'4.
:

Aux—u.
add

a .,
35""

37}

:r
I:r{_:

:

‘I‘
m”,
«w.

.v
man-a
MW:
1-41..
41,,
I

”ml“:

.‘1.

. :I,"r‘\

.‘r.

v“;-

,.
..,,,,,,,..

II," ".N'

.. “.3 0's
“-1"
I

r.
‘(

‘.. '7‘:
V ,..., ..

 

31,“: ' -'

K»

r} .‘l
.97. J; .1 k _
u.
in.
want-r;

«M

.4
v4. “mg,“

.21.?“ u.
r 4.4

w .
”A“

x:

-.n.r~ F

I»: u 1 R4.
21m: w.
.u‘c‘

‘-
I ...

.1. A». a.
n ._l
r.

13;“

x " ‘.
. .V\I!'<}~:,
u: mun-.u
”aim: .uu .
e- . (wan “—1.1
my“. L a

v‘

71
“L :u.
yaw“

r x
xfl'm
2,, ~ .

r

e

' thflC“
_ ”.54.“: ' ~

1‘ - . J
1 Jr: 1
' 4: ’x" I‘ll
H“, H”,
' {IV-157$
4. ‘ .
‘ ' I
“fry-f,
1'11“: A}.
« J-X /

1‘; I

’
P f.-‘.';;,!,w.".€ ‘
.u - v ,

. , .. ”H A

‘ .. . , ..‘

. , . :, 5,-1.3

“ > :H r-

ump”. U‘ u. . ‘ .., , N _

x I ‘ wt- ,.,.. ‘w‘a p x, ”qr“! .

, ,, .. , .,.,,..'.,.‘ .. 1. .-. .. 1-; . p.,,. "7.511.- r

1': . " W4 “"17 I‘ ""YN‘{‘I-'.~':""\ "5"“ I
. ‘.., ,. ..,. u . . ‘ ..

‘ I ' "r _ _ :u‘ , j! 1. .J,.,',V_:‘J, y _'-.~ I ,

{/11 'v’v 7‘ ' _',l L‘L‘hi"! 1.1:,

. punk}; - i": , (i. uiyr K, At...

. ~_ ‘ v I ,L... 41.1... "I "-

‘I

‘ -,
V All" n

n

,

l
u. ,m..
mun...”

“m “u. .I'Lh
,

w;

-. ... l

2. . ,‘U ,‘
~' 7 ¢
. {5 V

f

,l“““.’.§=;"

i":

.r
.4 ..

r -_ e..-

 

\4-‘1 :'

‘. 4,..1n

"'" ‘65???

SITY LIBRAR E

\llllllllllll x All

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled
CRITICAL STRUCTURAL ELEMENTS OF THE VP16

TRANSCRIPT IONAL ACT IVAT ION DOMAIN

presented by
William Douglas Cress, Jr.

has been accepted towards fulfillment
of the requirements for

Ph . D . degree in Biochemistry

 

 

 

4 ’ ,
/Major§ofessor fl
Date W?

MS U is an Affirmative Action/Equal Opportunity Insriturion 0-12771

 

t.

i . _ m i
; LEERAfiV i
gMichigsn Rica, ‘1

i univa r$éfigz i

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Institution
c:\circ\detedue.pm3-p.1

 

CRITICAL STRUCTURAL ELENIENTS OF THE VP16
TRANSCRIPTIONAL ACTIVATION DOMAIN

By

William Douglas Cress, Jr.

A DISSERTATION
Submitted to
Michigan State University ‘
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Biochemistry

1991

ABSTRACT

CRITICAL STRUCTURAL ELENIENTS OF THE VP16
TRANSCRIPTIONAL ACTIVATION DOMAIN

By

William Douglas Cress, Jr.

VP16 is a virion protein of herpes simplex virus type 1 which enters the host
cell upon infection and potently and specifically activates transcription of the viral
immediate early genes. The domain of VP16 which confers the ability to activate
transcription is one of the strongest transcriptional activation domains yet identified.
The activation domain of VP16, rich in acidic amino acids, has been modeled as a
random coil and alternatively as an amphipathic alpha-helix. However, neither of
these models has previously been experimentally tested.

In the present work, I have used a combination of site-directed mutagenesis
and biophysical analysis in an attempt to determine which elements of structure are
critical to the function of the VP16 transcriptional activation domain. Mutational
analysis demonstrated that: 1) the VP16 activation domain activates transcription
overall as a function of its net negative charge, 2) substitution of helix-breaking

proline residues within the predicted amphipathic alpha-helix of the VP16 activation

domain does not affect its activity, and 3) a phenylalanine residue at VP16 position
442 (Phe 442) is critical to its function. These observations are inconsistent with
proposed random coil and amphipathic alpha-helix models.

Ultraviolet spectroscopy revealed that VP16 Phe 442 is buried within the
activation domain structure. Furthermore, comparison of the amino acid sequences of
a variety of transcriptional activation domains revealed that a pattern of bulky
hydrophobic residues is somewhat conserved. These observations suggest that
intramolecular hydrophobic interactions may be critical to the structure and function
of the VP16 activation domain and of transcriptional activation domains in general.
The elements of secondary structure of the VP16 activation domain were directly
measured using Fourier-transform infrared and circular dichroism spectroscopies.
Spectral analysis indicates that the VP16 activation domain is composed primarily of

coil and beta-sheet structures and possesses very little alpha-helix.

Whatever you do,
whether in word or deed,
do it in the name of the Lord Jesus,
giving thanks to God the Father through Him.

The Apostle Paul in Col. 3:17

iv

 

ACKNOWLEDGNIENTS

I thank Steve Triezenberg for being an excellent graduate advisor and role
model and I acknowledge his great contribution to the work described herein. I also
thank my wife, Andrea Cress, for her tolerance as fellow student and collaborator and
I acknowledge her input. I thank Jeff Regier for painstakingly correcting many
versions of this dissertation. I also thank Drs. Lee Velicer, Lee McIntosh, Zach
Burton and Bob Hausinger for their service as members of my guidance committee. I
acknowledge the important contributions of my collaborators, including Drs. Shelley
Berger and Leonard Guarente (MIT), Drs. Jim Ingles, Michael Shales and Jack
Greenblatt (Univ. of Toronto) and Mark Prairie and Dr. Bill Kreuger (Upjohn Co. ,
Kalamazoo, MI). Those who have given advice or have loaned equipment have been
explicitly referenced within the text. I thank those who took the time to help me
learn molecular biology early in my graduate career: Jianli Cao, Don Lorimer,
Matthew Morrel, Mark Bloom and Dave Schwab. I thank Steve Triezenberg, Dave
Pepper], Andrea Cress, Jeff Regier, Rath Pichyangkura, and Lisa Ortquist for making
522 a pleasant place to work. I acknowledge the Westside Deli, the Korea House and
the Thai Kitchen for their contributions to intellectual and not-so-intellectual

communication during Thursday Noon. Finally, I thank my parents for their years of

support.

 

TABLE OF CONTENTS

 

PAGE

LIST OF TABLES ...................................... viii

LIST OF FIGURES ...................................... ix

LIST OF ABBREVIATIONS ................................ xi

CHAPTER I: INTRODUCTION ............................. 1

Eukaryotic Transcription ............................... l

Trans-activator Proteins ................................ 3

Mechanisms of Action of Trans-activator Proteins ................ 6

VP16 ........................................... 9

Acidic Transcriptional Activation Domains .................... 11

Overview ......................................... 18
CHAPTER II: ANALYSIS OF THE VP16 ACTIVATION DOMAIN

BY SITE-SPECIFIC MUTAGENESIS ............... 20

Introduction ....................................... 20

Construction of VP16 Derivatives .......................... 20

Correlation Between Net Negative Charge and Activation ........... 28

Correlation Between Predicted Amphipathy and Activation .......... 28

Role of Putative Alpha-helical Structure in Activation .............. 33

A Critical Role for Phe 442 in Activation ..................... 36

Conservation of Bulky Hydrophobic Residues Among
Various Activation Domains ......................... 43
Conclusions ....................................... 48

vi

TABLE OF CONTENTS (cont’d)

PAGE

CHAPTER III: ANALYSIS OF THE VP16 ACTIVATION DOMAIN
BY SPECTROSCOPIC AND BIOPHYSICAL METHODS . . 50

Introduction ...... , ................................. 50
GAL4-VP16 Fusion Proteins ............................. 50
Composition of GAL4-VP16 Fusion Proteins ................... 52
Purification of GAL4-VP16 Fusion Proteins ................... 55
Purification of VP16 Activation Domain Peptides ................ 5 8
Ultraviolet Spectroscopy ............................... 5 9
Fourier-Transform Infrared Spectroscopy ..................... 73
Circular Dichroism ................................... 81
Native Gel Electrophoresis .............................. 83
Differential Scanning Calorimetry .......................... 86
Conclusions ....................................... 90
CHAPTER IV: DISCUSSION ............................... 91
Introduction ....................................... 91
Mechanism of Action of the VP16 Activation Domain ............. 91
Critical Structural Elements of the VP16 Activation Domain ......... 96
Future Studies ...................................... 99
LIST OF REFERENCES .................................. 100

vii

 

 

LIST OF TABLES
PAGE
Chapter II

Table 1. Mutagenic oligonucleotides ............................ 25

viii

LIST OF FIGURES

Chapter I

Figure l. A schematic representation of a pattern purported to be

diagnostic of acidic transcriptional activation domains ........

Chapter 11

Figure 1. Nucleotide and amino acid sequences of the minimal

activation domain of VP16 .........................

Figure 2. Relationship between net negative charge and the

relative activities of VP16 derivatives ..................

Figure 3. Schematic models depicting the predicted amphipathic
alpha-helix of the minimal VP16 transcriptional activation domain

Figure 4. Schematic helical-wheel model representing the putative
amphipathic alpha-helical structure of the minimal VP16 activation

domain .....................................

Figure 5. The VP16 minimal activation domain does not activate

transcription as a function of its predicted amphipathy ........

Figure 6. VP16 Phe 442 is critical in transcriptional activation . .

Figure 7. VP16 derivatives having substitutions at Phe 442 are

expressed at levels indistinguishable from the parental protein . . .

Figure 8. Substitution of Phe 442 does not affect the binding

of VP16 to host DNA-binding proteins .................

Figure 9. Amino acids proximal to Phe 442 are critical in

transcriptional activation ..........................

ix

PAGE

.......... 17

.......... 22

.......... 3O

.......... 31

.......... 32

.......... 35

.......... 37

.......... 38

.......... 40

.......... 42

LIST OF FIGURES (cont’d)

PAGE
Chapter II (cont’d)
Figure 10. Conservation of bulky hydrophobic residues among
activation domains ........................................ 46
Chapter HI

Figure 1. GAL4 1-147 and GAL4—VP16 fusion proteins ................. 54
Figure 2. UV spectra of free tyrosine at different pHs .................. 62
Figure 3. UV spectra of Tyr 442 at different pHs .................... 65
Figure 4. Difference spectra of GAL4-del456 and GAL4-FY 442 ........... 68
Figure 5. Difference spectrum of GAL4-VP16 ...................... 71
Figure 6. FTIR spectra of model proteins ......................... 76
Figure 7. FTIR spectra of GAL4-VP16 proteins .................... 78
Figure 8. CD analysis of the full length VP16 activation domain ........... 82
Figure 9. Native gel electrophoresis of GAL4-VP16 proteins .............. 85
Figure 10. DSC scans of GAL4 1-147 and GAL4-VP16 ................ 89

BSA

CAPS

CAPSO

CD

DE

DSC

FPLC

HEPES

HSV-l

ICP

LIST OF ABBREVIATIONS

absorbance at 230 nm

absorbance at 280 nm

absorbance at 600 nm

amino acid

amphipathic alpha-helix

adenosine triphosphate

basepair

bovine serum albumin

3-[cyclohexylamino]- l-propanesulfonic acid
3-[cyclohexylamino]-2-hydroxyl-1-propane sulfonic acid
circular dichroism

delayed early

differential scanning calorimetry

fast protein liquid chromatography

Fourier-transform infrared

N —2-hydroxyethylpiperiazine-N ’-2-ethanesulfonic acid
herpes simplex virus type-1

infected cell protein

immediate early

xi

LIST OF ABBREVIATIONS (Cont’d)

kb kilobases

kd kilodaltons

L late

LTR long terminal repeat

MSV murine sarcoma virus
NMR nuclear magnetic resonance
NTP nucleotide triphosphate

psi pounds per square inch
SDS sodium dodecyl sulfate

SDS-PAGE sodium dodecyl sulfate-polyacrylamide gel electrophoresis

TFII transcription factor of RNA polymerase II

tk thymidine kinase

TRIS tris(hydroxymethyl)aminomethane hydrochloride
UAS upstream activating sequence

UV ultraviolet

VP16 virion protein number 16

Single letter abbreviations for the amino acids: A, Ala; C, Cys; D, Asp; E, Glu;
F, Phe; G, Gly; H, His; 1, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R,
Arg; 8, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

xii

CHAPTERI

INTRODUCTION

Eukaryotic Transcription

An important question in biology is how eukaryotic cells regulate the
expression of their protein coding genes during development and in response to
environmental stimuli. A key regulatory step in the expression of protein coding
genes is the initiation of transcription by RNA polymerase II (RNA pol II) which
transcribes cellular mRNAs. Unlike the RNA polymerase holoenzyme of eubacteria,
catalytic RNA pol H does not recognize promoters by itself (reviewed by Roeder,

1976), but rather requires additional factors (Matsui et a1. , 1980; Davidson et a1. ,

 

 

1983; Samuels et a1., 1982; Dynan and Tjian, 1983; Dignam et al., 1983). These

 

factors can be classified into two categories; the basal transcription factors which by
definition are required for recognition of all class II promoters, and the specific
transcription factors which operate at promoters containing specific regulatory
sequences.

The RNA pol II transcriptional regulatory elements can be divided (somewhat
artificially) into two classes; namely the basal elements which are directly recognized
by the basal transcription factors and the promoter-specific regulatory elements which
are recognized by promoter-specific regulatory proteins. Two basal regulatory

elements are the initiator element which includes or lies very close to the transcription

2
start site (Corden et al., 1980; Smale and Baltimore, 1989; Nakatani et al., 1990) and

 

the TATA-box which is generally positioned around 30 nucleotides 5’ to the start of
transcription in mammalian cells (reviewed by Breathnach and Chambon, 1981). The
initiator and the TATA box element both appear to interact with the basal
transcription factor TFIID (see below) and apparently function to establish the start
site (Smale e_tjlé, 1990; Nakatani ML, 1990) and direction (Carcamo et_aL, 1990)
of transcription. RNA pol II promoters may not require both of these basal elements,
since functional promoters lacking each have been identified or constructed (Azizkhan
fl, 1986; Myers et_a_l,_, 1986; Smale fl, 1990; Nakatani gal, 1990). In cases
where both an initiator sequence and a TATA-box are present, they work in
cooperation to yield a stronger basal promoter than is observed with either site alone

(Smale and Baltimore, 1989; Nakatani et al., 1990; Conaway et al., 1990).

 

The first step in the assembly of an initiation complex is the step of template
commitment (reviewed by Sawadogo and Sentenac, 1990) during which the basal
transcription factor TFIID binds weakly to the TATA-box (and / or initiator element)
(Fire et_al,, 1984; Reinberg gal” 1987; Buratowski fl, 1989; Smale and
Baltimore, 1989; Smale e_t_al._, 1990). Stable binding of TFIID appears to be
facilitated by TFIIA (Davidson e_tA” 1983; Fire fl, 1984; Reinberg gal, 1987).
Subsequent to stable complex formation, the remaining basal transcription factors and
RNA pol II assemble to form the complete preinitiation complex (Davison 1g,
1983; Reinberg etal, 1987; Reinberg and Roeder, 1987; Buratowski e131,, 1989)
which is thought to consist of TFIID, TFIIA, TFIIB, RNA Pol II, TFIIE (Ohkuma e_t

a, 1990), TFIIF (Flores et al., 1989) or its RNA pol II associated equivalent

 

 

3
RAP30/74 (Burton et al., 1988) and the recently discovered TFIIG (Sumimoto et al.,

1990). Once the preinitiation complex has formed, there is a requisite hydrolysis of
the beta-gamma bond of ATP (Bunick eLaL, 1982) followed by the formation of the
first phosphodiester bond. Following initiation, TFIID appears to remain bound at
the transcription start site, whereas the other transcription factors must reassemble to
initiate transcription a second time (Lin and Green, 1991).

In addition to the basal promoter elements, many RNA pol II promoters
possess specific regulatory sequences. Many of these promoter-specific elements,
such as the CCAAT or GC elements, exert their control within a few hundred
nucleotides from the transcription start site, thus they are generally considered part of

the proximal promoter (McKnight and Kingsbury, 1982; Myers et al., 1986; Maniatis

 

egg; 1987). Other promoter-specific regulatory sequences can exert control from
many thousands of bases away from the start site of transcription. Such promoter
elements which increase the rate of transcriptional initiation are termed enhancers
(reviewed by Khoury and Gruss, 1983) and the proteins which bind to them are
termed Ens-activators (McKnight and Tjian, 1986; Mitchell and Tjian, 1989;
Johnson and McKnight, 1989). Promoter-specific regulatory elements which decrease
the rate of transcriptional initiation are termed repressors (reviewed by Levine and

Manley, 1989).

Trans-activator Proteins
Many tran s—activator proteins appear to be modular, in that they are composed

of at least two separable domains, one of which binds to specific DNA elements and

4

the other of which confers the ability to activate transcription (Brent and Ptashne,
1985; Hope and Struhl, 1986). There are several well defined structural classes of
DNA-binding domains, as well as several more poorly defined classes (reviewed by
Mitchell and Tjian, 1989; Johnson and McKnight, 1989). The well-characterized
classes of DNA-binding domains have recognizable conserved sequence motifs which
are often given imaginative names. One such motif is the "Zinc finger" which was

originally identified as the DNA—binding structure of TFIIIA (Miller et al., 1985).

 

The Zn-binding domains of TFIIIA are composed of approximately 30 amino acids
with two invariant cysteines and two invariant histidines which stabilize the domain by
tetrahedrally coordinating a single Zn2+ ion. A second type of zinc binding domain,
present in the DNA-binding domains of the steroid hormone receptors, utilizes two
pairs of cysteines instead of the cysteine/histidine combination (Freedman gm,
1988). A third type of Zn-containing DNA-binding domain is the binuclear cluster
exemplified by the DNA-binding domain of GAL4 (Pan and Coleman, 1990), in
which four cysteine residues coordinate two Zn2+ ions.

A second type of DNA-binding domain is the "bZip" motif which was first

identified as the dimerization/DNA-binding domain of C/EBP (Johnson et al., 1987;

 

Landscultz et_al., 1988). The bZip motif comprises two regions; a basic region (the b
in bZip) composed of thirty or so amino acids characterized by net positive charge,
and the "leucine zipper" region (the Zip in bZip) which is composed of a stretch of
amino acids having four leucine residues positioned at intervals of seven amino acids.
Biophysical and genetic characterizations clearly suggest that the leucine zipper region

forms a coiled-coiled intermolecular dimer of helices (O’Shea et al., 1989; Hu et al.,

 

5
1990). In complex with DNA, the basic regions of the bZip dimer are thought to

surround the DNA with each basic helix malcing direct contact with the major groove

of DNA (Vinson et al., 1989). A DNA—binding motif which appears similar to the

 

bZip domain is the helix-loop-helix domain (Murre et al., 1989). The

 

helix-loop-helix domain contains a basic region which appears to be involved in DNA
binding and a dimerization region composed of two short putative amphipathic helices

separated by a short intervening region which might form a loop (Davis et al., 1987;

 

Voronova and Baltimore, 1990).

A third well defined class of DNA-binding domains is the homeodomain which

 

encompasses about 60 amino acids (Scott et al., 1989). The crystal structure of the
D. melanogaster protein engrailed homeodomainzDNA complex has recently been

solved (Kissinger et al., 1990). The structure of this domain is composed of three

 

helices (1-3). Helices l and 2 pack against each other in an antiparallel arrangement;
helix 3 is perpendicular to l and 2. The hydrophobic face of helix 3 makes contact
with l and 2, while the exposed hydrophilic face makes direct contact with the major
groove of the DNA.

The DNA-binding domains which have been extensively characterized appear
to form very highly ordered three-dimensional structures and have been found to have
invariant sequence elements. In contrast, the activation domains of gaps-activator
proteins do not appear to possess clear sequence homology to each other, but rather
have been classified into three general categories; acidic, glutamine-rich, and
proline—rich, based upon richness of particular amino acids (Mitchell and Tjian,

1989). Activation domains have typically been identified by deletion analysis and

6

have been somewhat characterized using genetics and site-directed mutagenesis. The
best studied of these activation domains are those of the acidic class which will be
reviewed in detail in a following section. In no case has the three-dimensional

structure of a transcriptional activation domain been reported.

Mechanisms of Action of Trams-activator Proteins

When enhancer regulatory elements were first identified they posed something
of a dilemma. How could a regulatory protein bound thousands of nucleotides away
from a transcription start site affect the rate of productive transcriptional initiation by
RNA pol II? A number of possible mechanisms were proposed to explain the
activation phenomenon, which included models termed twisting, sliding, oozing and
looping (reviewed by Ptashne, 1986). Presently, the looping model is generally
favored. In this model the tran_s-activator binds to its particular fi-regulatory element
through its DNA-binding domain, the transcriptional activation domain makes contact
with some target protein or proteins involved in transcription and the long stretch of
intervening DNA simply loops out (Ptashne, 1988; Ptashne and Gann, 1990). An
important and unanswered question in regard to the looping model is what is the
target (or targets) of the activation domains of trans—activator proteins. Numerous
targets have been postulated (reviewed by Ptashne, 1988; Ptashne and Gann, 1990)
and it is becoming evident that indeed a number of mechanisms of tr_an_s-activation
may occur.

Perhaps the simplest mechanism of activation demonstrated so far is termed

 

antirepression (Croston et al., 1991). This mechanism of transcriptional activation

 

7

involves the fact that transcriptional regulation occurs in the context of genomic DNA
packaged into chromatin structure. It is clear that such packaging is inhibitory of
transcriptional initiation in vitrg (Workman et al., 1991; Croston et a1. , 1991) and

presumably in vivo (reviewed by Elgin, 1988). It has recently been shown that

 

mans-activator proteins from the glutamine-rich and acidic classes of activators are
able to overcome the inhibition of initiation by nucleosomes or histone H1 Mtg;
(Workman et_al., 1991; Croston et_al” 1991). This observation indicates that at least
one mechanism of activation utilized by gag-activators involves overcoming
inhibition by histones, presumably by making a promoter template more accessible to
transcription factors.

A second demonstrated mechanism of action of a m-activator protein also
relies on relieving a repression; in this case not a repression of initiation, but a
repression of elongation. This mechanism of activation is exemplified by the hsp 70
promoter of D. melanogaster. Prior to the induction of heat shock, a single RNA pol
II molecule is transcriptionally engaged at the hsp 70 promoter, it has formed a
nascent RNA chain of approximately 25 nucleotides, but it is apparently arrested at
that point (Gilmour and Lis, 1986; Rougvie and Lis, 1988). Upon heat shock, the
heat shock transcription factor (I-ISF) , which previously was not bound to DNA, is
activated such that it binds to the hsp 70 heat shock response element and through an
unknown mechanism frees the arrested RNA pol 11.

Activation of transcription by gang-activator proteins, however, is not limited
to relieving repression, since transcriptional activation 'mv_itr_o has been observed on

naked DNA in the absence of histone proteins using relatively pure preparations of

 

8

basal transcription factors (Peterson et a1. , 1990). Thus it is possible that some

 

m-activator proteins may activate transcription by recruiting basal factors to the
promoter site. For instance, genetic experiments (Allison and Ingles, 1989) and in
m experiments using ms-activator proteins as affinity column ligands (Bde and
Struhl, 1989) suggest that yeast Ens-activators may interact directly with RNA pol
II. Similarly, the TATA element-binding protein TFIID is clearly a candidate for the
target of transcriptional activation domains. Experimental evidence that TFIID could
be a target of transcriptional activators came from the observation that the
transcription factor ATF interacts with the TFHD protein to facilitate the

establishment of a preinitiation complex (Horikoshi et a1. , 1988; Hai et al., 1988).

 

 

Further evidence that TFIID might be a direct target of m-activators came from an
experiment using cloned recombinant yeast TFHD (Stringer et_al., 1990). In this
experiment the acidic activation domain of VP16 (see below) was used as a column
ligand in affinity chromatography. They found that the VP16 affinity column bound
very selectively to human TFIID and purified recombinant yeast TFIID.

Lin and Green (1991) have recently discovered evidence that the VP16
activation domain specifically interacts with yet another basal transcription factor,
TFIIB. In this work, Lin and Green used the VP16 activation domain as a
chromatography affinity matrix in a manner similar to Stringer _et_al_. (1990). They
passed Hela cell nuclear extracts over the VP16 affinity column and observed binding
of TFIID to the column. However, Lin and Green went on to demonstrate that
TFIID is not the only basal transcription factor bound to the VP16 column. TFIIB

was also bound and appeared to be bound with greater affinity than is TFIID (with

 

8

basal transcription factors (Peterson et a1. , 1990). Thus it is possible that some

 

m—activator proteins may activate transcription by recruiting basal factors to the
promoter site. For instance, genetic experiments (Allison and Ingles, 1989) and 'm
\_'_i_tr_'Q experiments using m-activator proteins as affinity column ligands (Bde and
Struhl, 1989) suggest that yeast m—activators may interact directly with RNA pol
11. Similarly, the TATA element-binding protein TFIID is clearly a candidate for the
target of transcriptional activation domains. Experimental evidence that TFIID could
be a target of transcriptional activators came from the observation that the
transcription factor ATF interacts with the TFIID protein to facilitate the

establishment of a preinitiation complex (Horikoshi et al. , 1988; Hai et al., 1988).

 

 

Further evidence that TFIID might be a direct target of trans-activators came from an

experiment using cloned recombinant yeast TFIID (Stringer et al. , 1990). In this

 

experiment the acidic activation domain of VP16 (see below) was used as a column
ligand in affinity chromatography. They found that the VP16 affinity column bound
very selectively to human TFIID and purified recombinant yeast TFIID.

Lin and Green (1991) have recently discovered evidence that the VP16
activation domain specifically interacts with yet another basal transcription factor,
TFIIB. In this work, Lin and Green used the VP16 activation domain as a
chromatography affinity matrix in a manner similar to Stringer gal. (1990). They
passed Hela cell nuclear extracts over the VP16 affinity column and observed binding
of TFIID to the column. However, Lin and Green went on to demonstrate that
TFIID is not the only basal transcription factor bound to the VP16 column. TFIIB

was also bound and appeared to be bound with greater affinity than is TFIID (with

9
respect to the concentration of KCl used as eluant). Stringer et_g, (1990) did not

eliminate the possibility that TFHB was bound to their columns.

VP16

Herpes simplex virus type I is a widespread human pathogen (reviewed by
Whitley, 1990) which has been used as a model system for the study of gene
expression in mammalian cells. HSV-l possesses a large double-stranded DNA
genome encoding approximately 70 genes which are transcribed by host RNA pol H
(reviewed by Roizman and Sears, 1990). During lytic HSV-l infection there is
temporal expression of three major classes of coordinately regulated genes known as
immediate-early (IE), delayed early (DE), and late (L) genes (reviewed by Wagner,
1990). Temporal HSV-l gene expression is regulated by a mechanism whereby
accumulated IE gene products suppress IE expression and activate the expression of
DE genes. DE gene products result in the expression of L genes by initiating replica-
tion of the viral genome upon which L gene expression is dependent. A late gene
product and component of the HSV-l virion, termed VP16 (also known as Vmw65,
alpha-TIF and ICP25), then specifically and potently activates IE gene expression
during the next round of infection (Batterson and Roizman, 1983; Campell et_g,
1984; reviewed by Roizman and Spector, 1991).

Activation of IE gene expression by VP16 depends upon specific sequence
elements present (often in several copies) in all IE 9i; regulatory regions (Mackem
and Roizman, 1982; Cordingley et_aL, 1983; Preston fl, 1984; Bzik and Preston,

1986). The sequence motif 5’-TAATGARAT (where R indicates a purine) is the

10

major sis-regulatory factor in VP16-mediated tans-activation (Kristie and Roizman,

1984; Triezenberg et al., 1988a). VP16 is an atypical member of the class of

 

gang-activator proteins since, apparently, VP16 does not have a high affinity for
DNA (Marsden fl, 1987 ; Kristie and Sharp, 1990), but rather is tethered to the
TAATGARAT element by binding to the ubiquitous octamer-binding protein Oct-l
which in turn binds the TAATGARAT (or the overlapping sequence, 5 ’-ATGCTAAT)
sequence (Gerster and Roeder, 1988; Preston iii, 1988; O’Hare and Goding,

1988). At least one other host factor appears to be involved in the VP16:Oct-1zDNA
complex (McKnight et_al., 1987 ; Kristie and Sharp, 1990; Katan e_tA, 1990), but its
role and identity are unknown. A second motif, rich in G and A residues, is also
important for maximal VP16-mediated m-activation (Triezenberg et_al., 1988a;

Spector et al. , 1990). The host protein, termed IEF that binds to this element has

ga’
been characterized (LaMarco and McKnight, 1989), but a direct interaction with
VP16 has not been demonstrated.

VP16 is present at an estimated 500-1000 copies per virion (Spear and
Roizman, 1972; Hiene fl, 1974), but it is unknown how VP16 accompanies the
viral DNA to the cell nucleus. The VP16 polypeptide is composed of 490 amino acid
residues as inferred from DNA sequence (Pellett et_al, 1985; Dalrymple girl, 1985;
Triezenberg et_al., 1988b) with a calculated (Devereux fl, 1984) molecular weight
and isoelectric point of 54,316 daltons and 4.7, respectively. VP16 purified from

virus particles is a phosphoprotein with an apparent molecular weight of

approximately 65 kd as measured by SDS-PAGE (Gibson and Roizman, 1974).

11

Despite the apparent discrepancy between calculated and predicted molecular weights
the VP16 polypeptide does not appear to be glycosylated (Hiene et_al_., 1974).
Molecular genetic studies of VP16 have divided the polypeptide into two
functional domains relevant to its function as a tr__a_n_§-activator. The IE specificity
function of VP16 (i.e. the ability of VP16 to interact with Oct-1) is conferred by a
poorly defined domain (or domains) contained within the N-terminal 400 amino acids

of VP16 (Triezenberg et al. 1988b, Ace et al., 1988, Greaves and O’ Hare 1989,

 

 

Werstuck fl” 1990). The transcriptional activation function is carried within a
domain composed of the C-terminal 80 amino acids of VP16 (Triezenberg et_al_.,
1988b; Sadowski et_al., 1988; Greaves and O’ Hare, 1989; Cousens et_al., 1989;
Werstuck and Capone, 1989). The 80 amino acid C—terminal domain of VP16 has
been shown to be sufficient to function as an activation domain (independent of other
VP16 polypeptide sequences) by attaching it to an unrelated DNA—binding domain
from the yeast GAL4 protein. This chimeric protein has been found to be among the

strongest transcriptional activators tested both in vivo (Sadowski et al., 1988; Cousens

 

et al., 1989) and in vitro (Chasman et al., 1989; Carey et al., 1990a).

 

Acidic Transcriptional Activation Domains

Inspection of the amino acid sequence of the 80 amino acid VP16 activation
domain reveals that this domain is rich in acidic residues. Yeast transcriptional
activator proteins GAL4 (Ma and Ptashne, 1987a and b), GCN4 (Hope and Struhl,
1986), and HAP4 (Forsburg and Guarente, 1989), and a variety of mammalian

hormone receptors (Hollenberg and Evans, 1988; Godowski et al., 1988; reviewed by

12

Evans, 1988; Zhu fl, 1990) also contain activation domains which possess net
negative charge, suggesting that negative charge is a critical component of at least one
class of eukaryotic transcriptional activators (reviewed by Mitchell and Tjian, 1989) .
GAL4 is a yeast protein of 881 amino acids which is required for the
expression of the Saccharomyces oerevisiae genes needed for galactose metabolism
(Douglas and Hawthorne, 1964). GAL4 binds to upstream activating sequence
(U AS G) of galactose metabolizing genes and activates their transcription. GAL4
binds to UASG DNA through an N-terminal domain (amino acids 1-147) and
possesses two acidic transcriptional activation domains I and II, encompassed by
amino acids 148-196 and 768-881, respectively (Ma and Ptashne, 1987a). In an
effort to study the critical elements of an acidic activation domain of GAL4, Gill and
Ptashne (1987) chemically mutagenized the GAL4 DNA region I activation domain.
The authors characterized seven mutations that increased the activity of the GAL4
region 1 activation domain, five which retained partial activity, and twenty which
were completely inactive. All of the mutations which increased the activity of the
mutated GAL4 protein increased the net negative charge of the domain by replacing
positively charged residues with neutral counterparts. However, no improved
activators were obtained resulting from the addition of negative charge into the
domain, suggesting that indiscriminant placement of negative charge into the domain
was not an easy way to increase its strength as an activator. Of the five partially
active mutants characterized, two neutralized acidic residues, two replaced hydrophilic
residues with hydrophobic counterparts and one introduced a stop codon which

severely truncated the domain. The observation that two mutations at noncharged

13

residues decreased the activation function of GAL4 suggests that some structural
requirement in addition to net negative charge is important to the function of the
domain. Of the 20 completely inactive GAL4 derivatives which were isolated none
encoded the entire GAL4 gene protein. The authors suggest that their inability to
isolate single (or even multiple) amino acid substitutions that completely inactivated
the domain was evidence of its functional redundancy.

In an accompanying paper, workers from the same lab reported a novel type
of genetic experiment aimed at understanding what elements are critical to the
function of acidic activation domains (Ma and Ptashne, 1987b). In this experiment,
the authors fused "random" 8J1; 3A fragments of M genomic DNA to the gene
encoding the GAL4 DNA-binding domain (GAL4 amino acids 1-147). They then
screened for fusion proteins that functioned as transcriptional activators in yeast. The
strength of the activating hybrids roughly correlated with net acidity of the amino
acids coded by the random m DNA fragments. The activating E43111 sequences
showed no extended homology to each other (particularly in relation to distribution of
negative charge) or to the native GAIA activation domains.

Hope, Mahadevan and Struhl (1988) characterized an acidic activation domain
from the yeast GCN4 protein (Hope and Struhl, 1986) by high resolution deletion
mutagenesis. In this experiment the authors fused twenty-six carboxyl-terminal
deletions of the GCN4 activation domain to the 100-residue GCN4 DNA-binding
domain and measured the transcriptional activation function of each derivative in an i_n
yi_vo assay. The activities of these deletion derivatives fit a simple pattern; a series of

small stepwise decreases in activity were observed as amino acids (and net negative

 

14

charge) were truncated. The authors argue from this data that the activation domain
of GCN4 must be unlike rigid structural elements, such as active sites of enzymes,
where one would expect some point where there would be a sudden loss of activity.
The authors suggest that the GCN4 activation domain represents a "repeated structure
composed of small units acting additively" and that it might possibly function as "a
dimer of alpha-helices from two GCN4 monomers ".

The results of the previously described work on GAL4 and GCN4 suggest that
yeast transcriptional activation domains require some particular distribution of
negative charge (Gill and Ptashne, 1987 ; Ma and Ptashne, 1987b), but are
functionally redundant (Hope gal, 1988). Unfortunately these studies did not
identify any sequence homologies that would suggest structural similarities among the
various activating peptides. This observation is not a totally novel one in that, as
pointed out by Ma and Ptashne (1987b), there are other examples of protein
sequences which bear little sequence homology, but which possess identical
biochemical activities. For instance, the sequences which identify proteins for
secretion bear little sequence homology (reviewed by Briggs and Gierasch, 1986), nor
do the signal sequences of mitochondrial proteins (von Heijne, 1986; Schatz, 1987).
Indeed, it has been demonstrated that both the signal sequences of secreted (Kaiser e_t
al, 1987) and of mitochrondrial proteins (Baker and Schatz, 1987) can be functionally
replaced by a large fraction of random peptide sequences.

The apparent lack of sequence homology among various acidic transcriptional
activation domains led X-ray crystallographer Paul Sigler (1986) to propose that

acidic activation domains may function, more or less, as random coils and that a

15

specific secondary structure is not a critical requirement. This model is not
inconsistent with mest of the results just described. However, the "acid blob" or
"negative noodle" model (as it has been dubbed) is difficult to reconcile with the
observation that mutations not affecting charge of the GAL4 region I activation
domain can nonetheless have significant effects on activity.

A second model (Gininger and Ptashne, 1987; Ptashne, 1988; Hope et_all,
1988) proposed that acidic activation domains form amphipathic alpha-helices (AAHs)
possessing opposing hydrophobic and acidic "faces" (a tendency toward a third
hydrophilic face was also suggested, as well as dimerization mediated by the
hydrophobic faces). This model stems from the observation that many of the acidic
activation domains identified would indeed form such structures if they folded into
alpha-helices. To test this model Gininger and Ptashne (1987) constructed a DNA
fragment encoding a peptide which could potentially form a four-tum, negatively
charged, alpha-helix having alternating hydrophobic, hydrophilic and acidic faces. In
addition, a second DNA fragment encoding a peptide which might also form a
four—tum, negatively charged, AAH, but in which the hydrophilic face of the first
peptide (glutamine residues) was replaced by a hydrophobic face (valine residues).
As controls, two peptides of identical composition were constructed which would not
form amphipathic structures if folded into alpha-helices. Each of these peptides were
fused to the DNA-binding domain of GAL4 (amino acids 1-147), the fusion proteins
were expressed in yeast and assayed for activation of a GALl—ng fusion gene. Only
the two peptides which could form AAHs were functional. The peptide proposed to

have alternating hydrophobic, hydrophilic and acidic faces (termed AH) functioned

 

16

more than half as well as the region I activation domain of GAL4 while the peptide
which could potentially have only acidic and hydrophobic faces functioned only 5 % as
well. The control peptides which could not form amphipathic structures were not
functional.

The suggestion that acidic activation domains might function as negatively
charged AAHs prompted a group led by Temple Smith and James Figge to use a
supercomputer system [ARIEL (Lathrop mg, in press)] to compare the sequences of
various known acidic activation domains in order to determine if a common
sequence/ structural motif pattern among these activator proteins could be identified if
predictions of secondary structure were taken into account rather than simple sequence
homology (Zhu ML, 1990). Using such a procedure the group identified a pattern
diagnostic of acidic transcriptional activation domains. This pattern (represented
schematically in Figure 1) consisted of nine components: (N to C-terminal) a local
hydrophobic minimum (often matched to a region containing proline), a spacer of -1
to 7 residues, followed by a predicted alpha-helix overlapped with the following
charged template; charge-charge—any residue-any residue-charge—any residue-any
residue-charge—charge, a spacer of —l to 8 residues, a local hydrophobic maximum
containing a valine, leucine or isoleucine, a spacer of 4-8 amino acids and finally a
predicted beta-turn with an aspartic acid residue between 5 and 0 amino acids from
the N-terminal end of the beta-tum. The element most heavily weighted in this
analysis was the predicted alpha-helix having a negatively charged face. Various
acidic activation domains were found to fit this pattern including activation domains

from viral activators (such as the VP16 activation domain), steroid hormone

 

(4.8) B-lum (5.010
[v1.11

’1

Local Hydrophobic profile
maximum wuh

 

('1-8)

 

Schematic

Elements
alpha-helix with
acid faced segment

(~1.7)u

 

 

Local Hydrophobic
profile minimum

Pattern

figure 1. A schematic representation of a pattern purported to be diagnostic of
acidic transcriptional activation domains. A. Ribbon model representation of an
amphipathic alpha helix. B. The nine elements of the diagnostic pattern refined by
ARIEL (Lathrop et al., in press). This figure was taken from Zhu et al., 1990.

 

 

 

l... l .

 

18

receptors, yeast transcriptional activator proteins and some of the acidic activation

domains isolated from E. coli (Ma and Ptashne, 1987b). However, five of the nine

 

strongest activating sequences isolated from E. coli did not fit this pattern (B17, B7,

 

Bl, B32, and B3; Ma and Ptashne, 1987b).

Overview

The observations of Zhu et al. (1990) suggest that various acidic

 

transcriptional activation domains may have some structural similarities including an

AAH. Unfortunately the structural similarities identified by Zhu et al. are based

 

completely upon structural predictions which are known to be unreliable. Thus it
becomes imperative to test this model experimentally. The VP16 acidic activation
domain appears to be the strongest yet identified, thus, it is an ideal candidate for
studying acidic activators in general and is the topic of this dissertation. In the next
chapter of this dissertation I describe work in which I sought to test the various
models regarding the structure of the VP16 activation domain using mutagenesis. In
that work I made the following observations: 1) the VP16 activation domain activates
transcription overall as a function of its net negative charge, 2) the predicted AAH of
the VP16 activation domain did not appear to be very important, 3) Phe 442 of the
VP16 activation domain was critical to its function, and 4) there is some conservation
of bulky hydrophobic residues among the sequences of various activation domains

from different classes. My observations support the suggestion of Zhu et al. (1990)

 

that various activation domains do have structural similarities (similarities which

probably center around a hydrophobic core), but demonstrate that an AAH is probably

 

19

not a necessary component of their structure.

In chapter 111, I describe work which was aimed at directly examining the
structural elements of the VP16 activation domain. In that work, I used ultraviolet
spectroscopy to demonstrate that the critical phenylalanine, Phe 442, of the VP16
activation domain is buried within the activation domain structure. This observation
supports a model that hydrophobic interactions are very important in maintaining the
structure of the VP16 activation domain. I then utilized infrared and circular
dichroism spectroscopies to demonstrate that the VP16 activation domain contains
virtually no helical structure; a result consistent with the observation from chapter 11
that alpha-helix was not a critical element of the activation domain structure.

In chapter IV, I discuss alternative models regarding the structure of the VP16
activation domain and propose future studies. Furthermore, I discuss collaborations
and other investigations in which a number of the VP16 mutants described in this
dissertation have been used as reagents to address the question of how the VP16
activation domain may interact with other transcription factors to increase the rate of
productive initiation. This section ties the relevance of my work to the larger arena

of transcriptional control.

 

CHAPTER II
ANALYSIS OF THE VP16 ACTIVATION DOMAIN
BY SITE-SPECIFIC MUTAGENESIS

Introduction

The eighty amino acid VP16 activation domain is rich in acidic amino acid
residues, as are the activation domains of the yeast proteins GCN4 (Hope and Struhl,
1986), GAL4 (Ma and Ptashne, 1987a), and HAP4 (Forsburg and Guarente, 1989),
and numerous mammalian hormone receptors (reviewed by Evans, 1988). The
preponderance of negative charge among these activation domains suggests that
negative charge is a critical component of activation domain structure. I began my
characterization of the VP16 activation domain by measuring the correlation between
the ability of the VP16 activation domain to activate transcription and its total net
negative charge. For this purpose, I generated a series of activation domain
mutations in which the acidic amino acids aspartate and glutamate of the VP16

activation domain were replaced with uncharged asparagine and glutamine residues.

Construction of VP16 Derivatives
Figure 1 shows the region of the minimal VP16 activation domain targeted for
mutagenesis (VP16 codons 427 to 451), which includes ten acidic amino acid

residues. This minimal activation domain, previously identified by deletion

20

21

Figure 1. Nucleotide and amino acid sequences of the minimal activation domain
of VP16. The amino acids targeted for mutagenesis are highlighted in bold letters;
the acidic residues are numbered 1-10. This 135 basepair E I/ BlamH I DNA
fragment [which originates from pMSVPl6 del456 (Triezenberg et al., 1988b)] was
cloned into the phage vector M13 mp19 (Yanisch-Perron gal, 1985) to generate a
single-stranded template for oligonucleotide—directed mutagenesis. The codons for the
ten acidic amino acids were mutagenized in four clusters using mixed-sequence
oligonucleotides (Table I) to generate all combinations of desired substitutions within
each cluster. Aspartate codons (GAT or GAC) were altered to asparagine codons
(AAT or AAC), and the glutamate codon GAG was altered to the glutamine codon
CAG. The Ng I and m I restriction sites were introduced using translationally
silent nucleotide changes, to facilitate construction of intercluster combinations.
Altered activation domain DNA fragments were reintroduced into the parental VP16
expression plasmid, pMSVPl6 del456, and mutant plasmid identities were confirmed
by sequence determination (Sanger et al., 1977).

 

 

 

22

EEmm
05.3000
sz m 0 m

omv ovv
H00 000 OO._. .20 000 O<O 000 0C. 0.5x O<O OFO ._.<0 OtE. H<O O<O
O a m o O o O ._ E . Q ._ a m D o

o— o m n o m
>_ ”.9920 :_ Gaza
_ 3:2 _ ooz omv

Enadolqno 80 .20 80.94.30 90 98 .98 000 95 EL 05
._ < o < I < 2 < > o m o a 4 z

a m N F

__ 530 final

omv _ _mm 0 So
O._.O O<O O<O 000 9.0 OO< O._.O ._.<O OO< 000 000 000 OO< OOH GPO
._ m o O ._ m > o ._. d n. < ._. m 4

Figure 1

23

mutagenesis, retains about one-half of the activity of the full-length domain both i_n_
m (Triezenberg et_al., 1988b) and mm; (Berger gal, 1990). Use of this
minimal activation domain reduced the target size for mutagenesis, and in principle
permitted the detection of variants with increased activity, since the truncated domain
possessed only half the full-length activity.

The minimal activation domain of Figure 1 was divided into four "clusters"
which were mutagenized using mixed-sequence oligonucleotides (Table 1). These
mixed-sequence oligonucleotides were synthesized by incorporating equal proportions
of wild-type and mutagenic nucleotides at each of several positions selected to give
desired mutations. In addition these oligonucleotides generated an N_c_o I site between
clusters I and II and an _1V_11_u I site between clusters H and III to allow intercluster
recombination to generate activation domain derivatives with as many as ten negative
charges replaced. The details of the constructions of these derivatives are as follows.
The plasmid pMSVP16 del456 was used as a source of VP16 DNA for mutagenesis
and has been described (Triezenberg fig, 1988b). The VP16 minimal activation
domain sequence (from Figure 1) was prepared for mutagenesis by cloning the 135 bp
Sell 1/ M I fragment of pMSVPl6 del456 into the phage vector M13 mp19 to
generate mpDCl.

The system of nomenclature used to refer to mutants emerging from
mutagenesis utilizes the numbers assigned to the ten negative residues highlighted in
Figure 1 and the single letter codes for the amino acids aspartate (D) and asparagine
(N). A mutant changing negative residues 1, 2 and 5 from acidic to neutral is termed

DN 125, while a mutant changing all ten negative residues is termed DN 1-10. For

 

 

24

Table 1. Mutagenic oligonucleotides. The mixed oligonucleotides used to alter the
VP16 gene are represented in this table. Parentheses indicate mixed positions. All
oligonucleotides were purchased from the MSU Department of Biochemistry
Macromolecular Facility and were used without further purification. Mutagenesis
reactions were performed according to standard protocols (Zoller and Smith 1982;
Kunkel, 1985; Bio-Rad Catalog #170-3571). Single-stranded uracil-containing
template DNA was prepared by growing M13 phage in E, coli host CI 236 (gitgl,
u_ngl, rm, relAl; pCJ105(Cm') which incorporates uracils at about 1% of the normal
thymidine residues. Uracil containing phage were purified by equilibrium
sedimentation in CsCl, phage DNA was extracted by treatment with phenol/chloro-
form, and circular template was purified by electrophoresis. Synthesis of the second
strand was performed i_n m using a mutagenic oligonucleotide as primer.

Mutagenic primers were annealed to templates in 20 mM TRIS-HCI (pH 7.4), 2 mM
MgCl, 5.0 mM NaCl by heating to 65 °C and cooling 1 C/minute to 4 °C using a
standard primer/template ratio of 20:1 (Bio—Rad Catalog #170-3571). Synthesis of the
complementary DNA strand was done in 23 mM TRIS-HCl (7 .4), 5 mM MgC12, 35
mM NaCl, 1.5 mM dithiothreitol, 0.4 mM deoxynucleotide triphosphates, 0.75 mM
ATP, 0.2 Units/p1 T4 DNA ligase, and 0.1 Units/pl T4 DNA polymerase. The
synthesis reactions were carried out five minutes at 4 C, five minutes at 25 °C and 90
minutes at 37 °C and were terminated by a 20:1 dilution in water followed by
freezing. Mutagenic synthesis reactions were then used to transform competent dut+
ung” MV1193 cells {(Alac-pro AB), thi, supE, A(srl-rec A)306::Tn10(tet’)[F’:tra
D36,pro.AB lac IqZAM151}. Transformation of this duplex DNA into dut+ ung+ E,
poll results in selection against the uracil-containing wild-type strand due to the action
of uracil glycosylase. Transformant MV1193 plaques were screened for mutations by
dideoxy sequencing (Sanger gal, 1977). Controls in mutagenesis procedures
included uracil-containing template without added primer, uracil-containing template
plus a control primer having no mismatches, and uracil-containing template with both
control and mutagenic primers. Control reactions without added primer indicated that
less than 2% of mutagenic transformants resulted from endogenous priming and from
template molecules not inactivated by incorporation of uracil.

 

25

Table 1

Mutagenic oligonucleotides

 

ST 1a
ST lb
ST 2
ST 3
ST 4
ST 9
ST 10
ST 11
ST 14
ST 15
ST 16

5’-CATGGCCACGT(t/c)CT(g/c)G(t/c)CGT(t/c)TAAGTG-3’
5’-CATGGCCACGT(t/c)CTGG(t/c)CGT(t/c)TAAGTG-3’
5’~GTCTAACGCGT(t/c)G(t/g)CA(t/c)(c/g)CGC-3’
5’-CATGT(t/c)CAGAT(t/c)GAAAT(t/c)GT(t/c)TAACGCG-3’
5’-GGAAT(t/c)C(t/c)CGT(t/c)C(t/c)CCAA-3’
5’—GCCGTCTAAGGGGAGCTCGT-3’
5’-CGCGTCGGGATGCGCCATGGGCACGTC-3’
5’-GTCCAGATCGGGATCGTCTAA-3’
5’-GTCCAGATCG(t/g)(a/c)ATCGTCTAA-3’
5’-CAACATGGCCAGAGCGAAAGCGGCTAACGC-3’
5’-CAACATITCCAGTTCGAATTC'I'TCTAACGC-3

 

 

p" ’3

 

26

simplicity this nomenclature ignores the fact that acidic residue number two is
glutamate not aspartate.

Mutagenesis of cluster I was done using mpDCl as single-stranded template
and the mixed—oligonucleotide primers ST 1a and b (see Table 1). Cluster I
mutagenesis resulted in the isolation of the pseudo-wildtype mutant mpDC2 (which
contains an artificial Nco I site between clusters I and II) and the mutants mpDC DN
1, 2, 3, 12, 13, 23 and 123. Mutagenesis of mpDC2, using ST 2 as mutagenic
primer, generated the mutant mpDC DN 4 and a second pseudo-wildtype mutant,
mpDC3, which possessed artificial Ng I and m I sites, between clusters I and II,
and, II and 111, respectively (see Figure l). The pseudo-wildtype mutant mpDC3 was
used as the template for all subsequent mutagenesis. The cluster III sequence was
altered using ST 3 which generated the mutants mpDC DN 5, 6, 7, 8, 56, 57, 58, 67,
68, 78, 567, 568, 578, 678, and 5678. Mutagenesis of cluster IV was accomplished
using the oligonucleotide ST 4 which generated the mutants mpDC DN 9, 10 and
910. Combinations between HI and IV were made by taking the two cluster IV single
mutants, DN 9 and DN 10, and performing ST 3-mediated mutagenesis upon them.
All fifteen possible combinations from each template were isolated. Thus, I was able
to construct all necessary combinations of the ten targeted acidic residues simply
using the Ngo I site between clusters I and II, the ltflu I site between clusters II and
III and a unique Bgl 11 site within the M13 mp19 vector.

Minimal activation domain derivatives were reintroduced into the pMSVPl6
de145 6 vector which drives the expression of VP16 del45 6 using the MSV-LTR

(which includes a transcriptional enhancer, promoter and mRN A cap site) within the

 

27

vector 5’ to the VP16 gene (Graves gl 1985). Termination codons and
polyadenylation signals were also provided by the vector (Triezenberg et_al, 1988b).
Plasmid pMSVP16 C +119bp (which has two S_a_l_ I sites and two BgmH I sites) was
partially digested with Slal I and single-cut, linearized plasmid was purified by agarose
gel electrophoresis. The E I linearized vector was methylated using Msp I
methylase (New England Biolabs) to protect one BgmH I site of the pMSVPl6
C +119 bp vector while the other EmH I site was digested to completion with M
I. The resulting 6.1 kb pMSVP16 B_arr_1H I/ E I vector was purified by agarose gel
electrophoresis. Mutant 135 bp MH I/S_al_ I activation domain fragments were
excised from the various mpDC clones by digestion with Ba_mH I and fl 1, electro-
phoretically purified and cloned into the 6.1 kb pMSVPl6 BamH I/fl I fragment.
The ability of VP16 derivatives to activate IE gene transcription was measured
by a transient transfection assay (Triezenberg gal, 1988a) in which a plasmid
expressing the wildtype or a mutant VP16 protein was cotransfected with both an
indicator and an internal control plasmid. The indicator plasmid pSJT-703 contained
the IE upstream regulatory sequences of the HSV~1 ICP4 gene, fused to the body of
the HSV-l tk gene. The second plasmid, pMSV-tk, served as an internal control for
transfection efficiency, RNA recovery and primer extension assays. Transcription of
tk-specific RNA from this plasmid is not activated by VP16 (Triezenberg gal,
1988a). The IE—tk and MSV-tk RNAs yielded primer extension products of 81 and 55
bases, respectively, which were separated by electrophoresis, detected by

autoradiography and quantitated by scintillation spectroscopy.

28

Correlation between net negative charge and transcriptional activation.

From the relative activities of a collection of 21 such mutants, summarized in
Figure 2, I infer that negative charge is an important feature of the VP16 activation
domain. Replacement of increasing numbers of acidic residues with uncharged
residues led to a progressive decrease in transcriptional activation. Removal of seven
or more negative charges was required to inactivate the protein beyond the level of
detection of this assay. However, net negative charge is not the sole determinant of
activity, since some derivatives with identical net negative charge had activities which
were clearly distinguishable. For instance, the VP16 derivative DN 2 is less active
than DN 6; DN 1234 is less active than DN 5678; DN 1-4,910 is less active than DN
5-10; and DN 123,5-8 is less active than DN 4-10. In each of these cases, mutations
of negative residues 1, 2 and 3 tend to have a greater effect on activity than mutations
at the other positions. These observations suggest that either charge distribution

and/or secondary structure also contribute to activity.

Correlation Between Predicted Amphipathy and Activation
It has been proposed that AAHs (Gininger and Ptashne, 1987; Ptashne, 1988;

Zhu et al. , 1990) are critical structural elements of acidic activation domains.

 

Furthermore, secondary structure predictions for the VP16 amino acid sequence
suggested that the domain could form an AAH (Cousens et_al, 1989). Schematic
models representing the minimal VP16 activation domain, should it form an
alpha-helix, are shown in Figures 3 and 4. I tested the role of predicted amphipathy

by constructing a series of VP16 derivatives in which I removed blocks of

 

L:

 

29

Figure 2. Relationship between net negative charge and the relative activities of
VP16 derivatives. The relative activities of twenty-one VP16 mutants are plotted as
a function of the number of remaining acidic amino acids. The derivative
designations indicate the amino acid substitutions made (DN means Asp to Asn) and
the amino acid positions affected (numbered 1 through 10; see Figure 1). For
simplicity this nomenclature ignores the fact that acidic residue number two is
glutamate not aspartate. Bars represent two standard deviations about the mean value
obtained from at least six transient expression assays. Method. Transfections were
performed essentiall as described by Triezenberg et al. 1988b. One day prior to
transfection, 5 x 1 mouse L cells (tk', aprt') were plated onto 60 mm culture dishes
in Dulbecco’s modified Eagle’s medium, supplemented with 10% fetal calf serum.
CsCl-purified DNAs were transfected into the cells using the DEAE-dextran—mediated
method (Lopata gal, 1984). Each plate received two pg of the ICP4—gt fusion
plasmid (pSJT-703), two pg of the internal control plasmid (pMSV-tlg) and 50 ng of
activating plasmid (pMSVP16 de145 6 or derivative thereof). Control plates did not
receive inducing plasmid. Forty-two hours after transfection, total RNA was
harvested by the proteinase K/ DNAse I method (Eisenberg gal, 1985). A 32P
labeled synthetic oligonucleotide which hybridizes to 1k mRN A was extended by
reverse transcriptase in presence of deoxynucleotides. Extension products were
separated by electrophoresis, detected by autoradiography and quantitated by
scintillation spectroscopy. The relative abilities of VP16 mutants to activate IE
transcription was calculated from the ratio of indicator signal (pSJT-703) to internal
control signal (MSV-tk), normalized to the parental clone, pMSVPl6 del456.

 

30

 

3:3 9.23 222 a .3532

 

 

 

 

 

 

 

O _ N n .v m o h m m
07:3 2.6.3.3 o-.zo 90.3.5 _ _ _ _ _
94 2o
231.20
a-.. 20
07¢ 20 3&5
.
o3 3:3 3
1.
2.»... 2o 8:3
aoeza
232° «.3
czo
.L J
0320 nzo
r .E
oz

 

 

 

 

o\oON

$o¢

n3.0m

$00

$00.

eSON.

Kumov alumnae

Figure 2

31

 

 

 

 

 

 

 

Figure 3. Schematic models depicting the predicted amphipathic alpha-helix of
the minimal VP16 transcriptional activation domain. A. The upper half of this
flattened-helix model for VP16 amino acids 425 to 451 shows the preponderance of
negatively charged residues [boldface, numbered (by superscript) as in Figure 1.] on
one surface of the putative alpha-helix. The lower half displays a predominantly
hydrophobic surface. B. A ribbon model of the VP16 minimal activation domain if
it is helical. Negative residues are highlighted as circles and are numbered 1-10.

32

 

Figure 4. Schematic helical-wheel model representing the putative amphipathic
alpha-helical structure of the minimal VP16 activation domain. Numbers within
smaller circles identify positions of amino acids within the VP16 polypeptide chain.
Acidic amino acids are represented by filled circles.

33

four negatively charged residues circularly permuted about the putative AAH. I
reasoned that activation domains lacking charged residues in the center of the charged
face of the AAH would likely be strongly affected in transcriptional activation,
whereas mutants lacking charged residues on the periphery of the charged face might
be only slightly affected. Helical wheel models and relative activities of this series of
derivatives are shown in Figure 5 . Although the activities of these mutants display a
four-fold range, no correlation between transcriptional activation and predicted
amphipathy is observed. For instance, the mutants DN 34610 and DN 34510 each
lack negatively charged residues in the center of the charged face of the putative
AAH, yet their activities are higher than the mutants DN 1789 and DN 2789 which

retain negative charges on the putatively exposed surface.

Role of Putative Alpha-helical Structure in Activation

Because I observed no relationship between predicted amphipathy and
transcriptional activation, I questioned whether the function of VP16 depends on the
predicted helicity between amino acids residues 421 and 444. The cyclic side-chain
of proline cannot be accommodated into an alpha-helix; therefore, proline
substitutions in the activation domain should alter local helical structure. Therefore, I
replaced selected amino acids with proline residues to disrupt the predicted
alpha-helical structure in three regions across the activation domain. This series
included two constructs with single substitutions at His 425 (HP 425) or Phe 442 (PP
442), and one construct with substitutions at Ala 432 and Ala 436 (AP 432/6). This

double mutation was constructed because this region possessed the greatest predicted

 

34

Figure 5. The VP16 minimal activation domain does not activate transcription as
a function of its predicted amphipathy. In the schematic helical-wheel models that
represent the predicted structures of ten VP16 derivatives, filled circles represent
wildtype acidic amino acids and numbered circles represent altered sites. The mean
relative activity of each derivative is presented along with standard deviations
determined from at least six transient expression assays. Method. Transient
expression assays were performed as described in the legend of Figure 2. These
derivatives were constructed in the context of the M13 vector by intercluster
recombination among appropriate derivatives utilizing the artificial N_co_ I and MI! I
restriction sites generated between clusters I and II and H and HI and a unique Bgl 11
site within the M13 mpl9 vector. Following recombination activation domain
fragments were reintroduced into the pMSVPl6 background as described in
"Construction of VP16 Derivatives" section.

 

 

35

 

 

o\o w “NV

O_mvN 20

fig

o\om H8
89 20

 

 

 

 

Figure 5

 

 

 

 

36

helix-forming tendencies. Figure 6 shows the effects of these changes on activation
of IE gene expression. Surprisingly, neither the single substitution HP 425 nor the
double substitution AP 432/ 6 had a detectable effect on activation by VP16,
suggesting that the predicted helical structure is not critical in activation. The
substitution of proline at Phe 442 however, essentially abolished transcriptional

activation.

A Critical Role for Phe 442 in Activation

To determine whether the effect of the proline substitution of Phe 442 was due
to local disruption of helical structure or to the elimination of a critical side-group, I
replaced Phe 442 with three helix-compatible amino acids, serine (F S 442), alanine
(PA 442) and tyrosine (FY 442). The activation of IE gene expression by these three
derivatives is shown in Figure 6. Only tyrosine was able to functionally replace Phe
442, producing an activation domain with about one-third (35 j; 15%) the wildtype
activity. In contrast, substitution of Phe 442 with the non-aromatic residues alanine
and serine potently inactivated VP16. These results demonstrate that the side-group
of amino acid 442 has a critical role in VP16 transcriptional activation.

To be certain that the effects of the Phe 442 substitutions were due to
structural changes specific to the activation domain, it was important to demonstrate
that these substitutions affected neither the stability of the derivative polypeptides nor
the function of the VP16 IE specificity domain. Western blot analysis (Figure 7)
indicated that the loss of activity of the PP 442 derivative could not be accounted for

by diminished accumulation of the derivative polypeptides in transfected cells.

37

'E'“‘ In --

Control --~

HP 425
AP 432/6
FP 442
FS 442
FA 442
FY 442
no VP16

wt

Figure 6. VP16 Phe 442 is critical in transcriptional activation. Autoradiogram of
a transient expression assay of several VP16 derivatives designed to test the role of
predicted helicity in VP16 activation. The positions of primer extension products
corresponding to transcripts from the reporter (IE-g) and control plasmids are
indicated. Method. This transient expression assay was performed as described in
the legend of Figure 2. The mutants HP 425, AP 432/6 and FF 442 were generated
using the oligonucleotides ST 9, ST 10 and ST 11 of Table 1, respectively. The
mutants FS 442, FA 442 and FY 442 were generated using the mixed oligonucleotide
ST 14 (Table l).

 

38

97kd—~
68kd—\
45kd—\ ‘59 =rwww
29kd—\
to
a.
a 8
(ommNNNN
.ia‘WHH;
0.0.
EggI<EuUEEEE

Figure 7. VP16 derivatives having substitutions at Phe 442 are expressed at
levels indistinguishable from the parental protein. Polyclonal antisera raised
against gel-purified VP16 (S. Triezenberg, unpublished) was used in a Western blot
analysis to probe extracts from L cells transfected with VP16 expressing plasmids.
Lanes corresponding to each VP16 derivative are indicated as are the positions and
sizes (in kilodaltons, kd) of molecular weight markers. Method. For each VP16
derivative a 60 mm plate of L cells, transfected with 10 pg of plasmid, was lysed in
two ml of 10 mM TRIS-HCl (pH 8), 5 mM EDTA and 1% SDS. The lysate was
sonicated and precipitated by addition of 8 ml of cold acetone. After centrifugation
(3000 X g, 5 min) protein pellets were resuspended into 0.2 ml SDS-PAGE sample
buffer. One-tenth of each sample (20 pl) was loaded onto a 10% polyacrylamide
minigel (Laemmli, 1970). Following electrophoresis, proteins were transferred to
nitrocellulose (Towbin et al. 1970) and probed using rabbit antiserum raised against
gel-purified VP16 (S. Triezenberg) and an avidinzbiotinzperoxidase conjugate
detection system (Vector Labs).

 

39

Furthermore, the phenylalanine to proline substitution did not disrupt the IE
specificity function of the VP16 polypeptide. This was demonstrated using an assay
which tests the ability of VP16 derivatives to interfere with activation by the wildtype
protein (Triezenberg gall, 1988b). It has been shown that a truncated VP16
polypeptide containing only the 413 N-terminal amino acids of VP16 (pMSVPl6 delta
413—490) can interfere with transcriptional activation by the wildtype protein. This
interference is presumably due to interaction with host DNA binding proteins. To
determine if the proline substitution at Phe 442 unexpectedly disrupted IE specificity
interactions, FP 442 was compared with VP16 derivatives, pMSVP16 delta 413—490
and pMSVP16 delta 393-490, in an interference assay shown in Figure 8. VP16 delta
393-490, like delta 413—490, is a C-terminal VP16 deletion which is unable to

interfere with activation by wildtype VP16 (Triezenberg et al., 1988b) presumably

 

due to the disruption of a domain which interacts with Oct—l (Greaves and O’Hare,
1990). The first four lanes of this autoradiogram (corresponding to uninfected cells)
reveal that PP 442, delta 413-490 and delta 393-490 do not activate IE transcription.
The fifth lane shows activation of IE transcription by the wildtype VP16 supplied by
the infecting virus. The sixth and eight lanes reveal that FP 442 and delta 413—490
greatly inhibit activation of IE expression by wildtype VP16, whereas delta 393-490
shows reduced inhibition (seventh lane). This observation demonstrates that the PP
442 protein is present in the cell nucleus with an intact capacity to interact with host
DNA-binding proteins that mediate interaction with IE cis-regulatory sequences.

The side-group of phenylalanine is not expected to be chemically reactive, so

its role in VP16 activation is likely to be structural. I suggest that Phe 442 (and

4O

lE-tk " . p"-

 

(O (D

‘— 0‘) co ‘- N 0‘) CO
0- g or .- 11 st or 1-
> 1 co <1- > V co <2-
0 o. a) E o n. a: m
t: u. '5 '5 C U. '0 '0

not
infected infected

Figure 8. Substitution of Phe 442 does not affect the binding of VP16 to host
DNA-binding proteins. In this activation interference assay mouse L cells were
transfected with 3 pg of plasmid expressing VP16 derivatives as indicated below each
lane along with the indicator and internal control plasmids described in the legend of
Figure 2. Two days later cells were infected (the first four control lanes were not
infected) with HSV-l at a multiplicity of infection of 10 plaque forming units per cell
in the presence of 100 pg/ ml cycloheximide. 'vao hours postinfection RNAs were
isolated and analyzed by primer extension.

41

probably other hydrophobic residues) is involved in a hydr0phobic interaction either
in self-folding of the activation domain or in contact with the molecular target of
VP16. The role of hydrophobic interactions in protein self-folding (Lesk and
Chothia, 1980) and in protein: protein interactions such as antigen-antibody complexes
(Amit et_al_._, 1986; Standfield gg, 1990) is well established. Disruption of that
hydrophobic interaction, either by addition of a polar hydroxyl group onto the
aromatic ring (as in the case of FY 442) or by removal of the bulky hydrophobic
side-group (as for PA 442), resulted in a dramatic reduction of transcriptional

activation by VP16.

 

The environment around the critical residue Phe 442 may also be important for
activation. I noted that the VP16 mutant DN 567 8, in which the four aspartic acid
residues closest to Phe 442 (Figure 1) have been replaced by asparagine, is relatively
active (62 i 11%). If secondary structure is important in that local region, then the
substitution of asparagine for aspartic acid must conserve that structure. To
determine whether making more radical changes in the vicinity of Phe 442 would
have stronger effects on activity, I changed the four proximal aspartic acid residues to
glutamate (DE 567 8) or alanine (DA 567 8). Glutamate substitution was chosen as a
conservative change which is helix favorable, conserves negative charge, and has only
a slightly different side-chain structure than aspartate. Alanine substitution was
chosen to be compatible with alpha-helical structure, but conserving neither charge,
polarity nor side-chain structure.

The autoradiogram shown in Figure 9 reveals a hierarchy of relative activities;

mutant DN 5678 (62 j; 11%) is stronger than DE 5678 (40 i 10%), which is

42

IE-tk .a.

Control “m

no VP16
DE 5678
DN 5678
DA 5678

wt

Figure 9. Amino acids proximal to Phe 442 are critical in transcriptional
activation. Autoradiogram of a transient expression assay of three VP16 derivatives
designed to test the importance of the four acidic residues in the local environment of
Phe 442. Method. This transient expression assay was performed as described in
the legend of Figure 2. The VP16 derivatives DE 5678, and DA 5678 were
generated using the oligonucleotides ST 15 and 16 of Table 1, respectively.

43

stronger than DA 5678 (15 j; 5%). Thus, asparagine is the best replacement for
aspartate in the structure of the activation domain despite the fact that it does not have
negative charge. Glutamate is a slightly poorer replacement than asparagine, even
though it conserves negative charge and is very helix-favorable (Chou and Fasman,
1978). Finally, alanine, although compatible with helical structure, is a very poor
substitute for aspartate. I conclude that polar or charged amino acids flanking Phe
442 are also important to the VP16 activation domain structure.
Conservation of Bulky Hydrophobic Residues Among Various
Activation Domains

Two models have been previously put forth to describe the structure of acidic
transcriptional activation domains. The first model proposes that such domains exist
as poorly structured polypeptides whose strengths are simply related to the net charge
(Sigler, 1988). This model is inconsistent with the results reported here, which
indicate that the acidic amino acids are important but not sufficient for activation by
VP16. The second model suggests that acidic activation domains form AAHs
(Gininger and Ptashne, 1987; Ptashne, 1988; Zhu girl, 1990). This model, too, is
inconsistent with my results. Disruption of predicted amphipathy or helicity did not
generally correlate with activity. Furthermore, the failure of mutant FA 442 to
activate transcription argues strongly that the predicted acidic AAH is not sufficient
for the function of VP16.

I have reexamined the primary sequences of VP16 and other transcriptional

activators in an effort to identify some common features consistent with the results

44

presented in this report. In Figure 10, the amino acid sequences of several activation
domains from three recognized classes of transcriptional activators (Mitchell and
Tjian, 1989) are aligned using as a guide the six bulky hydrophobic residues of the
minimal activation region of VP16. Remarkably, I observe great similarity between
the acidic activation domain of VP16 and the glutamine-rich transcriptional activation
domains of Spl (Courey and Tjian, 1988). For instance, all six bulky hydrophobic
residues of VP16 can be aligned with bulky hydrophobic residues in the activation
domains A and B of Spl. In addition, eight of the ten VP16 acidic amino acids can

be directly aligned with glutamine residues in the Spl (A) domain. I have shown that

 

uncharged, carbonyl-containing residues can functionally replace aspartate residues of
the VP16 activation domain, with some decrease in transcriptional activation. VP16
is a much stronger activator than Spl; for instance, the IE upstream regulatory region
has five Spl binding sites (Jones and Tjian, 1985), yet I barely detect IE-tk transcripts
in absence of VP16. It is possible that the Spl activation domains are uncharged (and
thus less active) analogs of the VP16 activation domain which nonetheless utilize
similar hydrophobic contacts. The hydrophobic matches between VP16 and the other
acidic or proline-rich activators shown in Figure 10 are not as striking as for Spl.
However, shorter regions of similarity are observed. These short regions of
similarity are characterized by bulky hydrophobic residues flanked by
carbonyl-containing residues.

When I began this project (1988) a critical role for hydrophobic contacts was
not anticipated. The analysis of Zhu gg (1990), however, included a region of

local hydrophobic maximum containing leucine, valine or isoleucine as one of nine

 

45

Figure 10. Conservation of bulky hydrophobic residues among activation
domains. Alignments are based upon visual inspection aided by sequence comparison
programs (Devereux M... 1984), but do not necessarily represent the most
parsimonious alignment. Spl (Courey and Tjian, 1988) and CTF (Mermod et_al.,
1989) are mammalian transcriptional activators of the glutamine-rich and proline-rich
classes, respectively. GAL4 (Ma and Ptashne, 1987a), GCN4 (Hope and Struhl,
1886) and HAP4 (Forsburg and Guarente, 1989) are yeast acidic transcriptional
activator proteins. B17 (Ma and Ptashne, 1987b) is encoded by a randomly-cloned ;
99h DNA fragment, which when fused to the GAL4 DNA binding domain activates
transcription in yeast. AH (Gininger and Ptashne, 1988) is an artificial sequence
designed to form an AAH.

46

QOZWZZ

(3132401211112

oo>,amom<

(DQaOiDUJQOaE-tHOi

 

 

 

QHQDJZOQH>O

 

 

Iqquqzqzqql

 

a0.

m>

Zm

D0100

[uHthNi—IAMBi—Il

 

 

 

DZBOMMDZE—tm

QOaODZEuEIJMDOiOl

 

 

AEZI—‘lEi—Ii—JIEEAQ

 

 

<>mmm9m>om
QO>QE4ILDZUIOI
«mommmommq

mammogmmmm

40228

NZHr—I
(1).-1&0

mmm

mO
BA
2d

|:>:>:>.—itutu|:nom

 

 

DOE-immbdmmz
momHmquo
OZE—tE—ttxJDNKCm
oomm><moo

ma
sflm
Amvvgaw
mac
Aavvmmm
«zoo
Aavvqao
Amvamm
Lavadm
mam>

AQHZ>AHU§<
IEZHQKCBKch-I

Figure 10

47

diagnostic components of acidic activation domains. Although Phe 442 does not lie
within the region of hydrophobic maximum identified by Zhu et_al (1990) it is
possible that the hydrophobic maximum identified by Zhu ml, (1990) may be a
critical structural element of acidic activation domains. In all cases of activators
aligned in Figure 10 the region of hydrophobic maximum identified by Zhu et_al,
(1990) lies just C-terminal to the regions of activation domains similar to the region
Phe 442 (for instance Leu 447 of VP16 was highlighted by Zhu e_t_al, 1990).

The comparisons in Figure 10 predict that other bulky hydrophobic amino

acids, such as leucine or tryptophan, may function at position 442 of VP16, and that

 

other bulky hydrophobic residues may be important in transcriptional activation. Jeff
Regier has characterized a number of VP16 mutants which substitute Phe 442 and
flanking hydrophobic residues. Jeff has found that tryptophan, like tyrosine, can
replace Phe 442 with retention of about one-third activity. On the other hand, leucine
was dysfunctional at position 442. This observation strongly suggests that the residue
at position 442 must be aromatic. Jeff has also altered two bulky hydrophobic
residues flanking Phe 442, Leu 439 and Leu 444. Jeff has found that substitution of
these leucine residues with other bulky hydrophobic residues such as valine and
phenylalanine preserve activity, whereas substitutions of these residues with alanine or
serine severely affect activation. Jeff’ s observations clearly support a critical role for

hydrophobic residues in the activation by VP16.

48

Conclusions

The results of studies described in this chapter suggest the following
conclusions: 1) the VP16 activation domain activates transcription as a function of its
net negative charge, but negative charge is not sufficient for activation, 2) the
predicted AAH of the VP16 activation domain does not appear to be a critical
structural component of the VP16 activation domain, 3) Phe 442 of the VP16
activation domain is critical to its function, and 4) there is some conservation of
bulky hydrophobic residues among the sequences of various activation domains from

different classes.

 

These results suggest that both ionic and hydrophobic interactions are critical
to the function of the VP16 activation domain. It seems reasonable to assume that
this domain binds to a target protein which couples VP16 to some component of the
basal transcriptional machinery. I suggest two possible models which may describe
the structure of the VP16 activation domain and which may account for the
importance of Phe 442 within that structure. In the first model I propose that Phe
442 is involved in hydrophobic contacts within the folded structure of the activation
domain itself. Phe 442 and other hydrophobic residues interact hydrophobically to
determine the internal structure of the domain, whereas hydrophilic and acidic
residues seek the surface of the domain. Charged residues on the surface can then
interact with the target of the VP16 activation domain by electrostatic interactions. If
appropriate hydrophobic contacts are disrupted (as when Phe 442 is replaced) the
internal structure of the activation domain is altered, which in turn alters the charged

surface of the folded molecule which can no longer interact appropriately with a

49
charged surface of the presumed target. In the second model, I propose that Phe 442

may instead be involved in a hydrophobic contact with a target protein. For
instance, charged residues could bring the VP16 activation domain into close
proximity to its target (presumed to be positively charged) by long-range electrostatic
attraction. Once the target is in close proximity, hydrophobic contact could prevail
leading to intimate binding. Phe 442 for instance, might fit into an hydrophobic cleft
within the structure of the target molecule. Such an interaction is not unprecedented.
For example, the anticoagulant hirudin which binds to the exosite of thrombin

possesses a critical phenylalanine residue which inserts itself into a hydrophobic cleft

 

of the thrombin molecule upon binding (Rydel et al., 1990). In fact, the amino acid
sequence of the hirudin polypeptide flanking its critical phenylalanine residue is

similar to the amino acid sequences flanking Phe 442 of VP16.

CHAPTER III

ANALYSIS OF THE VP16 ACTIVATION DOMAIN
BY SPECTROSCOPIC AND BIOPHYSICAL NIETHODS

Introduction
In the preceding chapter I described a mutational analysis aimed at

understanding the critical structural elements of the VP16 activation domain. That

 

 

analysis suggested that net charge and some structural component were critical to the
function of the VP16 activation domain. Data emerging from mutagenesis further
demonstrated that position 442 was important to the function (and presumably
structure) of the activation domain. Unfortunately, mutagenesis cannot be used to
define protein structure m. Thus, I initiated a biophysical characterization aimed
at directly defining the structural elements of the VP16 activation domain. This
characterization has utilized five methods of structural analysis: ultraviolet (UV)
spectroscopy, Fourier-transform infrared (FTIR) spectroscopy, circular dichroism
(CD) spectroscopy, native gel electrophoresis, and differential scanning calorimetry

(DSC).

GAL4-VP16 Fusion Proteins
Many methods of biophysical analysis require milligram quantities of the

sample polypeptide to be characterized. Unfortunately, it would be very difficult to

50

51
adequately purify the amounts of virion-borne VP16 necessary for such analysis (D.

Lewis, S. Lui, R. Pichyangkura, J. Regier and A. Cress, unpublished experiments).
Biophysical analysis of the VP16 activation domain in the context of the virion-born
protein is further complicated in that the activation domain represents less than
one-fifth of the entire VP16 polypeptide (the secondary structure of which is totally
undefined) making it very difficult to define structural elements specific to the
activation domain.

To circumvent these problems I have utilized GAL4-VP16 fusion proteins for

structural analysis. GAL4-VP16 fusion proteins are composed of the C-terminal

 

activation domain of VP16 (or derivatives thereof) fused to the yeast GAL4
DNA-binding domain (Sadowski gg, 1988). There are a number of advantages to
using these proteins for structural analysis of the VP16 activation domain: 1)
milligram quantities of these proteins can be prepared from Egli at greater than
90% purity, 2) the GAL4 DNA-binding domain has been extensively studied and its
structural elements are well defined, 3) VP16 activation domain peptides containing
only a few non-VP16 residues at their N -termini could be separated from fusion
proteins by proteolysis with trypsin, 4) the DNA-binding activity of the GAL4
component of the fusion protein offered a simple assay to monitor for the presence of
the fusion protein, and 5) GAL4-VP16 fusion proteins can activate transcription in i_n
1mg assays (Chasman mil, 1988; Berger gal, 1990), thus offering a way to

determine the affect of various treatments upon activity (there is no such in vitro

 

assay for transcriptional activation by full-length VP16).

52
Composition of GAL4~VP16 Ersion Proteins

The first bar of Figure 1 represents the GAL4 DNA-binding domain (GAL4
1-147) which is composed of the 147 N-terminal amino acids of the yeast GAL4
protein. CD analysis suggests that the GAL4 1—147 protein is approximately 40%
helical, 40% coil and less than 20% beta-sheet (Pan and Coleman, 1989). GAL4
1-147 has three important subregions, a Zn-binding region (Pan and Coleman, 1989),
a DNA-specificity region (Corton and Johnston, 1989) and a dimerization region

(Carey et al. , 1990a). The Zn-binding region forms a Zn(II)2Cys6 binuclear cluster

 

as determined using 113Cd NMR (Pan and Coleman, 1990) and the complete

 

secondary structure of the Zn-binding domain (residues 7-49) has been determined
using 1H NMR (Gadhavi e_tg, 1990). The entire three-dimensional structure of
GAL4 1-147 should be forthcoming in the near future.

The second bar in Figure 1 represents GAL4-VP16 in which the 78 C-terminal
amino acids of VP16 (the full-length activation domain) are fused to the GAL4 1-147
polypeptide (Sadowski e_t_g, 1988). This fusion protein is a powerful transcriptional

activator from promoters containing GAL4 binding sites in mammalian cells in vivo

 

(Sadowski g, 1988) and in yeast and mammalian cell extracts in_vm (Chasman et
al., 1989; Carey e_t_al, 1990a) demonstrating that activation domain structural
elements are preserved in these fusion proteins. Recent mutational studies by Jeff
Regier (unpublished data) indicate that the full—length activation domain appears to
encode two activation regions, termed regions I and II. The third bar in Figure 1
shows GAL4-del456 in which only activation region I is fused to GAL4 1-147. In the

previous chapter, region I was referred to as the minimal (or truncated) activation

 

53

Figure 1. GAL4 1-147 and GAL4-VP16 fusion proteins. GAL4 1-147 is the
DNA-binding domain of GAL4 (stippled). GAL4-VP16 is GAL4 1-147 fused
in—frame to the full—length VP16 activation domain (V P16 amino acids 413-490).
GAL4—del456 is GAL4 1-147 fused in-frame to the region I activation domain (VP16
amino acids 411-456). The various mutants described in this work are derivatives of
GAL4—del456, as indicated above. The GAL4-VP16 fusion protein contains a seven
amino acid linker (PEFPGIW) between GAL4 Ser 147 and VP16 Ala 413. The
GAL4-del456 fusion protein contains a two amino acid linker (GS) between GAL4
Ser 147 and VP16 Ser 411.

54

 

mmnm 20-3420

9va $.35 mm mm
«3. E35 Ti
«3. $-35
N3. E420
gag

/

 

 

 

 

 

 

o _ 28mm 2 838-420 1
EVE: mm N W.
:8 8&3 8a. 8... «5. E

 

 

.............................................

 

o ___8_am Eosmm z oE>-3<o

 

 

 

EVE: mm A

as 429 5; mm mm P

 

o z 5;; 35

 

 

 

 

 

/

56?. 50mm coamm
828855 2368329 86:53

55
domain. GAL4-del45 6 activates transcription about half as well as GAL4-VP16 i_n_

 

mg (Triezenberg e1; g, 1988b) and in vitro (Berger et _al, 1990). All of the mutants

described in the previous chapter are derivatives of this truncated domain, as shown

(Figure 1).

Purification of GAL4-VP16 Fusion Proteins
The proteins represented in Figure 1 were expressed in E. coli XA90 cells
under control of the tac promoter by plasmids pTGH21 (GAL4 1-147), pJL2

(GAL4-VP16) and pAC-del456 (and mutant derivatives). Constructions of pTGH21

 

(Lin gal, 1988) and pJL2 (Chasman gt_al_., 1989) have been described. The
pAC-del45 6 series of plasmids was derived from pJL2 (A. Cress and S. Triezenberg,
unpublished data). An inconvenient Sal I site in the pGEM polylinker of the pJL2
vector was eliminated by digestion with Sal I, fill-in of the Sal I site with T4 DNA
polymerase, followed by ligation to yield the intermediate pAC 1627 . A Sal I linker
was then inserted into the four QR I sites present in the pAC 1627 as follows: pAC
1627 was partially digested with EQQR I, single-cut plasmid was purified by
electrophoresis, Ec_oR I ends were made blunt by treatment with mung bean nuclease,
Slal I linkers were attached by ligation and trimmed with M 1, linear DNA was
purified by electrophoresis and religated to form four possible constructs each
replacing one of four EQQR I sites with Sal I. The desired construct having replaced
the EQQR I site of pJL2 within the GAL4-VP16 encoding region was identified by
restriction mapping and the vector termed pAC 1779. The pAC 1779 vector allowed

easy subcloning of activation domain DNA sequences directly from pMSVP16 vectors

5 6
as S_a,l I-Hind III fragments and removed an unnecessary linker encoding the amino

acids (PGFPGIW) from between the GAL DNA-binding domain and the VP16
activation domain.

In general, one liter cultures of XA90 cells were used for preparation of fusion
proteins as has been previously described (for GAL4 1-147 see Lin et_al, 1988; for
GAL4-VP16 see Chasman fig, 1989; for GAL4—del456 and derivatives see Berger
e_t_g, 1990). Yields of protein from one liter of late log-phase XA90 cells were
typically about 20 mg. Ten liter cultures were also employed requiring significant

modifications of original protocols. Ten liter cultures were grown in a 12 liter New

 

Brunswick Microferm Fermentor (courtesy of Dr. G. Zeikus). LB medium (including
0.2 % antifoam agent) was autoclaved in the fermentation chamber the day prior to
inoculation and allowed to cool to 37 °C overnight. The following morning
ampicillin was added to 50 mg/l and the fermentation system fully assembled.
Fermentation parameters were: temperature, 37 °C; incoming air pressure, 6 psi;
drive rate, 200 rev/ min. The ten liters were inoculated with 200 m1 stationary phase
cells and growth was monitored by measuring A600. When the A600 reached 0.7
(about 2-3 hours) expression of fusion proteins was induced by the addition of
isopropylthiogalactoside to 1 mM and cells were grown an additional 3 hours. Cells
were harvested by centrifugation, washed in 800 ml 20 mM HEPES, pH 7.5, 0.2 M
NaCl and resuspended and lysed by sonication in 400 ml ice cold buffer A (20 mM
HEPES [pH 7.5], 20 mM 2-mercaptoethanol, 10 pM zinc acetate, 2 pg/ml each of
leupeptin, pepstatin, and tosylsulfonyl phenylalanyl chloromethyl ketone, 20 pg/ ml of

benzamidine and 200 pg/ ml of phenylmethylsulfonyl fluoride) containing 0.2 M NaCl.

57
The lysate was then cleared by a 20 min centrifugation at 10,000 X g. GAL4 fusion

proteins were selectively precipitated from the crude lysate by addition of
polyethyleneimine to 0.25 % (wt/vol) and the pellets washed with 200 ml buffer A
containing 0.2 M or 0.4 M NaCl for GAL4-del456 and GAIA-VP16, respectively.
Proteins were then resuspended in 200 ml buffer A containing 0.75 M NaCl, and
precipitated by addition of solid ammonium sulfate to 40% saturation for

GAL4-del45 6 or 35 % for GAL4-VP16. Ammonium sulfate pellets were resuspended
in buffer A to a conductivity equal to buffer A containing 0.1 M N aCl (approximately

50 m1) and were loaded on a 50 ml (bed volume) Whatman DE-52 column

 

preequilibrated in buffer A containing 0.1 M NaCl. The columns were eluted with a
50 ml gradient of NaCl from 0.1 to 0.4 M. Peak fractions (A280) were diluted with
buffer A to a conductivity equal to buffer A containing 0.1 M NaCl and loaded on a
100 ml (bed volume) Whatman P-ll column preequilibrated in buffer A containing
0.1 M NaCl. The column was then eluted with a 100 ml gradient of 0.1 to 1.0 M
NaCl, peak fractions (A280) were pooled and adjusted to conductivity equal to buffer
A containing 0.2 M NaCl (sample volume at this point was as much as 100 m1) and
loaded on a Pharmacia heparin-Sepharose CL—6B column preequilibrated in buffer A
containing 0.2 M NaCl. Protein was step eluted from the heparin-Sepharose column
using 0.6 M NaCl. The protein peak from this column represented 90-95 % pure
fusion protein. Yield was typically about 200 mg (from 10 liters) in a final volume of

10-15 ml buffer A containing 0.6 M NaCl.

58
Purification of VP16 Activation Domain Peptides

Peptides corresponding to the full-length and truncated activation domains
could be separated from the fusion proteins GAL4-VP16 and GAL4-del456 by tryptic
cleavage, since the activation domain contains no lysine or arginine residues, whereas
there are 25 such trypsin sites in GAL4 1-147 including a most C-terminal site at
position 142. Complete tryptic digestion of GAL4-VP16 yielded 25 short GAL4
peptides and a 90 amino acid peptide containing the 78 amino acid VP16 activation
domain and 12 non-VP16 amino acids (QLTVSPEFPGIW) at its N-terminus.
Complete tryptic cleavage of GAL4-del45 6 yielded the numerous GAL4 peptides and
a 54 amino acid peptide containing the 46 amino acid minimal VP16 activation
domain and eight non-VP16 amino acids (QLTVSPGS) at its N-terminus.

Optimized trypsinization conditions were: 1 mg/ ml fusion protein, 2 pg/ ml
I-IPLC-purified trypsin (courtesy of Joe Leykam), 20 mM Tris-HCl, pH 8.5, 1 hr at
37 °C. Trypsinization reactions were terminated by addition of phenylmethylsulfonyl
flouride (0.3 mg/ ml) and the reactions were filtered through a 0.2 pm filter (Gelman)
prior to chromatography. For the full-length activation domain the trypsinization
reactions were loaded on a Whatman DE-52 column and the column washed with 20
mM HEPES at pH 7.5 containing 0.1 M NaCl. The tryptic peptides originating from
GAIA eluted from this column in the wash while the negatively charged full-length
activation domain was retained. The activation domain peptide was eluted with a
linear gradient from 0.1 to 0.5 M NaCl, in 20 mM HEPES, pH 7.5 and the elution
monitored (A230). For the truncated activation domain it was necessary to use FPLC

to resolve the activation domain from contaminating GAL4 peptides. GAL4-del456

59

trypsinization reactions were loaded on a Pharmacia FPLC Mono-Q column (courtesy
of Drs. J. Hosler and S. Ferguson-Miller) and the column washed with 20 mM
HEPES at pH 7.5 . Peptides were then eluted from the column with a linear gradient
from 0 to 0.5 M NaCl in 20 mM HEPES, pH 7 .5 and the elution monitored by
reading A230 of collected fractions. Four absorbing peaks were resolved from the
Mono-Q column. The first two peaks contained unresolved tryptic peptides from the
GAL4 specificity domain and the fourth apparently contained trypsin as judged by
mobility on SDS-PAGE. The third peak contained a single peptide corresponding to
the minimal activation domain. Identity and purity of the full-length and minimal
activation domain peptides were confirmed by amino acid sequence analysis (Joe
Leykam). Typical yields of the activation domain peptides from 10 mg of starting
protein was 1.3 mg for the full-length domain and 0.7 mg for the minimal domain as
measured using A230 (assuming 1 mg/ml protein has an A230 of 2.5). Dilute peptide

preparations were concentrated using a Centricon-3 microconcentrator.

Ultraviolet Spectroscopy

In the previous chapter, I determined that Phe 442 of the minimal VP16
activation domain was critical to the function of VP16 and suggested two possible
models as to how Phe 442 might be important. In the first model I proposed that Phe
442 is involved in hydrophobic contacts within the folded structure of the activation
domain itself. In the second model I proposed that Phe 442 may instead be involved
in a hydrophobic contact with a target protein. To distinguish between these two

possible models I sought to determine if the amino acid residue at position 442 was

 

60
exposed to solvent (which would support the model favoring Phe 442 contacting

another protein) or if position 442 was buried within the VP16 structure (which would
support a role for Phe 442 in self-folding of the domain).

Ultraviolet spectroscopy combined with solvent perturbation is a useful method
for measuring the local environment of the polar aromatic residues tyrosine and
tryptophan in proteins. Fortunately, tyrosine is functional at position 442 of VP16 i_n

vivo in the context of pMSVP16 del456 (previous chapter) and in vitro in the context

 

of a GAL4-del456 derivative (GAL4-FY 442) (S. Berger and A. Cress, unpublished

results). Thus, determining the environment of tyrosine 442 within the GAL4-FY

 

442 protein should reflect the environment of the wildtype phenylalanine at that
position, taking into account, of course, that tyrosine is only one-third as active as
phenylalanine at that position.

The GAL4 DNA-binding domain has two polar aromatic residues which could
potentially interfere with UV/ solvent perturbation analysis of Tyr 442, a tryptophan at
position 36 and a tyrosine at position 40. Fortunately, both of these residues lie
within the Zn-binding domain of GAL4 1-147, the structure of which has been
determined by 1H NMR (Gadhavi et_al, 1990). Both Trp 36 and Tyr 40 are tightly
buried within that structure, thus they should not be strongly affected by the solvent
environment (unless that environment unfolds the GAL4 structure).

At neutral pH tyrosine has a local absorption maximum at 274 nm and little
absorption at 295 nm. However, upon dissociation of the phenolic proton the local
absorption maximum shifts to 295 nm with an approximately two-fold increase in

molar absorptivity. Figure 2 shows the UV spectra of free tyrosine at four different

61

Figure 2. UV spectra of free tyrosine at different pHs. The UV spectra of 0.16
mM tyrosine (Sigma) was taken at pHs 9, 10 and 11 in 0.2 M NaCl and 50 mM each
of TRIS, CAPS and CAPSO buffers at the indicated pHs. The spectrum of tyrosine
at pH 13 was taken in 0.2 M NaCl and 0.1 M NaOH. Spectra were obtained using a
Beckrnan DU 68 and a 50 pl quartz cell (pathlength 1 cm).

62

 

016'“

 

OJZ"

eoueqrosqv

008 '

 

 

 

 

 

I
266 278 290 302 314

Wavelength (nm)

memZ

 

63
pHs. The pKa of free tyrosine is 10.1, thus at pH 9 protonated tyrosine demonstrates

an absorption maximum at 274 nM. At pH 10, when free tyrosine would be expected
to be almost 50% dissociated a dramatic shift toward an absorption maximum of 295
nm is observed. At pHs 11 and 13, free tyrosine should be fully dissociated, thus the
absorption maximum of 295 nM is observed.

In order to obtain UV spectra specific to Tyr 442 of GAL4-FY 442,
GAL4—FA 442 was used as a spectral background. GAL4-FA 442 is identical in
composition to GAL4-FY 442 with the exception of having an alanine (which doesn’t
absorb in the UV) rather than a tyrosine at position 442. GAL4-FA 442 was used
instead of wildtype GAL4-del456 to eliminate the contribution of wildtype Phe 442 to
the UV background. Figure 3 shows the UV spectra of Tyr 442 obtained in this way
at four different pHs. Unlike free tyrosine, Tyr 442 shows virtually no shift toward
an absorption maximum at 295 nM at pH 10, and shows only a slight shift at pH 11,
whereas free tyrosine is completely dissociated (see Figure 2). At pH 13 Tyr 442
shows a complete shift to the dissociated form. The observation that Tyr 442 does
not dissociate like free tyrosine suggests that it is buried within the protein’s structure
and is not accessible to solvent until the protein denatures at extreme pH.

A second UV method for determining whether a tyrosine or tryptophan is
internal or external relies on the effect solvent polarity has upon the spectra of these
residues. In solvents less polar than water, such as aqueous 20% ethylene glycol, the
absorption maxima and molar absorptivity of tyrosine and tryptophan both increase
slightly relative to their spectra in water. This difference is generally presented as a

difference spectra in which the spectrum of the sample in water is subtracted from the

 

 

Figure 3. UV spectra of Tyr 442 at different pHs. The UV spectrum of Tyr 442
of GAL4-FY 442 was obtained by subtracting the spectrum of 0.16 mM GAL4-FA
442 from the spectrum of 0.16 mM GAL4-FY 442. Method. Proteins stored at -70
°C in buffer A containing 0.6 M NaCl were thawed and aliquots were dialsz (1 hr
in a microdialysis chamber) against 1 mM TRIS (pH 8) and 0.2 M NaCL at 10 °C to
remove salts and protease inhibitors. Following dialysis proteins were diluted into 1
mM TRIS (pH 8) and 0.2 M NaCl to the appropriate concentration and one-tenth
volume of 10X triple buffer at pH 9, 10 or 11 was added (1X triple buffer was 50
mM each of TRIS, CAPS and CAPSO; this buffer mixture provided high buffering
capacity in the range of pH used). The spectrum obtained at pH 13 was obtained by
adding NaOH to 0.1 M. Protein concentrations were determined using a
commercially available protein assay from Bio-Rad (catalogue #500-0001) which
utilized Commassie G—250 as dye reagent and bovine gamma globulin as a standard.
Spectra were obtained as in the legend of Figure 2.

65

 

 

0.16-

 

80.12-

aoueqrosqv

pH9 10 1 0'08 .-

\ 0.04 -

 

 

I
266 278 290 302 314

Wavelength (nm)

Figure 3

66

spectrum of the same sample in the perturbing solvent. Figure 4 shows difference
spectra of free tyrosine and tryptophan (each at 0.4 mM) using 20 % ethylene glycol
as perturbing solvent. Tryptophan reveals a difference spectrum with predominant
maximum at 295 nm, a secondary maximum at 281 nm, and tertiary maximum at 273
nm. Tyrosine has a significantly weaker difference spectra with a predominant
maximum at 285 nm and a secondary maximum at 279 nM.

As applied to protein structure, solvent perturbation measures whether or not a
given tyrosine or tryptophan is exposed to solvent. If a residue is on the surface of

the protein it will have a difference spectrum similar to the difference spectrum of the

 

free amino acid (Figure 4). If the residue is buried, it will not show a difference
spectrum since the residue in not in contact with the solvent. I first applied solvent
perturbation to GAL4-del456 (which does not have polar aromatic amino acids in the
activation domain) in order to determine what contributions Trp 36 and Tyr 40 of the
GAL4 DNA-binding domain would make to the difference spectra. In Figure 4, I
compare the 20 % ethylene glycol difference spectra of 0.4 mM tyrosine, 0.4 mM
tryptophan, a mixture of 0.4 mM each of tyrosine and tryptophan, and 0.4 mM
GAL4-del456. The relatively weak difference spectra of 0.4 mM GAL4-del456
reveals that neither Trp 36 nor Tyr 40 are very accessible to solvent. This
observation is in agreement with recent 1H NMR data which demonstrates that Trp 36
and Tyr 40 are buried (Gadhavi ml, 1990).

Given that GAL4 Trp 36 and Tyr 40 are buried within the GAL4 structure I
was able to perform UV analysis on activation domain derivatives possessing tyrosine

and tryptophan with little contamination. First, I compared the 20% ethylene glycol

 

67

Figure 4. Difference spectra of GAL4-del456 and GAL4-FY 442. A. 1:1 mixture
of 0.4 mM tyrosine and 0.4 mM tryptophan (this spectrum is offset by 0.04
absorbance units), B. 0.4 mM tryptophan, C. 0.4 mM tyrosine, D. 0.4 mM
GAL4-FY 442, E. 0.4 mM GAL4-del456. Method. Tyrosine and tryptophan were
obtained from Sigma. Proteins stored at -70 °C in buffer A containing 0.6 M NaCl
were thawed and aliquots were dialyzed (1 hr in a microdialysis chamber) against 20
mM TRIS (pH 7.5) and 0.2 mM NaCl at 10 °C to remove salts and protease
inhibitors. Following dialysis proteins were diluted into 20 mM TRIS (pH 7.5) and
0.2 M NaCl to the appropriate concentration. Samples were divided into two 80 pl
aliquots, 20 pl of ethylene glycol were added to one aliquot, 20 pl of water were
added to the other. The difference spectra were determined by using the aliquot
containing water as the blank. Protein concentrations were determined as in legend
of Figure 3. Spectra were obtained as in the legend of Figure 2.

68

 

 

 

 

 

 

0.16"
o.12 - >
O"
U)
A 3
O'
m
:3
O
(D
0.08 "
B 0.04 -
C D .
r ._L 1 r I E F5 I I
257 271 > 285 299 313

Wavelength (nm)

Figure 4

69
difference spectrum of 0.4 mM GAL4-FY 442 with 0.4 mM free tyrosine as shown in

Figure 4. The extremely weak difference spectrum of Tyr 442 relative to an equal
concentration of free tyrosine strongly suggests, in agreement with the previous
titration data, that Tyr 442 is buried.

I extended UV analysis to the more complex fusion protein containing the full-
length activation domain, GAL4—VP16. GAL4-VP16 has three polar aromatic
residues in addition to GAL4 Trp 36 and Tyr 40: a tryptophan residue which is
encoded by the cloning linker between the GAL4 and VP16 domains (Trp 154), a

natural tyrosine at VP16 position 465 (Tyr 465) and another natural tyrosine residue

 

at VP16 position 488 (Tyr 488) just two amino acids from the VP16 C-terminus.
Figure 5 compares the 20% ethylene glycol difference spectra of 0.4 mM
GAL4-VP16 with the spectrum of a mixture of 0.4 mM tryptophan and 0.8 mM
tyrosine which would be the expected spectrum of 0.4 mM GAL4-VP16 if both
tyrosines and the tryptophan were exposed. The similarity of the spectra compared in
Figure 5 suggest that Trp 154, Tyr 465 and Tyr 488 are exposed to solvent. The fact
that the real GAL4-VP16 difference spectrum is broader than the spectrum of
equivalent concentrations of the free amino acids may reflect interactions with other
groups within the protein. It is not surprising that Trp 154 is solvent exposed, since
it is encoded by an artificial linker which is not likely to form a specific structure that
would occlude solvent from a specific residue. Similarly it is not unexpected that Tyr
488 is exposed since it lies so close to the C-terminus of the protein. Tyr 465 lies
within the activating region H of the full-length VP16 activation domain proximal to a

phenylalanine residue thoughtto be homologous to Phe 442 of activating region I.

 

70

Figure 5. Difference spectrum of GAL4-VP16. A. 0.4 mM GAL4-VP16, B. 2:1
mixture of 0.8 mM tyrosine and 0.4 mM tryptophan (stippled). Method. These
spectra were obtained as described in the legend of Figure 4.

 

>cmoqcmzoo

 

0.16-

71

 

_ _
2 8
4| 0
0 0
508...... 0 CI l0
“8:0"00'09 0'
Chilleh
.0 I‘I‘COIIO
fiODe.‘

 

0.04 -
313

299

(nm)

Figure 5

285

Wavelength

271

257

 

72

Knowing that Tyr 465 is exposed to solvent will aid in future studies of the structure
of the full-length VP16 activation domain.

The results of UV analysis suggest strongly that Tyr 442 is buried within the
structure of GAL4-FY 442, and from this I infer that Phe 442 is buried. There are at
least three problems with applying conclusions from this data to the structure of the
VP16 activation domain. First, I do not know that Tyr 442 being buried is
functionally relevant. For instance, Tyr 442 may be buried in the context of the

GAL4 fusion protein by a fairly nonspecific hydrophobic interaction with the GAL4

 

DNA-binding domain. One observation which argues against this criticism is that the
GAL4-FY 442 protein activates transcription m (S. Berger and A. Cress,
unpublished results) with approximately the same relative activity as pMSVP16 FY
442 activates transcription ill—Vllg (the previous chapter). Second, I cannot be certain
that Tyr 442 accurately reflects the environment of Phe 442; however, it is very likely
that it does. For instance, if Phe 442 were solvent exposed it would be unlikely that
Tyr 442 (which is more polar than Phe 442) would seek a more nonpolar environment
in the same context. A third problem with drawing conclusions from this UV data is
a technical one, in that I do not know precisely the concentrations of proteins used in
these experiments. The protein concentrations used in these experiments are
estimated primarily from a commercially available protein assay (described in the
legend of Figure 3), since attempts to accurately determine protein concentrations
using quantitative amino acid analysis have been unsuccessful for an unknown reason

(J. Leykam, unpublished experiments). Although I do not expect a large inaccuracy

73

based upon this protein assay, an independent method of accurately determining the
protein concentration would be preferred and will be obtained prior to publication of
these results.

The finding that Phe 442 is likely buried is clearly in support of the model
proposing that Phe 442 is critical in the folding of the VP16 activation domain upon
itself. VP16 derivatives with substitutions such as proline, alanine and serine at
position 442 probably can not fold properly. Substitution of Phe 442 with tyrosine
allows folding of the domain into a structure with diminished activity. Diminished

activity is likely due to the fact that polar tyrosine doesn’t fit as well as phenylalanine

 

into the natural hydrophobic environment of Phe 442.

Fourier-Transform Infrared Spectroscopy

UV analysis suggests that the VP16 activation domain structure insulates Phe
442 from solvent, but doesn’t give information regarding secondary or tertiary
structure. Infrared absorption spectroscopy, on the other hand, has been used to
study the secondary structure of polypeptides. Protein structure analysis by FTIR is
generally done in D20 solution, instead of water, since D20 is relatively transparent
in the region of interest. In deuterium oxide solution the most useful absorption band
for secondary structure studies is the amide I band (1620 to 1690 cm‘l) which
corresponds to the C =0 stretching vibration of peptide groups. The C =0 stretch is
sensitive to secondary conformation, thus different secondary substructures give rise
to slightly different absorption maximum. Although the various C =0 stretching

Components of typical proteins have inherently broad line shapes, the spectral

74

contributions of various substructures can often be resolved using mathematical
techniques. For instance, Fourier self-deconvolved (Kauppinen gal, 1981) FTIR
spectroscopy has been used to examine the structures of 21 globular proteins (Byler
and Suzi, 1986). In that study, the amide I spectra of various proteins consisted of
six to nine components which were empirically assigned to various elements of
secondary structure such as alpha-helix, beta-sheets, coil regions and turns.

The power of FTIR spectroscopy to resolve between alpha-helical and
beta-sheet structural components of proteins in solution is particularly good, as

demonstrated in Figure 6 which presents the deconvolved FTIR spectra of

 

concanavalin A and hemoglobin. Hemoglobin is approximately 80% alpha-helical and
is without beta-structure or turns. Accordingly, its IR spectrum shows a single large
peak centered at 1651 cm‘1 which corresponds to the single helix-specific C=O
stretch identified by Byler and Suzi (1986). As expected, hemoglobin does not have
strong absorbances corresponding to beta-sheet or turn structures. Concanavalin A,
on the other hand, is about 60% beta-sheet, thus its IR spectrum shows a principle
peak at 1635 cm'1 and a weaker peak at 1623 cm‘1 corresponding to two beta-sheet
specific C=O stretches identified by Byler and Suzi (1986).

I have applied FTHI analysis in an attempt to measure the elements of
secondary structure within the VP16 activation domain. Figure 7 shows the amide I
region of the FTIR spectra of GAL4 1-147, GAL4-VP16, GAL4-del456, GAL4-PP
442 and the VP16 activation domains purified from GAL4-VP16 and GAL4-del456.
Each of these polypeptides have FTIR maxima in the spectral region corresponding to

alpha-helical (1650-1654 cm'l) and coil structures (1644-1648 cm‘l), but do not have

 

75

Figure 6. FTIR spectra of model proteins. A. Hemoglobin, B. Concanavalin A.
The absorbance scale is in arbitrary units. Method. Lyophilized proteins obtained
from Sigma were dissolved in 200 mM NaCl in deuterium oxide solution (50 mg/ml)
and allowed to equilibrate for 24 hrs. FTIR spectra were obtained and analyzed using
a Nicolet 7199 FTIR instrument (courtesy of Dr. R. Hollingsworth). Three thousand
co-added interferograms were obtained at 2 cm'1 nominal resolution. The sample
volume and pathlength were 0.05 ml and 0.075 mm, respectively. Deconvolution
parameters were VFO=20 cm‘1 and VF1=2.0 and Lorentzian line-shape was
assumed.

A
v

A
‘

Absorbance

 

V

76

  
   

L

 

 

1710.3

3705.2

{592.0 {375.9 {565.0 3652.7

Wavenu'm ber (cm" )

Figure6

£539.:

A
1626.4

5613.3

i600.2

 

 

 

 

 

 

77

Figure 7. FTIR spectra of GAL4-VP16 proteins A. Full-length (90 amino acids)
VP16 activation domain, B. truncated VP16 activation domain, C. GAL4-VP16,
D. GAL4-FF 442, E. GAL4-del456, F. GAL4 1-147. The absorbance scale is in
arbitrary units. Method. The activation domain peptides corresponding to the full-
length and truncated activation domains were separated from the fusion proteins
GAL4-VP16 and GAL4-del456 by tryptic cleavage and purified as described in the
text. Proteins were prepared for FTIR by dialysis against 5 mM HEPES (pH 7.5)
and 0.2 M NaCl. Samples were then lyophilized, resuspended in D20 and allowed to
equilibrate for 24-48 hr. FTIR spectra were then obtained as described in the legend
of Figure 6.

Absorbance

 

 

6710.3

78

A A
_‘

‘ A L ‘ A ‘ ‘
3705.2 {092.0 1070.0 1000.0 3002.7 1039.0 1020.4 10:3.3 1000.2

Wavenumber'lcm'1 )

Figure 7

79

major peaks in the spectral region corresponding to beta-sheet structure. These data
suggest that these proteins are composed predominantly of a mixture of helix and coil
substructures. This observation is consistent with published CD data which estimates
that GAL4 1-147 is 40% alpha-helix, 40% coil and 20% beta-sheet (Pan and
Coleman, 1989). The IR spectra of the purified activation domains are notably
broader than the spectra of the proteins containing GAL4. While the maxima of helix
and coil IR spectra are very close, the IR spectrum of coil structure is
characteristically broader than that of alpha helix (Figure 4 of Byler and Suzi, 1986).
The observation that the pure activation domain peptides have broad IR spectra
relative to GAL4 1-147 suggests that they possess a higher percentage of coil
structure.

The fusion proteins GAL4-VP16, GAL4-del456, GAL4-FP 442, and the
purified full-length and truncated activation domains appear to be predominantly a
mixture of helix and coil structures. Unfortunately, the IR resonances corresponding
to these substructures are highly overlapping making it very difficult to determine
relative contributions of these two substructures to the overall structure. Thus, I have
chosen to rely on a complementary method of structural analysis, circular dichroism
(CD), which is much more sensitive to differences in helix and coil substructures for
quantitative estimates (see the next section).

I have, however, used FTR qualitatively to study the affects of various
solvents upon the structures of GAL4 1-147 and the GAL4-VP16 fusion protein. One
important finding of these comparisons was the great sensitivity of the GAL4 1-147

DNA-binding domain to conditions of low salt and low pH. At pH 7 .5, the GAL4

80
1-147 protein is insoluble below 0.2 M NaCl, whereas the GAL4-VP16 fusion is

soluble down to 0.1 NaCl. Apparently the highly charged activation domain improves
the solubility of the fusion protein, perhaps by charge repulsion between the
negatively charged domains. Neither GAL4 1-147 nor GAL4-VP16 show significant
changes in the their FTIR spectra in the range of NaCl concentration from 0.2 M to
1.0 M, however the GAL4-VP16 protein does show an overall shift toward lower
wavenumbers at 0.1 M N aCl (where GAL4 1-147 precipitates), probably reflecting
the structural changes leading to the precipitation of GAL4 1-147 in low salt. At 0.2
M NaCl both GAL4 1-147 and GAL4-VP16 precipitate at pH less than 7.0, although
no structural changes are evident in the pH range from 7 to 9. Negatively charged
residues are important in the cation-binding domains of proteins such as calmodulin
and troponin C. Upon divalent-cation binding both calmodan and troponin C
undergo conformation rearrangements which are evident using FTIR spectroscopy
(Trewhella gg, 1989). Although the VP16 activation domain does not have striking
sequence homology to the cation-binding domains of calmodulin or troponin C it was
considered a possibility that the VP16 activation domain might interact with such
divalent cations. Thus, the effects of the cations K+, Ca++, Mg++, Mn++, Fe++,
and Zn++ upon the structure of GAL4-VP16 were evaluated using FTIR.
GAL4-VP16 (0.2 mM) was dialyzed against 2 mM EDTA, 0.2 M NaCl, and 5 mM
HEPES buffer (pH 7.5) to remove bound cations, then against the same buffer
without EDTA. The sample was lyophilized and resuspended in D20. Specific salt
solutions (KCl, CaClz, MgClz, MnClZ, FeC12 and ZnC12 in D20) were then added

and FTIR spectra obtained after 24-48 hours. Surprisingly, 2 mM FeClz or 2 mM

81
ZnC12 precipitated 0.2 mM GAL4-B58 (equimolar concentrations did not). This

effect was clearly specific to the GAL4 DNA-binding domain, since GAL4 1-147 was
also precipitated by equimolar FeClz, ZnC12 or 113CdC12. This phenomenon is
reflected in the protocols of investigators who have previously studied the GAL4
Zn-binding domain (Pan and Coleman, 1989 and 1990), but an adequate explanation
for its mechanism is not available. The cations K+, Ca++, Mg”, and Mn++ (in
ten-fold excess) did not have any strong effects on the FTHI spectra of GAL4-B58
(0.2 mM) nor did equimolar concentrations of FeC12 or ZnClz. These result suggest
that the VP16 activation domain structure is not affected by interaction with cations
and probably suggests that the VP16 activation domain is not a specific cation-binding

protein.

Circular Dichroism

FTIR analysis is limited in its ability to resolve alpha-helix and coil structures,
however the VP16 activation domain appears to be composed of a mixture of these
two substructures. To estimate the relative contributions of helix and coil to the total
structure of the VP16 activation domain, circular dichroism (CD) was used. Because
a CD instrument is not available on campus, I collaborated with Dr. B. Krueger and
M. Prairie at the Upjohn company in Kalamazoo, M1 to obtain a CD spectrum of the
full-length VP16 activation domain. Figure 8 shows the CD spectrum of the full-
length VP16 activation domain as obtained in 10 mM sodium phosphate buffer (pH
7.2) containing 0.2 M NaCl. This CD spectrum clearly reveals that the activation

domain has little alpha-helical content as indicated by its relatively low negative

 

 

82

 

 

MOLAR ELLIPTICITY X ID
I

_‘a—l

 

 

’12 I

 

I I I I I I I I
175 195 215 235 255 275

WAVELENGTH (NM)

Figure 8. CD analysis of the full-length VP16 activation domain. Method. The
full-length VP16 activation was prepared for CD analysis by dialysis against 10 mM
sodium phosphate buffer (pH 7 .2) and 0.2 M NaCl. The CD spectrum above is the

average of 16 scans obtained in a 0.1 mm cell at room temperature using a Jasco
600C CD spectropolarimeter.

 

 

 

83

ellipticity at 222 nm. The CD spectrum does however suggest considerable coil
structure is present. Unfortunately, the CD spectrum of the full—length tail did not
yield acceptable results using secondary structure calculations (Compton and Johnson,
1986; Manavalan and Johnson, 1987). This observation suggests that the unlmown
structure is not well represented in the data base of proteins used (Dr. B. Krueger,
personal communication).

It is possible that the structure observed for the purified activation domain
peptide is not the natural structure of the domain, but that the activation domain needs
to be attached to the GAL4 DNA-binding domain (or to the specificity domain of
VP16) to form an active structure. I do not expect that this is the case, however,
since the isolated VP16 activation domain (with only an added nuclear localization
signal) has been expressed in Hela cells and shown to inhibit transcription from
reporter promoters (Tasset gg, 1990) suggesting that no other sequences are

required for it to form an active structure.

Native Gel Electrophoresis

UV analysis suggests that Phe 442 is buried within the structure of
GAL4—VP16, thus Phe 442 may be critical in maintaining a particular protein
conformation. If Phe 442 is maintaining such a conformation then it is likely that
inactivating substitutions at position 442 might disrupt that structure. Furthermore, it
might be possible to detect this change of conformation using native-gel

electrophoresis, since the mobility of a native protein within an electric field is

 

84

inversely proportional to its frictional coefficient which is in turn a function of its
overall conformation.

I tested the possibility that substitutions at Phe 442 might alter the mobility of
GAL4-VP16 fusion proteins in the context of a DNA-binding/ mobility shift assay
(Fried and Crothers, 1981). In this assay, an oligonucleotide composing the
GAL4-binding site was incubated in the presence of various GAL4-VP16 fusion
proteins. The complexes were then electrophoresed on a nondenaturing gel and the
relative mobility of the complexes compared (see Figure 9). The first lane of Figure

9 shows the mobility of the GAL4 1-147:DNA complex which has the slowest

 

mobility of all of the GAL4-VPl6zDNA complexes despite being the smallest protein.
The second lane of Figure 9 shows the mobility of GAL4-AH which adds a fifteen
amino acid peptide designed to be an AAH to the GAL4 DNA-binding domain
(Gininger and Ptashne, 1988). The increased mobility of this proteinzDNA complex
relative to the GAL4:DN A complex is apparently due to the three negative charges
added by the AH peptide. The next seven lanes of Figure 9 compare the mobilities of
GAL4-del456 with six mutant derivatives. These lanes reveal that mutations of Phe
442 do not affect the mobility of GAL4-VP16 fusion proteinzDNA complexes in this
system, since GAL4-del456, -FP 442, -FA 442, and -FY 442:DNA complexes all
have indistinguishable mobilities. Ianes labeled DN 2578 and DN 2579 reveal the
importance of negative charge in this system since removal of four negative charges
considerably slows the mobility of these derivatives relative to GAL4-del45 6. The
final lane of Figure 9 shows the relative mobility of GAL4-VP16. Despite having

considerably more negative charge than GAL4-del456 derivatives the mobility of

85

'l
I

I
to

<2- " r~ LO
SI. “3, 3 <2- 3 ‘1' N N :3
1- I d) D.

v n. < 0- Z Z
‘3: T? u. it. It < o o >,
V V q. l “I. l I I . V
_. _n —I 3 _r 3 3 3. 3 _.
< < < < < < < < < <
(5 (5 (5 (5 (5 (5 (5 <5 (5 (5

Figure 9. Native gel electrophoresis of GAL4-VP16 proteins. The proteins assayed
are indicated by name below the lanes. Method. Binding assays (20 pl total volume)
contained 20 mM HEPES (pH 7.5), 50 mM KCl, 5 mM MgC12, 10 pM ZnClz, 6%
glycerol, 200 pg/ml BSA, 2 ng of 3ZZP—labeled double-stranded oligonucleotide
(5’—CGGACGAGAGTCTTCCG-3’) and 20 ng of each protein. Binding was allowed
to take place for 15 min at room temperature, then 4 pl of 5X loading dye (0.25 g/ml
ficoll with 0.01% bromophenol blue) was added and the sample loaded onto a
pre-electrophoresed 4.5% (w/v) polyacrylamide gel containing 90 mM TRIS base, 90
mM boric acid, 2 mM EDTA, and 1% glycerol. Electrophoresis was carried out in
gel buffer (minus glycerol) at 15 Volts/cm until the bromophenol blue had entered the
lower buffer tank. The gel was fixed for 30 minutes in 20% (v/v) methanol and 10%
(v/v) acetic acid, dried onto Whatman paper and an autoradiograph obtained by
exposure to X-ray film.

 

86
the GAL4-VP16 complex is only slightly greater suggesting that the GAL4-V1316

protein possesses a measurably greater frictional coefficient than its deleted
counterpart.

The observation that mutations at Phe 442 didn’t affect the mobility of
GAL4-del456:DNA complexes suggests that the presumed structural differences are
too small to detect in the context of this assay, but are otherwise inconclusive.
Because of the net negative charge of GAL4-VP16 fusion proteins it was also possible
to compare their mobilities in the absence of DNA using the same TRIS-Borate (pH
8.3) buffer system used to compare mobilities of proteinzDNA complexes.
Unfortunately, the mobilities of GAL4-del456 derivatives having substitutions at Phe
442 were also indistinguishable in absence of DNA ligand. It is possible that
structural changes are present, but that the relative mobilities of these proteins, at the
relatively high pH of the buffer system used, are simply dominated by their negative
charge. Unfortunately, GAL4 fusion proteins are insoluble below neutral pH, thus
low pH electrophoresis buffers which might maximize the contribution of protein
conformation toward electrophoretic mobility (while eliminating the dominant

contribution of so many acidic residues) could not be used.

Differential Scanning Calorimetry

UV analysis suggests that the VP16 activation domain folds into a specific
structure which insulates Phe 442 from solvent. I presume that mutations at Phe 442
in some way disrupt this structure, since they inactivate the domain, however, I was

not able to detect any differences in structure using native gel electrophoresis. This

 

87

observation may indicate that the overall shapes of inactive VP16 derivatives are not
very different than wildtype. However, the inactive derivatives might be less stable
thermodynamically, thus I attempted to measure the thermodynamic stabilities of
various activation domain derivatives using differential scanning calorimetry (DS C)
(reviewed by Krishnan and Brandts, 1978; Donovan, 1984).

Before DSC could be used to explore differences in activation domain stability
among various derivatives it was necessary to first determine if the VP16 activation
domain unfolded in a cooperative fashion such that it would yield a measurable
denaturation enthalpy. Figure 10 shows DSC scans of GAL4 1-147 and GAL4-VP16.
GAL4 1-147 did not tolerate heating. At temperatures above 47 °C GAL4 1-147
began to gradually precipitate, thus the DSC scan of GAL4 1-147 is not meaningful.
Varying solvent conditions (NaCl up to l M; and pH up to 9) did slightly raise the
temperature at which GAL 1-147 began to precipitate, but did not allow me to obtain
DSC data of GAL4 1-147 in solution. The GAL4-VP16 fusion protein on the other
hand remained soluble up to 100 °C; again, as in the case of low salt, the VP16
activation domain keeps the GAL4 DNA-binding domain in solution. Unfortunately
GAL4—VP16 did not yield any obvious peaks in heat capacity corresponding to the
unfolding of individual protein domains. Instead, a gradual increase in heat capacity
was ob served over the entire temperature range monitored. This observation probably
indicates that the GAL4-VP16 protein is not composed of one or two major
cooperatively folded units, but rather is composed of several small units that unfold

independently with overlapping melting temperatures. These hypothetical units could

 

88

Figure 10. DSC scans of GAL4 1-147 and GAL4-VP16. A. 2 mg/ml GAL4 1-147
B. 2 mg/ml GAL4-VP16 Method. Proteins were prepared for DSC by dialysis
against 20 mM HEPES (pH 7.5), 0.1 M NaCl, 20 mM monothioglycerol, and 10 pM
zinc acetate. Samples and reference buffers were deaerated with stirring under
vacuum. DSC experiments were conducted at a scan rate of 90 °C/hr in a Microcal
MC-2 scanning calorimeter (courtesy of Dr. J. Wilson).

 

meal/degree K

meal/degree K

89

 

 

 

 

 

 

 

 

ll 1 l l l l ' l
20 30 4O 50 60 70 80 90 100
Temperature (degrees C)
l—
l L l l l 1 LI
20 30 40 50 60 70 80 90 100

Temperature (degrees Cl

Figure 10

 

 

 

90
correspond to the Zn—binding, DNA-specificity, and dimerization domains of GAL4

1-147 and the two activating regions of the VP16 activation domain.

Conclusions

The work in this chapter represents the first biophysical characterization of a
transcriptional activation domain. The results of this chapter suggest two conclusions.
First, Phe 442 of the VP16 activation domain is probably buried within the tertiary
structure of the protein, whereas Tyr 465 and Tyr 488 are exposed to solvent. This
observation suggests that the critical functional role of Phe 442 discovered in the
previous chapter is to maintain the internal structure of the domain through
hydrophobic interactions. The second conclusion that can be made from this chapter
is that the VP16 activation domain contains very little alpha—helix. This observation
is consistent with the results obtained using mutagenesis with regard to the functional
importance of predicted alpha-helix.

The data presented in this and the preceding chapter support a model in which
internal hydrophobic interactions involving Phe 442 are important in maintaining the
structure of the VP16 activation domain. More detailed information regarding the
structure of the VP16 activation domain will require that the structure of the domain

be determined using methods such as 1H NMR or X—ray crystallography.

CHAPTER IV

DISCUSSION
Introduction

An important question, related to the central topic of this dissertation, is how

the VP16 activation domain works. When I began this project the mechanism of
activation by VP16, or any other activator, was primarily a matter of speculation.
Since that time however, the VP16 activation domain has been extensively studied and
has became somewhat of a paradigm for transcriptional activators. During the course
of my investigation I generated a number of VP16 derivatives which have become
important reagents in addressing the identity of the VP16 target. In this section I will
summarize several investigations which have utilized several of my VP16 activation

domain derivatives toward identifying the activation domain target(s).

Mechanism of Action of the VP16 Activation Domain.

It is becoming clear that the VP16 activation domain may utilize a number of
mechanisms to activate transcription. For instance, the VP16 activation domain can
clearly activate transcription by the mechanism of antirepression (Croston _e_t_g,
1991). The VP16 activation domain may also interact with the basal transcription
factor TFHD. Stringer, Ingles and Greenblatt (1990) performed the first experiments

that indicated that the VP16 activation domain might interact directly with TFHD. In

91

92

their experiments, they utilized the VP16 activation domain as a column ligand in
affinity chromatography. They found that the VP16 affinity column selectively
retained human TFHD and purified recombinant yeast TFHD. This result was
consistent with the hypothesis that TFHD might be a target of the VP16 activation
domain, however, it was open to the criticism that the interaction between the two
proteins could be relatively nonspecific due to the net negative charge of the VP16
activation domain and the net positive charge of the TFHD protein [(pKa 9.5 (Hahn Qt
g, 1989)].

Substitutions of Phe 442 of the minimal VP16 activation domain with alanine,
serine and proline abolish activation without affecting negative charge. If the
interaction of TFHD with VP16 was simply the result of charge attraction then these
inactivating substitutions should have no effect on the mutual affinity between TFHD
and VP16. To test the specificity of the interaction between TFHD and the VP16
activation domain, I sent the Ingles group a collection of VP16 activation domain
derivatives. In this collaboration (Ingles girl, in press) we measured the relative
affinities of the VP16 derivatives del456, FP 442, FS 442, FA 442, FY 442, DN
2789 and DN 2579 for yeast recombinant TFHD.

The deleted version of the VP16 activation domain del45 6 had a ten-fold lower
affinity for TFHD than did full-length VP16, despite the fact that it 10Ses only 50%
activity of the full-length activation domain. However, when VP16 derivatives having
substitutions at Phe 442 were compared in the context of the deleted version, their
affinities for TFHD correlated perfectly with their relative activities. The inactive

VP16 derivatives FP, FS and FA 442 had little measurable affinity for TFHD,

93
whereas the partially active derivative FY 442 had an affinity for TFIID partially

reduced from that of del456. This observation clearly rules out the possibility that the
affinity of TFHD to VP16 is purely a nonspecific ionic attraction. Surprisingly, the
derivatives DN 27 89 and 25 79 which possess 20 and 60 % relative activity,
respectively, both have lower affinities for TFHD than FY 442 (despite the fact that
DN 2579 is more active in transcription). The observations that affinity for TFHD
does not correlate perfectly with measured relative activities of VP16 derivatives
might suggest that TFHD binding may not be the sole mechanism of activation.
Alternatively, the imperfect correlation between affinity for TFHD and activation may
simply reflect that net negative charge is important in attracting and binding TFIID,
but that hydrophobic interactions are critical in the activation phenomenon (which
might involve a change in conformation). Thus, the mutant DN 2579 may activate
transcription better than FY 442 because DN 2579 better preserves hydrophobic
interactions critical in activation, whereas FY 442 binds TFHD better because it
preserves negative charges critical in attracting TFHD.

Lin and Green (1991) have recently discovered evidence that the VP16
activation domain specifically interacts with yet another basal transcription factor,
TFHB. In this work, Lin and Green used the VP16 activation domain as a
chromatography affinity matrix in a manner similar to Stringer _elgl (1990). They
passed Hela cell nuclear extracts over the VP16 affinity column and observed binding
of TFHD to the column. However, Lin and Green went on to demonstrate that
TFHD is not the only basal transcription factor bound to the VP16 column. TFHB

was also bound and appeared to be bound with greater affinity than is TFIID (with

94
respect to the concentration of KCl used as eluant). Stringer e1; gl, (1990) did not

eliminate the possibility that TFHB was bound to their columns. Lin and Green
recognized the possibility that the affinity of TFHB to the VP16 column could be due
to non-specific charge interactions. Thus, they requested the VP16 activation domain
derivatives del45 6 and FP 442 which they used as control affinity matrices. They
found that TFHB bound to the del45 6 protein and not to the FP 442 derivative under
identical conditions, thus the affinity of TFHB (like that of TFHD, described above)
for VP16 was specific to transcriptionally active versions of VP16 suggesting that the

ob served interaction is biologically relevant.

 

Activation of transcription by VP16 may not be limited to antirepression and
recruitment of basal factors. In a collaboration with S. Berger and L. Guarente, we
(A. Cress, S. Triezenberg and I) sought to identify the VP16 target using yeast i_n
21m transcription reactions and GAL4-VP16 fusion proteins (Berger gal, 1990). In
this collaboration we compared the effects of GAL4-VP16 on three promoter
templates: one having only a TATA box and two having upstream activating
sequences, GAL4 UAS or dAsz UAS. The basal level of transcription in these
experiments was defined as the level of transcription obtained from the promoter
containing only a TATA box. The GAL4 UAS promoter gave a basal level of
transcription in the yeast extracts in the absence of exogenous GAL4-VP16, whereas
the dAsz UAS template was activated to a level ten-fold basal (probably due to the
presence of a dAsz activating protein in the yeast extract). When GAL4-VP16 was
added to the GAL4 UAS template transcription was activated roughly 100-fold over

the basal level. However, GAL4-VP16 abolished transcription from templates

95

containing basal and dA:dT promoters. Mutations within the VP16 activation
domain, such as FP 442, which abolished activation by VP16 also abolished its ability
to inhibit, thus the mechanism of this inhibition presumably involves the ability of
VP16 to bind to some transcription factors (or factors) and titrate them away from
heterologous promoters. Addition of an oligonucleotide containing a GAL4-binding
site which would prevent GAL4-VP16 from binding nonspecifically to DNA,
however, was found to relieve inhibition of basal transcription of both the basal
template and the dA:dT template, but could not restore transcription to the activated
level. This observation suggested that GAL4-oligozGAL4-VP16 must be able to
titrate some factor involved in activated transcription, but not required for basal
transcription. An apparently identical activity which partially copurifies with TFIID
(presumably due to their association) has been identified by other workers (Kelleher e_t

g, 1990; Hoey et al., 1990; Pugh and Tjian, 1990; Peterson et al., 1990).

 

The list of factors through which VP16 has the potential to mediate its effect is
long. Is it possible that VP16 interacts with many factors to activate transcription? It

has been noted that transcriptional activators work synergistically when multiple

 

regulatory elements are present (Lin et al., 1988; Carey fl, 1990b). In
experiments which have studied synergistic action by acidic activation domains it has
been estimated that up to ten molecules of activator can synergize to activate
transcription (Carey et_al., 1990b). If the mechanism of activation required
interaction with a single factor then this would suggest that up to ten activation
domains could contact the target simultaneously. As pointed out by Ptashne and Gann

(1990), it is very difficult to imagine that ten domains could simultaneously contact a

96 1

single target protein such as TFHD, for instance, since its molecular weight is only 38
kD and TFHD also appears to interact in some manner with other basal factors,
TFHB and TFHA. Perhaps a better explanation would be that the acidic activation
domains indeed can recruit several components of the basal transcriptional machinery,
in addition to its ability to relieve repression by histones. This would explain how
multiple activators can function synergistically at a single promoter and would be
consistent with the observation that strong activator proteins, such as the full-length
VP16 activation domain, often have multiple activating regions (J. Regier and S.
Triezenberg, unpublished results).

The final answer to the question of how the VP16 activation domain functions
will clearly require further study. As clones corresponding to the various

transcription factors become available for use in in vitro transcription reactions they

 

will undoubtably provide new insights into the mechanisms of transcriptional
activation by VP16. Furthermore, the VP16 mutants described in this dissertation
will most likely be valuable as reagents in the characterization of the activation

mechanism.

Critical Structural Elements of the VP16 Activation Domain

The hypothesis that acidic transcriptional activation domains function as AAHs
(Gininger and Ptashne, 1987) has stood untested for a number of years. To my
knowledge, my work is the first in depth test of that hypothesis and the first structural
characterization of a eukaryotic transcriptional activation domain. In this work, I

believe that I have convincingly demonstrated that the AAH model is incorrect.

 

 

 

97

Unfortunately, I can not suggest a simple model to replace the AAH. In other words,
I have not been able to identify the "leucine zipper" or " zinc finger" or any other
telling anatomical feature of acidic activation domains, although I do observe some
conservation of bulky hydrophobic residues among transcriptional activation domains.
I propose two alternative models which are somewhat more general than the
AAH, but which take into account the measured structural elements of the activation
domain and seek to explain the critical role observed for hydrophobic residues. One
model proposes that Phe 442 may make direct contact with the VP16 activation

domain target. Phe 442 for instance, might fit into an hydrophobic cleft within the

 

structure of the target molecule. This model is made particularly attractive when the
similarities between the VP16 activation domain and the acidic tail of the
anticoagulant hirudin are considered. The hirudin acidic tail binds to the exosite of
thrombin and possesses a critical phenylalanine residue which inserts itself into a
hydrophobic cleft of the thrombin molecule upon binding (Rydel e_t_al, 1990). The
possibility that hirudin and the VP16 activation domain are structurally similar is
appealing because the crystal structure of hirudin in complex with thrombin has been
solved (Rydel §t_a_l,, 1990). The relevant sequences of hirudin and the VP16
activation domain compare as follows:

VP16 -Asp-Ala-Leu-Asp-Asp-m-Asp-Leu-Asp-Met-LeuM7

I-Iirudin -His-Asn-Asn-Gly-Asp-le-Glu-Glu-Ile-Pro-Glu61

Although the direct comparison of sequences is not tremendously convincing, Dr. A.

Tulinsky pointed out that the positions of negative charges and hydrophobic residues

98

relative to the central phenylalanine are identical if one of the sequences is considered
in reverse (i. e. the sequence read C-terminal to N—terminal):

VP16 N—Asp-Ala-Leu—Asp-ASP-P_hC-Asp-Leu-Asp-Met-Leu447
Hirudin C-Glu-Pro-11e—Glu-Glu-Pl'le—Asp-Gly-Asn-Asn-Hi551 .

Comparing these sequences in reverse orientation is reasonable in the case of hirudin,
since all hirudin/thrombin contacts in this region are side-chain contacts and thus do
not necessarily involve the polarity of the polypeptide mainchain.

The results of mutagenesis analyses of hirudin—like peptides further suggest
that a comparison between hirudin and the VP16 activation domain is relevant. For
instance, Stone fl (1989) determined that the affinity of hirudin-like peptides to
thrombin was a function of net negative charge. Furthermore, it was found that the
phenylalanine of hirudin could be replaced by tyrosine, but not by alanine, glutamate,
or leucine or by a variety of nonprotein phenylalanine analogs including
D-phenylalanine (Krstenansky et_al., 1987; Owen et_al., 1988). These results are
very similar to results that J. Regier (unpublished) and I have observed for Phe 442 of
the VP16 activation domain. Finally, prior to the determination of its crystal
structure with thrombin (Rydel et_al., 1990) the acidic tail of hirudin was modeled as
an AAH (Krstenansky et_al., 1987).

A simple test which might strengthen the hirudin/VP16 connection would be to
demonstrate that the VP16 activation could bind to the thrombin exosite or that the
hirudin tail could activate transcription. The first of these test has been done.
Unfortunately, (from a model building perspective) L. Graham and S. Triezenberg

Could find no experimental evidence that the VP16 activation domain bound to

99

thrombin (unpublished data) using native gel-electrophoresis experiments. It has not
yet been determined if the hirudin polypeptide can activate transcription.

The alternative model proposes that Phe 442 is important in hydrophobic
contacts involved in the self-folding of the VP16 activation domain. UV spectroscopy
data supports this model, in that Tyr 442 of GAL4-FY 442 is buried. This model is
also consistent with the observation that various activation domains have a conserved

pattern of bulky hydrophobic residues.

Future Studies

 

The final determination of the three-dimensional structure of the VP16
activation domain will clearly distinguish between these two models. Fortunately, I
have been able to initiate two projects with the potential to elucidate the structure of
the VP16 activation domain. These investigations will clearly be aided by the

availability of GAL4-VP16 fusion proteins purified from E. coli and the availability

 

of various mutants to correlate structural elements with function. The first project, a
collaboration with Dr. G. Wagner at Harvard University, will seek to determine the
solution structure of the VP16 activation domain using NMR. Preliminary NMR
analysis of the purified activation domain agrees with CD analysis in that the purified
activation domain appears to contain virtually no helical structure in solution. The
Second project, in collaboration with Dr. A. Tulinsky here at MSU, will seek to
crystallize the GAL4-VP16 fusion protein and then to determine its crystal structure

using X-ray diffraction.

LIST OF REFERENCES

Ace, C. I., Dalrymple, M. A., Ramsay, F. H., Preston, V. G. and Preston C. M.
(1988). J. Gen. Virol. 69:2595-2605.

 

Allison, L. A. and Ingles, C. J. (1989). Proc. Natl. Acad. Sci. USA 86:2794-2798.

Amit, A. G., Mariuzza, R. A., Phillips, S. E. V. and Poljak, R. J. (1986). Science
233:747-753.

 

Azizkhan, J. C., Vaughn, J ., Christy, R. J. and Hamlin, J. H. (1986). Biochemistry
25:6228-6236.

Baker, A. and Schatz, G. (1987). Proc. Natl. Acad. Sci. USA 84:3117-3121.

Batterson, W. and Roizman, B. (1983). J. Virol. 46:371-377.

Berger, 8., Cress, W. D., Cress, A. P., Triezenberg, S. J. and Guarente, L. (1990).
Cell 61:1199-1208.

Brandl, C. I. and Struhl, K. (1989). Proc. Natl. Acad. Sci. USA 86:2652-2656.
Breathnach, R. and Chambon, P. (1981). Annu. Rev. Biochem. 50:349-393.
Brent, R. and Ptashne, M. (1985). Cell 43:729-736.

Briggs, M. and Gierasch, L. (1986). Adv. Protein Chem. 38:109-180.

Bunick, D., Zandomeni, R., Ackerman, S. and Weinmann, R. (1982). Cell
29:877-886.

Burton, Z. R, Killeen, M., Sopta, M., Ortolan, L. G. and Greenblatt, J. (1988).
Mol. Cell. Biol. 8:1602-1613.

Buratowski, S., Hahn, S., Guarente, L. and Sharp, P. A. (1989). Cell 56:549-561.

Byler, D. M. and Suzi, H. (1986). Biopolymers 25:469-487.

100

101
Bzik, D. J. and Preston, C. M. (1986). Nucleic Acids Res. 14:929-943.

Campbell, M. E. M., Palfreyman, J. W. and Preston, C. M. (1984). J. Mol. Biol.
180:1-19.

Carey, M., Leatherwood, J. and Ptashne, M. (1990a). Science 247:710-711.

Carey, M., Lin. Y.-S., Green, M. R. and Ptashne, M. (1990b). Nature
345:361-364.

Carcamo, J., Lobos, S., Merino, A., Buckbinder, L., Weinmann, R., Venkatachala,
N. and Reinberg, D. (1989). J. Biol. Chem. 264:7704-7714.

Carcamo, J., Maldonado, E., Cortes, P., Ahn, M.-H., Hho-Ha, Kasai, Y., Flint, J.
and Reinberg, D. (1990). Genes Dev. 4:1611—1622.

Chasman, D. I., Leatherman, J., Carey, M., Ptashne, M. and Kornberg, R. D.
(1989). Mol. Cell. Biol. 9:4746-4749.

 

Chou, P. Y. and G. D. Fasman. (1978). Ann. Rev. Biochem. 47:251-276.
Compton, L. A. and Johnson, Jr., W. C. (1986). Anal. Biochem. 155:155-167.

Conaway, J. W., Travis, E. and Conaway, R. C. (1990). J. Biol. Chem.
265:7564-7596.

Courey, A. J. and Tjian, R. (1988). Cell 55:887-898.

Cousens, D. J., Greaves, R. Goding, C. R. and O’Hare, P. (1989). EMBO J.
8:2337-2342.

Corden, J., Wasylyk, B., Buchwalder, A., Sassone-Corsi, P., Kedinger, C. and
Chambon, P. (1980). Science 209:1405-1414.

Cordingley, M. G., Campbell, M. E. M. and Preston, C. M. (1983). Nucleic Acids
Res. 11:2347-2365.

Corton, J. C. and Johnston, S. A. (1989). Nature 340:724-727.

Croston, G. E., Kerrigan, L. A., Lira, L. M., Marshak, D. R. and Kadonaga, J. T.
(1991). Science 251:643-649.

Dalrymple, M. A., McGeoch, D. J., Davison, A. J. and Preston, C. M. (1985).
Nucleic Acids Res. 13:7865-7979.

102

Davidson, B. L., Egly, J .-M., Mulvihill, E. R. and Chambon, P. (1983). Nature
301:680-686.

Davis, R. L., Weintraub, H. and Lassar, A. B. (1987). Cell 51:987-1000.
Devereux, J ., Haeberli, P. and Smithies, O. (1984). Nuc. Acids Res. 12:387-395.

Dignam, D. L., Levobitz, R. M., and Roeder, R. G. (1983) Nuc. Acids Res.
11:1475-1485.

Douglas, H. D. and Hawthorne, D. C. (1964). Genetics 49:837-844.

Donovan, J. W. (1984). TIBS:340-344.

Dynan, W. S. and Tjian, R. (1985). Cell 35:79-87.

Eisenmann, D. M., Dollard, C. and Winston, F. (1989). Cell 58:1183-1191.
Elgin, S. C. R. (1988). J. Biol. Chem. 263:19259-19262.

Evans, R. M. (1988). Science 240:889-895.

Fire, A., Samuels, M. and Sharp, P. A. (1984). J. Biol. Chem. 259:2509-2516.

Flores, O., Maldonado, E. and Reinberg, D. (1989). J. Biol. Chem.
264:8913-8921.

Forsburg, S. L. and Guarente, L. (1989). Genes Dev. 3:1166-1178.

Freedman, L. P., Luisi, B. F., Korszun, Z. R., Basavappa, R., Sigler, P. B. and
Yamamoto, K. R. (1988). Nature 334:543-546.

Fried, M. and Crothers, D. M. (1981). Nucleic Acids Res. 9:6505-6525.

Gadhavi, P. L., Raine, A. R. C., Alefounder, P. R. and Laue, E. D. (1990). FEBS
276249-53.

Gill, G. and Ptashne, M. (1987). Cell 51:121-126.
Giniger, E. and Ptashne, M. (1987). Nature 330:670-672.

Gerster, T. and Roeder, R. G. (1988). Proc. Natl. Acad. Sci. U.S.A.
85: 6347-635 1 .

Gibson, W. and Roizman, B. (1974). J. Virol. 13:155-165.

 

103
Gilmore, D. S. and Lis, J. T. (1986). Mol. Cell. Biol. 6:3984-3989.

Godowski, P. J ., Picard, D. and Yamamoto, K. R. (1988). Science 241:812-816.

Graves, B. J ., Eisenman, R. N. and McKnight, S. L. (1985). Mol. Cell. Biol.
5: 1948-1958.

Greaves, R. and O’Hare, P. (1989). J. Virol. 63:1641-1650.
Greaves, R. and O’Hare, P. (1990). J. Virol. 64:2716-2724.

Hahn, S., Buratowski, S., Sharp, P. A. and Guarente, L. (1989). Cell
58:1173-1181.

Hai, T., Horikoshi, M., Roeder, R. G. and Green, M. R. (1988). Cell
54:1043-1051.

Heine, J. W., Honess, R. W., Cassai, E. and Roizman, B. (1974). J. Virol.
14:640-651.

Hoey, T. Dynlacht, B. D., Peterson, M. G. Pugh, B. F. and Tjian, R. (1990). Cell
61:1179-1186.

Hollenberg, S. M. and Evans, R. M. (1988). Cell 55:899-906.
Hope, I. A. and Struhl, K. (1986). Cell 46:885-894.
Hope, I. A., Mahadevan, S. and Struhl, K. (1988). Nature 333:635-640.

Horikoshi, M., Hai, T., Lin, Y.-S., Green, M. R. and Roeder, R. G. (1988). Cell
54: 1033-1042.

Hu, J. C., O’Shea, E. K., Kim, P. S. and Sauer, R. T. (1990). Science
250: 1400-1403.

Ingles, C. J ., Shales, M., Cress, W. D., Triezenberg, S. J. and Greenblatt, J. (in
press). Nature.

Jones, K. A. and Tjian, R. (1985). Nature 317:179-182.

Johnson. P. F., Landshultz, W. H., Graves, B. J. and McNight, S. L. (1987).
Genes Dev. 1:133-146.

Johnson, P. F. and McNight, S. L. (1989). Annu. Rev. Biochem 58:799-839.

 

 

104
Kaiser, C., Pruess, D., Grisafi, P. and Botstein, D. (1987). Science 235:312-317.

Katan, M., Haigh, A., Verrijzer, C. P., Vandervliet, P. C. and O’Hare, P. (1990).
Nucleic Acids Res. 18:6871-6880.

Kauppinen, J. K., Moffat, D. J., Mantsch, H. H. and Cameron, D. G. (1981).
Applied Spectroscopy 35:271-276.

Kelleher, R. J., Flanagan, P. M. and Kornberg, R. G. (1990). Cell 61:1209-1215.
Khoury, G. and Gruss, P. (1983). Cell 33:313-314.

Kissinger, C. R., Liu, B., Martin-Blanco, E., Kornberg, T. B. and Pabo, C. O.
(1990). Cell 63:579-590.

Klapper, M. H. (1977). Biochem. Biophys. Res. Comm. 78:1018-1024.
Krishnan, K. S. and Brandts, J. F. (1978) Methods Enzymol. 49:3-14.
Kristie, T. M. and Roizman, B. (1984). Proc. Natl. Acad. Sci. USA 81:4065—4069.
Kristie, T. M. and Roizman, B. (1987). Proc. Natl. Acad. Sci. USA 84:71-75.
Kristie, T. M., LeBowitz, J. H. and Sharp, P. A. (1989). EMBO J. 8:4229-4238.
Kristie, T. M. and Sharp, P. A. (1990). Genes Dev. 4:2383-2396.

Krstenansky, J. L., Owen, T. J., Yates, M. T. and Mao, S. J. T. (1987). J. Med.
Chem. 30:1688-1691.

Kunkel, T. A. (1985). Proc. Natl. Acad. Sci. USA 82:488-492.
Laemmli, U. D. (1970). Nature 227:680-685.
LaMarco, K. L. and McKnight, S. L. (1989). Genes Dev. 3:1372-1383.

Landschulz, W. H., Johnson, P. F. and McKnight, S. L. (1988). Science
240: 1759-1764.

Lathrop, R. H., Webster, T. A., Smith, T. F. and Winston, P. H. (in press). In
Artificial Intelligence at MIT:Expanding Frontiers (Winston, P. H. and Shellard,
S. A. eds.), pp. 70-103, IVHT Press, Cambridge, MA.

Levine, M. and Manley, J. L. (1989). Cell 59:405-408.

 

105
Lesk, A. M. and Chothia, C. J. (1980). J. Mol. Biol. 136:225-270.
Lin, Y.-S., Carey, M. F., Ptashne, M. and Green, M. R. (1988). Cell 54:659—664.
Lin, Y.—S. and Green, M. R. (1991). Cell 64:971-981.
Ma, J. and Ptashne, M. (1987a). Cell 48:847-853.
Ma, J. and Ptashne, M. (1987b). Cell 51:113-119.
Mackem, S. and Roizman, B. (1982). J. Virol. 44:939-949.
Manavalan, P. and Johnson, Jr., W. C. (1987). Anal. Biochem. 167276-85.
Maniatis, T., Goodboum, S. and Fischer, J. A. (1987). Science 236:1237—1244.
Mardsen, H. S., Campbell, M. E. M., Haarr, L., Frame, M. C., Parris, D. S.,
Murphy, M., Hope, R. G., Muller, M. T. and Preston, C. M. (1987). J. Virol.
61:2428-2437.

Matsui, T., Segall, J., Weil, A. P. and Roeder, R. G. J. Biol. Chem. 257:14419-
14427.

McKnight, J. L. C., Kristie, T. M. and Roizman, B. (1987). Proc. Natl. Acad. Sci.
USA 84:7061-7065.

McKnight, S. L. and Kingbury, R. (1982). Science 217:616-324.
McKnight, S. L. and Tjian, R. (1986). Cell 46:795-805.

Mermelstein, F. H., Flores, O. and Reinberg, D. (1989). Biochemica and
Biophysica Acta 1009: 1—10.

Mermod, M., O’Neill, E. A., Kelly, T. J. and Tjian, R. (1989). Cell 58:741—753.
Miller, J., McLachlan, A. D. and Klug, A. (1985). EMBO J. 4:1609—1644.
Mitchell, P. J. and Tjian, R. (1989). Science 245:371-378.

Murre, C, McCaw, P. S., Vaessin, H., Caudy, M., Jan, L. Y. ,,Jan Y. N.,
Cabrera, C., Buskin, D. C., Lassar, A. B. ,Weintraub, H. and Baltimore, D. (1989).
Cell 58:537-544.

Myers, R. M., Tilly, K. and Maniatis, T. (1986). Science 232:613—618.

106

Nakatani, Y., Horikoshi, M., Brenner, M., Yamamoto, T., Besnard, F., Roeder, R.
G. and Freese, E. (1990). Nature 348:86—88.

O’Shea, E. K., Rutkowski, R. and Kim, P. S. (1989). Science 243:538—542.
O’Hare, P. and Goding, R. C. (1988). Cell 52:435-445.
O’Hare, P., Goding, C. R. and Haigh, A. (1988). EMBO J. 7:4231-4238

Ohkuma, Y., Sumimoto, H., Horikoshi, M. and Roeder, R. G. (1990). Proc. Natl.
Acad. Sci. USA 87:9163-9167.

Owen, T. J., Krstenansky, J. L., Yates, M. T. and Mao, S. J. T. (1988). J. Med.
Chem. 31:1009—1011.

Pan, T. and Coleman, J. (1989). Proc. Natl. Acad. Sci. USA 86:3145-3149.
Pan, T. and Coleman, J. (1990). Proc. Natl. Acad. Sci. USA 87:2077—2081.

Pellet, P., McKnight, J. L. C., Jenkins, F. and Roizman, B. (1985). Proc. Natl.
Acad. Sci. USA 82:5870-5874.

Peterson, M. G., Tanese, N., Pugh, B. F. and Tjian, R. (1990). Science
248:1625-1630.

Post, L. E., Mackem, S. and Roizman B. (1981). Cell 24:555—565.

Preston, C. M., Cordingley, M. G. and Stow, N. D. (1984). J. Virol. 50:708-716.
Preston, C. M., Frame, M. C. and Campbell, M. E. M. (1988). Cell 52:425—434.
Ptashne, M. (1986). Nature 322:697-701.

Ptashne, M. (1988). Nature 355:683—689.

Ptashne, M. and Gann, A. A. F. (1990). Nature 346:329-331.

Pugh, B. F. and Tjian, R. (1990). Cell 61:1187—1197.

Reinberg, D. and Roeder, R. G. (1987). J. Biol. Chem. 262:3310—3321.

Reinberg, D., Horikoshi, M. and Roeder, R. G. (1987). J. Biol. Chem.
262:3322-3330.

 

 

 

107

Roeder, R. G. (1976). In RNA Polymerases (Losick, R. and Chamberlin, M., eds.),
pp. 285-300, Cold Spring Harbor, NY.

Roizman, B. and Spector, D. (1991). In Herpesvirus Transcription and its
Regulation (Wagner, E. ed.), pp. 17-28. CRC Press, Boca Raton, FL.

Roizman, B. and Sears, A. E. (1990). In Virology (Fields, B. and Knipe, D. eds.),
pp. 1795-1842. Raven Press, New York.

Rougvie, A. E. and Lis, J. T. (1988). Cell 54:795-804.

Rydel, T. J ., Ravichandran, K. G., Tulinski, A., Bode, W., Huber, R., Roistsch, C.
and Fenton H, J. W. (1990). Science 249:277-280.

Sadowski, 1., Ma, J., Triezenberg, S. J. and Ptashne, M. (1988). Nature.
335:563-564.

 

 

Samuels, M., Fire, A. and Sharp, P. A. (1982). J. Biol. Chem 257:14419-14427.

Sanger, F., Nicklen, S. and Coulson, S. A. (1977). Proc. Natl. Acad. Sci. USA
74:5463-5467.

Sawadogo, M. and Sentenac, A. (1990). Annu. Rev. Biochem. 56:711-754.
Schatz, G. (1987). Eur. J. Biochem. 165:1—6.

Scott, M. P., Tamkun, J. W. and Hartzell HI, G. W. (1989). Biochim. Biophys.
Acta 989:25-48.

Sigler, P. (1988). Nature 333:210-212.
Smale, S. T. and Baltimore, D. (1989). Cell 57:103-113.

Smale, S. T., Schmidt, M. C., Berk, A. J. and Baltimore, D. (1990). Proc. Natl.
Acad. Sci. USA 87:4509-4513.

Spear, P. G. and Roizman, B. (1972). J. Virol. 9:431-439.

Spector, D., Purves, F. and Roizamn, B. (1990). Proc. Natl. Acad. Sci. USA
87:5268-5272.

Stanfield, R. L., Fieser, T. M., Lerner, R. A. and Wilson, I. A. (1990). Science
248:712-719.

Stern, S., Tanaka, M. and Herr, W. (1989). Nature 341:624-630.

108
Stone, S. R., Dennis, S. and Hofsteenge, J. (1989). Biochemistry 28:6857—6863.

Stringer, K. F., Ingles, C. J. and Greenblatt, J. (1990). Nature 344:783-786.

Sumimoto, H., Ohkuma, Y., Yamamoto, T., Horikoshi, M. and Roeder, R. G.
(1990). Proc. Natl. Acad. Sci. USA 87:9158-9162.

Tasset, D., Tora, L., Fromental, C., Scheer, E. and Chambon, P. (1990). Cell
62:1177-1187.

 

Triezenberg, S. J ., Kingsbury, R. C. and McKnight, S. L. (1988a). Genes Dev.
2:718-729.

Triezenberg, S. J ., LaMarco, K. L. and McKnight, S. L. (1988b). Genes Dev.
21730—742.

Trewhella, J ., Liddle, W. K., Heidom, D. B. and Strynadka, N. (1989).
Biochemistry 28:1294-1301.

Vinson, C. R., Sigler, P. B. and McKnight, S. L. (1989). Science 246:911-916.
von Heijne, G. (1985). J. Mol. Biol. 184:99-105.
Voronova, A. and Baltimore, D. (1990). Proc. Natl. Acad. Sci. USA 87:4722-4726.

Wagner, E. K. (1990). In Herpesvirus Transcription and its Regulation (Wagner,
E. K., ed.), pp 1-16. CRC Press, Boca Raton, FL.

Werstuck, G. and Capone, J. P. (1989). Gene 75:213-224.
Werstuck, G. and Capone, J. P. (1990). J. of Virology 63:5509-5513.

Whitley, R. J. (1990). In Virology (Fields, B. and Knipe, D., eds.), pp. 1843-1888.
Raven Press, New York.

Workman, J. L., Taylor, I. C. A. and Kingston, R. E. (1991). Cell 64:533-544.
Yanisch-Perron, C., Vieira, J. and Messing, J. (1985). Gene 33:103-119.
Zhu, Q., Smith, T. F., Lathrop, R. H. and Figge, J. (1990). Proteins 82156-163.

Zoller M. J. and Smith, M. (1982). Nucleic Acids Res. 10:6487-6500.

      

   

CI V.

”Influlfilfllljlfllfllumll

91

 

111i“