siRNA shows depletion of
Vps34 protein levels. α-tubulin was probed as a loading control. (B) U2OS-EGFP-2xFYVE cells
were transfected with MTMR6 siRNA, fixed, and imaged at 60x by fluorescent microscopy as in
Fig. 1. A bar, 10 µm. (C) MTMR6 mRNA was depleted by 96% following transfection with
MTMR6 siRNA for 48 hours. RNA extracted from control- or MTMR6- siRNA transfected cells
was converted to cDNA and MTMR6 levels determined by qRT-PCR using gene-specific
primers. Values were normalized to GAPDH. (D-I) U2OS-EGFP-2xFYVE cells were transfected
with control (D) or PTPRS siRNA [E, siRNA-1, F, siRNA-2, G, siRNA-3, H, siRNA-4, I,
siRNA-pool (1-4)] for 48 hours, fixed, and imaged by fluorescent microscopy (green: PI(3)P,
EGFP-2xFYVE; blue: nuclei). Insets are 2x magnifications of boxed regions, highlighting the
abundant vesicles caused by PTPRS siRNA transfection. Bars, 10 µm. (J) EGFP-2xFYVE
punctae were quantified from cells following PTPRS knockdown with four unique siRNAs.
Values represent means and bars represent s.e.m. (K) PTPRS mRNA knockdown following 48
hour transfection with four unique siRNAs (individually or pooled) was determined by qRT-PCR
using gene-specific primers and GAPDH normalization, as described above.

65

Figure 2.6 (cont'd)

B

siRNA

MTMR6 siRNA

Control Vps34
Vps34
α-tubulin

D

Control siRNA

G

PTPRS siRNA-3

E

H

PTPRS siRNA-1

PTPRS siRNA-4

K
300
250

Relative PTPRS
Expression

Mean EGFP-2xFYVEpositive vesicles per cell

J

200
150
100
50
0

C

Relative
MTMR6 Expression

A

Control 1 2 3 4
siRNA PTPRS siRNA

66

1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0

F

I

1.2
1.0
0.8
0.6
0.4
0.2
0

Control MTMR6
siRNA siRNA
PTPRS siRNA-2

PTPRS siRNA
Pool (1-4)

Control 1 2 3 4 (1-4)
Control
siRNA
siRNA PTPRS siRNA

Figure 2.7. PTPsigma knockdown increases the abundance of autophagic, but not
endocytic, vesicles. (A) U2OS cells were transfected with control or PTPRS siRNAs, fixed, and
immunostained with anti-EEA1 antibodies (see Fig. 1H). EEA1-positive vesicles were quantified
using image analysis software. Bars represent s.e.m. (B) U2OS cells transfected with control (left
panels) or PTPRS (right panels) siRNAs were cultured for 1 hour with nutrient-rich medium (top
panels) or 50 nM rapamycin (bottom panels). Cells were stained with anti-Atg12 antibodies and
imaged by fluorescent microscopy at 60x (green: Atg12; blue: nuclei). Bars, 10 µm. (C) U2OS
cells transfected with control or PTPRS siRNAs were cultured for 1 hour with normal growth
media (full nutrients; left), 25 μM chloroquine in normal growth media (middle), or 50 nM
rapamycin and 25 μM chloroquine in normal growth media (right). Cells were fixed,
immunostained with anti-LC3B antibodies, and imaged by fluorescent microscopy at 60x. LC3positive punctae were quantified images using image analysis software (black: control siRNA;
white: PTPRS siRNAs). Bars represent s.e.m., *p < 0.05, **p < 0.01.

67

Figure 2.7 (cont'd)

B

120

Mean EEA1-positive
punctae per cell

120
100
100

80

60

40

20

80
60
40
20
0

Control PTPRS

0
NT si

300

Mean LC3-positive
punctae per cell

C

80

Control siRNA

Full Nutrients

140

Rapamycin

A

siRNA
Control siRNA
PTPRS siRNA
**

60
40
20
0

Full Rapamycin
Full
Nutrients Nutrients
Chloroquine

68

PTPRS siRNA

Figure 2.8. FL-PTPsigma colocalization with mRFP-LC3 and mock control for FLPTPsigma immunofluorescence. (A) U2OS-EGFP-2xFYVE cells were mock transfected (with
transfection reagent but no DNA) for 24 hours, stained with PTPsigma (anti-D1) antibodies, and
imaged as in Fig. 4A-C, Fig. 5, and (B). Absence of signal in the red channel demonstrates
specificity of FL-PTPsigma expression captured in the above figures [green: PI(3)P; red: antiPTPsigma (D1-targeted antibodies); blue: nuclei]. Bars, 10 µm. (B) FL-PTPsigma was
transiently expressed in U2OS-mRFP-LC3 cells and LC3 and PTPsigma imaged by fluorescent
microscopy following 2 hour incubation with full nutrient media (top panels) or amino acidstarvation media (lower panels) [red: mRFP-LC3; green: PTPsigma (D1-targeted antibodies);
blue: nuclei]. Insets are 2x magnifications of boxed regions. White arrows indicated punctae
positive for both PTPsigma and LC3.

69

Figure 2.8 (cont'd)

EGFP-2xFYVE

FL-PTPsigma:
anti-D1 domain

Overlay (with Hoechst)

mRFP-LC3B

FL-PTPsigma:
anti-D1 domain

Overlay (with Hoechst)

Mock Transfection

A

AA-Starvation

Full Nutrients

B

70

Figure 2.9. PTPsigma dephosphorylates phosphotyrosine, but not PI(3)P, in vitro. (A)
Recombinant GST-fusions of PTP1B (left), PTPsigma (middle) and MTMR6 (right) were
incubated with a phosphotyrosine peptide at the indicated concentrations for 15 minutes at 37°C
and released phosphates measured by malachite green quenching and 650 nm absorbance. (B)
Recombinant GST-fusions of PTP1B (left), PTPsigma (middle) and MTMR6 (right) were
incubated with water-soluble PI(3)P substrate at the indicated concentrations and released
phosphates measured by malachite green quenching and 650 nm absorbance. Bars represent
s.d.m.

71

Figure 2.9 (cont'd)

B
2000
1600
1200

p-Tyr (µM)
0
10
25
100

800
400
0

PTP1B PTPsigma MTMR6
Enzyme

72

Picomoles phosphate released
per min per μg (mean)

Picomoles phosphate released
per min per μg (mean)

A

2000
1600
1200

PI(3)P (µM)
0
25
50
200

800
400
0

PTP1B PTPsigma MTMR6
Enzyme

CHAPTER 2 SECTION 2

A potential role for PTPsigma as a Vps34 complex effector and the discovery of autophagyrelevant processing events

Martin KR, MacKeigan JP

73

ABSTRACT

Phosphatidylinositol-3-phosphate (PI(3)P) is an intracellular signaling lipid required for
dynamic membrane trafficking pathways including endocytosis and autophagy(Funderburk,
Wang et al. 2010). As such, it is essential that PI(3)P be generated and maintained on membranes
in a spatially and temporally controlled manner. In a search for phosphatases which contribute to
this exquisite regulation, we recently reported that PTPsigma, a dual-domain protein tyrosine
phosphatase (PTP), functions as a negative regulator of PI(3)P signaling and autophagy (Martin,
Xu et al. 2011). Following this initial identification, several important issues remained
concerning the regulation and function of this phosphatase. Specifically, we found that loss of
PTPsigma increases PI(3)P-positive vesicle abundance and enhances autophagy; however, the
mechanism underlying this loss-of-function phenotype was not established. Further, we
determined that PTPsigma localizes to PI(3)P-positive vesicles in a dynamic manner; however,
potential processing events responsible for targeting this receptor-like PTP from the cell surface
to its vesicular location were not investigated.
Here, we address these key outstanding issues to uncover important features of PTPsigma
function and regulation as it pertains to PI(3)P signaling. First, we determine that loss of
PTPsigma increases the catalytic activity of Vps34, the lipid kinase which generates PI(3)P, and
further, that PTPsigma is capable of interacting with Vps34 and one of its binding partners,
Rubicon. Second, loss of PTPsigma induces changes in tyrosine phosphorylation detected in the
Vps34 complex as well as whole cell extracts. Future identification of the targets for this
phosphorylation will likely reveal substrates of PTPsigma that mediate its role in PI(3)P
signaling. Lastly, we detect the formation of distinct PTPsigma processing fragments during

74

normal growth conditions as well as upon autophagy induction. We discuss potential
mechanisms by which PTPsigma may function to downregulate PI(3)P and the proteolytic
processing events which control its internalization and presence on PI(3)P membranes.

75

INTRODUCTION

Balanced PI(3)P signaling is crucial to maintain homeostatic function of both endocytosis
and autophagy, two convergent membrane-trafficking processes. Vps34, the sole member of the
mammalian class III phosphoinositide-3-kinase (PI3K) family, produces PI(3)P and has been
found to reside within large complexes of regulatory proteins which control its function in these
processes (Sun, Fan et al. 2008; Matsunaga, Saitoh et al. 2009; Zhong, Wang et al. 2009). After
formation, turnover of PI(3)P through dephosphorylation can be executed by myotubularinrelated lipid phosphatases (Blondeau, Laporte et al. 2000; Vergne, Roberts et al. 2009; TaguchiAtarashi, Hamasaki et al. 2010). In addition to these regulatory mechanisms, we hypothesized
that novel protein phosphatases exist which control PI(3)P, especially through downregulation of
that autophagic machinery. To explore this hypothesis, we performed a cell-based screen of
human phosphatase genes using RNA interference and found that loss of PTPsigma
hyperactivates PI(3)P and autophagy (Martin, Xu et al. 2011).
In our initial report, we showed that PTPsigma resides on PI(3)P membranes in a
dynamic and nutrient-responsive manner. We also observed through an RNAi rescue experiment
that overexpressed PTPsigma can reduce cellular PI(3)P levels. Although we first hypothesized
that PTPsigma may directly dephosphorylate PI(3)P, after extensive investigation, we could not
detect robust lipid phosphatase activity in vitro. Additionally, when we depleted cells of PI(3)P
(by inhibiting Vps34), the dynamic localization of PTPsigma to PI(3)P-positive membranes was
not affected, suggesting that its vesicular presence occurs independently of this lipid. These
findings led us to postulate that PTPsigma has protein substrates involved in PI(3)P signaling and
that alterations of their activity upon PTPsigma knockdown explain this loss-of-function

76

phenotype. Because of the robust PI(3)P production in the absence of PTPsigma and evidence
for tyrosine phosphorylation of Vps34 and several of its binding partners, we hypothesized that
PTPsigma may function as a Vps34-complex effector. In this study, we explore this hypothesis
and report our findings from the following aims: (1) to determine if PTPsigma has the ability to
interact with proteins comprising Vps34 complexes; (2) to establish the control, if any,
PTPsigma exerts on Vps34 catalytic activity; and (3) to identify phosphorylation changes which
occur upon PTPsigma knockdown. Indeed, we found that PTPsigma is capable of interacting
with Vps34 and its endocytic binding partner, Rubicon. Further, loss of PTPsigma enhanced
phosphorylation of unknown targets associated with Vps34 and increased lipid kinase activity as
measured in vitro.
An additional question arising from our initial identification of PTPsigma as a regulator
of autophagy concerns the processing of PTPsigma. As a cell-surface receptor-like molecule,
PTPsigma is capable of undergoing defined proteolytic processing whereby its large ectodomain
is shed and a membrane-tethered C-terminal fragment (CTF) is internalized inside the cell
(Aicher, Lerch et al. 1997). LAR, a close relative of PTPsigma, has been characterized to
undergo an additional cleavage event, mediated by gamma secretase and presenilin, which
results in the formation of an intracellular domain (ICD) which is liberated from the membrane
(Ruhe, Streit et al. 2006; Haapasalo, Kim et al. 2007). Based on close homology to LAR, it is
possible that PTPsigma also forms an ICD although to our knowledge, this has yet to be studied.
We raised the possibility that processing events may be regulated to control the function of
PTPsigma in autophagy. Specifically, we hypothesized that PTPsigma undergoes distinct
processing events which lead to the formation of a CTF or ICD and it is in this form that
intracellular PTPsigma controls PI(3)P signaling. We determined that under normal growth

77

conditions (that is, in full nutrients), an abundant PTPsigma P-subunit is processed into a CTF
which accumulates kinetically upon lysosomal inhibition. This suggests a continuous formation
and turnover of the CTF within the lysosome. Intriguingly, additional processing occurs during
starvation-induced autophagy to produce a potential PTPsigma ICD; however, this fragment does
not appear to be a major target of the lysosome. The lysosomal turnover of a CTF and further
nutrient-responsive processing represent novel findings for PTPsigma that may be important for
regulation of its function in PI(3)P signaling.

78

RESULTS

PTPsigma and Vps34 can co-immunoprecipitate with one another
Loss of PTPsigma appears to hyperactivate autophagy from a very early step (PI(3)P).
Thus, we filtered a publically available database, PhosphoSite Plus, for evidence of tyrosine
phosphorylation on proteins functioning early in autophagy (www.phosphosite.org). In our
search of proteins including Vps34 and its key binding partners, the ULK1 complex, and
autophagic PI(3)P effectors, we found six potentially phosphorylated tyrosine residues (Table
2.2). Intriguingly, all but one of these were found on Vps34-related proteins. From this finding,
together with the robust effect loss of PTPsigma exerts on PI(3)P and the localization of
PTPsigma on membranes enriched in PI(3)P, we postulated that PTPsigma may regulate Vps34
or an associated protein.
To determine if PTPsigma could interact with Vps34, we used a combination of
exogenous and endogenous protein detection in co-immunoprecipitation (co-IP) experiments.
PTPsigma expression is low and undetectable at the endogenous protein level by immunoblot
using currently available resources. Endogenous Vps34 can be readily detected, however,
antibodies which can immunoprecipitate Vps34 are not available. Accordingly, we began by
immunoprecipitating exogenously expressed V5-PTPsigma-CTF (residues including and Cterminal to the transmembrane domain) with a substrate trapping mutation in the D1 active site
(C1589S) from 293FT cells. We found that under normal growth conditions (that is, full serum,
glucose, and amino acids), endogenous Vps34 was pulled down in the precipitate (Figure
2.10A). Interestingly, this interaction was not apparent after 1 hour amino acid starvation
(autophagy) (Figure 2.10A). To see if we could detect this interaction in the reverse orientation,

79

we expressed and immunoprecipitated V5-Vps34 from 293FT cells which co-expressed its
regulatory proteins, Vps15 and Beclin1, as well as untagged full-length PTPsigma. When probed
with a D1-targeted antibody, PTPsigma was readily detected in all Vps34 immunoprecipitates
(Figure 2.10B). While the substrate-trapping mutation appeared to enhance the interaction, it
cannot be excluded that expression levels contributed to this difference. Importantly, substitution
of the putatively phosphorylated tyrosine, Y517, with an alanine, did not alter the interaction
with PTPsigma (Figure 2.10B). Taken together, these data suggest that PTPsigma and Vps34 are
capable of interacting together in cells and this interaction does not depend on phosphorylation
of Vps34 at Y517.

PTPsigma and Rubicon may interact in cells
In an effort to further elucidate the potential interaction with Vps34, we tested the ability
of PTPsigma to interact with several components of the Vps34 complex (Vps15, Beclin1,
UVRAG, and Rubicon) using a similar co-IP approach. We discovered that PTPsigma was
capable of interacting with Rubicon independent of Vps34 overexpression, although this did not
appear to be as robust as the interaction with Vps34 (Figure 2.10C). Rubicon is part of the Vps34
complex that localizes to endomembranes and functions along the endocytic system (Matsunaga,
Saitoh et al. 2009; Zhong, Wang et al. 2009). It is unique among the Vps34 interactors in that
rather than supporting the function of Vps34, it suppresses its activity and downregulates the
maturation of endosomes and autophagosomes (Matsunaga, Saitoh et al. 2009; Zhong, Wang et
al. 2009). Given this interaction from cell lysates, we analyzed the location of Rubicon and
PTPsigma by fluorescent microscopy and found that their subcellular localization largely

80

overlapped (Figure 2.10D). Specifically, both proteins localize to peripheral vesicles as well as
perinuclear membranous networks (Figure 2.10D).

Loss of PTPsigma is characterized by specific cellular changes in tyrosine phosphorylation
After establishing the potential for PTPsigma to reside within a Vps34 complex, we
sought to address if PTPsigma controls the phosphorylation of these proteins. To this end, U2OS
cells were transfected with control siRNAs or siRNAs targeting PTPsigma for 48 hours (this
knockdown condition results in a substantial increase in cellular PI(3)P and enhances autophagy;
See Chapter 2 Section 1). For the final 24 hours of knockdown, we exogenously expressed the
Vps34 core complex (Vps34-Vps15-Beclin1). After cells were lysed and Vps34immunoprecipitates collected, we assayed tyrosine phosphorylation by immunoblot.
Remarkably, we discovered phosphorylation of a prominent band near 100 kDa which was only
present in the absence of PTPsigma (Figure 2.11A). Acute treatment of control cells with the pan
PTP inhibitor, pervanadate, prior to lysis did not generate phosphorylation of this band
suggesting it is generated through a more time-intensive process (Figure 2.11A). The molecular
weight of this band does not correlate with any of the three overexpressed Vps34-complex
proteins but rather, it may represent an endogenous protein interacting and precipitating with this
complex. A protein of slightly smaller size appears to also be phosphorylated and incorporated
into this complex although its phosphorylation state was not substantially changed by loss of
PTPsigma (Figure 2.11A).
We chose to complement this finding with an unbiased approach to observe global
changes in phosphorylation of proteins elicited by PTPsigma knockdown. We once again
transfected U2OS cells with control or PTPsigma siRNAs for 48 hours. Following, the lysates

81

were immunoprecipitated with phospho-tyrosine-specific antibodies and captured proteins were
probed by immunoblot using the same phospho-tyrosine antibody. We discovered at least three
bands with increased phosphorylation in the absence of PTPsigma, each corresponding to a
unique molecular weight (Figure 2.11B). A band just beneath the 115 kDa marker showed
considerable phosphorylation in cells transfected with PTPsigma siRNAs although it was also
basally phosphorylated in the control cells. A faint band close to the 180 kDa marker was
phosphorylated in PTPsigma-depleted cells but apparently not in the control. Lastly, and relevant
to the Vps34 complex phospho-analysis, a phosphorylated band near 100 kDa was present only
upon PTPsigma knockdown (Figure 2.11B). The key next step is to use mass spectrometry to
identify these phosphorylated proteins.

Loss of PTPsigma increases the in vitro kinase activity of Vps34
Based on the observation that loss of PTPsigma drives changes in phosphorylation and
these alterations may be related to the Vps34-complexes, we next determined whether PTPsigma
could affect the kinase activity of Vps34 in vitro. It has been reported that full activity of Vps34
in vitro requires immunoprecipitation from cells co-expressing functional binding partners
(Backer 2008). We verified that three exogenous proteins comprising the Vps34 core complex
(V5-Vps34, Flag-Vps15, and Beclin1) were forming a functional complex in cells. First, a series
of co-IP experiments revealed that Vps34 interacts with both Vps15 and Beclin1 when expressed
in pairs and also, when all three are expressed together (Figure 2.12A). Further, we tested the
ability of immunoprecipitated Vps34 to catalyze the production of PI(3)P in vitro using synthetic
phosphatidylinositol substrate and radiolabeled ATP. Consistent with the previous data, we
found that co-expression of binding partners enhanced the catalytic activity of Vps34 (Figure

82

2.12B). A portion of this increase may be attributed to more stable expression of Vps34
promoted by co-expression of Vps15 and/or Beclin1 (Figure 2.12A). When this activity was
tested from cells transfected with PTPsigma siRNAs, an increase of approximately 60% was
observed (Figure 2.12C). When activity was measured from cells expressing a larger Vps34
complex which included the core, Rubicon, and UVRAG, PTPsigma knockdown similarly
increased kinase activity, by approximately 75% (Figure 2.12C). Although not shown here, the
increase in activity is likely not due to increased expression as a previous experiment under the
same general conditions showed similar expression and immunoprecipitation of the Vps34
complex from control or PTPRS siRNA-transfected cells (Figure 2.11A).

PTPsigma is processed to a CTF and degraded in the lysosome
This revelation that PTPsigma may function as an effector of the Vps34-complex,
together with prior knowledge that PTPsigma localizes to PI(3)P-vesicles in a dynamic manner,
led us to wonder whether its function in this axis is characterized by unique proteolytic
processing events. PTPsigma has been previously reported to shed its ectodomain and internalize
from the cell surface as a membrane-bound C-terminal fragment (CTF) (Aicher, Lerch et al.
1997). The intracellular destination of this CTF, and any further processing to a soluble
intracellular domain (ICD), has not been studied. We aimed to determine if PTPsigma exists as a
CTF and/or ICD in our cell model and if nutrient-starvation triggers proteolytic changes in
PTPsigma, much like it triggers a redistribution to smaller, peripheral vesicles.
To this end, we expressed full-length untagged exogenous PTPsigma in U2OS cells and
used immunoblotting to determine the presence of all potential fragments– the pro-subunit, Psubunit, CTF, and ICD – each distinguished by a unique predicted molecular weight of ~170

83

kDa, 85 kDa, 76 kDa, and 72 kDa, respectively. Because processing into each of these forms
occurs upstream of the D1 domain, we used a D1-directed antibody for detection. First, we
analyzed basal processing of PTPsigma which occurs during normal growth conditions. We
found that PTPsigma exists predominantly as a full-length pro-subunit and as a cell surfaceexpressed P-subunit (Figure 2.13A). When lysosomes were inhibited with bafilomycin A1
(BafA1), a smaller band of a molecular weight consistent with a CTF appeared within 15 to 30
minutes and accumulated kinetically (Figure 2.13A). To determine if this fragment was in fact
processed from the P-subunit, we inhibited extracellular metalloproteases responsible for Psubunit-to-CTF processing using a small molecule inhibitor, Batimastat (Ruhe, Streit et al.
2006). In contrast to the vehicle control, cells pre-treated with Batimastat failed to produce this
fragment, supporting its identity as a processed P-subunit fragment, likely a CTF (Figure 2.13B).

PTPsigma is further processed during starvation-induced autophagy
To determine whether PTPsigma processing is altered during autophagy induction, we
kinetically starved cells of amino acids and probed for PTPsigma fragments. Starvation induced
the generation of a novel PTPsigma fragment of lower molecular weight than the putative CTF
(Figure 2.13C). When cells were concurrently starved of amino acids and treated with BafA1,
the putative CTF accumulated once again, but the smaller fragment did so only slightly, if at all
(Figure 2.13C). This suggests that although the membrane-bound CTF is degraded in the
lysosome, this novel fragment is likely not regulated in the same manner.

84

DISCUSSION

In sum, the results of experiments summarized here begin to address some of the
questions raised in our initial report that PTPsigma suppresses autophagy. First, through co-IPs,
we detected interactions between PTPsigma and both Vps34 and Rubicon. While it will need to
be further characterized and verified endogenously, it appears that a Vps34-PTPsigma interaction
may be relevant during normal growth conditions but relieved during autophagy induction. This
is an intriguing notion given that the robust PI(3)P phenotype elicited by PTPsigma knockdown
was initially discovered and subsequently characterized in cells cultured in full nutrient media.
Accordingly, if PTPsigma in fact functions as a negative regulator of Vps34, we predict it exerts
this control during normal growth because upon depletion, cells exhibit inappropriately active
PI(3)P signaling and autophagy.
In this light, normal expression of PTPsigma could be viewed as a brake pedal in Vps34
activity, functioning in a similar role to Rubicon (Matsunaga, Saitoh et al. 2009; Zhong, Wang et
al. 2009). This function could have great purpose given the bimodal manner in which autophagy
contributes to cell fate (Levine 2007). At one extreme, a cell defective in autophagy is rendered
susceptible to stress and starvation and more prone to cell death induced by these. Conversely, a
cell which engages autophagy in excess is likely to suffer a premature death resulting from
detrimental degradation of vital cytosolic content (Levine 2007 Nature). Consequently, it would
benefit a cell to have checkpoints in place, such as PTPsigma suppression of PI(3)P production,
which would contribute to a homeostatic level of autophagy.
In agreement with an inhibitory function, PTPsigma knockdown appears to increase the
catalytic activity of Vps34 as measured in vitro. This enhanced kinase activity likely explains, at

85

least in part, the abundance of PI(3)P previously observed under these same conditions.
Intriguingly, we observed an endogenous protein of approximately 100 kDa which coprecipitated with exogenous Vps34 and was tyrosine-phosphorylated in response to PTPsigma
knockdown. While the identity of the phospho-protein is unknown, it is an exciting possibility
that PTPsigma exerts control over Vps34 through its regulation. Future studies will seek to
identify this protein and determine whether it is directly dephosphorylated by PTPsigma. If so,
the contribution of its phosphorylation status to Vps34 function will be investigated.
In contrast to the Vps34 interaction, the ability of PTPsigma to precipitate Rubicon was
relatively weak. Despite this, their subcellular localization on endomembranes was in close
alignment, suggesting they function within the same subcellular compartment. Rubicon has been
found to localize to membranes of both early and late endocytic compartments, including those
enriched with PI(3)P (Sun, Fan et al. 2008; Matsunaga, Saitoh et al. 2009; Zhong, Wang et al.
2009; Sun, Zhang et al. 2011). This suggests that the PI(3)P-positive vesicles where we observe
PTPsigma are likely also endocytic in origin. The phenotype reported for RNAi-mediated
knockdown of Rubicon was very similar to that which we observed for PTPsigma knockdown
and included an increased abundance and flux of autophagic vesicles (Matsunaga, Saitoh et al.
2009; Zhong, Wang et al. 2009). Because both of these proteins downregulate autophagy,
localize to the same vesicles, and can interact to some degree, it is possible that PTPsigma and
Rubicon cooperate with one another to suppress Vps34-PI(3)P signaling. Interestingly, Rubicon
has at least one potentially phosphorylated tyrosine residue and is 109 kDa in size, similar to that
of the unidentified phosphoprotein discussed above, raising the possibility that Rubicon is a
substrate of PTPsigma. This hypothesis remains to be tested.

86

We found that PTPsigma was processed from its P-subunit into a putative CTF. This
processing occurs under normal growth conditions and results in CTF targeting and turnover in
the lysosome. Further, we presented evidence that PTPsigma is processed into a C-terminal
fragment smaller than the CTF during starvation-induced autophagy. This fragment does not
appear to be targeted to the lysosome and while it is logical to assume it is an ICD, its molecular
weight is smaller than that reported for a LAR ICD (Haapasalo, Kim et al. 2007). PTPsigma does
in fact have residues within and just C-terminal to its transmembrane domain similar to those in
Notch that are required for gamma-secretase-mediated ICD formation (Gupta-Rossi, Six et al.
2004). However, mutation of the analogous residues in LAR did not impair the ability of gamma
secretase to generate an ICD so they may not be critical for a PTPsigma-ICD either (Haapasalo,
Kim et al. 2007). If this detected fragment is in fact an ICD, it may be generated through a
unique mechanism or from a specific site downstream of the transmembrane domain, resulting in
a smaller size. Regardless, because PTPsigma exerts its inhibitory effect on autophagy in the
presence of nutrients, it is likely that the CTF is the fragment relevant to PI(3)P suppression. The
fact that the CTF is membrane-bound and turned over in the lysosome supports a model where
PTPsigma is processed and internalized from the cell surface as part of the endocytic pathway. It
would likely be during this trafficking that PTPsigma resides on PI(3)P-positive endocytic
vesicles and suppresses Vps34 signaling to impede the maturation of both endosomal and
autophagic vesicles, akin to Rubicon.

87

MATERIALS AND METHODS

Co-immunoprecipitation experiments
10 cm dishes of 80% confluent U2OS or 293FT cells were transfected with the indicated
plasmids for 24 hours prior to cell lysis using a 1:3 µg DNA to µl FuGeneHD (Roche) ratio in
Optimem and McCoy’s 5A (U2OS) or DMEM (293FT) supplemented with 10% FBS. Cells
were lysed under conditions previously described in Chapter 2 Section 1. V5-fusions were
precipitated from cleared lysates (normalized for total protein by Bradford assay) using V5
antibodies (Invitrogen) for 2 to 24 hours at 4°C. Protein G was added for the final 1h at 4°C.
Beads were collected and washed three times with lysis buffer. After the final wash, beads were
resuspended in 2x sample loading buffer and resolved by SDS-PAGE. Western blotting was
performed as described in Chapter 2 Section 1 using the antibodies indicated in figures.
Antibodies used included Vps15 (Abcam), Beclin1 (BD), V5 (Invitrogen), GFP (Abcam), Vps34
(Invitrogen), PTPsigma (kind gift from Michel Tremblay), and phosphotyrosine (CST).

Phosphotyrosine analyses
Pervanadate was prepared for 10 min in the dark from sodium orthovanadate and
hydrogen peroxide in 1xPBS and used on cells at a concentration of 100 µM in normal media for
30 min prior to cell lysis. For Figure 2.11A, U2OS cells expressing Vps34 complexes were
immunoprecipitated with anti-V5 antibodies and proteins captured were probed by western blot
with anti-phosphotyrosine antibodies. For Figure 2.11B, U2OS cells transfected with PTPsigma
or control siRNAs for 48 hours (as described in Chapter 2 Section 1) were lysed,

88

immunoprecipitated with anti-phosphotyrosine antibodies, and proteins captured were probed by
western blot with anti-phosphotyrosine.

PTPsigma and GFP-Rubicon fluorescent microscopy
Fluorescent microscopy of exogenously expressed proteins was performed as described
in Chapter 2 Section 1. Briefly, U2OS cells were co-transfected with EGFP-Rubicon (Addgene
plasmid 28022; Sun et al 2011 JBC) and full-length PTPsigma (see Chapter 2 Section 2) for 24
hours. Cells were stained with anti-PTPsigma D1 antibodies and AF546-conjugated secondary
antibodies (red channel). Cells were imaged with an 100x/oil objective using a Nikon Ti Eclipse
fluorescent microscope.

PTPsigma processing experiments
U2OS cells were transfected with full-length PTPsigma (see Chapter 2 Section 2 details)
for 24 hours. Following, cells were treated with BafA1 (100 nM) and/or starved of amino acids
(see Chapter 2 Section 1 for media composition) for the indicated times. Cells were lysed,
protein content normalized, and lysates probed by western blot with anti-PTPsigma D1 targeted
antibodies. In Figure 2.13B, cells were pre-treated with vehicle (DMSO) or batimastat (Tocris,
10 µM) for 4 hours then treated with BafA1 (100 nM) for the indicated times.

Vps34 in vitro kinase assays
Vps34 core complex proteins were cloned in full-length form into pRK7. Vps34
(NM_002647.2) was fused with an N-terminal V5 tag introduced in the forward PCR primer.
Vps15 (BC110318.1) was fused with an N-terminal Flag tag introduced in the forward primer.
Beclin1 (BC010276.1) was cloned without an epitope tag.
89

293FT cells were co-transfected with the Vps34 core complex (Vps15, Beclin1, and
Vps34) with or without EGFP-Rubicon and EGFP-UVRAG (Addgene plasmid 24296; (Itakura,
Kishi et al. 2008). Vps34 was immunoprecipitated as described above using V5 antibodies.
Beads were either collected and probed for complex proteins by western blot or washed and
subjected to in vitro kinase assays.
For kinase assays, beads were washed in 1% NP40 in 1xPBS, 100 mM Tris-HCl pH 7.5 +
500 mM lithium chloride, and TNE (50 mM Tris-HCl pH 7.5, 140 mM sodium chloride, 5 mM
EDTA). After the final wash, beads were resuspended in 50 µl Vps34 kinase buffer (50 mM
Tris-HCl pH 7.5, 150 mM sodium chloride), 100 µM short chain phosphatidylinositol (diC8PtdIns), and 10 mM manganese chloride. After a 10 min pre-incubation at room temperature, 10
to 20 µCi

32

P-ATP was added with 50 µM non-labeled ATP. Reactions proceeded at room

temperature for 25 min while shaking, stopped with 80 µl 1N HCl, and lipids extracted using 160
µl chloroform:methanol (1:1). Lipid phases were spotted on silica TLC plate s(Fisher) and
resolved for 2 to 4 hours in CHCl3:MeOH:NH4OH:H2O (86:76:10:14). Plates were exposed to
film (4 hours at -80°C and 3 hours at room temperature) for autoradiography (Figure 2.12B) or
quantified on the phosphorimager (Figure 2.12B, Figure 2.12C).

90

TABLES

Table 2.2. Potentially phosphorylated tyrosine residues of the early autophagic machinery.
The publically available database of post-translational protein modifications, PhosphoSite Plus,
was filtered for proteins involved in early autophagy (grouped by Vps34 and key binding
partners, ULK1 complex proteins, and autophagic PI(3)P effectors below). Predicted molecular
weights (also obtained from PhosphoSite Plus) and tyrosine residues with evidence of
phosphorylation, if any, are listed in the right-hand column. “None” indicates that no
phosphorylated residues have been identified and reported. Original citations for mass
spectrometry data identifying these residues are included in the right column.

91

Table 2.2 (cont’d)
Reference(s)

Protein

Molecular Weight
(kDa)

pY site

VPS34 (PIK3C3)

102

Y517

VPS15 (PIK3CR4)

153

Y6732

BECLIN1

52

None

1

3,4

UVRAG

78

Y516

RUBICON

109

Y4495-8

ATG14L (BARKOR)
ULK1

55
113

Y37
None

ULK2
ATG13
FIP200 (RB1CC1)
DFCP1 (ZFYVE1)
WIPI1 (ATG18)
WIPI2

113
57
183
87
49
49

Y331
None
None
None
None
None

92

9

10

Imami, Sugiyama et al.
2008
Brill, Xiong et al. 2009
--Oppermann, Gnad et al.
2009; CST Curated
Dataset 4609, 2008
CST Curated Dataset
871, 2005; CST Curated
Datasets 1648, 1658,
1652, and 1655, 2006;
CST Curated Dataset
2484, 2658, 2007; CST
Curated Dataset 3766,
2008
Iliuk, Martin et al. 2010
--Li, Ren et al. 2009
-----------

FIGURES

Figure 2.10. PTPsigma potentially functions as a Vps34 effector. (A-C) PTPsigma is capable
of interacting with Vps34 and Rubicon. (A) 293FT cells were transfected with V5-PTPsigmaCTF C1589S mutant and incubated for 1 hour in the presence (+) or absence (-) of amino acids
and lysates immunoprecipitated (IPed) with V5 antibodies. Whole cell lysates (input) and
immunoprecipitates (IP) were probed by immunoblot (IB) with V5 and Vps34 antibodies. (B)
293FT cells were transfected with a wild-type (WT) or Y517A mutant (Y/A; generated by sitedirected mutagenesis) V5-Vps34 (along with Flag-Vps15 and untagged Beclin1) and full-length
wild-type (WT) PTPsigma (FL-PTPsigma) or PTPsigma with a double C1589S/C1880S
mutation (C/S). Lysates were IPed with V5 antibodies and probed by western blot with V5 and
D1-targeting PTPsigma antibodies. (C) U2OS cells expressing GFP-Rubicon were co-transfected
with wild-type (WT) or C1589S-mutated (C/S) V5-PTPsigma-CTF and lysates IPed with V5
antibodies. Whole cell lysates (input) and immunoprecipitates were probed by immunoblot with
GFP and V5 antibodies. (D) U2OS cells were co-transfected with FL-PTPsigma and GFPRubicon and localization imaged at 100x magnification by fluorescent microscopy. FLPTPsigma was detected with D1-targeted antibodies and AF-546 secondary antibodies. Nuclei
stained with Hoechst. Insets represent 2x magnifications of boxed regions.

93

Figure 2.10 (cont'd)

B

A

Input
(WCL)

V5-Vps34: - - WT WT Y/A Y/A
FLAmino Acids:
PTPsigma: WT C/S WT C/S WT C/S
+
kDa
IB: V5 (V5IB: V5 (Vps34)
115 PTPsigma-CTF)

IP: V5 (V5PTPsigma
CTF)

IB: Vps34
IB: VPS34

IB: V5 (V5PTPsigma-CTF)
IB: Vps34
IB: VPS34

180 IB: PTPsigma
(D1)

115 82 IP: V5 (Vps34)

C

IP: V5
(V5-PTPsigma
CTF)
GFP-Rubicon: + + +
V5-PTPsigma-CTF: (-) WT C/S

Input
(WCL)
+ + +
(-) WT C/S

D
FL-PTPsigma
anti-D1 domain

IB: GFP (Rubicon)
IB: V5
(PTPsigma-CTF)

94

GFP-Rubicon

Overlay (with
Hoechst)

Figure 2.11. Phosphotyrosine analyses following PTPsigma knockdown. (A) Vps34-V5,
Flag-Vps15, and Beclin1 were expressed in U2OS cells transfected with control or PTPsigma
siRNAs. Cells were treated with (+) or without (-) pervanadate (VO4) for 30 min prior to cell
lysis. V5-immunoprecipitates were probed by immunoblot with V5 or phospho-tyrosine (pTyr)
antibodies. Heavy chain (HC) shown as an IP control. (B) U2OS cells transfected with control or
PTPsigma siRNAs were immunoprecipitated with pTyr antibodies and probed with pTyr
antibodies by western blot. Heavy chain (HC) indicated as an IP control.

95

Figure 2.11 (cont'd)

49 -

IP: V5 (V5-Vps34)

m
ig
Ps

l
tro

kDa
180 115 82 -

HC

PT

siRNA: Control PTPsigma
VO4 : +
+
kDa
IB: V5 (V5115 Vps34)
115 IB: pTyr

on

B

C

A

a

siRNA:

IP: pTyr
IB: pTyr

64 49 -

96

HC

Figure 2.12. In vitro kinase activity of functional Vps34 complexes. (A-B) Vps34 complexes
were tested for functional interactions and activity. (A) Combinations of V5-Vps34, Flag-Vps15,
and Beclin1 were transfected into 293FT cells and immunoprecipitated with V5 antibodies.
Whole cell lysates (input) and immunoprecipitates (IP) were probed by western blot with V5,
Vps15, and Beclin1 antibodies. (B) V5-immunoprecipitates were analyzed for in vitro kinase
activity. IPs were incubated with phosphatidylinositol (PtdIns) and
buffer, lipids extracted, and

32

32

P-ATP in kinase assay

PI(3)P products visualized by autoradiography (bottom panel) or

quantified by phosphorimaging (plot). (C) U2OS cells transfected with control or PTPRS
siRNAs were further transfected with the Vps34 core complex (V5-Vps34, Flag-Vps15, and
Beclin1) or the core complex plus GFP-Rubicon and GFP-UVRAG. In vitro kinase assays were
performed as in (B) and

32

PI(3)P quantification shown.

97

Figure 2.12 (cont'd)

V5-Vps34 Flag-Vps15 Beclin1 -

+
-

+
+
-

+
+

B

+
+
+

anti-V5 (Vps34)
WCL
(input)

anti-Vps15
anti-Beclin1
anti-V5 (Vps34)

IP: V5
(V5Vps34)

anti-Vps15
anti-Beclin1

PI(3)P Phosphosignal

A

5000
4000
3000
2000
1000

0
V5-Vps34 Flag-Vps15 Beclin1 -

+
-

+
+
-

+
+

+
+
+

*PI(3)P

C
PI(3)P Phosphosignal

5000

Control siRNA
PTPsigma siRNA

4000
3000
2000
1000
0

Vps34
core

Vps34
core +
Rubicon +
UVRAG

98

Figure 2.13. Proteolytic processing of PTPsigma. (A-B) A PTPsigma C-terminal fragment
(CTF) is generated and processed in a lysosomal pathway. (A) U2OS cells expressing full-length
untagged PTPsigma (FL-PTPsigma) were treated with bafilomycin A1 (BafA1), a lysosomal
inhibitor, for 0 to 240 minutes. Lysates were probed by western blot with D1-targeted PTPsigma
antibodies. (B) U2OS cells expressing FL-PTPsigma were pre-treated with Batimastat (Bati), an
extracellular metalloprotease inhibitor, or DMSO (vehicle) for 4 hr then supplemented with
BafA1 for the indicated times. Lysates were probed by western blot with D1-targeted PTPsigma
antibodies. (C) U2OS cells expressing FL-PTPsigma were amino acid-starved to induce
autophagy for the times indicated in the absence (left) or presence (right) of BafA1. Lysates were
probed by western blot with D1-targeted PTPsigma antibodies.

99

Figure 2.13 (cont'd)

A

B

BafA1
(full nutrients)

min: 0 15 30 60 120 240
kDa
-Pro-Protein
180 -

DMSO Pre-treat
DMSO + BafA1 (min)
Ctrl 0

kDa

15

30 165 Ctrl 0

30 165

115 -

82 -

15

-Pro-Protein

180 -

115 -

Bati Pre-treat
Bati + BafA1 (min)

-P-subunit
82 -CTF (putative)
IB: anti-PTPsigma (D1)

AA-Starvation
15 30 60 120 240

kDa
180 -

C

on

0

tro

l

C

-P-subunit
-CTF (putative)
IB: anti-PTPsigma (D1)

AA-Starvation +
BafA1
15 30 60 120 240
-Pro-Protein

115 -

82 -

-P-subunit
-CTF (putative)
-Processed fragment

100

CHAPTER 3

In silico-based identification of small molecule inhibitors targeting PTPsigma

Martin KR, Xu Y, Narang P, Petit J, Meurice N, Xu E, MacKeigan JP

101

INTRODUCTION

Tyrosine phosphorylation is a key mechanism by which cells exert exquisite control of
signaling processes. Protein tyrosine kinases (PTKs) and phosphatases (PTPs) work in concert to
control these cascades and alterations in the expression or activity of these enzymes hallmark
many human diseases (Tonks 2006; Lahiry, Torkamani et al. 2010). Given their generally
positive regulation of signaling, protein kinases have long been the focus of both extensive
research and drug development efforts. In contrast, the molecular characterization of PTPs trailed
that of PTKs by ten years, and their role as critical mediators of signal transduction was initially
underappreciated (Tautz, Pellecchia et al. 2006). Only recently has the PTP field reached the
forefront of disease research and as justification for this, half of PTP genes are now implicated in
at least one human disease (Tautz, Pellecchia et al. 2006).
The development of phosphatase-modulating compounds is critical not only for potential
therapeutic benefit, but more fundamentally, for use as molecular probes which will be useful for
discerning the complex functions of these enzymes. Unfortunately, phosphatases have
historically been perceived as “undruggable” for several important reasons (Tautz and Mustelin
2007). The first is that a single phosphatase often controls multiple signaling pathways and thus,
inhibition would not yield a specific desired result. Second, signaling cascades are generally
controlled by multiple phosphatases and accordingly, blocking the activity of one may not be
sufficient to yield a desired pathway effect. Finally, and most importantly, the active site of
phosphatases displays high conservation which hinders the ability to develop catalysis-directed
inhibitors with any degree of selectivity (Tautz and Mustelin 2007). Despite these pitfalls, the
emerging role of phosphatases as mediators of vital cell processes and disease etiology has

102

necessitated a solution to these drug development issues. Largely through use of structure-based
drug design, several PTPs are now promising targets for disease treatment. Most notably,
bidentate inhibitors of PTP1B, implicated in type II diabetes and obesity, have been developed
which span both the catalytic pocket and a second substrate binding pocket discovered adjacent
to the active site (Shen, Keng et al. 2001; Zhang 2002; Sun, Fedorov et al. 2003).
Several studies have uncovered physiologically important functions for PTPsigma, a
dual-domain receptor type PTP, which highlight its attractiveness as a biological target. It has
been well established, primarily through knockout animal studies, that loss of PTPsigma
expression enhances axon guidance and neurite outgrowth (Elchebly, Wagner et al. 1999;
Wallace, Batt et al. 1999). Further, it was recently reported that loss of PTPsigma facilitates
nerve regeneration following spinal cord injury (SCI), owing to the interaction of its ectodomain
with chondroitin sulfate proteoglycans (CSPGs) (Shen, Tenney et al. 2009). In addition to its
neural function, PTPRS has been implicated in two cancer paradigms, chemoresistance and
metastatic disease. First, RNAi-mediated knockdown of PTPsigma in cultured cancer cells was
found to confer resistance to several chemotherapeutics (MacKeigan, Murphy et al. 2005).
Additionally, loss of PTPRS expression in metastatic prostate cancer was uncovered through a
study of laser-captured patient tissues encompassing progressive stages of prostate malignancy
(Tomlins, Mehra et al. 2007). Finally, we have discovered that loss of PTPsigma hyperactivates
autophagy, a cellular recycling program that may contribute to chemoresistance of cancer cells
(Martin, Xu et al. 2011). Taken together, it is apparent that modulation of PTPsigma may have
therapeutic potential in a range of diseases including SCI, neuronal diseases, and cancer. Aside
from therapeutic potential, a better understanding of the cellular function of PTPsigma and its
substrates would be aided by use of a specific molecular inhibitor.

103

Several approaches can be taken for the identification of small molecule inhibitors of
phosphatases. The first method involves high-throughput screening (HTS) of thousands of
compounds in vitro (Tierno, Johnston et al. 2007). This approach has become plausible with the
advent of automated HTS instrumentation and has been successfully utilized to discover
inhibitors of LAR (PTPRF), PTP1B, SHP2, CD45, and others (Mattila and Ivaska 2011). While
this approach is beneficial in that it directly measures enzyme inhibition by a variety of scaffolds,
the technical and physical investment is considerable as is the potential for experimental artifacts
leading to false negatives and positives.
An alternative approach to an initial in vitro screen, and the one adopted for this study,
involves a primary screen completed entirely in silico. This method entails computationally
docking small molecules into the crystal structure of a phosphatase active site and selecting hits
as those molecules which bind favorably, akin to a natural substrate (Kitchen, Decornez et al.
2004). Subsequently, the chosen lead scaffolds can be screened for phosphatase inhibition in
vitro. This approach has gained popularity as the number of enzymes with solved crystal
structures has increased and it is advantageous in many ways. First, utilization of the phosphatase
structure allows for the exclusion of molecules which have little chance of interacting with the
active site, greatly reducing the number of scaffolds to be physically screened. This manageable
number of compounds contributes to improved in vitro screen quality and confidence in its
results. Second, an understanding of the unique structural features and residues comprising the
active site as well as potential proximal folds or binding pockets can guide the selection or
refinement of an inhibitor. Further, an in silico approach is incredibly efficient in that it allows
tens of thousands to millions of compounds to be virtually screened (via computer software
programs) in a matter of weeks with no wet lab undertaking.

104

In this study, we sought to identify small molecule inhibitors targeting the active site of
PTPsigma. We employed virtual library screening (VLS) in conjunction with in vitro
phosphatase inhibition assays, to identify structural determinants of PTPsigma inhibitor activity,
selectivity, and potency (Figure 3.1). From over one million compounds screened in silico, we
identified the top-ranking 200 based on predicted optimal binding energies, which were further
filtered for diversity of scaffolds and verified by visual inspection. This VLS approach identified
66 lead scaffolds with promising binding potential. These PTPsigma target leads, as well as 88
additional compounds identified by substructure search in ChemBridge, were tested for
inhibition of PTPsigma in vitro. While we discovered several compounds with low micromolar
potency against PTPsigma, follow-up investigation revealed little to no selectivity, as most
scaffolds inhibited PTP1B in a similar manner. We discuss a future refinement of this approach
which we predict will lead to the discovery of inhibitors with greater selectivity for PTPsigma
and which could be applied to drug discovery of other classes of phosphatases as well.

105

RESULTS

In silico screening identifies small molecules targeted to the D1 active site of PTPsigma
The tandem phosphatase domains of PTPsigma have been crystallized in their apo form
(Almo, Bonanno et al. 2007). We retrieved this structure from the protein data bank (PDB 2FH7)
and verified its utility by molecularly docking a phosphotyrosine peptide. While the D1 domain
exhibited favorable binding with this natural substrate, we were intrigued to observe that the
active site harbored a uniquely wide conformation (Figure 3.2A). The specificity for tyrosine
residues observed by classic PTPs is due in large part to a binding cleft which is deep, to
accommodate a phosphotyrosyl ring, but narrow, to occlude bulkier species like phospholipids
(Begley, Taylor et al. 2006). We performed a similar docking of phosphotyrosine into the active
site of PTP1B and confirmed the narrowness of a typical PTP pocket (Figure 3.2B). While some
of the difference in conformation may stem from the fact that the PTPsigma crystal is in an open
form (that is, the WPD loop has not closed over a substrate), we hypothesized that the PTPsigma
active site could be exploited in the development of inhibitors selectively targeted to PTPsigma.
To this end, we used ZINC to screen a library of ChemBridge compounds (totaling
1,100,000) in silico for their ability to dock into the D1 domain of PTPsigma (Irwin and Shoichet
2005). From the top scoring 200 compounds which were most favorably bound by the active site,
we chose 66 compounds representing unique scaffolds to test in vitro (Figure 3.2C,D).

In silico-identified compounds inhibit PTPsigma activity in vitro
We measured the ability of compounds to inhibit the catalytic activity of PTPsigma using
the generic phosphatase substrate, para-nitrophenyl phosphate (pNPP). The dephosphorylated

106

product, para-nitrophenol (pNP), yields an intense yellow color under alkaline conditions
measurable at 405 nm absorbance on a spectrophotometer. Briefly, compounds or controls were
incubated (at a final concentration of 10 µM) with purified recombinant PTPsigma prior to the
addition of pNPP substrate. The reactions were stopped, products measured, and activity
calculated. We found that 19 compounds conferred inhibition of PTPsigma; the most potent
being compounds 6, 46, 48, and 49 which reduced PTPsigma activity by greater than 60%
(Figure 3.3A-F). We performed a library search for compounds similar to these effective
scaffolds and discovered 88 additional candidates which were tested in vitro in a slightly
modified manner. In this secondary screen, compounds were screened at a higher concentration
(100 µM) but with a shortened pre-incubation period to favor direct inhibitors. We found that the
majority of compounds inhibited PTPsigma by greater than 60% (Figure 3.4A-D). However,
compounds similar to 46 represented less than 10% of these effective inhibitors and in fact,
compound 46 itself reduced PTPsigma activity by only 30% under these modified conditions
(Figure 3.4B). Accordingly, we narrowed our focus to compounds similar to compounds 6, 48,
and 49 for subsequent investigation.

PTPsigma-targeted compounds lack selectivity
We next sought to establish the relative specificity of these compounds for PTPsigma
over other phosphatases. To this end, we measured the ability of compounds to inhibit PTP1B, a
classic PTP in a distinct subfamily. When biochemically validating competitive inhibitors, it is
critical that enzymes be tested using a substrate concentration at or below the Km (Tierno,
Johnston et al. 2007). We determined apparent Km values for both PTPsigma and PTP1B to be
approximately 250 µM pNPP and then selected an optimal substrate concentration of 200 µM
107

(Figure 3.5B-C,E-F). Next, we selected an enzyme amount (2 µg) for each phosphatase that
yielded linear product formation for the duration of the reaction while producing a maximal
signal of at least five-fold above background (Figure 3.5A,D). We began by measuring inhibition
with the three lead scaffolds as well as the pan PTP inhibitor, sodium orthovanadate. The
compounds were each tested across a range of doses from 0 to 500 µM. Unexpectedly, we
determined that not only did the lead compounds inhibit PTP1B, they did so slightly more
potently than PTPsigma (Figure 3.6A-D).
Retrospectively, we hypothesized that because the primary screen conditions included a
several hour pre-incubation period of enzyme and compound, the lead scaffolds may have been
inhibiting PTPsigma via an indirect mechanism. In particular, because PTP active sites are
maintained in a reduced state for preservation of the nucleophilic cysteine and primed for
optimal activity, these enzymes are extremely sensitive to oxidation (Tonks 2005). Thus,
oxidative species, such as hydrogen peroxide (H2O2), generated in assay buffers is a common
culprit for compromised phosphatase activity (Tautz and Mustelin 2007). To determine whether
the reaction conditions were favoring H2O2-mediated inhibition of PTPsigma, we repeated
compound experiments in the presence or absence of catalase, an enzyme which converts H2O2
into water and oxygen. In fact, catalase negated all inhibition conferred by compounds 48 and 49
(Figure 3.6E). In all reactions, even the DMSO vehicle control, catalase increased the activity of
PTPsigma (Figure 3.6E). This effect is likely explained by catalase-mediated removal of other
oxidative species formed in the reaction buffer, notably from DMSO, which is a mild oxidant.
To determine if any of the original primary compounds were true competitive inhibitors
of PTPsigma, we again altered the screening conditions to significantly diminish the potential for

108

H2O2 generation. We used a low dose of compound (10 µM) and reduced the pre-incubation
period to only 10 minutes, as H2O2 inhibition is time-dependent (Tautz and Mustelin 2007).
Under these conditions, we discovered that two compounds, 36 and 38, inhibited PTPsigma by
approximately 40%, slightly more potent than the equivalent dose of sodium orthovanadate
(Figure 3.6F). Following, we tested their strength of inhibition against PTPsigma compared to
PTP1B using serially diluted doses from 0 to 100 µM. Despite inhibition of PTPsigma with IC50
concentrations between 5 to 10 µM, we once again observed more enzymatic inhibition of
PTP1B (Figure 3.6G,H). In fact, when all 154 compounds used in this study were
comprehensively tested for inhibition of PTP1B under these improved conditions, we discovered
an essentially identical profile of inhibition as that of PTPsigma (data not shown).

109

DISCUSSION

Taken together, this virtual screening approach led to the identification of several small
molecule inhibitors of PTPsigma with modest potency in vitro. Computational docking
demonstrated that these compounds were molecularly accommodated by the D1 PTP domain of
PTPsigma, similar to a natural phosphotyrosine substrate. Accordingly, we predicted that these
compounds function as competitive inhibitors. Using conditions optimized for this mode of
inhibition, we found at least two compounds, 36 and 38, which inhibited PTPsigma potently at
IC50 concentrations between 5 and 10 µM. Unfortunately, we found that these compounds
inhibited PTP1B with similar or slightly better efficacy than PTPsigma.
A partial explanation for this promiscuity lays in the probable generation of H2O2 species
by at least a subset of these compounds (as was confirmed for compounds 48 and 49). Oxidation
and inhibition of PTP active sites by H2O2 has been well established as a physiological mode of
regulation (Salmeen, Andersen et al. 2003). A number of compounds, in particular those
containing quinones, have been documented to inhibit phosphatases through the generation of
H2O2 species (Urbanek, Suchard et al. 2001; Bova, Mattson et al. 2004; Tautz and Mustelin
2007). While none of the compounds we tested contained quinones, they may contain other
properties and side groups which participate in oxidation (Figure 3.2C). Although the precise
mechanism was not characterized, the ablation of inhibition by 48 and 49 achieved through cotreatment with catalase provides evidence that for at least these compounds, inhibition is partially
mediated through H2O2 generation. Despite this, compounds 36 and 38 inhibited both PTPsigma
and PTP1B under conditions where oxidation was not a major contributor. This suggests that
110

despite functioning as predicted competitive inhibitors designed to dock into the distinct
PTPsigma active site, these compounds do not contain selectivity for PTPsigma. They likely
have sufficient abilities to bind both PTP domains studied here as well as many others.
In response to this unexpected outcome, we propose an alternative strategy for the
development of small molecule inhibitors of PTPsigma with improved selectivity. First, despite
our findings, we believe an in silico docking approach is an effective method to identify unique
scaffolds that can lead to the development of both potent and selective inhibitors. However,
selectivity could be improved by implementing a secondary in silico screen which filters out
first-hit compounds that are predicted to have PTP1B binding. Scaffolds with the greatest
differential in binding energies, meaning favorable docking into PTPsigma but unfavorable
binding into PTP1B, would then be prioritized for in vitro validation. Because of this coupled
primary and secondary screening and more stringent hit requirements, the primary scaffold hit
list should be of greater scope than documented here and include several hundred leads from
disparate classifications.
In addition, a detailed active site analysis of PTPsigma should be undertaken. Especially
when sequence and structural conservation is high, as is the case for PTPs, the identification of
key residues or structural determinants comprising and surrounding the active site can facilitate
the identification or refinement of a selective inhibitor. For example, the undesirable druggability
of PTPs stems from the fact that residues forming the active site predominantly lie within highly
conserved motifs showing little sequence variability across the entire PTP family (Andersen,
Mortensen et al. 2001). After performing a structural alignment of PTPsigma and PTP1B, we
found that of five regions facing the active site, four were bona fide conserved PTP motifs
(Figure 3.7A-B). The remaining region contained residues linking two hydrophobic core motifs

111

(Figure 3.7A). In fact, of the four residues facing the active site which differed from PTPsigma
to PTP1B, we found that only two showed divergent orientations: PTPsigma residues R1498
(corresponding to S118 in PTP1B) and H1558 (corresponding to F182 in PTP1B) (Figure 3.7A).
In our initial docking experiment, R1498 did not contribute to p-Tyr binding directly although it
faces the active site. H1558 directly follows the WPD loop, a flexible hinge motif that folds
proximal to the active site in response to substrate binding. It is possible that choosing or
modifying a scaffold to incorporate molecular interaction with these residues could increase
selectivity for PTPsigma. However, if the sequence conservation proves to prevent selectivity
from being reasonably achieved with an active site-directed compound, targeting a less
conserved region of PTPsigma, such as the ectodomain, could be pursued. Alternatively, further
exploration of the crystal structure to identify adjacent binding pockets may prove productive.
Use of bidentate inhibitors which bind both the active site and a proximal pocket has proven
successful for inhibiting PTP1B (Zhang 2002). This dual binding mechanism is proposed to
improve affinity while the interaction with less-conserved adjacent residues improves selectivity
(Shen, Keng et al. 2001; Zhang 2002; Sun, Fedorov et al. 2003).
The potential utility of a PTPsigma inhibitor is evidenced by the significant finding that
loss of PTPsigma increases regeneration following spinal cord injury (Shen, Tenney et al. 2009).
Further, our discovery that loss of PTPsigma hyperactivates autophagy suggests that a small
molecule inhibitor of this enzyme would function as an autophagy agonist. As such, it could
serve as a useful molecular probe for research in this emerging field and potentially, also have
therapeutic utility for a host of diseases such as neurodegenerative diseases. Unfortunately, the
field of phosphatase drug discovery is plagued with issues and past setbacks, such as the
difficulty in achieving selectivity. Despite this, the critical role of PTPs in normal physiology as

112

well as disease etiology demands a solution to these problems. More rationally designed drug
discovery strategies, such as those discussed here for PTPsigma, may inform the development of
potent and selective PTP-targeted compounds for lead optimization.

113

MATERIALS AND METHODS

Structural modeling and substrate docking
The crystal structures of PTPsigma (PDB 2FH7) and PTP1B (PDB 1SUG) were retrieved
from the Protein Data Bank. The initial conformations of p-Tyr peptide were extracted from the
CD45-p-Tyr peptide complex structure (PDB 1YGU). The ICM program was used for protein
and substrate preparation (MolSoft, La Jolla, CA). p-Tyr peptide was docked into the active site
of PTPsigma and PTP1B with default parameters implemented in the ICM program. Structural
docking was completed by Yong Xu.

Virtual Library Screening (VLS)
We used the ZINC library (version 8; University of California San Francisco) of
ChemBridge compounds for virtual screening with the D1 active site of PTPsigma (PDB 2FH7).
GOLD program was used for virtual docking and ChemScore scoring function was used to rank
the top 200 hits with favorable binding energies (Cambridge Crystallographic Data Centre,
Cambridge). We used ICM clustering analysis (MolSoft) to identify 66 representative
compounds from unique clustering groups. Substructure similarity searching based on
compounds 6, 46, 48, and 49 within ChemBridge compounds identified 88 additional leads. VLS
was completed by Yong Xu.

In vitro phosphatase assays
The 66 in silico-identified compounds were purchased from ChemBridge and diluted to 5
or 10 mM in DMSO. GST-tagged recombinant PTPsigma containing all residues C-terminal to

114

the transmembrane domain (BC104812 cDNA; aa 883-1501) was generated in pGEXKG (Guan
and Dixon 1991). GST-tagged recombinant full-length PTP1B (BC018164) was generated with a
6xHIS tag in pGEXKG. Proteins were purified from BL21 Escherichia coli after isopropyl β-D1-thiogalactopyranoside (IPTG) induction and purity was confirmed by SDS-PAGE and
coomassie blue staining. Compounds were pre-incubated with recombinant enzymes in freshly
prepared phosphatase buffer (50 mM sodium acetate, 25 mM Tris-HCl, 3 mM DTT, pH 6.5) for
10 to 120 minutes, as indicated in figure legends. Following, para-nitrophenyl phosphate (pNPP;
Sigma S0942), initially diluted in assay buffer, was added to reactions for a final volume of 100
µl and reactions were carried out in a 37°C water bath for 15 to 30 minutes. Reactions were
quenched with 100 µl 1N sodium hydroxide (NaOH) and 180 µl was transferred to flat-bottom
clear 96-well plates. Absorbance of pNP product at 405 nm was measured on a
spectrophotometer and plotted. Background absorbance values of compound-only wells were
subtracted from the corresponding reactions. DMSO was included as a vehicle control.

Structural sequence alignment and analysis
SiteFinder (MOE) was used to predict the binding pocket residues of PTPsigma using the
crystal structure (PDB 2FH7). PTPsigma and PTP1B sequences were aligned based on structure
using MOE. Alignment was performed by Pooja Narang and Nathalie Meurice.

115

FIGURES

Figure 3.1 Workflow overview for PTPsigma inhibitor search. Virtual library screening
(VLS) was performed using the D1 active site of PTPsigma as a target. 66 target leads were
screened in vitro to hypothesize about structural determinants of potency using 4 chosen lead
compounds. 88 additional compounds based on the 4 leads were used for structure-activity
analysis. 2 non-selective but potent inhibitors resulted leading to the refinement of a model
which will include consideration of selectivity constraints and a revisit of the VLS.

116

Figure 3.1 (cont'd)

PTPsigma D1 domain
(PDB 2FH7)

+

Structure-based VLS

66 lead scaffolds
Hypothesis on the
structural determinant of
PTPsigma inhibitor ACTIVITY

4 PTPsigma inhibitors
identified
Structure Activity Relationship
on 88 PTPsigma compounds
Hypothesis on the
structural determinant of
PTPsigma inhibitor SELECTIVITY
2 nonselective PTP
sigma inhibitors
Refine protein
pharmacophore model
(include selectivity
constraints)

117

Figure 3.2. In silico screen for compounds which dock into PTPsigma. (A-B) Analysis of
phosphatase active sites and interactions with a natural substrate. (A) The D1 domain of
PTPsigma docked a phosphotyrosine (p-Tyr) substrate with favorability, although it revealed a
uniquely wide conformation. (B) The PTPsigma pocket was contrasted with that of the classic
PTP, PTP1B, which has a deep yet narrow active site to accommodate p-Tyr. Surface resonance
of the active sites are displayed (upper panels) with negatively (red) and positively (blue)
charged residues shown and substrates drawn in ball-and-stick form. Active site cross-sections
are shown with bound substrates (lower panels). All structures were generated with ICM
software (MolSoft). (C) In silico docking identified 66 lead scaffolds which dock into
PTPsigma. A ZINC library of compounds was virtually screened for the ability to bind the D1
PTP pocket of PTPsigma. From 200 top scoring hits, 66 uniquely-structured scaffolds were
chosen as leads. Chemical structures (from ChemBridge) are shown.

118

PTPsigma (D1)
A
+ p-Tyr

B PTP1B + pTyr

Cross Section

Surface Resonance

Figure 3.2 (cont'd)

C

1

12

2

3

13 14

4

5

6

7

8

9

10

11

15 16

17

18

19

20

21 22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

119

Figure 3.3. Primary in vitro screening of lead scaffolds filters for potency. (A) 66 lead
scaffolds were purchased from ChemBridge and tested in vitro for the ability to inhibit
PTPsigma phosphatase activity. Compounds were incubated with recombinant PTPsigma at a
final concentration of 10 µM for two hours prior to the addition of pNPP substrate.
Dephosphorylated product was measured by its specific absorbance at 405 nm as a readout for
PTPsigma activity. Relative activity expressed as a percent (normalized to DMSO) is plotted for
each compound. A threshold of 60% inhibition distinguished 4 leads from the 66 screened. Only
the 19 compounds which inhibited PTPsigma in any degree are shown. (B) The chemical
structures of the 4 lead compounds – 4, 46, 48, and 49- are highlighted. (C-F) The D1 active site
of PTPsigma docked with each of the 4 leads is shown (C-6; D-46; E-48; F-49). Surface
resonance of the active site is displayed with negatively (red) and positively (blue) charged
residues shown and compounds drawn in ball-and-stick form. All docking and poses were
completed with ICM software (MolSoft).

120

A

0
Compound

C
DMSO
18
34
1
15
43
13
11
12
32
45
17
36
21
19
28
46
48
6
49

Relative PTPsigma Activity
(percent DMSO)

Figure 3.3 (cont'd)

100

B

80

60

40

20

6

6

D
46

E

A

121
48

46
48

F
49

49

Figure 3.4. In vitro screen of additional compounds with structural similarities to the 4
leads. (A-D) 88 compounds were identified by a substructure search of ChemBridge compounds
for structural features relating to the 4 leads (A- similar to 6; B- similar to 46; C- similar to 48;
D- similar to 49). These were tested for potency of inhibition of PTPsigma activity as described
in Figure 3.3 with the following modifications: compounds were used at a final concentration of
100 µM and pre-incubated with PTPsigma for only 30 minutes. Sodium orthovanadate
(Na3VO4) is a pan inhibitor of PTPs and was included as a positive control. Original scaffolds
are indicated with asterisks.

122

Relative PTPsigma Activity
(percent DMSO)
100

120

100

120
0
Compounds similar
to 46
100

80

60

40

20

Relative PTPsigma Activity
(percent DMSO)

*

DMSO
126
125
106
117
120
104
107
109
102
113
128
127
115
101
105
110
112
118
119
122
124
103
111
6
121
Na3VO4
123
114
116

B
0

*

DMSO
130
133
129
46
138
131
137
139
132
Na3VO4
136
135

D
120

DMSO
150
146
151
142
141
143
154
Na3VO4
149
153
152
140
148
48
145
147

Relative PTPsigma Activity
(percent DMSO)

A

165
DMSO
181
186
178
158
180
185
166
169
176
157
170
168
160
164
162
184
159
188
167
171
49
175
179
Na3VO4
174
155
172
177
182
187
183
161
163
173

Relative PTPsigma Activity
(percent DMSO)

Figure 3.4 (cont'd)
Compounds similar to 6

80

60

40

20

C

123

120
Compounds similar to 48

80

60

40

20

0

*

Compounds similar to 49

100

80

60

40

20

0

*

Figure 3.5. Optimization of in vitro screening conditions for selectivity analysis. (A-C)
Analysis of the enzymatic properties of PTPsigma. (A) The linear formation of product by
various quantities of recombinant GST-PTPsigma was observed through time-course reactions.
pNPP-phosphatase assays were completed as described in Materials and Methods. A saturating
dose of 1 mM pNPP was used and background-corrected absorbances of dephosphorylated
product are plotted by time of reaction. Each plot stems from the legend-indicated quantities of
PTPsigma. (B) 2 µg enzyme was chosen from (A) for analysis of activity with varying doses of
pNPP substrate. Each plot represents a unique dose of pNPP (indicated in the legend).
Background-corrected absorbance values of dephosphorylated product are plotted by time of
reaction. (C) From the slopes of the lines of (B), initial velocities (Y-axis) were calculated for
PTPsigma phosphatase activity at each of the indicated substrate concentrations (X-axis). From
this, an approximate Km of 250 µM is observed. (D-F) The analyses outlined in (A-C) were
repeated with recombinant GST-PTP1B. (G). 20 µg purified proteins (GST-PTPsigma-CTF or
full length GST-PTP1B) were resolved by SDS-PAGE and stained with coomassie blue. Purity
of products is shown with free GST in the fractions seen at ~26 kDa.

124

G

GSTkDa PTPsigma
115 -

30

00

20

00

0

10

F
0.25
0.20
0.15

00

50

00

40

30

00

0.10
0.05
0

00

pNPP (µM)
1.2
4000
1.0
2000
0.8
1000
500
0.6
250
0.4
125
0.2
62.5
0
0
0 5 10 15 20
Reaction Time (min.)

00

E

pNPP substrate (µM)

20

PTP1B

pNP

3.0
2.5
2.0
1.5
1.0
0.5
0

Enzyme (ng)
4000
2000
1000
500
250
125
62.5
0
0 5 10 15 20
Reaction Time (min.)

0.200
0.175
0.150
0.125
0.100
0.075
0.050
0.025
0

0

pNP Absorbance

D

0

C

10

0.5

pNPP (µM)
4000
2000
1000
500
250
125
62.5
0
0 5 10 15 20
Reaction Time (min.)

pNP Absorbance
(405nm) per min.

1.0

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

pNP Absorbance
(405nm) per min.

1.5

pNP Absorbance

2.0

PTPsigma

pNP

Enzyme (ng) B
4000
2000
1000
500
250
125
62.5
0
0 5 10 15 20
Reaction Time (min.)

pNP Absorbance

2.5

pNP Absorbance

A

0
40 0
00
50
00

Figure 3.5 (cont'd)

pNPP substrate (µM)

GSTPTP1B

82 64 49 37 -

125

Figure 3.6. Relative inhibitions of PTPsigma and PTP1B by lead compounds. (A-D) As an
initial assay, the inhibition of PTPsigma and PTP1B phosphatase activity was measured using
the pan PTP inhibitor, sodium orthovanadate (A), and the three most potent original leads (B-6;
C-48; D-49) by standard pNPP assay outlined in Materials and Methods. Briefly, 2 µg
recombinant PTPsigma (block squares) or PTP1B (gray circles) was incubated with the doses of
compound shown for 30 min at 37°C. Then, 200 µM pNPP was added for a 10 (PTPsigma) or 25
(PTP1B) min reaction at 37°C (reaction time monitored to yield sufficient signal-to-background
ratio while still linear). Dephosphorylated product was measured and relative PTP activity
(percent of DMSO vehicle control) is plotted. Approximate IC50 values can be derived from the
plots and are discussed in Results. (E) PTPsigma activity towards pNPP was measured in the
presence of DMSO vehicle, compound 48, or compound 49 as described previously. Bovine liver
catalase (50 units per reaction) was included (+) to degrade hydrogen peroxide. Relative
PTPsigma activity was plotted (percent of activity in DMSO without catalase). (F) The original
66 VLS-identified scaffolds were re-screened in vitro under conditions which minimized the
possibility of hydrogen peroxide generation and non-specific inhibition. Here, 10 µM compound
was pre-incubated for 10 min at room temperature followed by a 15 min reaction with pNPP.
Compounds 36 and 38 inhibited PTPsigma by >40% (dotted line threshold), more than sodium
orthovanadate at the same dose. Relative activity (percent DMSO) was plotted by all compounds
tested. (G-J) Dose response-inhibition of PTPsigma (black square) and PTP1B (gray circles) was
measured for compounds 36 and 38 using a 10 min room temperature pre-incubation period and
15 min 37°C reaction time with 200 µM pNPP. Relative activity (percent DMSO) was plotted by
the indicated concentrations.

126

Compound (µM)

127
Compound (µM)

0

10

25

3

120
100
80
60
40
20
0

6.

C

6

36
38

Compounds
Compound 38

Na3VO4

A

Inhibition PTPsigma
Phosphatase Activity
(% vehicle)

50
40
30
20
10
0

1.

4

Phosphatase
Activity (% vehicle)

Compound 36

0.

0

10

25

3

6.

6

1.

120
100
80
60
40
20
0

4

B

0.

Phosphatase
Activity (% vehicle)

Figure 3.6 (cont'd)

Figure 3.7. Methods for building specificity to small molecule inhibitors of PTPsigma. (A)
Using the crystal structure of PTPsigma (PDB 2FH7), SiteFinder (MOE) was used to predict the
D1 binding pocket. 23 residues were found and are highlighted in yellow. PTPsigma (sequence
(1)) and PTP1B (1SUG; sequence (2)) were aligned based on their structures (using MOE).
Residues which differ between the two are highlighted in green. The binding pocket is made up
of highly conserved PTP motifs: (1) the p-Tyr recognition (KNRY) motif; (2) the least conserved
region spanning two conserved hydrophobic motifs; (3) WPD loop; (4) cysteine-based catalytic
site (for sequence reference, the yellow cysteine within this motif corresponds to residue 1589 in
PTPsigma); (5) water/Q-loop. (B) The PTPsigma-D1 and PTP1B structures were aligned to
highlight their structural conservation. Structures are depicted in a ribbon diagram (PTPsigma:
turquoise; PTP1B: purple). A pTyr peptide docked into the active site is depicted by the red and
gray spheres.

128

Figure 3.7 (cont'd)

A
1
2

MLSHPPIPIADMAEHTERLKANDSLKLSQEYESIDPGQQFTWEHSNLEVNKPKNRYANVI
----------EMEKEFEQIDKS-GSWAAIYQDIRHEASDFPCRVAKLPKNKNRNRYRDVS

1
2

AYDHSRVILQPIEGIMGSDYINANYVDGYRRQNAYIATQGPLPETFGDFWRMVWEQRSAT
PFDHSRIKLHQED----NDYINASLIKMEEAQRSYILTQGPLPNTCGHFWEMVWEQKSRG

1
2

IVMMTRLEEKSRIKCDQYWPNRG--TETY--GFIQVTLLDTIELATFCVRTFSLHKNGSS
VVMLNRVMEKGSLKCAQYWPQKEEKEMIFEDTNLKLTLISEDIKSYYTVRQLELENLTTQ

1
2

EKREVRQFQFTAWPDHGVPEYPTPFLAFLRRVKTCNP--PDAGPIVVHCSAGVGRTGCFI
ETREILHFHYTTWPDFGVPESPASFLNFLFKVRESGSLSPEHGPVVVHCSAGIGRSGTFC

1
2

VIDAMLERIKPEK---TVDVYGHVTLMRSQRNYMVQTEDQYSFIHEALLEAVGCG----LADTCLLLMDKRKDPSSVDIKKVLLEMRKFRMGLIQTADQLRFSYLAVIEGAKFIMGDSS

1
2

------------VQDQWKELSHEDL

B

129

CHAPTER 4

Mathematical modeling of autophagic vesicle dynamics

Martin KR, Barua D, Chaudhury S, Sinitsyn N, Stites E, Posner R, Hlavacek W, MacKeigan JP
(2011). Mathematical modeling of autophagic vesicle dynamics.
Prepared for submission to PLOS Biology

130

INTRODUCTION

Macroautophagy (autophagy) is an evolutionary conserved cellular recycling program
whereby cytosolic contents are sequestered in double-membrane vesicles and delivered to the
lysosome for degradation (Klionsky 2007). The breakdown of autophagic cargo generates basic
biochemical building blocks, such as fatty acids and amino acids, which can be exported back to
the cytosol for reuse. This process is utilized by the cell to rid itself of long-lived or damaged
proteins and organelles in an effort to maintain homeostasis. In addition, autophagy is
dramatically upregulated during bouts of stress or starvation where it serves to generate an
internal nutrient pool, an energetically favorable alternative to de novo synthesis.
Autophagy is executed in four stages: initiation, nucleation, maturation, and completion
(Figure 4.1A). Nutrient-activated mTORC1 (mammalian target of rapamycin complex 1)
controls autophagy induction by inhibiting the gatekeeper complex, ULK1-mAtg13-FIP200
(Ganley, Lam du et al. 2009; Hosokawa, Hara et al. 2009; Jung, Jun et al. 2009). When
mTORC1 activity is low (i.e. during starvation), ULK1 is functional and permits the nucleation
of an isolation membrane, or phagophore. The synthesis of this cup-shaped double-membrane
structure is promoted in large part by Vps34, a lipid kinase which incorporates PI(3)P onto
autophagic membranes, and Atg9, a transmembrane protein and putative lipid-carrier involved in
membrane growth (Young, Chan et al. 2006; Webber and Tooze 2010). Expansion of the
phagophore and eventual closure into a mature autophagosome is executed by two ubiquitin-like
conjugation systems involving Atg5-Atg12-Atg16L and LC3 (Atg8 in yeast) (Ohsumi and
Mizushima 2004). The completion of this process is marked by the fusion of the autophagosome

131

with a lysosome (or endocytic compartment destined for the lysosome) and degradation of
sequestered material (Dunn 1990; Berg, Fengsrud et al. 1998; Klionsky 2007).
Despite its initial discovery 50 years ago, essential questions about autophagy remain
unanswered (Chen and Klionsky 2011). Elegant studies in yeast and mammalian systems have
identified over 30 proteins required for autophagy, however, their molecular mechanisms and
regulation are generally uncharacterized. Further, autophagy contributes to cell fate in a
complicated manner that is incompletely understood. In a fundamental sense, it functions as a
survival mechanism to delay or prevent apoptosis in response to stress, however, it can also
participate in cell death when activated in excess or for prolonged periods (Codogno and Meijer
2005). Disregulated autophagy has been found to contribute to the pathology of several diseases,
namely cancer and neurodegeneration (Rosenfeldt and Ryan 2009; Wong and Cuervo 2010).
Therefore, greater knowledge of the regulation, molecular underpinnings, and cellular
consequences of autophagy is critical not only for an understanding of normal physiology but
also for the comprehension of disease etiology and the rational design of therapeutics.
An effective strategy for studying complicated cellular processes, such as autophagy,
involves the construction of computational or mathematical models. These models, developed
and refined from experimental knowledge, can serve as tools used to interrogate signaling
pathways, formulate novel hypotheses about a system, and make predictions about cell signaling
changes induced by specific interventions (i.e. genetic changes, treatment with targeted
compounds) (Hopkins 2008). Here, we present the development of a novel mathematical model
describing autophagic vesicle dynamics in a mammalian system. We utilized live-cell kinetic
microscopy to quantify the synthesis and turnover of autophagic vesicles in response to various
perturbations in single cells. The stochastically simulated model was tested by both chemical and

132

genetic modulation of autophagic machinery and found to accurately predict vesicle dynamics
observed experimentally. Further, the model predicted a positive correlation of LC3
concentration and vesicle size which has been previously reported in publications from
independent groups. Taken together, we believe this model to be useful and accurate and as such,
it will serve as the foundation for a more comprehensive model of autophagy in the future.

133

RESULTS

Construction of a conventional model of autophagic vesicle synthesis and turnover
As an initial attempt at modeling autophagy, we chose to construct a relatively simple
model which described autophagic vesicle dynamics from formation to turnover (Figure 4.1B).
Using knowledge of molecules integral for this process, we outlined a system which began with
Vps34-catalyzed PI(3)P production. Input to autophagy triggers the activation of Vps34
(activated form depicted as Vps34*, below). The activation occurs through a step-wise function
with unique rate-constants representing times prior to 0 (t < 0; Kact I1) and times after 0 (t > 0;
Kact I2) (1-2). This distinction allows for a controlled pre-incubation of cells experimentally
followed by exposure to specific autophagy-relevant conditions at time 0. Vps34 can also be
deactivated by a first-order process (3). Active Vps34 catalyzes phosphorylation of PtdIns to
form PI(3)P (PI3P, below) (4).

Vps34

Vps34*

(1)

I(t) = kact I1 t < 0; kact I2 t > 0
Vps34*
Vps34* + PtdIns

Vps34

(Vps34* · PtdIns)

(2)
(3)

Vps34* + PI3P

(4)

A PI(3)P-phosphatase was included to counterbalance Vps34 activity as there have been
multiple reports providing evidence for the existence of such an enzyme functioning in
autophagy (5) (Vergne, Roberts et al. 2009; Taguchi-Atarashi, Hamasaki et al. 2010).

134

PI3P

PtdIns

(5)

PI(3)P functions as a second messenger which recruits lipid-binding proteins to specific
subcellular compartments, in this case the isolation membrane (IM). We included WIPI (there
are two isoforms in mammals, WIPI-1 and WIPI-2), a mammalian protein which reversibly
binds PI(3)P during autophagy, to serve as a PI(3)P effector in our model with the complex
represented by (WIPI · PI3P) (6) (Proikas-Cezanne, Waddell et al. 2004; Polson, de Lartigue et
al. 2010).

WIPI + PI3P

(WIPI · PI3P)

(6)

The cycling of Atg9, that is its movement from peripheral locations to the isolation
membrane assembly site, is important for membrane growth at this stage and it is hypothesized
that Atg9 functions as a lipid-carrier (Webber and Tooze 2010). Although the mechanism is not
known, it has been shown that in mammals, Atg9 cycling is dependent upon Vps34 activity, and
in yeast, Atg9 function requires Atg18, a PI(3)P-binding protein (Reggiori, Tucker et al. 2004;
Young, Chan et al. 2006). Further, Atg5 fails to be recruited to the IM in yeast void of Atg9
(Suzuki, Kirisako et al. 2001). Accordingly, we placed Atg9 downstream of PI(3)P formation
and as a requisite for IM synthesis. Atg9 has been shown to self-oligomerize and was also
detected in an asymmetric distribution on autophagic membranes (He, Baba et al. 2008; Gao,
Kang et al. 2010). Therefore, we included a multistep cascade whereby PI(3)P-activated WIPI
engages Atg9 (7) and only after a threshold of this multimeric complex is reached (generated in
(8)), is an IM formed and Atg9 deactivated (10). Below, X0 represents the (Atg9.WIPI.PI3P)

135

complex and Xi represents the complex modified n times. Because the IM is formed de novo, it
is depicted as (Ø  IM) and is catalyzed by the complex at the required threshold, Xn (10). The
complex can dissociate when Atg9 exits the complex (10).

Atg9 + (WIPI · PI3P)
X0
Xi

X1

···

X0
Xn-1

(7)
Xn

Atg9 + (WIPI · PI3P), i = 0, …, n

Ø

IM, Xn

X0

(8)
(9)

(10)

Once nucleated, the Atg5-Atg12-Atg16 complex (Atg5-12-16 below) at the IM catalyzes
LC3-II formation from LC3-I (representing knowledge that the complex functions like an E3like enzyme controlling LC3-II lipidation to autophagic membranes) (11) (Fujita, Itoh et al.
2008). The IM gives way to a freely diffusing autophagic vesicle (Vfree) once LC3-II reaches a
threshold on the membrane (12,13). Blinking of vesicles between visible and invisible accounts
for system noise, discussed later (14)

Atg5-12-16 + LC3-I
LC3-II0

(Atg5-12-16 · LC3-I)

LC3-II1

···

LC3-IIn-1

IM
Vfree

Vfree
V*free
136

Atg5-12-16 + LC3-II
LC3-IIn

(11)
(12)

(13)
(14)

Turnover is marked by the conversion of vesicles from free to lysosomal-deposited
(Vlyso) (15).

Vfree

Vlyso, V*free

Vlyso,

(15)

Because this dynamical model was focused simply on key players involved in vesicle
dynamics, we chose to solve using conventional mechanics. Specifically, we constructed a
stochastic model which was best suited to reflect the large dynamic range of vesicle
measurements observed experimentally using single cells.

Experimental design for measuring autophagic vesicle dynamics
We implemented a cell system which allowed us to monitor both the synthesis and
turnover of positive autophagic vesicles (AVs). We generated a monoclonal U2OS cell line
stably expressing LC3 fused to a fluorescent tag (U2OS-ptfLC3), similar to the line previously
used by our laboratory (Kimura, Noda et al. 2007; Martin, Xu et al. 2011). While this reporter
contains both an mRFP and GFP sensor, we utilized only the GFP dynamics for the purposes of
this model. Accordingly, it is referred to as U2OS-GFP-LC3 from this point forward for
simplicity. Subcellular GFP-positive punctae can be captured by fluorescent microscopy and
serve as an accurate marker of LC3-positive AVs. We designed an image processing protocol,
including deconvolution and intensity thresholding, which allowed for the accurate
quantification of AVs from single cells using fluorescent images (Figure 4.2A).
GFP-LC3 has several properties which we exploited for measurement of vesicle
formation and turnover rates. First, while LC3 incorporated on the outer membranes of AVs is
137

recycled back to the cytosol, LC3 embedded on the inner membrane is carried into the lysosome
and degraded along with vesicle cargo (Tanida, Minematsu-Ikeguchi et al. 2005; Klionsky,
Abeliovich et al. 2008). Therefore, LC3 turnover is an effective measure of autophagic flux
(Tanida, Minematsu-Ikeguchi et al. 2005). Specifically, the rate at which LC3 accumulates in
response to lysosome inhibition, measured at either the level of protein or vesicle abundance,
correlates with the rate of autophagy. Further, the GFP moiety of the GFP-LC3 fusion is pHsensitive and quenched by the acidity of the lysosome. Thus, GFP-LC3 selectively labels
autophagosomes, but not autolysosomes (Kimura, Noda et al. 2007).
We used this knowledge to construct a simple system for calculating rates of autophagic
vesicle production (the synthesis of a newly observed GFP-LC3 vesicle) and turnover (that is,
the deposition of a GFP-LC3 vesicle in the lysosome). When lysosomal function is
uninterrupted, a GFP-LC3-positive punctae represents an AV at a discrete stage of maturation
between production and degradation (an IM or mature autophagosome). In the absence of stress
or starvation, this number is generally low and stable, representing the steady state of autophagy.
When lysosomal fusion is blocked by treatment with bafilomycin A1, a lysosomal proton pump
inhibitor, the turnover of vesicles in the lysosome is prevented. Because their production
continues unimpeded, vesicles then accumulate at a rate that corresponds to the level of
autophagic flux (Figure 4.2B). Over a two hour time-course, we observed a substantial increase
in vesicle abundance when lysosomes were inhibited, but not when they were active (Figure
4.2C,D).
With this concept in place, we sought to parameterize a simple model of autophagy. We
imaged and quantified vesicle dynamics in cells cultured in full nutrient media (to determine the
level of basal autophagy) and in cells treated with the autophagy-inducing mTOR inhibitor,

138

rapamycin (to measure the level of induced autophagy). Again, we observed an increase in
vesicle abundance upon lysosomal inhibition and found that this increase was exacerbated with
rapamycin treatment (Figure 4.3A-D).
In order to calculate rates of vesicle synthesis and turnover from these experiments, we
implemented several simple equations. First, the number of vesicles synthesized (Vsyn) from
time t to t+1 was calculated using the following equation, where Vli(t) = the vesicle count at time
t in the presence of lysosome inhibition (bafilomycin treatment) and Vli(t+1) = the vesicle count
at time t+1 in the presence of lysosome inhibition (16).

Vsyn = Vli(t+1) – Vli(t)

(16)

Second, the number of vesicles turned over during that same time span was calculated by
coupling the formulation above with vesicle measurements obtained in cells with active
lysosomes (vehicle-treated). Here, the number of vesicles turned over (Vdeg) from time t to t+1
can be calculated using the equation below where Vla(t) is the number of vesicles at time t in the
presence of active lysosomes (i.e. vehicle treatment) and Vla(t+1) is the number of vesicles at
time t+1 in the presence of active lysosomes (i.e. vehicle treatment) (17). An example is
illustrated in Figure 4.2B. There, by observing cells with inhibited lysosomes, we can determine
that 4 vesicles were formed during the time interval captured (from 0 min to 8 min). However, in
cells with active lysosomes, we saw no change in vesicle number during that time interval,
indicating that 4 vesicles were normally turned over.
139

Vdeg = Vsyn – [Vla(t+1) – Vla(t)]

(17)

Employing these calculations, we plotted the total number of vesicles synthesized or
turned over with time basally or upon rapamycin treatment (Figure 4.3E,F). From the slopes of
these lines, we determined the mean rate of vesicle synthesis during basal autophagy was 0.43
vesicles per minute while vesicles were turned over at a rate of 0.38 per minute. Rapamycin
increased both vesicle synthesis and turnover to 0.68 and 0.66 vesicles per minute, respectively
(Figure 4.3E,F).

Initial parameterization of the model
To begin parameterization of the model, we made estimates for starting concentrations of
molecules which were consistent with levels measured in a gene expression dataset obtained
from U2OS cells (Jeff Kiefer, unpublished data). Next, we used experimentally-derived vesicle
counts and dynamic rates from the rapamycin and nutrient data sets collected under lysosomal
inhibition to fit the model (Figure 4.4A,B). We tuned the adjustable parameters of the model so
that theory curves, generated from the average of many stochastic simulations, matched the
experimentally derived data (Figure 4.4A; solid and dotted lines). As one example of model
tuning, we introduced positive feedback such that as the number of free vesicles (Vfree)
increases, so does the rate of turnover (Vfree  Vlyso). This was in response to our observation
that when lysosomes are active, vesicle counts always remain relatively small, even during

140

rapamycin-induced autophagy. This indicates that when vesicles are produced, they are
efficiently turned over, a feature of the system then incorporated into the model specification.
Reassuringly, the simulated curve for the pre-incubation period (time -90 to 0 minutes)
closely matched that of the experimental data, which was not used to tune the model (Figure
4.4C). To further support accuracy of the model, we uncovered experimental data curves from
individual cells which closely matched theory curves generated from individual stochastic
simulations (Figure 4.4D-G).
During our analysis of the experimental data, we observed fluctuations in the vesicle
counts that could not be explained biologically given our experimental assumptions. For
example, during bafilomycin A1-induced lysosomal inhibition, the vesicle count should never
drop from one time point to the next as that would indicate vesicle degradation; however,
oscillations in the data, including such drops, were observed (see curves in Figure 4.3D).We
hypothesized that this noise stems from imperfections of the image analysis procedure as
multiple vesicles observed juxtaposed to one another could incorrectly be counted as a single
object. Additionally, it is possible for a vesicle to be hidden from view during image capture (i.e.
in the perinuclear region where the cell is thicker) then reappear in a later image at a new visible
cellular location. With relatively small observed vesicle counts, these occurrences could result in
noticeable data fluctuations. To account for this noise, regardless of the ultimate source, we
parameterized the model to include randomness generated by the ability of a freely diffusing
vesicle (Vfree) to merge with another, then split (captured in vesicle blinking; reaction (14)).
Taking this noise into account, we could extrapolate true vesicle counts from visible
(experimentally measured) vesicle counts (Figure 4.5A-D).

141

After the model was tuned, it predicted a relatively low abundance of true isolation
membranes (IMs) compared to freely diffusing vesicles (Vfree). To provide data to support this
low number of IMs in our cell system, we immunostained cells using antibodies targeting Atg12,
a marker of nascent IMs, and quantified on average 0.2 Atg12-punctae per cell in full nutrients,
and roughly 2.0 Atg12-punctae per cell following rapamycin treatment ((Figure 4.4E;
(Mizushima, Yamamoto et al. 2001)).
In sum, the tuned model produced simulated data in close alignment with experimentallyderived measurements. Thus, we next aimed to test the model by making logical predictions
about the requirement for key autophagic machinery specified in the model.

The model accurately predicts output of genetic and chemical modulation of autophagic
machinery
As Vps34 is required to produce PI(3)P for IM and vesicle formation, outlined in the
model reactions, the model predicted that reduction of Vps34 would substantially reduce the
formation and turnover of vesicles from control cells (Figure 4.5F). To test this prediction, we
treated cells with wortmannin, a potent pan PI3K inhibitor which ablates Vps34 activity (Figure
4.6A). In agreement with the model prediction, we found that wortmannin treatment reduced the
abundance of vesicles in the presence of active lysosomes and also reduced their accumulation
upon lysosomal inhibition (Figure 4.6B,C). Employing formulas (16) and (17), we determined
that wortmannin reduced vesicle synthesis and turnover rates to approximately 27% and 29%
that of vehicle-treated cells, respectively (Figure 4.6D,E).
Next, we sought to formulize and test a prediction about Atg9, whose precise function
and requirement in autophagy, especially in this cell type, is less clear than that of Vps34. Again,
142

the model predicted that a 90% reduction in Atg9 content would cause a substantial reduction in
vesicle synthesis and turnover, owing to its requirement outlined in the model reactions (Figure
4.7F). To test this experimentally, we transfected U2OS-GFP-LC3 cells with siRNAs targeting
Atg9 for 48 hours prior to analysis and live cell imaging. Similar to Vps34 inhibition, we
determined that loss of Atg9 substantially augmented vesicle synthesis and turnover rates to
approximately one-third that of wild-type cells (Figure 4.7A-E).

The model predicts a positive correlation of LC3 concentration and autophagic vesicle size
Finally, we used simulations from the tuned model to form the hypothesis that levels of
LC3 would positively correlate with autophagic vesicle size. When input parameters were
modulated to incrementally increase LC3 concentrations, the model predicted a concurrent
increase in the mean vesicle size (Figure 4.8). The small nature of autophagic structures and the
resolution of objects which can be achieved by conventional fluorescent microscopy prevented
us from accurately testing this prediction in our model system. However, we found reports in the
literature by independent groups which supported this phenomenon. In studies in both yeast and
mammalian cells, it has been observed that increased LC3 levels correlated with increased size
of autophagic vesicles (Nakagawa, Amano et al. 2004; Xie, Nair et al. 2008).

143

DISCUSSION

In summary, we have generated a mathematical model which accurately simulates
autophagic vesicle dynamics, including synthesis and turnover, under conditions of both basal
and induced autophagy. Model predictions of Vps34 and Atg9 depletion were confirmed with
experimental data and the model was further tuned to reflect these results.
Intriguingly, our model revealed that, at least in our cell system, once an autophagic
vesicle forms, it is efficiently turned over. That is, during rapamycin-induced autophagy, we
observed increased vesicle synthesis as measured kinetically during lysosomal inhibition.
However, we did not see a significant rise in the number of GFP-LC3 vesicles present in cells
with active lysosomes (vehicle/DMSO treatment), suggesting that those vesicles being
synthesized were normally efficiently turned over. This indicates a short half-life of autophagic
vesicles in this experimental system, although, because our sensor is GFP-LC3, a vesicle is no
longer detected upon entry to the acidic lysosome (due to GFP fluorochrome quenching)
(Kimura, Noda et al. 2007). Therefore, tracing autophagic vesicles with a more stable marker,
such as endogenous LC3 or RFP-LC3, would allow insight into the entire lifespan of an
autophagic vesicle from nucleation to complete degradation inside the lysosome.
We chose to model autophagy using stochastic specification given the considerable
dynamic range of biological data captured from single cells. Discrete stochastic models are an
ideal choice when randomness, or noise, is present in a system. In addition to detection of a wide
range of raw vesicle counts and dynamic rates of synthesis and turnover across the population of
cells studied, we observed fluctuations of vesicle dynamics within single cells over time. The
variability observed in our system may have a number of underlying explanations from

144

variations in cell-cycle status (our cell populations were asynchronous) to differences in genetic
makeup or protein concentrations (Spencer and Sorger 2011). Regardless of the source, this
dynamic variability may be a physiologically important feature of autophagy signaling within a
larger network (i.e. a population of cells). In the absence of single-cell measurements, this
heterogeneity would not be been observed and thus, our results contribute to accumulating
evidence that resolution at the level of single cells is critical for a complete understanding of
cellular processes (Spiller, Wood et al. 2010; Tay, Hughey et al. 2010; Spencer and Sorger
2011).
Experimentally captured data agreed well with model predictions of decreased vesicle
production and turnover upon loss of the key autophagy proteins, Vps34 and Atg9. Further, the
model led us to generate the hypothesis that LC3 concentrations correlated with increased size of
autophagic vesicles. While we did not directly test this prediction, we found evidence from other
published reports which supports this hypothesis. Studies from both yeast and mammalian cells
have reported that increased LC3 levels correlate with increased size of autophagic vesicles,
consistent with a proposed function for LC3 in phagophore elongation (Nakagawa, Amano et al.
2004; Xie, Nair et al. 2008).
The model presented here, while accurate and useful, is simple and includes only
minimal molecular details. To comprehensively model autophagy, an expansion of this
framework is required. Models which recognize and elucidate the vast inputs to autophagy,
including contributions that feed downstream through mTORC1 and also inducers that function
independent of mTOR (e.g. IP3 depletion), will be important (Sarkar, Floto et al. 2005).
Additionally, models built on a slow time scale will be useful for understanding the
consequences of autophagy on cell fate. Resolution of the apparent dichotomy whereby
145

autophagy can confer a survival advantage to cells under stress yet also contribute to cell death
will be aided by an informative computational model. Finally, a model based on the one
presented here, will be improved with the inclusion of all known molecular machinery involved
in vesicle dynamics. The increased complexity of such a model will necessitate the use of rulebased mechanics, a modeling approach optimized for reaction networks comprising large
numbers of proteins with complicated interactions, activities, and post-translational
modifications (Hlavacek, Faeder et al. 2006). This rule-based model, driven experimentally
through the measurement of several key nodes along the pathway, will be the focus of our future
efforts.
Arguably the most important feature of a mathematical model is that it be predictive,
allowing the generation of novel hypotheses which can be explored experimentally. Our model
of autophagy vesicle dynamics, presented here, proved to be predictive through a series of tests,
and its accuracy and utility demonstrate that it can serve as the solid foundation of a more
comprehensive model of autophagy in the future.

146

MATERIALS AND METHODS

Experimental design and live-cell imaging
A monoclonal U2OS cell line was generated which displayed moderate expression
(easily detected fluorescent punctae without saturation of signal) of ptfLC3 plasmid (Addgene
plasmid 21074) (Kimura, Noda et al. 2007). Although this construct expresses both an mRFP
and GFP tag fused to LC3, we only measured GFP-LC3 dynamics and thus, for simplicity, we
refer to GFP-LC3 and U2OS-GFP-LC3 cells within the text.
U2OS-GFP-LC3 cells were at a density of 150,000 cells on 35 mm dishes with a number
1.5 coverglass bottom in 2 ml normal cell maintenance media (McCoy’s 5A (Invitrogen)
supplemented with 10% fetal bovine serum (FBS) (CellGro)). 24 to 48 hours later, cells were
switched to fresh insulin-containing media (McCoy’s 5A supplemented with 20 mM HEPES (to
buffer the pH; Invitrogen) and 100 nM insulin (Invitrogen)) for 1 to 3 hours prior to treatments
(henceforth called “full nutrient media”). Cells were subjected to a 90 min pre-incubation period
with 2ml full nutrient media and either 125 nM bafilomycin A1 (BafA1; A.G. Scientific) or an
equivalent amount of vehicle (DMSO at 1.25 µl per 2 ml; Sigma). Data from these two preincubation periods was captured only once with cells imaged every 4 min for 90 min (see
below).
For remaining experiments, the pre-incubation period was completed in an incubator and
not imaged. Following, cells were removed from the incubator, media aspirated, and replenished
with the desired condition media with BafA1 (for inhibited lysosomes) or DMSO (for active
lysosomes). Treatments included full nutrient media (DMSO or BafA1), 50 nM rapamycin

147

(Calbiochem) in full nutrient media (DMSO or BafA1), and 100 nM wortmannin (SigmaAldrich) in full nutrient media (DMSO or BafA1).
For live-cell imaging, GFP-LC3 was imaged in the FITC channel using a 60x oil
objective and a Nikon Ti Eclipse fluorescent microscope. Cells were imaged live by maintaining
a humid environment at 37°C and 5% CO2 in an environmental chamber fixed around the
microscope stage. For imaging, ten fields of view were chosen for each experiment and software
set to automatically image each position every 4 min using a perfect focus function to maintain
the desired focal plane. Fields of view were chosen for their inclusion of healthy cells which
were adherent, at the periphery of a cluster, and moderately expressing GFP-LC3. All
representative images are of the FITC channel displayed in black-and-white for easier
visualization of punctae.

siRNA-mediated Atg9 knockdown
U2OS-GFP-LC3 cells were seeded at a density of 75,000 cells on 35 mm dishes with a
number 1.5 coverglass bottom in 2 ml normal cell maintenance media (McCoy’s 5A with 10%
FBS). The next day, cells were transfected with either control (non-targeting) siRNAs or a pool
of four siRNAs targeting Atg9 (Atg9A: SI04364675, SI04162781; Atg9B: SI04364535,
SI04309389) at a final concentration of 25 nM using 2 µl oligofectamine (Invitrogen) in 0.2 ml
Optimem (Invitrogen) and 0.8 ml normal cell maintenance media. Image capture and
quantification was completed 44 to 48 hours post-transfection. Knockdown was measured by
qRT-PCR using Atg9A-specific primers and an endogenous GAPDH control. Delta delta Ct
method was used to determine relative copy numbers from control and Atg9 siRNA samples.

148

Atg12 immunofluorescence
Atg12 immunofluorescence was performed as described previously and imaged with a
60x oil objective on a Nikon Eclipse Ti microscope (see Chapter 2 Section 1).

Image processing and vesicle quantification
To quantify, images were deconvolved using a 2D blind deconvolution with three
iterations and settings of normal cell thickness and normal noise level. Following, regions of
interest were drawn around the borders of each cell. If nuclei or the perinuclear region had high
background fluorescence, this region was omitted from the region of interest. Intensity thresholds
were set to include all pixels equal to and greater than 500 units of intensity above the mean
background fluorescence from the cell (to control for background fluorescence and minor
variation in fluorescent expression level). Objects were quantified using an automated object
count function from this thresholded region and exported to excel. For this model, vesicle count
was the most utilized parameter although other parameters (size and intensity) were collected.

Mathematical modeling
The model was first described as a series of reactions (outlined in Results). Starting
5

concentrations of all molecules was set to 10 molecules per cell, consistent with standard values
for signal transduction proteins. Log2 transformed gene expression (microarray) data from U2OS
cells (Jeff Kiefer, unpublished data) confirmed this assumption to be fair. Most kinetic
parameters were assumed to be constant (typical forward constants (kf) =1 µM s-1 and reverse
constants (kr) = 0.1 µM s-1). Other system-specific parameters (e.g. IM nucleation rates, IM
lifetimes, and vesicle synthesis and degradation rates) were constrained by fitting the model to
149

experimental data. The model was coded and implemented using programming language C. All
simulation results were obtained by applying a kinetic Monte Carlo algorithm. Mathematical
modeling and simulations were completed by the joint efforts of William Hlavacek, Dipak
Barua, Srabanti Chaudhury, Nikolai Sinitsyn, Ed Stites, and Richard Posner (of TGen and Los
Alamos National Laboratory).

150

FIGURES

Figure 4.1. Overview of autophagy and key molecules involved in vesicle dynamics. (A)
Standard cartoon depiction of the autophagy process. Autophagy is executed in four stages: 1)
mTORC1 controls autophagy initiation through inhibition of the ULK1/Atg13/FIP200 complex;
2) ULK1 activity permits nucleation of the double-membrane phagophore which is largely
executed by the Vps34 complex, PI(3)P-binding effectors (i.e. WIPI proteins), and the
transmembrane protein, Atg9; 3) membrane maturation into an enclosed autophagosomes is
accomplished by two ubiquitin-like conjugation events involving LC3 and Atg5-Atg12-Atg16;
4) autophagy is completed when the autophagosome fuses with a lysosome (or with an endocytic
compartment destined for the lysosome) to form an autolysosome which leads to the degradation
of sequestered cargo. (B) A simplified diagram depicting the steps and molecules comprising our
model of autophagic vesicle dynamics. Vps34 catalyzes the production of PI(3)P while being
counterbalanced by a PI(3)P-phosphatase. PI(3)P is bound by WIPI and together, they are
engaged by Atg9. This complex undergoes n modifications (forming a multimer) to permit
nucleation of the isolation membrane (IM). The Atg5-Atg12-Atg16 complex promotes LC3-I to
LC3-II conversion. LC3-II at a threshold (depicted with n) triggers IM progression into a freely
diffusing autophagic vesicle (V-free). Following, V-free fuses with the lysosome to generate a
lysosome-deposited vesicle (V-lyso). Proteins and organelles are encased by rounded rectangles.
Other species are in non-bordered text boxes. Double-headed arrows connect non-covalent
binding partners. The PI3P-WIPI-Atg9 complex is encompassed with a thick-line rectangle and
Xn indicates its multimeric modification. Open circles depict catalysis. Molecular species were
diagrammed using OmniGraffle Pro (The Omni Group).
151

Figure 4.1 (cont'd)

A

mTORC1

Atg13
ULK1 FIP200

proLC3
Atg4
LC3-I

PE

Vps34

LC3-II

PE

Beclin1

Atg7/3
Vps15

Atg14

Atg5

3P

Atg16

PI

Atg4

phagophore

PI3P

2

g1

g5

Atg9

At

I1

At

PI3

At

P

6
g1

Atg12

WIP

LC3

-II

Atg23

Atg7/10
Atg12

LC3-I

autophagosome

B

152

autolysosome

Figure 4.2. Experimental design for measuring GFP-LC3 vesicle dynamics. (A) GFP-LC3
vesicles were imaged in U2OS cells by fluorescent microscopy and subjected to an image
processing protocol including deconvolution, intensity thresholding, and object quantification
(see Materials and Methods for details). Cell shown was cropped from a 60x-captured image.
(B) Simplified model of lysosomal inhibition. Cells treated with DMSO (vehicle control) have
functional (active) lysosomes and GFP-LC3 vesicles are continually synthesized and turned over
(fluorescence quenched) in the lysosome. Bafilomycin A1 treatment inhibits lysosome function
to cause the accumulation of GFP-LC3 vesicles which are protected from fluorescent quenching.
In this model, two vesicles form from 0’ to 4’ and two more form from 4’ to 8’. The fact that
these vesicles were made but are not seen in DMSO-treated cells indicates 2 vesicles were also
turned over during 0’ to 4’ and also from 4’ to 8’. (C) Example plots of GFP-LC3 vesicle counts
from a single cell with each active (gray circles) and inhibited (black squares) lysosomes during
a 120 min treatment period with full nutrient media. (D) Snapshots from areas imaged within
single cells with active (top panels) and inhibited (bottom panels) lysosomes at 20 min intervals
for 120 min treatments. Insets are from deconvolved 60x-captured images.

153

Figure 4.2 (cont'd)

Vehicle Control
Bafilomycin A1
Lysosomes Active Lysosomes Inhibited

Treatment Time

V V

V

L

V V

L

0’
4’
8’

D

28’

48’

quantification

80
Inhibited
Active
70
60
50
40
30
20
10
0
0 20 40 60 80 100 120 140
Treatment Time (min.)

Treatment Time
68’
88’

Inhibited

Active

8’

C
GFP-LC3-positive punctae

B

intensity
threshold

deconvolution

A

154

108’

128’

Figure 4.3. Initial GFP-LC3 data collection. (A-B) GFP-LC3 vesicle dynamics in full nutrient
media (basal autophagy) and rapamycin media (induced autophagy). Following 90 min preincubation period with full nutrient media containing DMSO (top panels, active lysosomes) or
BafA1 (bottom panels, inhibited lysosomes), U2OS-GFP-LC3 cells were treated for 120 min in
fresh media of the same composition (A) or supplemented with 50 nM rapamycin (B). Cells were
imaged and quantified every 4 min. An image of several cells at 0 min and 120 min is shown.
Insets are 2x magnifications of boxed regions. Note the accumulation of vesicles upon lysosomal
inhibition. (C-D) Single cell traces of GFP-LC3 vesicle counts from full nutrient media cells
with active (C) or inhibited (D) lysosomes. Each line represents a single cell. (E-F) Total GFPLC3 vesicles synthesized (gray circles) and turned over (black squares) in full nutrient media (E)
and rapamycin media (F) from time 0 shown. Vesicle synthesis and turnover calculated using
formulas (16) and (17) described in the text. Note, lines largely overlap so synthesis plot
sometimes hidden by turnover plot. Bars represent standard deviation from synthesis rate
calculations.

155

Figure 4.3 (cont'd)

B Rapamycin

A Full Growth Media
120’

0’

Inhibited

Inhibited
Treatment Time
(min.)

Treatment Time
(min.)

Treatment Time
(min.)

156

140
120
100
80
60
40
20
0

Synthesis
Turnover

20
40
60
80
100
120

Synthesis
Turnover

Vesicle Count

140
120
100
80
60
40
20
0

Rapamycin

F
Vesicle Count

Full Growth Media

E

20
40
60
80
100
120

140
120
100
80
60
40
20
0

20
40
60
80
100
120

20
40
60
80
100
120

Vesicle Count

140
120
100
80
60
40
20
0

Inhibited

D
Vesicle Count

Active

C

120’

Active

Active

0’

Treatment Time
(min.)

Figure 4.4. Model simulations after fitting of experimental data. (A-B) Average plots of
experimental GFP-LC3 vesicle counts measured from cells during the 90 min pre-incubation
period in full nutrient media with lysosomes inhibited (open squares, -90 min to 0 min) and the
120 min treatment with rapamycin media and inhibited lysosomes (black circles, 0 min to 120
min). Bars represent standard deviation. The dotted curve in (A) was generated to have the same
slope as the best-fit line going through averaged data from a 120 min treatment with full nutrient
media and inhibited lysosomes in (B). A theory curve (solid line) was generated for the
rapamycin media condition as the average of many stochastic simulation runs. (C) Averaged
plots of GFP-LC3 vesicle counts from cells during the 90 min pre-incubation period with full
nutrient media and active lysosomes (open triangles, -90 min to 0 min) and 120 min treatment
with rapamycin media (black diamonds, 0 min to 120 min). Bars represent standard deviation. A
theory curve (solid line) was generated from the model tuned with parameters from (A) and (B)
but not experimental value from (C). Note the agreement of experimental data with model
simulation. (D-G) Single stochastic traces generated from the model with no parameter-tuning
matched single cell GFP-LC3 vesicle traces found within the data sets. GFP-LC3 vesicle counts
shown from cells under the following conditions: (D) 90 min pre-incubation period with full
nutrient media and lysosomes inhibited, (E) 120 min treatment with full nutrient media and
lysosomes inhibited, (F) 120 min treatment with rapamycin media and active lysosomes, (G) 120
min treatment with rapamycin media and inhibited lysosomes. These traces do not represent
mean or average data but were actively selected from highly variable experimental and stochastic
simulation datasets to demonstrate that single simulations match actual cell measurements.

157

Figure 4.4 (cont'd)

E

50
0
-90 -60 -30 0 30 60 90 120
Time (min.)

B
Vesicle Count

100

Experimental
(Lysosomes Inhibited)

F

50
0
-90 -60 -30 0

30 60 90 120

Time (min.)

Vesicle Count

C
20

Vesicle Count

100

Vesicle Count

Vesicle Count

150

D

Simulations
(Lysosomes Active)

10
0
-90 -60 -30 0

30 60 90 120

Vesicle Count

200

Simulations
(Lysosomes Inhibited)

G
Vesicle Count

A

Time (min.)

158

200
150
100
50
0
-90 -70 -50 -30 -10
Time (min.)
200
150
100
50
0
0

40
80 120
Time (min.)

200
150
100
50
0
0

40
80 120
Time (min.)

200
150
100
50
0
0

40
80 120
Time (min.)

Figure 4.5. Simulations take into account observed system noise. (A-D) Theory curves
generated from the tuned model are depicted for 120 min treatment periods with full nutrient
media (A-B) or rapamycin media (C-D) and inhibited (A, C) or active (B, D) lysosomes.
Fluctuations were observed in the experimental data and this noise was accounted for by fitting
model parameters to account for visible and invisible vesicle states (see reaction (12) in text).
This accounts for vesicle aggregates being miscounted as a single vesicle or vesicles
disappearing from the focal plane. Resulting total vesicle counts are in red while visible vesicle
counts are in black. Blue lines depict isolation membranes (IMs). Parameters were tuned to
assume few IMs compared to large free vesicle numbers. (E) Wild-type U2OS cells were
immunostained for endogenous Atg12, a marker of IMs which dissociates from the free vesicle,
and supported the assumption of few IMs compared to free vesicles. AF488-conjugated
secondary antibodies were to detect primary Atg12 staining and images captured at 60x. Inset
represents a 2x magnification of the boxed region.

159

Figure 4.5 (cont'd)

A

Full Nutrients (Inhibited)

150

150

100

100

50
0

0

B

0
0

D

9
6
3
0

20 40 60 80 100 120
Time (min.)

20 40 60 80 100 120
Time (min.)
Rapamycin (Active)

12
Vesicle Count

Vesicle Count

E

50

20 40 60 80 100 120
Time (min.)
Full Nutrients (Active)

12

0

Rapamycin (Inhibited)

200
Vesicle Count

Vesicle Count

200

C

9
6
3
0

anti-Atg12

160

0

20 40 60 80 100 120
Time (min.)

Figure 4.6. Model prediction and test: Vps34 inhibition. (A) The PI3K inhibitor, wortmannin,
inhibits Vps34 activity. U2OS cells stably expressing an EGFP-2xFYVE construct which
specifically binds PI(3)P show an absence of PI(3)P-positive vesicles following 30 min treatment
with 100 nM wortmannin (bottom) compared to vehicle-treated cells (top). (B-C) Following 90
min pre-incubation period with full nutrient media containing DMSO (top panels, active
lysosomes) or BafA1 (bottom panels, inhibited lysosomes), U2OS-GFP-LC3 cells were treated
for 120 min in fresh media of the same composition but supplemented with 100 nM wortmannin.
Images captured and quantified every 4 min. An image of several cells at 0 min and 120 min is
shown in (B). Insets are 2x magnifications of boxed regions. Note the lack of accumulation of
vesicles upon lysosomal inhibition. Mean vesicle counts from cells with inhibited (gray circles)
or active (black squares) lysosomes are plotted in (C). Bars represent s.d.m. (D-E) Wortmannin
reduces GFP-LC3 vesicle synthesis and turnover rates as calculated from experimental data
using formulas (16) and (17) (see text). Total vesicles synthesized (gray circles) and turned over
(black squares) beginning at time 0 shown. (D) Average rates of synthesis and turnover (in mean
GFP-LC3 vesicles per min) during the 120 min treatments were derived from linear lines-of-best
fit of the plots for wortmannin in (C) and full nutrient media in Figure 4.3E. Black bars =
synthesis, gray bars = turnover. (F) Model-predicted simulation of 90% Vps34 inhibition when
lysosomes are inhibited (solid line) or active (dotted line).

161

Figure 4.6 (cont'd)

PI(3)P (Vps34
product)

Wortmannin

B 0’

120’

Inhibited

Wortmannin

Active

Control

A

C
200
150
100
50

135
115
95
75
55
35
15

Synthesis
Turnover

115
95
75
55
35
15

0
0
20
40
60
80
100
120
0 20 40 60 80 100 120
Time (min.)
0.50
0.5
0.40
0.4

Synthesis Rate
Turnover Rate

0.30
0.3
0.20
0.2
0.10
0.1

0

0.00

Full
WortGrowth mannin
Media

0

20

40

60

80

100

120

0
20
40
60
80
100
120
140

-5

Time (min.)

F
200
Vesicle Count

E
Vesicles per Min.

Experimental
(Vps34 Inhibition)

135

Inhibited
Active

Vesicle Count

Vesicle Count

D

Experimental
(Vps34 Inhibition)

Simulation Prediction
(Vps34 Inhibition)

150
100
50
0
0
20
40
60
80
100
120
0 20 40 60 80 100 120
Time (min.)

162

Figure 4.7. Model prediction and test: Atg9 depletion. (A-B) 48 hours prior to imaging and
quantification, U2OS-GFP-LC3 cells were transfected with siRNAs targeting Atg9. Following
90 min pre-incubation period with full nutrient media containing DMSO (top panels, active
lysosomes) or BafA1 (bottom panels, inhibited lysosomes), cells were treated for 120 min in
fresh media of the same composition. Images captured and quantified every 4 min. Images of
several cells at 0 min and 120 min are shown in (A). Insets are 2x magnifications of boxed
regions. Note the lack of accumulation of vesicles upon lysosomal inhibition (and lack of
efficacy of the pre-incubation because Atg9 is knocked down for the entire experiment). 120 min
panels from Figure 4.3A are shown as a reference for wild-type cells dynamics. Mean vesicle
counts from cells with inhibited (gray circles) or active (black squares) lysosomes are plotted in
(B). Bars represent s.d.m. (C-E) Atg9 siRNA-mediated knockdown reduces GFP-LC3 vesicle
synthesis and turnover rates as calculated from experimental data using formulas (14) and (15)
(see text). (B) Total vesicles synthesized (gray circles) and turned over (black squares) beginning
at time 0 shown. (C) Average rates of synthesis and turnover (expressed as mean GFP-LC3
vesicles per min) during the 120 min treatments were derived from linear lines-of-best fit of the
plots for Atg9 knockdown in (B) and full nutrient media in Figure 4.3E. Black bars = synthesis,
gray bars = turnover. (D) Quantitative real-time PCR (qRT-PCR) using Atg9A-specific primers
demonstrated 89% knockdown of mRNA with Atg9 siRNAs versus control (non-targeting)
siRNAs. (F) Model-predicted simulation of 90% Atg9 depletion when lysosomes are inhibited
(solid line) or active (dotted line).

163

Figure 4.7 (cont'd)

A Full Growth Media (Atg9 siRNA)
120’

Inhibited

Inhibited

Active

Active

0’

Positive Control
120’

C

Experimental
(Atg9 Depletion)
Inhibited
Active

100
100

50
50

D

E
0.5
0.45

80
60
40

0
0

20

40

60

80

F

0.40

0.35

0.30
0.30
0.3
0.25
0.20
0.20

0.2
0.15
0.10
0.10

0.1
Full
ATG9
Growth siRNA
Media

120

140

Simulation Prediction
(Atg9 Depletion)

140

0.40
0.4

100

Time (min.)

140
120

Relative Atg9A
Expression

Vesicles per Min.

0.50
0.50

0

100

20

0
0
0
20
40
60
80
100
120
0 20 40 60 80 100 120
Time (min.)

0.05
0.00
0.00

Synthesis
Turnover

120

120

100
80
60
40
20
0
Control Atg9
siRNA siRNA
100

80
60
40
20
0

164

200

Vesicle Count

150
150

140

Vesicle Count

Vesicle Count

200
200

140
120
100
80
60
40
20
0

Experimental
(Atg9 Depletion)

0
20
40
60
80
100
120
140

B

150
100
50
0

0 20 40 60 80 100 120
Time (min.)

Figure 4.8. Model prediction: LC3 concentration and vesicle size. The model was used to
generate predictions for several hypothetical LC3 concentrations and resulting outputs of
autophagic vesicle size (in average units). Simulated conditions are 120 min of full nutrient
media with active lysosomes. Each data point represents vesicle size with respect to time of
4

5

treatment. LC3 concentrations correspond to 10 /cell (red diamonds), 10 /cell (default value
6

7

used in the model) (black circles), 10 /cell (blue triangles), and 10 /cell (magenta squares).

165

Figure 4.8 (cont'd)

Simulation - LC3 Concentration

Mean Vesicle Size
(average units)

50
40
30
20
10
0

0

20

40

60

80

100 120

Treatment Time (min.)

166

CHAPTER 5

Summary and Future Directions

167

SUMMARY AND FUTURE DIRECTIONS

PTPsigma as a new effector of Vps34-PI(3)P signaling
The foundation for this entire body of work was an interest in Vps34-PI(3)P signaling
and the aim of identifying novel phosphatases which control PI(3)P signaling. At the time this
project originated, Vps34 had been extensively characterized for its role in endocytosis,
beginning with its identification in yeast and followed with a detailed mammalian
characterization (Herman and Emr 1990; Schu, Takegawa et al. 1993; Stack, Herman et al. 1993;
Simonsen, Lippe et al. 1998; Christoforidis, Miaczynska et al. 1999; Futter, Collinson et al.
2001).
More recently, publications had suggested that Vps34-PI(3)P signaling was also
important for two additional cell processes, mTOR signaling and autophagy (Petiot, Ogier-Denis
et al. 2000; Kihara, Noda et al. 2001; Byfield, Murray et al. 2005; Nobukuni, Joaquin et al.
2005). While the role in mTOR signaling has not been well substantiated in the time since, it has
becoming increasingly evident that Vps34-PI(3)P signaling is essential for autophagy (Itakura,
Kishi et al. 2008; Juhasz, Hill et al. 2008; Obara, Noda et al. 2008);(Proikas-Cezanne, Waddell et
al. 2004). Given the pivotal role of autophagy in cell survival and its implications in the etiology
of several diseases, it was the role of PI(3)P in this process that was, and continues to be, our
primary interest.
We hypothesize that phosphatases exist which regulate Vps34-PI(3)P signaling during
autophagy. In support of this notion, two myotubularin phosphatases have been functionalized as
direct PI(3)P phosphatases with important roles in autophagy regulation (Vergne, Roberts et al.
2009; Taguchi-Atarashi, Hamasaki et al. 2010). Given the complexity of autophagy execution

168

and the number of proteins involved in the Vps34 complexes alone, we believed it likely that
additional protein phosphatases exist to control Vps34-PI(3)P signaling. Accordingly, we
designed and implemented a cell-based RNA interference screen where we monitored PI(3)P
dynamics following loss of expression of individual phosphatase genes. We uncovered a number
of interesting phenotypes and ultimately chose to focus primarily on one, that of PTPsigma, as
the PI(3)P phenotype elicited by knockdown closely matched that observed in autophagic cells.
This PTPsigma loss-of-function phenotype was intriguing because there was no clear
precedence for PTPsigma in PI(3)P signaling. As a receptor-like molecule highly expressed in
neurons, PTPsigma has been characterized for its role in mammalian development and the
control of neurite outgrowth and regeneration. These functions likely stem from interactions of
its ectodomains with other cells and the extracellular environment (Shen, Tenney et al. 2009;
Fry, Chagnon et al. 2010). In agreement with this, N-Cadherin and neutrophin receptors (TrkA,
TrkB, TrkC), proteins involved in cell adhesion and neurite outgrowth, are the only substrates
demonstrated to be directly dephosphorylated by PTPsigma and also supported in vivo (Faux,
Hawadle et al. 2007; Siu, Fladd et al. 2007).
There are, however, several interesting potential connections of the PTPsigma neuronal
phenotype with autophagy. First, over half of PTPsigma knockout mice die within two days of
birth (Elchebly, Wagner et al. 1999; Wallace, Batt et al. 1999). Autophagy has been shown to be
upregulated during this neonatal period to allow the newborn to survive removal from the
maternal-supplied nutrient source (Kuma, Hatano et al. 2004). In support of this, a similar pattern
of neonatal death has been reported in murine knockout models of Atg5 and Atg7, two important
components of the autophagy machinery (Kuma, Hatano et al. 2004; Komatsu, Waguri et al.
2005). Given the balanced level of autophagy required for homeostasis, it is possible that

169

-/-

defective autophagy, as in Atg5 or Atg7
-/-

PTPsigma

-/-

mice, as well as hyperactive autophagy, as in

mice, could both be detrimental during the neonatal period.

A second feature of PTPsigma neuronal function with potential implications for
autophagy concerns its role in nerve regeneration. PTPsigma knockout mice have been well
characterized for their enhanced regeneration in models of neuronal injury (McLean, Batt et al.
2002; Thompson, Uetani et al. 2003; Sapieha, Duplan et al. 2005). Autophagy has been
implicated as a survival program in response to axon injury and thus, it is possible that PTPsigma
loss promotes nerve regeneration by enhancing survival following insult (Matthews 1973;
Rubinsztein, DiFiglia et al. 2005; Sternberg, Benchimol et al. 2010)
As outlined in Chapter 2 Section 1, we proceeded to assign a role for PTPsigma in
autophagy, largely through characterization of the loss-of-function phenotype. In addition to
abundant PI(3)P-positive vesicles, we found that knockdown increased the presence of vesicles
positive for endogenous Atg12 and LC3B, two specific markers of autophagic vesicles, but did
not significantly change the abundance of EEA1-positive early endosomes. This suggested that
cells lacking PTPsigma had enhanced autophagy including a stage of vesicle positive for PI(3)P.
We verified that autophagy was functional in these cells using lysosomal inhibitors to gauge
autophagic flux, suggesting that the abundance of vesicles was not the product of a trafficking
defect.
To determine if there was a potential for PTPsigma to directly regulate PI(3)P, we
established its subcellular localization by tracking the D1-phosphatase domain by
immunofluorescence. We uncovered that not only was PTPsigma distributed on abundant
intracellular membranes, these membranes were also frequently positive for PI(3)P. When we
induced autophagy in cells by amino acid starvation, we found that the numerous peripheral
170

PI(3)P vesicles which form, presumably of autophagic origin, were also positive for PTPsigma.
In part, this close spatial association of PTPsigma and PI(3)P led us to hypothesize that
PTPsigma may function as a direct PI(3)P-phosphatase.
To explore this hypothesis, we first utilized an in silico approach. Elegant comparative
studies using crystallography and molecular docking have uncovered that the conformation of a
PTP active site is a key determinant of substrate specificity (Begley, Taylor et al. 2006). The
catalytic cleft of a lipid phosphatase must be uniquely deep and wide to accommodate bulky
lipid head groups. In particular, the active site of a phosphoinositide phosphatase must be not
only large enough to bind the hexameric inositol ring, but also wide enough to accommodate the
1′ phosphate that links the ring to a glycerol moiety (Begley, Taylor et al. 2006). To determine if
the conformation of either PTPsigma active site would allow PI(3)P binding, we performed a
structural docking experiment in which a PI(3)P molecule was inserted into the crystal structure
of the PTPsigma catalytic domains. We discovered that the membrane-proximal D1 domain
accommodated PI(3)P favorably, similar to a phosphotyrosine peptide (data not shown).
Fueled by this crystal structure revelation, we devoted considerable effort to testing the
catalytic activity of PTPsigma towards PI(3)P in vitro. While activity could be detected at times
in specific assays, we did not feel it was substantial or consistent enough to consider PTPsigma a
bona fide lipid phosphatase. However, given the ability of the D1 active site to virtually dock
PI(3)P, we cannot exclude the possibility that PTPsigma may function as a PI(3)P phosphatase in
vivo.
Given an apparent lack of PI(3)P-phosphatase activity in vitro but an active site which
apparently supports PI(3)P binding, we hypothesized that the active site may contribute to the
targeting of PTPsigma to PI(3)P-enriched membranes via protein-lipid binding. To test this, we

171

monitored PTPsigma localization in cells treated with wortmannin, a potent pan PI3K inhibitor
which effectively depletes PI(3)P from cells. We found that the vesicular presence of PTPsigma
and even its redistribution to abundant vesicles during autophagy was unaffected by wortmannin
treatment. This suggests that PI(3)P presence on membranes is not required for PTPsigma
targeting.
Lacking support for a PTPsigma-PI(3)P interaction, we chose to explore alternative
hypotheses that PTPsigma may function in autophagy through protein-mediated mechanisms.
Given the robust PI(3)P phenotype, we felt this point of regulation was likely to be close to
PI(3)P. A search of a public repository of post-translational modifications revealed that several
potentially phosphorylated tyrosine residues exist within Vps34 and its associated proteins. Thus,
we developed a new working hypothesis that PTPsigma regulates PI(3)P by functioning as an
effector of a Vps34 complex.
In support of this, we detected an interaction between PTPsigma and Vps34 as well as
Rubicon, a Vps34 binding partner functioning on endocytic compartments. The subcellular
localization of PTPsigma and Rubicon also largely overlapped. Upon PTPsigma knockdown, a
distinct band near 100 kDa was observed to be tyrosine-phosphorylated (not so in control cells)
which precipitated with the Vps34 core complex (Vps34-Vps15-Beclin1). A similar band was
observed in whole cell lysates, independently of Vps34 immunoprecipitation. It is an exciting
possibility that this protein may represent a phosphosubstrate of PTPsigma relevant to PI(3)P
signaling. In particular, Rubicon has at least one reported phosphotyrosine site and has a
predicted molecular weight of 109 kDa. Given that PTPsigma and Rubicon are capable of
interacting in cells, reside primarily at the same subcellular compartments, and display very
similar loss-of-function phenotypes, we generated a working model in which PTPsigma may

172

control Vps34-PI(3)P signaling through a mechanism involving Rubicon (Figure 5.1). Here, we
envision PTPsigma and Rubicon as cooperative partners functioning from the endocytic
compartment as a brake pedal or checkpoint of Vps34 function. In this manner, they quench
PI(3)P signaling to ensure it proceeds to a degree which is conducive to homeostasis and cell
viability.
Mass spectrometry is required to identify the phosphorylated protein(s) observed in our
experiments. Following, whether that substrate is Rubicon or another protein not yet considered,
a detailed characterization will be required to establish this new mechanism. Potentially
phosphorylated residues can be mutated by site-directed-mutagenesis in order to determine
which sites are regulated by PTPsigma and importantly, are relevant to autophagy.

Proteolytic processing of PTPsigma as it relates to an autophagy function
Aside from the identification of phosphosubstrates, much work remains to elucidate the
processing which controls PTPsigma and its internalization. In Chapter 2 Section 2, we
determined that PTPsigma is processed by extracellular metalloproteases to a C-terminal
fragment (CTF) which is targeted to and turned over within the lysosome. This CTF processing
occurs in a relatively constitutive manner although, based on observations from multiple
experiments, there may be a cell-density contribution to the extent of processing (data not
shown). Because the PTPsigma knockdown phenotype manifests in the presence of full nutrient
growth conditions and the membrane-bound CTF is most likely trafficked along the endocytic
pathway, it is plausible that this CTF functions in the regulation of autophagy. From an
endocytic compartment, PTPsigma could elicit control of autophagy at the point of convergence
of autophagosomes and endosomes. It has been determined that this convergence, which results

173

in the generation of hybrid organelles called amphisomes, is a frequent and important feature
involved in the efficient maturation of autophagic vesicles (Fader and Colombo 2009). In
agreement with this model, Rubicon knockdown was shown to elevate autophagosome
abundance and autophagic (and endocytic) flux and it functions in these processes from a
location on endocytic membranes (Matsunaga, Saitoh et al. 2009; Zhong, Wang et al. 2009; Sun,
Westphal et al. 2010).
Intriguingly, co-immunoprecipitation experiments revealed that the interaction between
PTPsigma and Vps34 is possible for not only the CTF, but the pro-protein and P-subunit as well.
This interaction may simply reflect an affinity detected in the cell lysate or could reflect a more
complex association of these proteins. Although less intuitive than a CTF-mediated interaction, it
is possible that PTPsigma in other forms could encounter Vps34. As a type I transmembrane
protein, PTPsigma is translated as a pro-protein in the ER with its phosphatase domains facing
the cytosol. Given recent evidence that at least a portion of autophagic vesicles originate from
PI(3)P-enriched microdomains cradled by the ER, it is possible that PTPsigma interacts with
Vps34 from the ER in its pro-protein form. Also, internalization of PTPsigma without prior
ectodomain shedding and CTF formation was reported and thus, PTPsigma could potentially be
in proximity of Vps34 on endomembranes in full-length or P-subunit form (Aicher, Lerch et al.
1997).
In addition to the basally produced CTF, we also observed a second smaller fragment
produced only during starvation-induced autophagy which did not appear to be a target of the
lysosome. We previously showed that PTPsigma redistributes to a pool of abundant peripheral
PI(3)P vesicles during starvation so it is possible that this unique processing accompanies the
targeting of PTPsigma to this location. Identification of the fragment, be it a typical ICD or more

174

unusual processing product, could lead to an understanding of PTPsigma activity during
starvation. A combination of chemical inhibition (i.e. inhibition of metalloproteases and gamma
secretase) and site-directed mutagenesis of potential sites of processing will aid in the
characterization of these processing events. Of note, we attempted to mutagenize at least three
sites of potential cleavage of PTPsigma and although the cloning effort was successful, we could
not detect significant alterations in PTPsigma processing (data not shown). The primary
difficulty in this approach is that cleavage sites are often large and not explicitly defined for
PTPsigma. Thus, an improvement of this mutagenesis strategy will need to be employed and will
likely require more complicated mutagenesis for efficient inhibition of processing.
Finally, identification of the signals which induce PTPsigma internalization and function
in autophagy will be important. While nutrient availability regulates autophagy induction,
precise mechanisms of amino acid sensing are lacking. The fact that PTPsigma knockdown cells
generally mimic cells starved of amino acids in terms of autophagic phenotype could suggest a
defect involving amino acid sensing. This is a relatively attractive hypothesis given PTPsigma
presence at the cell surface, the interface of extracellular nutrients and cellular import
(Goberdhan, Ogmundsdottir et al. 2009). However, mTOR signaling, which requires amino acids
for full activation, is fully functional in these cells (data not shown). Thus, if there is a defect in
amino acid sensing in the absence of PTPsigma, it would have to be impinging at more direct
point in the autophagy pathway, downstream or independent of mTOR. Alternatively, as is
common for receptor molecules, PTPsigma has been shown to be internalized at high cell
densities, a condition reported to activate autophagy (Kisen, Tessitore et al. 1993; Wu, Yang et
al. 2006). With respect to this, its regulation may involve the sensing of cell adhesion,
extracellular signals, or cell-to-cell contacts. This is a reasonable potential mechanism of

175

regulation given the large ectodomain of adhesion motifs which constitutes the E-subunit of
PTPsigma.

PTPsigma in cell survival
PTPsigma has been implicated in two cancer paradigms, chemoresistance and metastatic
disease. First, RNAi-mediated knockdown of PTPsigma in cultured cancer cells was reported to
confer resistance to several chemotherapeutics (MacKeigan, Murphy et al. 2005). Additionally,
loss of PTPsigma expression in metastatic prostate cancer was uncovered in a study of lasercaptured patient tissues encompassing progressive stages of prostate malignancy (Tomlins,
Mehra et al. 2007). Given the autophagic phenotype we established in the absence of PTPsigma,
we hypothesized that hyperactive autophagy following its loss may provide an intracellular
mechanism by which cells evade chemotherapeutic insult. In metastatic prostate cancer, this
autophagic proficiency may contribute to the survival of cancer cells and ultimate progression to
advanced therapy-refractory disease.
Autophagy has been shown to promote survival during stress and contribute to
chemoresistance (Amaravadi, Yu et al. 2007; Carew, Nawrocki et al. 2007; Wu, Chang et al.
2010). The microenvironment of a tumor, rapidly outgrowing its blood supply, is void of
nutrients and often hypoxic and this stress is exacerbated during chemotherapy. Cancer cells
uniquely proficient in autophagy are able to thrive despite this cytotoxic insult. Reports have
demonstrated that chemoresistant cell lines dependent upon autophagy for survival can be
sensitized to death when chemotherapeutic agents are supplemented with autophagy inhibitors
(Amaravadi, Yu et al. 2007; Carew, Nawrocki et al. 2007; Wu, Chang et al. 2010).

176

We attempted to support this model in a number of cell systems and contexts but
unexpectedly, could not (data not shown). Despite clearly activated autophagy, we could not
detect enhanced viability in U2OS cells transfected with PTPsigma siRNAs. Further, using HeLa
cells, the model established for assaying chemoresistance, we could not demonstrate that
PTPsigma knockdown favored survival in response to therapeutics. Further, using a combination
of chemotherapeutics and autophagy inhibitors, we could not provide evidence to support a role
for autophagy in the survival of these cells. We tested the kinetics of chemotherapeutic and
autophagy inhibitor dosing but could not detect a survival advantage under any circumstance.
These findings were unexpected but do not yet negate the model put forth. Given the
pivotal but bimodal function of autophagy in cell fate, this process must be exquisitely controlled
to provide the most benefit to a cell. While a cell incapable of undergoing autophagy is rendered
susceptible to stress and starvation -induced cell death, excessive autophagy can also contribute
to cell death through the detrimental degradation of cytosolic content (Levine 2007). Given this,
there is likely a tightly regulated threshold of autophagy which promotes survival versus
contributes to death. This threshold may be difficult to measure and given the considerable
activation of autophagy seen in PTPsigma knockdown cells, the cells may be so proficient in
autophagy that a survival advantage is difficult to capture. Further, the role of PTPsigma in vivo
with respect to autophagy signaling in a larger tissue or network is likely much more
complicated than seen in our in vitro cell model but may be the most physiologically relevant
means to testing this hypothesis. This concept is the focus of a new project in the lab and the
global role of autophagy in chemoresistance and cancer progression will be explored using
mouse models.

177

Small molecule inhibitors of PTPsigma
Although PTPsigma knockdown did not confer a survival advantage in our cell models,
we undertook a project to identify small molecule inhibitors of PTPsigma nonetheless and
outlined the results in Chapter 3. We felt it was useful to develop such inhibitors for several
reasons. First, it has been well established, primarily through knockout animal studies, that loss
of PTPsigma expression enhances neurite outgrowth and regeneration following injury (McLean,
Batt et al. 2002; Thompson, Uetani et al. 2003; Sapieha, Duplan et al. 2005). Notably, it was
recently found that loss of PTPsigma promotes neural regeneration following spinal cord injury
(SCI), owing to the interaction of its ectodomain with chondroitin sulfate proteoglycans (CSPGs)
(Shen, Tenney et al. 2009; Fry, Chagnon et al. 2010). Thus, inhibition of PTPsigma could
provide a therapeutic outlet for SCI by promoting bypass of barriers normally encountered in
axon regeneration. In addition, because neurodegenerative diseases are often hallmarked by the
formation of toxic aggregates which can be cleared by autophagy, reducing PTPsigma activity
could potentially be of use in these diseases as well (Hara, Nakamura et al. 2006; Komatsu,
Waguri et al. 2006). Along these lines, PTPsigma is highly expressed in the brain, making
PTPsigma an attractive central nervous system target (Pulido, Serra-Pages et al. 1995).
We coupled an in silico virtual ligand screening (VLS) approach with in vitro
phosphatase assays to identify a number of small molecules which favorably bind the PTPsigma
D1 active site and inhibit PTPsigma activity. Given the unique conformation of the PTPsigma
active site, described above, we were surprised to find that these compounds offered no
selectivity for PTPsigma when tested for inhibition of another classic PTP, PTP1B. This issue
with non-selectivity is a setback commonly encountered in phosphatase drug discovery projects
(Tautz, Pellecchia et al. 2006). Here, we proposed plans for a refined VLS approach which we

178

believe will lead to the identification of PTPsigma inhibitors with selectivity. This modified
method involves screening compounds in silico for both favorable binding to PTPsigma and
negligible binding to PTP1B. This should produce a lead compound shortlist to test in vitro
which offers an improved likelihood for selectivity.
It is possible that a selective inhibitor will not be identified easily through simple VLS
alone. Instead, we may need to utilize a detailed analysis of the PTPsigma crystal structure to
identify and exploit key residues involved in substrate, and inhibitor, binding. The high degree of
conservation of PTPs may necessitate this approach and will likely help drive the identification
of a selective compound. It may also be required that an inhibitor be directed to regions outside
the PTPsigma active site. For instance, effective and selective inhibitors of PTP1B have been
developed which target both the active site and an adjacent, less conserved, binding pocket
(Shen, Keng et al. 2001; Zhang 2002; Sun, Fedorov et al. 2003). It is possible that similar
adjacent regions exist within PTPsigma and could be exploited in the development of an
inhibitor.

Mathematical models of autophagy
The third aim of this project, and focus of Chapter 4 of this thesis, has undergone
considerable evolution since its origins as a mere thought-process. In fact, the plausibility that
this modeling effort would generate a useful and relevant framework for studying autophagy
seemed remote at times. Because the field of Systems Biology, founded on the acquisition of
comprehensive datasets and computation models to analyze such data, is still in its infancy,
efforts within this discipline are not always straightforward. Setbacks in this project, if only
mental, are reflected in the following quote from Hiroaki Kitano, a pioneer in Systems Biology:

179

“‘The toughness of systems approaches has constrained the field’s growth,’ says Kitano…
‘Genomics exploded, because if you buy a sequencing machine, anyone can do it,’ he says. ‘But
having to combine good biology with good mathematical modeling isn’t easy” (Macilwain
2011).
Our primary aim was to develop a model which was driven largely from data collected
experimentally. That is, instead of constructing a model based on data selectively chosen from
the literature and based entirely on assumptions, we aimed to build a model built and refined
using reliable cell-based data. While we initially envisioned a large rule-based model which
would encompass mTOR/PI3K/Akt pathway signaling and all known PI(3)P-dependent
processes, utilizing data captured from several pathway nodes, we quickly realized that this was
an impractical starting point. Instead, we chose to focus on our central interest, autophagic
vesicle dynamics, and drive the model with high-quality kinetic measurements of GFP-LC3
dynamics. Because this model was smaller than originally intended, we described it using a set
of chemical reactions and stochastic simulation instead of rule-based specification, which is
optimal for larger signaling networks.
This initial model of autophagy dynamics, captured with single cell resolution and across
a number of cellular conditions, accurately predicted responses to both Atg9 knockdown and
Vps34 inhibition. Further, the model produced a novel hypothesis that LC3 concentrations
positively correlate with autophagic vesicle size. While we did not test this prediction
experimentally, we found evidence in its support from both yeast and mammalian systems
(Nakagawa, Amano et al. 2004; Xie, Nair et al. 2008). This highlights the accuracy and potential
utility of this model outside our isolated cell system.

180

We intend to expand upon this simple model through generation of a more
comprehensive autophagy network framework. This model will include all known molecules
involved in autophagy, including more than 30 autophagy genes and numerous other molecules
involved in its function. The model will depict regulation upstream, including mTORC1, as well
as describe molecular details underlying vesicle dynamics (for example, the multiple processing
steps required to prime and process LC3 for vesicle conjugation). Inclusion of these details may
reveal features of the model not observed in its earlier version. With this goal in mind, we have
already constructed a contact map, a diagrammatic depiction of the entire autophagy process
with model features (e.g. proteins, complexes, modifiable residues, activations) drawn using
universally defined format (portion of the map shown in Figure 5.2).
A comprehensive model of autophagy dynamics which is predictive will have
considerable utility. As autophagy is an essential regulator of cell fate, contributing to both cell
survival during stress and participating in cell death in certain contexts, it is of great interest to
disease-centric translation research. In particular, being able to predict how a specific genetic
alteration (e.g. an oncogenic mutation defining a cancer subtype) or selective enzyme inhibition
will affect this process will be informative for the design of molecularly targeted therapies
(Hopkins 2008).

181

FIGURES

Figure 5.1. Working model of PTPsigma function. Given data presented here, we present a
potential model of PTPsigma function in PI(3)P signaling. Support for this concept includes 1)
co-immunoprecipitation of PTPsigma with Vps34 and Rubicon, 2) similar vesicular localization
pattern of PTPsigma and Rubicon, 3) similar loss-of-function phenotypes of PTPsigma and
Rubicon, and 4) and altered phosphorylation associated with the Vps34 complex in the absence
of PTPsigma. The Vps34 complex associated with Rubicon may include other binding partners
(Vps15, Rab5, Beclin1, UVRAG, Rubicon) although the inclusion of Beclin1 at the endosome
has not been established. The interactions between Beclin1, Vps34, Rubicon, and UVRAG
which participate in endocytosis and autophagy are dynamic and only beginning to be elucidated.
Evidence supports a model whereby PTPsigma is processed into a C-terminal fragment (CTF) in
the presence of nutrients and targeted to the lysosome. As the CTF is membrane tethered, it
presumably travels along endocytic vesicles where it can encounter the Vps34 complex (dashed
lines indicate evidence that exogenous PTPsigma can interact with Vps34 and Rubicon via coimmunoprecipitation experiments) and PI(3)P , be it at the endosome or amphisome (fusion of an
endosome and autophagosome), where it functions the regulation of this signaling axis. Proteins
with evidence of tyrosine phosphorylation are indicated with asterisks (outlined in Table 2.2).

182

Figure 5.1 (cont'd)

2

1

*

Rab5

*

Vps34
Beclin1

Vps15

D

D

Rubicon

PT

Ps

ig

m

a

endosome/
amphisome

*
*
UVRAG

183

Figure 5.2. Contact map of the mammalian autophagy network. Figure depicts a portion of
the complete contact map of all molecules and interactions included in a rule-based model of
autophagy signaling. Specifically, this portion includes the mTORC1 and ULK1 complexes.
Double arrowed lines indicate noncovalent binding, open circles point to targets of catalysis
(phosphorylation or nucleotide exchange; a slash indicates dephosphorylation), and small
squares indicate residues which can be modified within molecules [amino acids that can be
phosphorylated (tyrosine, threonine and serine), mutated, or used for covalent attachment].
“pS/T(a)” indicates activating phosphorylation of a serine or threonine residue; “pS/T(i)”
indicates inhibitory phosphorylation of a serine or threonine residue. Cellular locations are
indicated by squares attached to molecules: C = diffusing in cytoplasm, L = lysosome.

184

Figure 5.2 (cont'd)

mTORC1
Raptor
WD40

mTOR
HEAT

mLST8
FAT

FRB

Kin

FATC

WD40

C/L

ULK1
FIP200
C

ULK1
Kin

PS

CTD

mATG13
Kin

mATG13 Pase
Pase
C

pS/T(a)

pS/T(a)
pS/T(i)

185

pS/T(i) pS/T(a)

REFERENCES

186

REFERENCES
Abaan, O. D. and J. A. Toretsky (2008). "PTPL1: a large phosphatase with a split personality."
Cancer Metastasis Rev 27(2): 205-214.
Aicher, B., M. M. Lerch, et al. (1997). "Cellular redistribution of protein tyrosine phosphatases
LAR and PTPsigma by inducible proteolytic processing." J Cell Biol 138(3): 681-696.
Almo, S. C., J. B. Bonanno, et al. (2007). "Structural genomics of protein phosphatases." J Struct
Funct Genomics 8(2-3): 121-140.
Amaravadi, R. K., D. Yu, et al. (2007). "Autophagy inhibition enhances therapy-induced
apoptosis in a Myc-induced model of lymphoma." J Clin Invest 117(2): 326-336.
Andersen, J. N., O. H. Mortensen, et al. (2001). "Structural and evolutionary relationships among
protein tyrosine phosphatase domains." Mol Cell Biol 21(21): 7117-7136.
Araki, T., M. G. Mohi, et al. (2004). "Mouse model of Noonan syndrome reveals cell type- and
gene dosage-dependent effects of Ptpn11 mutation." Nat Med 10(8): 849-857.
Aricescu, A. R., I. W. McKinnell, et al. (2002). "Heparan sulfate proteoglycans are ligands for
receptor protein tyrosine phosphatase sigma." Mol Cell Biol 22(6): 1881-1892.
Axe, E. L., S. A. Walker, et al. (2008). "Autophagosome formation from membrane
compartments enriched in phosphatidylinositol 3-phosphate and dynamically connected
to the endoplasmic reticulum." J Cell Biol 182(4): 685-701.
Backer, J. M. (2008). "The regulation and function of Class III PI3Ks: novel roles for Vps34."
Biochem J 410(1): 1-17.
Begley, M. J., G. S. Taylor, et al. (2006). "Molecular basis for substrate recognition by MTMR2,
a myotubularin family phosphoinositide phosphatase." Proc Natl Acad Sci U S A 103(4):
927-932.
Berg, T. O., M. Fengsrud, et al. (1998). "Isolation and characterization of rat liver amphisomes.
Evidence for fusion of autophagosomes with both early and late endosomes." J Biol
Chem 273(34): 21883-21892.
Bird, I. M. (1994). "Analysis of cellular phosphoinositides and phosphoinositols by extraction
and simple analytical procedures." Methods Mol Biol 27: 227-248.
Blinov, M. L., J. R. Faeder, et al. (2004). "BioNetGen: software for rule-based modeling of
signal transduction based on the interactions of molecular domains." Bioinformatics
20(17): 3289-3291.

187

Blinov, M. L., J. R. Faeder, et al. (2006). "A network model of early events in epidermal growth
factor receptor signaling that accounts for combinatorial complexity." Biosystems 83(23): 136-151.
Blondeau, F., J. Laporte, et al. (2000). "Myotubularin, a phosphatase deficient in myotubular
myopathy, acts on phosphatidylinositol 3-kinase and phosphatidylinositol 3-phosphate
pathway." Hum Mol Genet 9(15): 2223-2229.
Bova, M. P., M. N. Mattson, et al. (2004). "The oxidative mechanism of action of ortho-quinone
inhibitors of protein-tyrosine phosphatase alpha is mediated by hydrogen peroxide." Arch
Biochem Biophys 429(1): 30-41.
Byfield, M. P., J. T. Murray, et al. (2005). "hVps34 is a nutrient-regulated lipid kinase required
for activation of p70 S6 kinase." J Biol Chem 280(38): 33076-33082.
Carew, J. S., S. T. Nawrocki, et al. (2007). "Modulating autophagy for therapeutic benefit."
Autophagy 3(5): 464-467.
Chan, G., D. Kalaitzidis, et al. (2008). "The tyrosine phosphatase Shp2 (PTPN11) in cancer."
Cancer Metastasis Rev 27(2): 179-192.
Chang, Y. Y. and T. P. Neufeld (2009). "An Atg1/Atg13 complex with multiple roles in TORmediated autophagy regulation." Mol Biol Cell 20(7): 2004-2014.
Christoforidis, S., M. Miaczynska, et al. (1999). "Phosphatidylinositol-3-OH kinases are Rab5
effectors." Nat Cell Biol 1(4): 249-252.
Codogno, P. and A. J. Meijer (2005). "Autophagy and signaling: their role in cell survival and
cell death." Cell Death Differ 12 Suppl 2: 1509-1518.
Domin, J., F. Pages, et al. (1997). "Cloning of a human phosphoinositide 3-kinase with a C2
domain that displays reduced sensitivity to the inhibitor wortmannin." Biochem J 326 (
Pt 1): 139-147.
Dromard, M., G. Bompard, et al. (2007). "The putative tumor suppressor gene PTPN13/PTPL1
induces apoptosis through insulin receptor substrate-1 dephosphorylation." Cancer Res
67(14): 6806-6813.
Dunn, W. A., Jr. (1990). "Studies on the mechanisms of autophagy: maturation of the autophagic
vacuole." J Cell Biol 110(6): 1935-1945.
Elchebly, M., J. Wagner, et al. (1999). "Neuroendocrine dysplasia in mice lacking protein
tyrosine phosphatase sigma." Nat Genet 21(3): 330-333.
Endy, D. and R. Brent (2001). "Modelling cellular behaviour." Nature 409(6818): 391-395.
Fader, C. M. and M. I. Colombo (2009). "Autophagy and multivesicular bodies: two closely
related partners." Cell Death Differ 16(1): 70-78.
188

Faeder, J. R., M. L. Blinov, et al. (2005). "Combinatorial complexity and dynamical restriction
of network flows in signal transduction." Syst Biol (Stevenage) 2(1): 5-15.
Fan, W., A. Nassiri, et al. (2011). "Autophagosome targeting and membrane curvature sensing
by Barkor/Atg14(L)." Proc Natl Acad Sci U S A 108(19): 7769-7774.
Faux, C., M. Hawadle, et al. (2007). "PTPsigma binds and dephosphorylates neurotrophin
receptors and can suppress NGF-dependent neurite outgrowth from sensory neurons."
Biochim Biophys Acta 1773(11): 1689-1700.
Fry, E. J., M. J. Chagnon, et al. (2010). "Corticospinal tract regeneration after spinal cord injury
in receptor protein tyrosine phosphatase sigma deficient mice." Glia 58(4): 423-433.
Fujita, N., T. Itoh, et al. (2008). "The Atg16L complex specifies the site of LC3 lipidation for
membrane biogenesis in autophagy." Mol Biol Cell 19(5): 2092-2100.
Funderburk, S. F., Q. J. Wang, et al. (2010). "The Beclin 1-VPS34 complex--at the crossroads of
autophagy and beyond." Trends Cell Biol 20(6): 355-362.
Futter, C. E., L. M. Collinson, et al. (2001). "Human VPS34 is required for internal vesicle
formation within multivesicular endosomes." J Cell Biol 155(7): 1251-1264.
Ganley, I. G., H. Lam du, et al. (2009). "ULK1.ATG13.FIP200 complex mediates mTOR
signaling and is essential for autophagy." J Biol Chem 284(18): 12297-12305.
Gao, W., J. H. Kang, et al. (2010). "Biochemical isolation and characterization of the
tubulovesicular LC3-positive autophagosomal compartment." J Biol Chem 285(2): 13711383.
Gaullier, J. M., A. Simonsen, et al. (1998). "FYVE fingers bind PtdIns(3)P." Nature 394(6692):
432-433.
Gillooly, D. J., I. C. Morrow, et al. (2000). "Localization of phosphatidylinositol 3-phosphate in
yeast and mammalian cells." EMBO J 19(17): 4577-4588.
Goberdhan, D. C., M. H. Ogmundsdottir, et al. (2009). "Amino acid sensing and mTOR
regulation: inside or out?" Biochem Soc Trans 37(Pt 1): 248-252.
Goldstein, B., J. R. Faeder, et al. (2004). "Mathematical and computational models of immunereceptor signalling." Nat Rev Immunol 4(6): 445-456.
Guan, K. L. and J. E. Dixon (1991). "Eukaryotic proteins expressed in Escherichia coli: an
improved thrombin cleavage and purification procedure of fusion proteins with
glutathione S-transferase." Anal Biochem 192(2): 262-267.
Gupta-Rossi, N., E. Six, et al. (2004). "Monoubiquitination and endocytosis direct gammasecretase cleavage of activated Notch receptor." J Cell Biol 166(1): 73-83.

189

Haapasalo, A., D. Y. Kim, et al. (2007). "Presenilin/gamma-secretase-mediated cleavage
regulates association of leukocyte-common antigen-related (LAR) receptor tyrosine
phosphatase with beta-catenin." J Biol Chem 282(12): 9063-9072.
Hanada, T., N. N. Noda, et al. (2007). "The Atg12-Atg5 conjugate has a novel E3-like activity
for protein lipidation in autophagy." J Biol Chem 282(52): 37298-37302.
Hara, T., K. Nakamura, et al. (2006). "Suppression of basal autophagy in neural cells causes
neurodegenerative disease in mice." Nature 441(7095): 885-889.
Hayashi-Nishino, M., N. Fujita, et al. (2009). "A subdomain of the endoplasmic reticulum forms
a cradle for autophagosome formation." Nat Cell Biol 11(12): 1433-1437.
He, C., M. Baba, et al. (2008). "Self-interaction is critical for Atg9 transport and function at the
phagophore assembly site during autophagy." Mol Biol Cell 19(12): 5506-5516.
Herman, P. K. and S. D. Emr (1990). "Characterization of VPS34, a gene required for vacuolar
protein sorting and vacuole segregation in Saccharomyces cerevisiae." Mol Cell Biol
10(12): 6742-6754.
Hlavacek, W. S., J. R. Faeder, et al. (2006). "Rules for modeling signal-transduction systems."
Sci STKE 2006(344): re6.
Hopkins, A. L. (2008). "Network pharmacology: the next paradigm in drug discovery." Nat
Chem Biol 4(11): 682-690.
Hosokawa, N., T. Hara, et al. (2009). "Nutrient-dependent mTORC1 association with the ULK1Atg13-FIP200 complex required for autophagy." Mol Biol Cell 20(7): 1981-1991.
Ichimura, Y., T. Kirisako, et al. (2000). "A ubiquitin-like system mediates protein lipidation."
Nature 408(6811): 488-492.
Imami, K., N. Sugiyama, et al. (2008). "Automated phosphoproteome analysis for cultured
cancer cells by two-dimensional nanoLC-MS using a calcined titania/C18 biphasic
column." Anal Sci 24(1): 161-166.
Irwin, J. J. and B. K. Shoichet (2005). "ZINC--a free database of commercially available
compounds for virtual screening." J Chem Inf Model 45(1): 177-182.
Itakura, E., C. Kishi, et al. (2008). "Beclin 1 forms two distinct phosphatidylinositol 3-kinase
complexes with mammalian Atg14 and UVRAG." Mol Biol Cell 19(12): 5360-5372.
Janes, K. A. and M. B. Yaffe (2006). "Data-driven modelling of signal-transduction networks."
Nat Rev Mol Cell Biol 7(11): 820-828.
Juhasz, G., J. H. Hill, et al. (2008). "The class III PI(3)K Vps34 promotes autophagy and
endocytosis but not TOR signaling in Drosophila." J Cell Biol 181(4): 655-666.

190

Jung, C. H., C. B. Jun, et al. (2009). "ULK-Atg13-FIP200 complexes mediate mTOR signaling
to the autophagy machinery." Mol Biol Cell 20(7): 1992-2003.
Kabeya, Y., N. Mizushima, et al. (2000). "LC3, a mammalian homologue of yeast Apg8p, is
localized in autophagosome membranes after processing." EMBO J 19(21): 5720-5728.
Kholodenko, B. N., O. V. Demin, et al. (1999). "Quantification of short term signaling by the
epidermal growth factor receptor." J Biol Chem 274(42): 30169-30181.
Kholodenko, B. N., J. F. Hancock, et al. (2010). "Signalling ballet in space and time." Nat Rev
Mol Cell Biol 11(6): 414-426.
Kihara, A., T. Noda, et al. (2001). "Two distinct Vps34 phosphatidylinositol 3-kinase complexes
function in autophagy and carboxypeptidase Y sorting in Saccharomyces cerevisiae." J
Cell Biol 152(3): 519-530.
Kimura, S., T. Noda, et al. (2007). "Dissection of the autophagosome maturation process by a
novel reporter protein, tandem fluorescent-tagged LC3." Autophagy 3(5): 452-460.
Kirisako, T., Y. Ichimura, et al. (2000). "The reversible modification regulates the membranebinding state of Apg8/Aut7 essential for autophagy and the cytoplasm to vacuole
targeting pathway." J Cell Biol 151(2): 263-276.
Kirschner, M. W. (2005). "The meaning of systems biology." Cell 121(4): 503-504.
Kisen, G. O., L. Tessitore, et al. (1993). "Reduced autophagic activity in primary rat
hepatocellular carcinoma and ascites hepatoma cells." Carcinogenesis 14(12): 2501-2505.
Kitano, H. (2002). "Computational systems biology." Nature 420(6912): 206-210.
Kitano, H. (2002). "Systems biology: a brief overview." Science 295(5560): 1662-1664.
Kitchen, D. B., H. Decornez, et al. (2004). "Docking and scoring in virtual screening for drug
discovery: methods and applications." Nat Rev Drug Discov 3(11): 935-949.
Klionsky, D. J. (2007). "Autophagy: from phenomenology to molecular understanding in less
than a decade." Nat Rev Mol Cell Biol 8(11): 931-937.
Klionsky, D. J., H. Abeliovich, et al. (2008). "Guidelines for the use and interpretation of assays
for monitoring autophagy in higher eukaryotes." Autophagy 4(2): 151-175.
Komatsu, M., S. Waguri, et al. (2006). "Loss of autophagy in the central nervous system causes
neurodegeneration in mice." Nature 441(7095): 880-884.
Komatsu, M., S. Waguri, et al. (2005). "Impairment of starvation-induced and constitutive
autophagy in Atg7-deficient mice." J Cell Biol 169(3): 425-434.

191

Kontaridis, M. I., K. D. Swanson, et al. (2006). "PTPN11 (Shp2) mutations in LEOPARD
syndrome have dominant negative, not activating, effects." J Biol Chem 281(10): 67856792.
Krueger, N. X., D. Van Vactor, et al. (1996). "The transmembrane tyrosine phosphatase DLAR
controls motor axon guidance in Drosophila." Cell 84(4): 611-622.
Kuma, A., M. Hatano, et al. (2004). "The role of autophagy during the early neonatal starvation
period." Nature 432(7020): 1032-1036.
Lahiry, P., A. Torkamani, et al. (2010). "Kinase mutations in human disease: interpreting
genotype-phenotype relationships." Nat Rev Genet 11(1): 60-74.
Ledig, M. M., F. Haj, et al. (1999). "The receptor tyrosine phosphatase CRYPalpha promotes
intraretinal axon growth." J Cell Biol 147(2): 375-388.
Levine, B. (2007). "Cell biology: autophagy and cancer." Nature 446(7137): 745-747.
Li, H., Z. Ren, et al. (2009). "Identification of tyrosine-phosphorylated proteins associated with
metastasis and functional analysis of FER in human hepatocellular carcinoma cells."
BMC Cancer 9: 366.
Liang, C., P. Feng, et al. (2006). "Autophagic and tumour suppressor activity of a novel Beclin1binding protein UVRAG." Nat Cell Biol 8(7): 688-699.
Liang, C., J. S. Lee, et al. (2008). "Beclin1-binding UVRAG targets the class C Vps complex to
coordinate autophagosome maturation and endocytic trafficking." Nat Cell Biol 10(7):
776-787.
Lorenzo, O., S. Urbe, et al. (2006). "Systematic analysis of myotubularins: heteromeric
interactions, subcellular localisation and endosome related functions." J Cell Sci 119(Pt
14): 2953-2959.
Macilwain, C. (2011). "Systems biology: evolving into the mainstream." Cell 144(6): 839-841.
MacKeigan, J. P., L. O. Murphy, et al. (2005). "Sensitized RNAi screen of human kinases and
phosphatases identifies new regulators of apoptosis and chemoresistance." Nat Cell Biol
7(6): 591-600.
Manning, B. D. and L. C. Cantley (2007). "AKT/PKB signaling: navigating downstream." Cell
129(7): 1261-1274.
Martin, K. R., Y. Xu, et al. (2011). "Identification of PTPsigma as an autophagic phosphatase." J
Cell Sci 124(Pt 5): 812-819.
Matsunaga, K., T. Saitoh, et al. (2009). "Two Beclin 1-binding proteins, Atg14L and Rubicon,
reciprocally regulate autophagy at different stages." Nat Cell Biol 11(4): 385-396.

192

Matthews, M. R. (1973). "An ultrastructural study of axonal changes following constriction of
postganglionic branches of the superior cervical ganglion in the rat." Philos Trans R Soc
Lond B Biol Sci 264(866): 479-505.
Mattila, E. and J. Ivaska (2011). "High-throughput methods in identification of protein tyrosine
phosphatase inhibitors and activators." Anticancer Agents Med Chem 11(1): 141-150.
McLean, J., J. Batt, et al. (2002). "Enhanced rate of nerve regeneration and directional errors
after sciatic nerve injury in receptor protein tyrosine phosphatase sigma knock-out mice."
J Neurosci 22(13): 5481-5491.
Meijer, A. J. and P. Codogno (2004). "Regulation and role of autophagy in mammalian cells."
Int J Biochem Cell Biol 36(12): 2445-2462.
Mizushima, N., T. Noda, et al. (1998). "A protein conjugation system essential for autophagy."
Nature 395(6700): 395-398.
Mizushima, N., A. Yamamoto, et al. (2001). "Dissection of autophagosome formation using
Apg5-deficient mouse embryonic stem cells." J Cell Biol 152(4): 657-668.
Mohi, M. G. and B. G. Neel (2007). "The role of Shp2 (PTPN11) in cancer." Curr Opin Genet
Dev 17(1): 23-30.
Mohi, M. G., I. R. Williams, et al. (2005). "Prognostic, therapeutic, and mechanistic implications
of a mouse model of leukemia evoked by Shp2 (PTPN11) mutations." Cancer Cell 7(2):
179-191.
Murray, J. T., C. Panaretou, et al. (2002). "Role of Rab5 in the recruitment of hVps34/p150 to
the early endosome." Traffic 3(6): 416-427.
Nakagawa, I., A. Amano, et al. (2004). "Autophagy defends cells against invading group A
Streptococcus." Science 306(5698): 1037-1040.
Nobukuni, T., M. Joaquin, et al. (2005). "Amino acids mediate mTOR/raptor signaling through
activation of class 3 phosphatidylinositol 3OH-kinase." Proc Natl Acad Sci U S A
102(40): 14238-14243.
Obara, K., T. Noda, et al. (2008). "Transport of phosphatidylinositol 3-phosphate into the
vacuole via autophagic membranes in Saccharomyces cerevisiae." Genes Cells 13(6):
537-547.
Obara, K., T. Sekito, et al. (2008). "The Atg18-Atg2 complex is recruited to autophagic
membranes via phosphatidylinositol 3-phosphate and exerts an essential function." J Biol
Chem 283(35): 23972-23980.
Ohsumi, Y. and N. Mizushima (2004). "Two ubiquitin-like conjugation systems essential for
autophagy." Semin Cell Dev Biol 15(2): 231-236.

193

Petiot, A., E. Ogier-Denis, et al. (2000). "Distinct classes of phosphatidylinositol 3'-kinases are
involved in signaling pathways that control macroautophagy in HT-29 cells." J Biol
Chem 275(2): 992-998.
Polson, H. E., J. de Lartigue, et al. (2010). "Mammalian Atg18 (WIPI2) localizes to omegasomeanchored phagophores and positively regulates LC3 lipidation." Autophagy 6(4).
Proikas-Cezanne, T., S. Ruckerbauer, et al. (2007). "Human WIPI-1 puncta-formation: a novel
assay to assess mammalian autophagy." FEBS Lett 581(18): 3396-3404.
Proikas-Cezanne, T., S. Waddell, et al. (2004). "WIPI-1alpha (WIPI49), a member of the novel
7-bladed WIPI protein family, is aberrantly expressed in human cancer and is linked to
starvation-induced autophagy." Oncogene 23(58): 9314-9325.
Pulido, R., C. Serra-Pages, et al. (1995). "The LAR/PTP delta/PTP sigma subfamily of
transmembrane protein-tyrosine-phosphatases: multiple human LAR, PTP delta, and PTP
sigma isoforms are expressed in a tissue-specific manner and associate with the LARinteracting protein LIP.1." Proc Natl Acad Sci U S A 92(25): 11686-11690.
Rashid-Doubell, F., I. McKinnell, et al. (2002). "Chick PTPsigma regulates the targeting of
retinal axons within the optic tectum." J Neurosci 22(12): 5024-5033.
Reggiori, F., K. A. Tucker, et al. (2004). "The Atg1-Atg13 complex regulates Atg9 and Atg23
retrieval transport from the pre-autophagosomal structure." Dev Cell 6(1): 79-90.
Rubinsztein, D. C., M. DiFiglia, et al. (2005). "Autophagy and its possible roles in nervous
system diseases, damage and repair." Autophagy 1(1): 11-22.
Ruhe, J. E., S. Streit, et al. (2006). "EGFR signaling leads to downregulation of PTP-LAR via
TACE-mediated proteolytic processing." Cell Signal 18(9): 1515-1527.
Rutherford, A. C., C. Traer, et al. (2006). "The mammalian phosphatidylinositol 3-phosphate 5kinase (PIKfyve) regulates endosome-to-TGN retrograde transport." J Cell Sci 119(Pt
19): 3944-3957.
Sajnani, G., A. R. Aricescu, et al. (2005). "PTPsigma promotes retinal neurite outgrowth noncell-autonomously." J Neurobiol 65(1): 59-71.
Salmeen, A., J. N. Andersen, et al. (2003). "Redox regulation of protein tyrosine phosphatase 1B
involves a sulphenyl-amide intermediate." Nature 423(6941): 769-773.
Samuels, Y., Z. Wang, et al. (2004). "High frequency of mutations of the PIK3CA gene in
human cancers." Science 304(5670): 554.
Sapieha, P. S., L. Duplan, et al. (2005). "Receptor protein tyrosine phosphatase sigma inhibits
axon regrowth in the adult injured CNS." Mol Cell Neurosci 28(4): 625-635.

194

Sarkar, S., R. A. Floto, et al. (2005). "Lithium induces autophagy by inhibiting inositol
monophosphatase." J Cell Biol 170(7): 1101-1111.
Schu, P. V., K. Takegawa, et al. (1993). "Phosphatidylinositol 3-kinase encoded by yeast VPS34
gene essential for protein sorting." Science 260(5104): 88-91.
Shen, K., Y. F. Keng, et al. (2001). "Acquisition of a specific and potent PTP1B inhibitor from a
novel combinatorial library and screening procedure." J Biol Chem 276(50): 4731147319.
Shen, Y., A. P. Tenney, et al. (2009). "PTPsigma is a receptor for chondroitin sulfate
proteoglycan, an inhibitor of neural regeneration." Science 326(5952): 592-596.
Simonsen, A., R. Lippe, et al. (1998). "EEA1 links PI(3)K function to Rab5 regulation of
endosome fusion." Nature 394(6692): 494-498.
Siu, R., C. Fladd, et al. (2007). "N-cadherin is an in vivo substrate for protein tyrosine
phosphatase sigma (PTPsigma) and participates in PTPsigma-mediated inhibition of axon
growth." Mol Cell Biol 27(1): 208-219.
Spencer, S. L. and P. K. Sorger (2011). "Measuring and modeling apoptosis in single cells." Cell
144(6): 926-939.
Spiller, D. G., C. D. Wood, et al. (2010). "Measurement of single-cell dynamics." Nature
465(7299): 736-745.
Stack, J. H., P. K. Herman, et al. (1993). "A membrane-associated complex containing the Vps15
protein kinase and the Vps34 PI 3-kinase is essential for protein sorting to the yeast
lysosome-like vacuole." EMBO J 12(5): 2195-2204.
Sternberg, C., M. Benchimol, et al. (2010). "Caspase dependence of the death of neonatal retinal
ganglion cells induced by axon damage and induction of autophagy as a survival
mechanism." Braz J Med Biol Res 43(10): 950-956.
Sun, J. P., A. A. Fedorov, et al. (2003). "Crystal structure of PTP1B complexed with a potent and
selective bidentate inhibitor." J Biol Chem 278(14): 12406-12414.
Sun, Q., W. Fan, et al. (2008). "Identification of Barkor as a mammalian autophagy-specific
factor for Beclin 1 and class III phosphatidylinositol 3-kinase." Proc Natl Acad Sci U S A
105(49): 19211-19216.
Sun, Q., W. Westphal, et al. (2010). "Rubicon controls endosome maturation as a Rab7 effector."
Proc Natl Acad Sci U S A 107(45): 19338-19343.
Sun, Q., J. Zhang, et al. (2011). "The RUN domain of rubicon is important for hVps34 binding,
lipid kinase inhibition, and autophagy suppression." J Biol Chem 286(1): 185-191.

195

Suzuki, K., T. Kirisako, et al. (2001). "The pre-autophagosomal structure organized by concerted
functions of APG genes is essential for autophagosome formation." EMBO J 20(21):
5971-5981.
Taguchi-Atarashi, N., M. Hamasaki, et al. (2010). "Modulation of local PtdIns3P levels by the PI
phosphatase MTMR3 regulates constitutive autophagy." Traffic 11(4): 468-478.
Tanida, I., N. Minematsu-Ikeguchi, et al. (2005). "Lysosomal turnover, but not a cellular level, of
endogenous LC3 is a marker for autophagy." Autophagy 1(2): 84-91.
Tautz, L. and T. Mustelin (2007). "Strategies for developing protein tyrosine phosphatase
inhibitors." Methods 42(3): 250-260.
Tautz, L., M. Pellecchia, et al. (2006). "Targeting the PTPome in human disease." Expert Opin
Ther Targets 10(1): 157-177.
Tay, S., J. J. Hughey, et al. (2010). "Single-cell NF-kappaB dynamics reveal digital activation
and analogue information processing." Nature 466(7303): 267-271.
Thompson, K. M., N. Uetani, et al. (2003). "Receptor protein tyrosine phosphatase sigma inhibits
axonal regeneration and the rate of axon extension." Mol Cell Neurosci 23(4): 681-692.
Tierno, M. B., P. A. Johnston, et al. (2007). "Development and optimization of high-throughput
in vitro protein phosphatase screening assays." Nat Protoc 2(5): 1134-1144.
Tomlins, S. A., R. Mehra, et al. (2007). "Integrative molecular concept modeling of prostate
cancer progression." Nat Genet 39(1): 41-51.
Tonks, N. K. (2005). "Redox redux: revisiting PTPs and the control of cell signaling." Cell
121(5): 667-670.
Tonks, N. K. (2006). "Protein tyrosine phosphatases: from genes, to function, to disease." Nat
Rev Mol Cell Biol 7(11): 833-846.
Uetani, N., M. J. Chagnon, et al. (2006). "Mammalian motoneuron axon targeting requires
receptor protein tyrosine phosphatases sigma and delta." J Neurosci 26(22): 5872-5880.
Uetani, N., K. Kato, et al. (2000). "Impaired learning with enhanced hippocampal long-term
potentiation in PTPdelta-deficient mice." EMBO J 19(12): 2775-2785.
Urbanek, R. A., S. J. Suchard, et al. (2001). "Potent reversible inhibitors of the protein tyrosine
phosphatase CD45." J Med Chem 44(11): 1777-1793.
Van Lieshout, E. M., I. Van der Heijden, et al. (2001). "A decrease in size and number of basal
forebrain cholinergic neurons is paralleled by diminished hippocampal cholinergic
innervation in mice lacking leukocyte common antigen-related protein tyrosine
phosphatase activity." Neuroscience 102(4): 833-841.

196

Vergne, I., E. Roberts, et al. (2009). "Control of autophagy initiation by phosphoinositide 3phosphatase jumpy." EMBO J 28(15): 2244-2258.
Wallace, M. J., J. Batt, et al. (1999). "Neuronal defects and posterior pituitary hypoplasia in mice
lacking the receptor tyrosine phosphatase PTPsigma." Nat Genet 21(3): 334-338.
Walsh, J. P., K. K. Caldwell, et al. (1991). "Formation of phosphatidylinositol 3-phosphate by
isomerization from phosphatidylinositol 4-phosphate." Proc Natl Acad Sci U S A 88(20):
9184-9187.
Webber, J. L. and S. A. Tooze (2010). "New insights into the function of Atg9." FEBS Lett
584(7): 1319-1326.
Wiley, H. S., S. Y. Shvartsman, et al. (2003). "Computational modeling of the EGF-receptor
system: a paradigm for systems biology." Trends Cell Biol 13(1): 43-50.
Wu, H., J. M. Yang, et al. (2006). "Elongation factor-2 kinase regulates autophagy in human
glioblastoma cells." Cancer Res 66(6): 3015-3023.
Wu, Z., P. C. Chang, et al. (2010). "Autophagy Blockade Sensitizes Prostate Cancer Cells
towards Src Family Kinase Inhibitors." Genes Cancer 1(1): 40-49.
Xie, Z., U. Nair, et al. (2008). "Atg8 controls phagophore expansion during autophagosome
formation." Mol Biol Cell 19(8): 3290-3298.
Yeo, T. T., T. Yang, et al. (1997). "Deficient LAR expression decreases basal forebrain
cholinergic neuronal size and hippocampal cholinergic innervation." J Neurosci Res
47(3): 348-360.
Young, A. R., E. Y. Chan, et al. (2006). "Starvation and ULK1-dependent cycling of mammalian
Atg9 between the TGN and endosomes." J Cell Sci 119(Pt 18): 3888-3900.
Zeng, X., J. H. Overmeyer, et al. (2006). "Functional specificity of the mammalian Beclin-Vps34
PI 3-kinase complex in macroautophagy versus endocytosis and lysosomal enzyme
trafficking." J Cell Sci 119(Pt 2): 259-270.
Zhang, Z. Y. (2002). "Protein tyrosine phosphatases: structure and function, substrate specificity,
and inhibitor development." Annu Rev Pharmacol Toxicol 42: 209-234.
Zhong, Y., Q. J. Wang, et al. (2009). "Distinct regulation of autophagic activity by Atg14L and
Rubicon associated with Beclin 1-phosphatidylinositol-3-kinase complex." Nat Cell Biol
11(4): 468-476.

197