MAKING HEADS AND TAILS OF MOLGULA: NEXT GENERATION SEQUENCING
ANALYSIS OF CLOSELY RELATED TAILED AND TAIL-LESS ASCIDIAN SPECIES
By
Elijah Kariem Lowe

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Computer Science - Doctor of Philosophy
2015

ABSTRACT
MAKING HEADS AND TAILS OF MOLGULA: NEXT GENERATION
SEQUENCING ANALYSIS OF CLOSELY RELATED TAILED AND
TAIL-LESS ASCIDIAN SPECIES
By
Elijah Kariem Lowe
Tunicates are invertebrate chordates and are the sister group to the vertebrates. Although tunicates have little morphological resemblance to vertebrates in their adult stage,
they do share several features in their larval stage: a hollow dorsal neural tube, gill slits,
and a post-anal tail, containing a notochord—a group of cells organized in a rod shaped
structure - the key features that classify the phyla. Within the tunicates, several ascidians
have undergone tail-loss, and many of them are Molgulidae. Hybrids have been produced
through the cross fertilization of two Molgula species (Molgula occulta and Molgula oculata),
and no other solitary Molgula species have been known to hybridize. Here we have sequenced
the transcriptomes of several developmental stages of both M. occulta and M. oculata—two
closely related, free-spawning ascidian species, and their hybrid, in order to study the mechanisms behind tail loss in M. occulta. We were first presented with the problem of identifying
the best approach for the de novo assembly of our transcriptomes. Here we determined that
processing reads through digital normalization, a redundancy reduction step, had less of an
e↵ect on assemblies than did the assembler used. We then sequenced and assembled the
genomes of M. occulta, M. oculata and a more distant species, M. occidentalis. This allowed
us to characterize the genomes, discovering that the species are more divergent then they
appear phenotypically, and also supporting better gene models. Through di↵erential expression analysis we determined that M. oculata and the hybrid appear to express many shared

transcripts that are up-regulated during the formation of the ascidian tail, and that these
genes are primarily upregulated in the tailed species and hybrid in relation to the tail-less
species. Furthermore, of those transcripts upregulated at the tailbud stage in the hybrid
but not in M. occulta, it appears that expression is being restored by the M. oculata allele.
This suggests that the relative lack of di↵erential gene expression in the neurula-to-tailbud
transition in M. occulta is due to loss-of-function of cis-regulatory elements controlling the
expression of key genes involved in tail and CNS formation.

ACKNOWLEDGMENTS

Where do I begin? I am a summation of all the people that I have come in contact with
throughout my life. So, I would like to dedicate this text to some of the most influential
people.
I start with my mother Patricia Lowe who always encouraged me to pursue my wildest
dreams. My father William Robertson, who taught me to question the world and to always
pay attention to the small things. My brother Fard Lowe, who consistently served as an
inspiration to me and was someone I could always count on. To my high school counselor,
Kathy Giles-Harris and high school math teacher, Christopher Reese who both saw something
special in me, and pushed me to pursue it.
My brothers from Morehouse—Terron Ferguson, Ryan Shepard, John-Marcus Philips,
Sean Brazier and Kelechi Kalu—who kept the laughs and good times rolling along my journey. You guys always made me feel cool, even though I’m a science nerd.
My MSU family, who are too many to name, but I’ll name a few—my brother and
sister, Cameron Khalfani Herman and Neem Serra, my best bud Ruby Carrillo, Dr. Barbara
Thelamour, my mentor Judi Brown-Clarke, my problem solvers Connie James and Darcie
Zubek, Carmel Martin-Fairey, Daniel Couvrtier, Luis Zaman, Chad Byers, Temple Smith,
James Kremer, okay I said a few, so I’ll stop there. I needed you all along the way, through
late nights of studying, endless laughs, awesome food, much needed pep talks and just being
there to help me survive Michigan.
Collaborators and friends, Billie Swalla, her lab, Lionel Christiaen, his lab, Alberto Stolfi
and the wonderful Princess Claudia Racioppi, who all helped me to develop as biologist,
experimentally and conceptually, and who advised me along the way. I look forward to
iv

working with you all for years to come. And a special thanks to Billie without whom this
project would not exist. We have traveled the world together (well at least a good part of
the US and Europe), and done what I think is some good science together.
My advisor, Titus Brown. I cannot thank you enough for guiding me along the way. I
truly lucked out when I landed in your lab. You are a great scientist, adviser, mentor, and
person, and this is the only time you’ll hear (or more like read) me say this.
My awesome committee—David Arnosti, Jin Chen, Ian Dworkin, and Yanni Sun—who
were always tough, but fair.
And to those who pursue knowledge not for the sake of notability but just to make the
world a better place.

v

TABLE OF CONTENTS

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

ix

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

x

Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

Chapter 2 Literature Review . . . . . . . . . . . . . .
2.1 Ascidian tail development . . . . . . . . . . . . . .
2.2 Notochord development as seen through Brachyury
2.3 Assembling and analyzing data . . . . . . . . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

Chapter 3 Evaluating a lightweight transcriptome assembly pipeline on
two closely related ascidian species . . . . . . . . . . . . . . . . . .
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 Sequencing preparation . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Assembly protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.3 Pre-assembly read trimming and normalization . . . . . . . . . . . .
3.2.4 Transcriptome assembly . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.5 Gene identification . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.6 Read mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1 Digital normalization reduces the resources needed for assembly . . .
3.3.2 Assembly statistics varied by preprocessing approach and assembler .
3.3.3 Trinity assemblies include more low-abundance k-mers than Oases assemblies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.4 Read mapping shows high inclusion of reads in the assembled transcriptomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.5 All assemblies recovered transcripts with high accuracy but varied
completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.6 Both unnormalized and normalized assemblies recovered many of the
same transcripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.7 Homology search against the Ciona proteome shows similar recovery
of ascidian genes across assemblies . . . . . . . . . . . . . . . . . . .
3.3.8 CEGMA analysis shows high recovery of genes . . . . . . . . . . . . .
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vi

4
4
9
12

14
14
16
16
16
17
18
19
19
20
20
21
22
25
27
29
29
32
33

3.4.1

3.5
3.6

Transcriptome assembly accurately recovers known transcripts and
many genes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.2 Digital normalization eases assembly without strongly a↵ecting assembly content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.3 Trinity assemblies are more sensitive to low-abundance k-mers but
contain no new conserved genes . . . . . . . . . . . . . . . . . . . . .
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 4 Genome assembly and characterization . . . . . . . . . . . . . . .
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Genomic DNA library preparation and sequencing . . . . . . . . . . .
4.2.2 Genome sequence assembly . . . . . . . . . . . . . . . . . . . . . . .
4.2.3 Gene identification and alignments . . . . . . . . . . . . . . . . . . .
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 Genome assemblies assessment . . . . . . . . . . . . . . . . . . . . . .
4.3.2 Gene complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.3 Divergence of Gene regulatory networks . . . . . . . . . . . . . . . .
4.4 Expression of the M. occidentalis Mesp gene marks the B7.5 cells . . . . . .
4.5 Divergence of Mesp cis-regulatory sequence function between M. occidentalis
and C. intestinalis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 5 Di↵erential expression analysis of tail loss in an invertebrate
chordate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 Sample collection, sequencing and assembly . . . . . . . . . . . . . .
5.2.2 Gene counts and di↵erential expression analysis . . . . . . . . . . . .
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1 M. occulta and M. oculata have strong overlap in gene expression . .
5.3.2 Notochord gene network . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.3 Di↵erential expression between neurula and tailbud appears to be key
factor in tail development . . . . . . . . . . . . . . . . . . . . . . . .
5.3.4 Overlap between hybrid and M. oculata alleles . . . . . . . . . . . . .
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 6 Conclusions and Discussion . . . . . . . . . . . . . . . . . . . . . .
6.1 Evaluating a lightweight transcriptome assembly pipeline on two closely related ascidian species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Change in gene function and cis-regulatory binding sites . . . . . . . . . . .
6.3 Di↵erential expression analysis of tail loss in an invertebrate chordate . . . .
vii

33
34
35
36
36
37
37
39
39
40
41
42
42
43
44
46
49
50
53

55
55
57
57
59
60
60
62
62
65
66
68
69
69
71
72

6.4

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

73

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix A: Supplemental Figures . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix B: Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

76
77
82

BIBLIOGRAPHY

84

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

viii

LIST OF TABLES

Table 3.1

Table 3.2

Table 3.3

Table 4.1

Table 5.1

Digitally normalized reads. The number of reads sequenced before and after digital normalization is shown for each lane of sequencing. The percentage of total reads kept after digital normalization is
shown in bold. M. occulta had approximately ⇠237 million reads and
was reduced to 91 million reads, a 60% reduction. M. oculata had
150 million reads and was reduced by 77% to ⇠50 million reads. . .

18

Transcriptome metrics. Several metrics used to assess the assembled transcriptomes. The N50, mean transcript length, total number
of transcripts and total number of base pairs are listed for each transcriptomes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

Multiplicity. The k-mer multiplicity shows uniqueness of each assembly. All k-mers with a multiplicity of one are unique. Trinity has
a higher percentage of unique k-mers when comparing assemblers.
The unnormalized Trinity had the highest number of unique k-mers
overall. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

Genome assembly statistics. The contig N50 length, mean contig length, total number of contigs, total number of base pairs and
CEGMA scores were collected for each draft assembly. The CEGMA
score is a metric of completeness measured against highly conserved
eukaryotic genes. Alignments of 70% or greater of the protein length
are called complete (C) and all other statistically significant alignments are called partial (P). . . . . . . . . . . . . . . . . . . . . . .

42

Di↵erential expression: Species vs time . . . . . . . . . . . . . . . .

63

ix

LIST OF FIGURES

Figure 2.1

Figure 2.2

Figure 2.3

Figure 3.1

Phylographic sketch. Phylogenetic placement of Cnidaria and the
Bilateria (Protostomata and Deuterostoma). Deuterostomes, which
in Greek means second mouth, are distinguished in the Bilateria by
forming their anus first. Tunicata and Cephalachordata are the sister
groups to Vertebrata. This subphyla shares several characteristics; a
notochord, a hollow dorsal nerve cord, and a post-anal tail at some
point in their life cycles. (Blue) branches are Deuterostomes. . . . .

6

Notochord cells.The primary notochord cells (red), also known as
the A-lineage, are specified at the 64-cell stage. There are a total
of 32 primary notochord cells that come from the A7.3 and A7.7
blastomere, and the intercalation of the cells happen in a stochastic
manner. The secondary notochord cells (blue) derive from the B8.6
blastomere and are specified at the 110-cell stage, one cell division
after the primary notochord cells. . . . . . . . . . . . . . . . . . . .

6

Notochord/nerve chord precursor at 32 cell stage. At the 32
cell stage the A6.2 and A6.4 (grey) cells of the ascidian embryo are
the notochord/nerve chord precursors. Notochord induction begins
at the 32-cell stage by FGF9/16/20. MAPK then actives brachyury
while repressing FoxB in the notochord precursor cells. MAPK is
repressed by Ephrin in the nerve chord precursors by neighboring
cells. ZicN actives FoxB which represses the expression of brachyury,
this occurs in the nerve chord cells because of the repression of MAPK
by Ephrin. Genes names in black are active and gene names in grey
are inactive. This figure was adopted from Hashimoto et. al [26] . .

9

Wall time and memory requirements for assemblies. Wall
time (a) in hours to complete the diginorm (DN) and raw read (RAW)
assemblies for both species and assemblers. Oases assembled multiple
k’s, 21  k  35 opposed to Trinity that uses only a single k. This is
one reason the assembly times di↵ered. (b) Shows the memory used to
assemble each of the transcriptomes. M. oculata (ocu) transcriptomes
assemble in less time than M. occulta (occ) because they have fewer
lanes of reads to assemble. In all cases diginorm required less time
and memory to complete the assembly. . . . . . . . . . . . . . . . .

21

x

Figure 3.2

Figure 3.3

Figure 3.4

Figure 3.5

K-mer distribution. The k-mer distribution is shown for each assembler and assembly condition, diginorm (DN) and unnormalized
reads. The k-mer distribution is the coverage of a given k-mer verses
how many k-mers of that coverage are incorporated in the respective assemblies. Both Oases and Trinity assemblies are shown for
(3.2a) the M. occulta k-mer distribution and (3.2b) the M. oculata
k-mer distribution. Trinity had a higher k-mer distribution for both
species, reflective of the inclusion of more low abundance reads into
the Trinity assemblies. . . . . . . . . . . . . . . . . . . . . . . . . .

24

Read mapping. Unnormalized reads were mapped back to each of
the assemblies to determine the inclusion of reads in the assembly.
M. occulta first round of gastrulation reads (f+3), showed the lowest
mapping quality for all assemblies, with the lowest being raw Oases
at 48.57%. M. occulta f+3 is the only case where mapping is less than
74% and the only case where DN Trinity mapped more reads than raw
Trinity. M. oculata unnormalized Oases performed the worst, with
the Trinity assembly having the best mappings. Trinity assemblies
have more mapped reads than Oases for all conditions, with at least
93% read mapping for both species. Raw Trinity typically mapped
slightly more reads than DN, and the opposite occurs for Oases, with
DN having more reads mapped to its assembly. Note that the Y axis
starts at 45%. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

Accuracy, completeness and recovery rate against known
Molgula sequences. The NCBI has 178 Molgula sequence in its
database. Transcripts were searched against these sequences using
BLASTN with a cut-o↵ of 1e-12. Trinity assemblies performed the
best, recovering all known sequences. The M. occulta unnormalized
assembly performed the worst, only recovering 79 (44%) of the transcripts. M. occulta tended to recover fewer of the known transcripts
as well. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

Gene recovery, raw reads versus normalized. Gene similarity
with C. intestinalis via BLAST for M. occulta (left) and M. oculata
(right). Each oval represent the total number of putative homologous
sequences recovered. In both species the Trinity assembler assembled
more homologous sequences. There was almost complete overlap in
homology for both assemblers and both assembly conditions. . . . .

31

xi

Figure 4.1

Figure 4.2

Figure 4.3

Adult ascidians. M. occulta (A) and M. oculata (B) are nearly
identical in their adult stage with the white pigment spot (red arrow).
Their tunic is covered in sand, since they are found on the sandy sea
bottoms. Under their sand covered tunic, the two species di↵er by the
color of their eggs—purple in M. oculata, pictured, and an orangeyellowish color in M. occulta—which are found just above the kidney
complex (C). C. intestinalis (D) is the best studied ascidian and has
a well-assembled and annotated genome [14, 102]. . . . . . . . . . .

39

Hox clusters for M. occulta, M. oculata and M. occidentalis
Eight hox genes were found in M. occulta and M. oculata, while nine
were found in M. occidentalis. Hox1, hox2, hox3-4, hox5, hox10 and
hox12-13 were found in all three Molgula species. Hox3-4 was found
on the same contig in all species, with hox12-13 found on the same
contig in M. occidentalis and M. oculata. *M. occulta hox12-13 is
not found on the same contig, but when aligned using mVista, there
is high sequence similarity, showing the possible placement of hox1213 in M. occulta. + M. occidentalis hox2 gene had a stop codon
found in the 3-4 helix. # numbers correspond to gene color, and
rearrangements have been found in Ciona and Molgula. . . . . . . .

45

The B7.5 lineage in M. occidentalis (A) Diagram comparing
M. occidentalis (top) and C. intestinalis (bottom) embryogenesis at
24C. Embryos were stained with Alexa Fluor dye-conjugated phalloidin to visualize cell outlines and DAPI to visualize cell nuclei. (B)
Diagram of mVISTA ([18]; genome.lbl.gov/vista/) alignment of M.
oculata Mesp (Moocul.Mesp) locus to orthologs in M. occulta, M. occidentalis, and C. intestinalis. Shaded peaks indicate sequence conservation above 70% over 100-bp windows (blue = protein-coding, pink
= non-coding). Arrows indicate direction of transcription of proteincoding genes. Non-coding sequences upstream of Mesp are only conserved between M. oculata and M. occulta. M. occidentalis and C.
intestinalis show considerable divergence even in protein-coding sequences. Note that microsynteny with SLC5A-related gene supports
the orthology of these sequences among the Molgulids. . . . . . . . .

47

xii

Figure 4.3

Figure 4.4

(C) In situ hybridization (ISH) for Moocci.Mesp in 110-cell stage embryo (vegetal view), showing mRNA detection (green) in B7.5 blastomeres. Nuclei were counterstained with DAPI (blue). Staging is
given by hours post-fertilization (hpf). (D) Vegetal view of a 110cell stage embryo electroporated with Moocci.Mesp>GFP reporter
construct. Reporter gene expression was detected by ISH for GFP
transcripts (green). Nuclei were stained with DAPI (blue). (E) Lateral view of a mid-tailbud stage embryo electroporated with Moocci.
Mesp>Histone2B::GFP reporter construct. GFP fluorescence reveals
B7.5 descendants on left side of embryo: two trunk ventral cells
(TVCs) and two anterior tail muscles (ATMs). (F) Diagram of B7.5
lineage divisions from 110-cell stage to mid-tailbud stage, inferred
from previous C. intestinalis studies. Cells are named according to
Conklin’s method (Conklin, 1905). The lineage is bilaterally symmetric, but only cells on the left side are indicated and named. Relative staging given for M. occidentalis (Mo.occi) and C. intestinalis
(Ci.inte). 110-cell and late gastrula: vegetal view. Initial tailbud:
dorsal view. Mid-tailbud: lateral view. Anterior pole is on the left in
all images and illustrations. DOI: 10.7554/eLife.03728.003 . . . . . .

48

The B7.5 lineage in M. occidentalis Cross-species reporter construct assays reveal multiple cases of cis-regulatory unintelligibility.
(A) C. intestinalis embryo electroporated with Ciinte.Tbx6-r.b>GFP
reporter construct [9], which drives GFP expression in tail muscles
and the B7.5 lineage cells (arrowheads), recapitulating endogenous
Ciinte.Tbx6-r.b expression. (B) M. occidentalis embryo electroporated with Moocci.Tbx6-r.b>GFP reporter construct, which recapitulates expression in tail muscle cells including B7.5 lineage cells (arrowheads). (C) C. intestinalis embryo electroporated with Moocci.Tbx6r.b>GFP, which drives expression in B-line tail muscle and mesenchyme cells but is excluded from the B7.5 lineage. (D) C. intestinalis embryo electroporated with Ciinte.Hand-r>H2B::GFP reporter (Davidson and Levine, 2003), which drives H2B::GFP expression in anterior endoderm, A7.6 lineage, and TVCs (arrowheads),
recapitulating endogenous Ciinte.Hand-r expression. (E) M. occidentalis embryo electroporated with Moccci.Hand-r>H2B::GFP construct, which recapitulates the same expression pattern. (F) C. intestinalis embryo electroporated with Moocci.Hand-r>H2B::GFP, which
drives expression in endoderm and A7.6 lineage, but not in B7.5. (G)
M. occidentalis embryo electroporated with Ciinte.Ebf neuron-specific
YFP (green) and H2B::mCherry (red) reporter constructs electroporated [114], which drive very weak expression in a limited subset of
motor ganglion neurons (green and red). . . . . . . . . . . . . . . . . 51

xiii

Figure 4.4

(H) C. intestinalis embryo electroporated with a Moocci.Ebf>YFP
reporter, which drives robust reporter gene expression in several brain,
motor ganglion, and nerve cord neurons. (I) C. intestinalis embryos
electroporated with Ciinte.Sox1/2/3 (left) and Moocci.Sox1/2/3 (right)
H2B::mCherry reporter constructs, both of which recapitulate Sox1/2/3
expression in ectoderm. Panels A-F are lateral views of tailbud embryos, panels G is a dorsal view of a tailbud embryo, panel H is a
dorso-lateral view of a hatched larva, and panel I shows vegetal views
of mid-gastrula stage embryos. Anterior is to the right, except in
Panel I, in which anterior is to the top. DOI: 10.7554/eLife.03728.023 52

Figure 5.1

Brachyury gene regulatory network. Bra is a key notochord
inducer, and without its expression neither the notochord nor the
chordate tail forms. Downstream genes have been identified in various
studies and of these 67 genes 11 were missing from the transcriptomes
of both species. The missing genes were cofilin, entactin (nidogen2-like), fibrinogen-like protein (FGL), fibronectin, multidom, noto1,
noto5, noto14, noto16, and tropomyosin. netrin, and noto9 are missing in M. occulta and Klf15 is missing in M. oculata. (Blue) missing in
M. occulta, (Red) missing in M. oculata, and (Purple) missing in both
species. This GRN was built from previous studies [31, 30, 32, 54, 53],
using the BioTapestry software [65]. . . . . . . . . . . . . . . . . . .

Figure 5.2

61

Di↵erential expression of homologous transcripts. Di↵erential
expression in M. occulta for (a) gastrula vs neurula, and (d) neurula
vs tailbud, M. oculata for (b) gastrula vs neurula, and (e) neurula vs
tailbud, and the hybrid for (c) gastrula vs neurula, and (f) neurula
vs tailbud. Genes in red are di↵erentially expressed with a FDR <0.05 64

xiv

Figure 5.3

Figure A.1

Figure A.2

Figure A.3

Figure A.4

Upregulated transcripts overlap between hybrid and M. oculata. When comparing the gastrula and neurula time point for
M. occulta, M. oculata, and their hybrid, both M. oculata and the
hybrid showed di↵erential expression in at least 7% of their transcripts with the majority being up-regulated. (5.3a) There is a 15%
overlap in overexpressed genes M. oculata and the hybrid, for the
transcripts overexpressed when comparing gastrula to neural expression, and there is no overlap with M. occulta. (5.3b) When looking
at the allelic express for the same condition, there is a strong skew
in expression coming from the tailed allele for the unregulated transcripts that overlap between M. oculata and the hybrid. The highest
percentage of transcripts of allelic expression in the hybrid also comes
from M. oculata for up-regulated transcripts that do not overlap with
M. oculata, but not to as great an e↵ect. For genes’ overall allelic expression, the majority comes from M. occulta, 36%, with 32% coming
from M. oculata. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

67

Alignment for hox12-13 in M. occulta, M. oculata and M.
occidentalis The contig containing hox12-13 for M. occidentalis and
M. oculata, along with the two contigs containing hox12 and hox13
for M. occulta. M. oculata was used as the anchor sequence because
it showed the most similarity between the three species. Outside of
the coding regions and its flanking area, there is very little sequence
similarity between the species, and M. occidentalis exclusively shows
similarity in coding regions. Grey arrows show the direction of the
contig. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

77

Alignment of M. occidentalis hox2 genes with Ciona show
premature stop codon. The M. occidentalis hox2 gene has a stop
codon in the 3’ region, inside of the 3/4 helix. hox2 knockdowns in
Ciona did not show any phenotypic di↵erence, so the function of hox2
may not be important in M. occidentalis . . . . . . . . . . . . . . .

78

Alignment of hox genes. Alignments of the aa homeobox sequences from all the Molgula species, show that they group with
their respective orthologs. All but one of the M. occidentalis cluster
properly, but M. occidentalis hox10 full homeobox sequence was not
fully assembled, so this is possibly the reason for poor clustering. .

79

Alignment of M. occidentalis duplicate hox10 genes Two
copies of hox10 were found in M. occidentalis ⇠12 kb apart on the
same contig. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

xv

Figure A.5

Figure A.6

Gel electrophoresis of cdc45, netrin and controls Netrin was
not found in M. occulta’s transcriptome but was recovered via PCR
from a cDNA library. However, this library has a wider range of
developmental stages, so it is still possible netrin is not expressed at
the right stage for tail development. . . . . . . . . . . . . . . . . . .

81

Whole Mount In Situ Hybridization of prickle in M. occulta
The PCP gene pk was shown to be expressed in the notochord in a
similar pattern to C. intestinalis . . . . . . . . . . . . . . . . . . .

81

xvi

Chapter 1
Introduction
The origin of body plans is an age-old question. Chordates are distinguished by their body
plan: most chordates form tadpole-shaped larvae that commonly have a hollow dorsal neural tube and a post-anal tail containing a central notochord flanked by bilateral muscle [86].
During embryogenesis, the notochord serves as the source of a patterning signal for neural
tube and paraxial mesoderm, an addition to the axial skeleton for the larval tail [42, 111].
Tunicates are one of the three subphyla of chordates and are so grouped because of their
outer covering known as a tunic, and during development form the typical tadpole larval
body plan [34]. Ascidians and larvaceans are two of the groups within the Tunicates, and
both develop typical chordate body plans. The ascidians exhibit their tailed body form during their larvae free-swimming stage, using the tail for locomotion before becoming sessile,
undergoing metamorphosis and filter feeding for the remainder of their lives. During the
free-swimming larval stage, the elongation and mobility of the tail is dependent upon the
proper formation of the notochord and muscle cells [98]. In ascidians and in lower vertebrates
the improper formation of the notochord leads to severely shortened larva that cannot swim
or feed properly [16, 43, 111]. Out of ⇠3000 ascidian species, 16 are known to have independently lost their larval tail, di↵erentiated notochord, muscle cell, and other chordate features,
changing their developmental body plan. The majority of these losses have happened within
the Molgula clade [2, 120], although the Styelidae also have two identified tail-less species
[34].
1

Ascidians are a simple system in which to study developmental processes: their cell
lineages have been traced starting at fertilization [84] to gastrulation [85, 82], they have
invariant early cell lineages and a small number of cells [59], and there has been no documentation of ascidians developing without an invariant cell lineage. Ascidian development
is nearly identical across distantly related solitary ascidians [60, 86]. They also have rapid
embryogenesis, compact genomes, and simplified larval body plans containing few larval tissue types [12, 40, 14]. In addition to the several Molgulids that have independently lost
their tail, two Molgulids, one tailed and one tail-less species, can be hybridized, o↵ering the
opportunity to study the mechanism behind evolutionarily divergent body plans [41].
Many genes in the notochord gene network have been identified by subtractive hybridization screening and microarrays [39, 30, 22, 50]. More recently, sequencing technologies such
as Ion Torrent, Roche 454 and Illumina have made genome or transcriptome wide analysis
more readily available for non-model species. These technologies have several advantages
over microarrays: they have a wider scope, are more precise and are able to find novel genes
[71]. With the advances in technology we have now sequenced the transcriptomes of both
species and their hybrid. This ability to sequence the entire transcriptome at a high resolution allows us to look at pivotal time points in tail development and compare across closely
related species to identify and study mechanisms that have been lost or modified during
evolution.
We present a comparative study of the tailed M. oculata and the tail-less M. occulta
through gene expression in order to understand the underlying factors behind tail development, tail loss and the origin of the chordate body plan. Although this study presents
the first assembled Molgula genomes, there are a number of sequenced tunicate genomes
available: in particular, we use the assembled and annotated genome of Ciona intestinalis,
2

which serves as the most documented and closest complete reference for the Molgula and
other ascidian species [14, 98, 100]. We began this project with RNA-seq data from several
time points from each of the species (M. occulta and M. oculata) and their hybrid. Whole
genome and transcriptome sequencing will not definitively identify the factors involved in
the development of the tailed chordate body plan since other processes are occurring, but
examining the correlation of gene expression patterns between the tailed, tail-less and interspecific Molgula hybrid allows us to filter genes associated with this process. Before we
can make biological inferences from our data we have to produce a quality assembly, and
because of this we first assessed the quality of an efficient low-memory assembly pipeline
for our RNA-seq data. We later obtained genomic DNA and assembled the genomes of M.
occulta, M. oculata and a more divergent species M. occidentalis. This allowed us to analyze
the homology between ascidian gene networks, and build more complete transcript modules
for di↵erential expression [133].

3

Chapter 2
Literature Review

2.1

Ascidian tail development

The notochord is one of the distinguishing characteristics of chordates: it is responsible for
the extension of the larval tail which is typical for chordate body plan. In their adult forms,
ascidians and their vertebrate cousins bear little resemblance to one another, however during
development they have similar body plans that include the notochord [40]. Ascidians are
known for their bilateral and invariant cell cleavage, and their development is well described
up to the gastrulation stage [84, 85, 82]. Like vertebrates such as Xenopus, ascidians depend on maternally localized determinants to regulate cell moments and division, while the
development of the early body plans are similar, the location and identity of these determinants are di↵erent in the ascidian and in vertebrates [60]. Solitary ascidian notochords
typically originate from two cell lineages, with the primary notochord deriving from the
“A” blastomere and the secondary notochord deriving from the “B” blastomere [84]. Both
“A” and “B” blastomeres can be identified at the 4-cell embryonic stage. At this stage the
blastomeres are labeled according to the Conklin convention: “a” and “A” label the anterior
animal and vegetal blastomeres, while “b” and “B” label the posterior animal and vegetal
blastomeres [11] . Although the notochords cells have been traced back to the 4-cell stage,
notochord induction does not occur until the 32-cell stage. By the 64-cell stage there are 10
notochord cell precursors, the 8 primary precursor notochord cells—A lineage—which are
4

no longer multipotent, and the 2 secondary notochord cells which are not restricted until
the 110-cell stage [85, 139, 140, 58]. Two additional stages of cell division occur, one at
gastrulation and one at neurulation, ending with the 40 notochord cells that are typical of
most solitary ascidian tadpole larvae [11]. At the onset of neurulation the notochord begins
to form, starting with the closing of the neural tube and posterior movement of the notochord and muscle cells, followed by the mediolateral convergence of the notochord cells to
the midline and then the polarization and intercalation of the cells through a process known
as convergence and extension [117]. At this point the larval tail is constructed of a notochord
flanked by 3 rows of muscles on each side, and both notochord and muscle cell derive from the
same blastomeres [85]. While the arrangement of the notochord cells is a stochastic process,
the anterior 32-cells—primary notochord cells—are always formed by the A7.3 and A7.7
blastomere and the posterior most 8—secondary—notochord cells are always formed by the
B8.6 blastomere; however, the ordering of the 32 most anterior is not determinate, as cells
from both the A7.3 and A7.7 intercalate in a random order (Figure 2.2)[84, 85, 78, 117, 51].
This process, along with muscle cell development, drives the formation of the larval tail
[78, 39, 117].
The ascidian tail is used for dispersal during the free-swimming larval stage, where the
ascidian locates a substrate to attach. After becoming sessile, the ascidian tail is absorbed
into the trunk region as the larvae metamorphose into the adult, the form in which they
filter feed for the duration of their lives. Although a tailed larvae is typical of most ascidians, several species within the Stolidobranchia order have individually undergone tailloss, and many of these species fall in the family Molgulidae [2, 42, 34, 69]. Species without tails tend to have lower speciation rates and smaller geographical ranges [68]. The
tail-less—anural—species develop in a similar manner and are indistinguishable from their
5

Vertebrata

Tunicata

Cephalochordata

Echinodermata

Protostomata

Cnidaria

Deuterostomes

Figure 2.1: Phylographic sketch. Phylogenetic placement of Cnidaria and the Bilateria
(Protostomata and Deuterostoma). Deuterostomes, which in Greek means second mouth,
are distinguished in the Bilateria by forming their anus first. Tunicata and Cephalachordata
are the sister groups to Vertebrata. This subphyla shares several characteristics; a notochord,
a hollow dorsal nerve cord, and a post-anal tail at some point in their life cycles. (Blue)
branches are Deuterostomes.

A7.3+A7.7

B8.6

Figure 2.2: Notochord cells.The primary notochord cells (red), also known as the Alineage, are specified at the 64-cell stage. There are a total of 32 primary notochord cells
that come from the A7.3 and A7.7 blastomere, and the intercalation of the cells happen in
a stochastic manner. The secondary notochord cells (blue) derive from the B8.6 blastomere
and are specified at the 110-cell stage, one cell division after the primary notochord cells.

6

tailed—urodele—counterparts up to late gastrulation [2, 120, 39]. When studying other
direct and indirect developers, such as the sea urchins Heliocidaris erythrogramma and Heliocidaris tuberculata, the body plans diverge at an earlier developmental stage. H. erythrogramma, a direct developer, forgoes the typical pluteus larval body plan and in doing
so changes its developmental plan [27]. In H. erythrogramma and other direct developers,
blastomeres are equivalently sized, cleavage orientation remains parallel and cell division is
synchronous. In contrast, after the 4 cell division an unequal three-tiered cleavage is established in H. tuberculata and other indirect developers along with the loss of synchronous
cell division. Cell fates also di↵ers between direct and indirect sea urchin species [138]. The
change in body plan in Molgula is a much more recent evolutionary occurrence in comparison
to Heliocidaris, and the comparison gives insight into early onsite of changing body plans,
and may demonstrate alternative methods to the changing of early cell fate. Additionally
studying the Molgula adds to the understanding of the ancestral chordate body plan.
Anural ascidians lack several urodele features including an intercalated and extended
notochord, di↵erentiated muscle cells and the otolith sensory organ. The absence of di↵erentiated muscles cells and intercalated notochord are the likely cause of tail-lessness in these
species [78, 120]. The development of several tail-less species has been studied in some detail. Molgula tectiformis notochord cells do not divide again after the 10 precursor cells are
formed and M. occulta stops dividing after 20 cells [42]. The same occurs in M. bleizi, however after the 20 notochord cells are formed, the embryo attempts to make a tail but never
completes the process [123]. It has also been shown that chordate embryos without fully
developed notochord and/or muscle cells do not fully elongate or fail completely to develop
a tail [42, 125, 111]. Vertebrates, cephalochordates and most ascidians have tailed larvae
and interspecific Molgula occulta/oculata hybrids can restore the urodele features, evidence
7

that the ancestral chordate had a tailed larvae and that the mechanism for tail development
was present in anural ascidians but was lost over evolutionary time [2, 39].
In order to study specific mechanisms of tail loss, we can study closely related anural and
urodele species. One such pair of species, M. oculata and M. occulta, both of the Roscovita
clade, have been shown to produce hybrids in lab conditions. Of the known Molgula species
M. occulta and M. oculata are the only two that can hybridize. Although M. occulta and
M. oculata have been found to dwell in the same habitat, hybrids have not been found in
nature and have only been produced in lab conditions. Fertilizing M. oculata eggs with
M. occulta sperm in most cases produce embryos with fully formed tails. The reciprocal
cross (M. occulta eggs X M. oculata sperm) produces a hybrid embryo with 20 notochord
cells like M. occulta, however the notochord cells converge and extend like M. oculata [120].
The ascidian tail has been shown to form in the presence of notochord and the absence of
muscles cells [78]. Tail development is similar in short-tailed hybrids, the notochord is not
flanked by muscle as in tailed species and the tail is only as long as the notochord [123].
Hybrid embryos that develop urodele features are batch specific, and tails develop only in
batches of M. occulta eggs that express the p58 protein which is associated with cytoskeleton
[119, 39]. Additionally, in hybrid embryos in which urodele features are restored, the number
of cells that express acetylcholinesterase (AChE) in a vestigial muscle cell lineage increased
in comparison to hybrids lacking urodele features and M. occulta [41]. This, along with
evidence that the ancestral notochord—the axochord—is muscle based [57], suggests the
need for both notochord and muscles cell lineages for the formation of the ascidian tail.

8

Figure 2.3: Notochord/nerve chord precursor at 32 cell stage. At the 32 cell stage
the A6.2 and A6.4 (grey) cells of the ascidian embryo are the notochord/nerve chord precursors. Notochord induction begins at the 32-cell stage by FGF9/16/20. MAPK then actives
brachyury while repressing FoxB in the notochord precursor cells. MAPK is repressed by
Ephrin in the nerve chord precursors by neighboring cells. ZicN actives FoxB which represses
the expression of brachyury, this occurs in the nerve chord cells because of the repression of
MAPK by Ephrin. Genes names in black are active and gene names in grey are inactive.
This figure was adopted from Hashimoto et. al [26]

2.2

Notochord development as seen through Brachyury

Brachyury, a T-box transcription factor, has been identified as essential for notochord development [140]. Bra was first discovered in mouse, where heterozygotes develop with shorten
tails, and homozygotes fail to form a tail and die in utero [28]. The notochord is induced
by bra in both vertebrates and ascidians, with consistent timing and expression of bra in the
notochord and mesoderm of mouse, xenopus, zebrafish and chicken [48], with bra expressed
exclusively in the notochord cells in ascidians [139]. Notochord induction is regulated by
the FGF/MAPK/Ets signaling cascade (Figure 2.3) [77]. In particular, the A6.2 and A6.4
notochord/nerve cord precursors are induced by FGF9/16/20 at the 32-cell stage, just after
the 7th cell cleavage [97]. It was observed from isolation experiments that notochord/nerve
cord precursors that lose FGF9/16/20 competence at the 32-cell stage assume the default
nerve cord cell fate, and the converse occurs for presumptive nerve cord blastomeres that are

9

introduce to FGF, they forgo their default nerve cord fate and become notochord [140, 77].
FGF9/16/20 activates MAPK which induces bra and represses FoxB [26]. FoxB represses
the activation of bra predominantly through the binding of Fox BS1. Without the repression
of FoxB TF, the notochord cell fate is not induced. FoxB is activated by ZicN and is present
in both nerve cord and notochord precursors, however FoxB is repressed by MAPK in the
notochord cell lineage at the 64-cell stage [26]. Ephrin ligand is expressed in the epidermis
and signals to the future nerve chord cells, inhibiting FGF/MAPK pathway. Notochord
cells, not being in contact with the epidermis, are free to activate FGF/MAPK, and activate
bra.
Although bra is necessary, its presence does not guarantee a tail. M. occulta and M.
tectiformis, two tailless Molgula, both express bra. In both cases bra expression stops earlier
than that of M. oculata, while notochord development is slightly di↵erent between the two
species. Bra is expressed in the 10 precursor notochord cells in M. occulta, which undergo
another round of cell division, while this final division does not occur in M. tectiformis. In
both M. occulta and M. tectiformis, larva-specific muscle actin genes have become pseudogenes, however the mutation in the muscle actin genes are not the same between the two
species [123, 42]. Manx, which encodes for a zinc-finger protein, is another gene involved
in tail development. Manx was identified through subtractive screening, and observed to
be absent or lowly expressed in M. occulta compared to M. oculata [123]. The expression
of manx was restored in the interspecific hybrids with urodele features, and demonstrated
to be necessary, by abolishing the zygotic expression through antisense phosphorothioate
oligodeoxynucleotides, which abolished the urodele features [121, 122].
After cell specification, the notochord cells must converge, intercalate and extend. The
Planar Cell Polarity (PCP) pathway is involved in cell movement during this process and
10

mutations in prickle—a known PCP gene—have been shown to cause a shortened ascidian
tail by a↵ecting both the mediolateral intercalation and the elongation of the ascidian tail
[43]. The pk mutant aimless produces a truncated tail, however the polarity of the nuclei
is established, showing that prickle does not establish polarity within the cell but polarity
between cells, acting in a local manner and perhaps as a global organizer [43, 51]. However,
even in the absence of the PCP pathway considerable convergence and elongation of the
notochord was observed in Ciona, driven by a presumed boundary e↵ect [132].
Many of the upstream genes and transcription factors that interact with bra have been
studied in detail, through knock-outs and cell isolation experiments. A larger scale subtractive screening was done to identify genes downstream of bra, in which 39 genes were initially
found [31]. A number of these genes have been characterized, identifying functions such as
extracellular matrix components (cadherin 8, entactin, fibronectin, laminin alpha1, alpha4,
and beta1, and thrombospondin), cell shape and polarity (pk, trop, ERM, ACL), and axon
guidance (netrin, semaphorin 3A), amongst a host of other biological processes [30, 32, 54].
Additionally, downstream genes regulated by bra have been examined by using ChIP-seq to
identify many known genes in the network, as well as to discover new genes [52, 45].
Larvaceans are pelagic tunicates that also develop in a typical chordate manner, featuring
a notochord. However, larvaceans notochords contain only 20 cells [106, 15]. The larvacean
Oikopleura dioica retains its tail during its adult life stage and at this point bra is not
expressed in the adult notochord, however, bra is expressed in the same manner in the
developing larval notochord as ascidians [1, 83]. When comparing gene networks for the
extent of variation, Oikopleura did not exhibit the same mechanism for tail development as
Ciona: of the 50 bra target genes previously identified in Ciona, only 26 of them had orthologs
in the Oikopleura genome, meaning that almost 50% of candidate bra target genes are not
11

present [53]. Of the genes that did show homology, expression ranged from notochord-specific
to tail-general, including expression in possible notochord to tissues that were clearly not the
notochord. From this we can infer that additional genes have gained function in notochord
formation in the C. intestinalis gene regulatory network, and the ancestral tunicate may
have had a small core set of genes for notochord formation.

2.3

Assembling and analyzing data

One of the major advances in biological sciences in the past 20 years has been the implementation of sequencing technologies. These technologies allow us to examine biological systems
genomically, with increasing ease. The first widespread sequencing technology was Sanger
sequencing in the 1986, but Sanger sequencing was not broadly used until 10 years later,
when automated sequencers became available. Another technology, microarrays, which became popular starting in the mid ’90s, allowed us to look at a wide spectrum of genes and
understand relative expression within a sample. For example, Kobayashi et al. [50] isolated
and analyzed gene expression in notochord (A7.3+A7.7) and nerve cord (A7.4+A7.8) precursors using microarrays. This study was able to identify 106 genes expressed in the notochord
precursor and 68 expressed in the nerve cord precursor at the 64-cell stage. Of these the
genes, 36 notochord genes and 25 nerve chord genes were confirmed via whole mount in situ
hybridization in the respective cells. This demonstrates the power of this technique, however,
prior knowledge of the gene sequences involved was needed. C. intestinalis was sequenced
using Sanger sequencing, and is the best assembled and annotated ascidian genome [14]. In
addition to long (Sanger) reads, sca↵olding was done using sca↵old-joining guided by pairedend expressed sequence tags and bacterial artificial chromosome (BAC) sequences, and BAC

12

chromosomal in situ hybridization data [102]. Sanger sequencing is able to sequence whole
genomes without the need of prior knowledge to identify novel genes but was costly and time
consuming[76, 63].
Sanger was the first generation of sequencing technologies, and currently both second and
third generation are in use, with Roche 454, Ion Torrent, Illumina and PacBio being the most
widespread. These technologies produce data far more easily and at a much lower cost than
Sanger sequencing [76]. There are many trade-o↵s for each of the technologies, including cost
per MB, sequencing time, prep cost, error rate and sequencing bias; in particular, 454 and
PacBio have longer reads than Illumina and Ion Torrent, 800 bp and 1+kbp, respectively.
However, both Illumina and Ion Torrent’s short reads are cheaper to generate, produce more
reads and are better for digital counting; in addition, PacBio has a very high error rate [20].
Illumina and Ion Torrent have the lowest error rates and while Ion Torrent is more sensitive
to single nucleotide polymorphisms, it also calls many more false positives. Illumina has
become the most used NGS technology because it is the most versatile and performs the best
in general [93]. The associated drop in sequencing price has yielded many of the assembled
genomes within the Tunicata phyla. Outside of this project there are ten tunicate genomes
assembled; C. intestinalis (species “Type A” and “Type B”), C. savignyi, Oikopleura dioica,
Botryllus schlosseri, Halocynthia aurantium, H. roretzi, Phallusia fumigata, P. mammilata,
and Didemnum vexillum, but no Molgula genomes. Molgula genomes and transcriptomes,
specifically the tailed M. oculata and the tail-less M. occulta would make great systems to
study the development of the chordate tail and chordate body plan.

13

Chapter 3
Evaluating a lightweight
transcriptome assembly pipeline on
two closely related ascidian species

3.1

Introduction

Next generation sequencing (NGS) has allowed us to study organisms with a broader lens,
looking at entire genomes and transcriptomes instead of single genes. This capability is particularly important for non-model organisms where little prior knowledge may be available,
and where NGS readily enables whole-transcriptome analyses [137], allowing us to study
organisms that are ecologically or evolutionarily interesting.
There are now several sequencing technologies, Illumina being one of the most versatile
[20], that can produce millions of short reads ranging from 75 to 150 bp in length at a low
cost [142]. As sequencing costs continue to drop, transcriptomes from multiple developmental
stages of non-model organisms can easily be sequenced. Various types of de novo assembly
algorithms and reference based assembly approaches have been developed to handle this
massive influx of transcriptomic data [92, 134, 110]. It has been shown in some cases that
0 This

chapter is in review at PeerJ: Lowe, Swalla, and Brown, 2014 (http://dx.doi.org/10.
7287/peerj.preprints.505v1).

14

mapping mRNA-seq reads to a reference genome yields better transcriptomes than de novo
assemblies, even if the genome is 5-15% divergent [133]. However, with many non-model
organisms, no closely related reference genome is available.
De novo is the only solution for transcriptome assembly of organisms with no evolutionarily close reference genome. Transcriptome assemblers such as Trinity [21] and Velvet/Oases
[141, 103] use De Bruijn-graph based de novo approaches which build graphs connecting
the reads based on k-mer overlap. These graphs are then traversed via an Eulerian path
algorithm to assemble transcripts. Because De Bruijn graphs are based on exact matches between DNA words, increasing numbers of sequencing errors result in an exponential number
of new paths, adding to the complexity of the graph and, in turn, increasing the assembly
time and memory requirements [92]. This is both time consuming and limits the ability to
assembly a transcriptome by the amount of available resources.
Here we have sequenced the transcriptomes of several developmental stages of Molgula
occulta and Molgula oculata—two closely related, free-spawning ascidian species, with no
available reference genome. Ciona intestinalis and Ciona savignyi are the closest related
ascidian species with well-assembled genomes, but are not close enough to use as a nucleotide
reference for transcriptome construction. In this paper, we describe an efficient, easy to follow
protocol for the transcriptome assembly of two Molgulid developmental transcriptomes. A
crucial part of this protocol is the use of a preprocessing step that normalizes read abundances
prior to assembly, called “digital normalization.” We study the e↵ect of digital normalization
on assemblies performed with both Trinity and Velvet/Oases. We compare our approach
to the results of running Trinity and Velvet/Oases without digitally normalized reads and
show that our approach recovers 99% the same gene content but has significantly reduced
requirements for time and memory. This reduction in time and memory lets us assemble
15

transcriptomes efficiently using cloud resources, making our results exceptionally easy to
reproduce [23], and more broadly enabling transcriptome assembly by researchers without
access to large computer resources.

3.2
3.2.1

Methods
Sequencing preparation

M. occulta and M. oculata were collected by dredging o↵ the shores of Rosco↵, France
near La Station Biologique. Swalla et al have previously described the maintenance [120]
and culturing [122] of the animals. The transcriptomes of M. occulta and M. oculata were
sequenced at Michigan State University (MSU) in the Research Technology Support Facility
on Illumina HiSeq 2000. Five lanes of sequences were generated for M. occulta, two lanes of
the gastrula stage (F+3), one of neurula (F+4), one of early tailbud (F+5), and one from the
tailbud (F+6) stage (Table 1). Three lanes of sequences were generated for M. oculata, one
each for the gastrula, neurula and tailbud stage. 10µg of RNA were sequenced for each stage
with the exception of M. occulta F+4, where 1.05µg of RNA was sequenced. On average
each embryonic stage yielded 48 million reads of 75 base pairs (bp) in length with paired-end
insert lengths of 250 bp. All reads can be found in the NCBI short read archive (SRA) under
accession number SRP040134.

3.2.2

Assembly protocol

Below is an overview of the steps used for the de novo assembly and annotation of our
transcriptomes.

16

1. Quality trimming and filtering of raw reads.
2. Apply digital normalization to decrease data size.
3. Assemble transcriptome.
4. Assess transcriptome quality.
5. BLAST (gene recovery/identification).
Scripts used to run these steps can be found in the following GitHub repository: https:
//github.com/ged-lab/2014-mrnaseq-cloud

3.2.3

Pre-assembly read trimming and normalization

Low quality bases were trimmed and low quality reads were removed using quality-trimpe.py found in the scripts directory of the repository. A hard trim is done at a Phred quality
score of 33 and reads less than 30 base pairs in length are discarded. This process creates a
paired and singleton fastq file for each library because of the removal of low quality reads.
The filtering of reads allows for better assembly and better mapping, although it may also
reduce sensitivity to low-expressed transcripts [64, 67]. The reads were initially 75 bp long,
and the average base pair (bp) length was 63 bp after quality trimming and filtering. After
quality trimming reads were either directly assembled, or first preprocessed with digital
normalization and then assembled.
Digital normalization (diginorm) is a technique that down samples reads from highly
abundant transcripts while retaining approximately the full sequence information content of
the reads [5]. Here, for each species, reads from all stages were normalized together to build
a common reference transcriptome; reads were normalized to a k-mer coverage of 20 with the
17

k-mer size set to 20 as well. The initial data set from M. occulta contained 237 million reads
from 5 lanes, and M. oculata contained 150 million total reads; after digital normalization,
the M. occulta dataset was reduced to 91.6 million reads and M. oculata was reduced to 50
million reads, a 60% and 77% reduction respectively (Table 3.1).
Sample
M. occulta F+3
M. occulta F+3.2
M. occulta F+4
M. occulta F+5
M. occulta F+6
M. occulta Total
M. oculata F+3
M. oculata F+4
M. oculata F+6
M. oculata Total

Table 1: Read counts
Number of reads Reads kept Percentage kept
42,174,510
50,018,302
44,948,983
53,692,296
45,782,981
236,617,072
91,316,419
38.6%
47,045,433
52,890,938
50,156,895
150,093,266
49,957,980
33.3%

Accession Number
SRR1197985
SRR1197986
SRR1199464
SRR1199259
SRR1199268
SRR1197522
SRR1197965
SRR1197972

Table 3.1: Digitally normalized reads. The number of reads sequenced before and after
digital normalization is shown for each lane of sequencing. The percentage of total reads
kept after digital normalization is shown in bold. M. occulta had approximately ⇠237 million
reads and was reduced to 91 million reads, a 60% reduction. M. oculata had 150 million
reads and was reduced by 77% to ⇠50 million reads.

3.2.4

Transcriptome assembly

We used the Trinity (r20140413p1) and Velvet/Oases (v1.2.08/v0.2.08) assembler packages,
both of which have been broadly used on other data sets [133, 21, 103]. Velvet was initially
developed to assemble genomes, and the Oases add-on package was developed for transcriptome assembly, since transcriptomes have variable coverage and many isoforms. Since Oases
cannot be run without Velvet, we refer below to transcriptomes assembled with Velvet and
Oases as Oases assemblies. Unlike Trinity, Oases requires the choice of a k-mer overlap for
assembly; we chose several k values ranging from k = 21 to k = 35, for odd values of k, with
sca↵olding turned o↵. After assembly, the Oases transcriptomes with the highest number

18

of blast hits to C. intestinalis were selected for further analysis. The Trinity assembler was
run with default parameters.
All assemblies were performed on the Michigan State University (MSU) High Performance
computing cluster (HPCC). All diginorm assemblies were repeated on Amazon EC2 machines
as a proof of concept. After assembly, transcripts shorter than 200 bp in length were removed,
and CD-HIT was used to eliminate small transcripts with 99% identity to longer transcripts
using the following command: “cd-hit-est -i <transcript file>-c 0.99 -o <output file>” [62].
To choose the best k-mer parameter for the Oases assemblies, C. intestinalis proteins
were searched with TBLASTN (e-value cuto↵ of 1e-6) against each Oases assembly and the
transcriptome with the most hits was selected for further analysis.

3.2.5

Gene identification

We used standalone BLAST to find reciprocal best hits (RBH) between the eight assembled
transcriptomes and the C. intestinalis proteome retrieved from NCBI under search term
“(ciona intestinalis) AND Ciona intestinalis [porgn: txid7719]”. At the time of retrieval
there were 16,123 sequences and they were downloaded and stored in the GitHub repository
under the file name “ciona transcriptome.fa” in case the sequences change on NCBI. An
e-value cuto↵ of 1e-6 was used as a minimum threshold for transcript identity. The findreciprocal-2.py script was used to identify the RBH.

3.2.6

Read mapping

To determine the inclusion of reads in the various transcriptome assemblies trimmed reads
were mapped to their respective species using bowtie2 v2.2.1 [56]. For both unnormalized

19

read and diginorm assemblies the full set of trimmed reads were used for mapping. Default
parameters were used, and both paired ends and singletons were mapped. Samtools v0.1.19
[61] was used for format conversion from SAM to BAM format, and also to calculate the
percentage of mapped reads. The BAM files were also used to calculate the coverage of
transcripts.

3.3
3.3.1

Results
Digital normalization reduces the resources needed for assembly

The M. oculata unnormalized read data set assembled with Oases used 44 CPU hours and
85 GB of RAM. The Oases assembly done with the digitally normalized reads took ⇠22 CPU
hours and 21 GB of RAM (Figure 3.1a); this includes the time and memory required to run
the digital normalization pipeline. M. occulta diginorm Oases assembly required over 100
GB of RAM, and the raw read Oases used 300 GB of RAM. The raw read Oases assemblies
for both species took twice as long and needed at least three times as much memory when
compared to the diginorm reads.
The di↵erence in assembly time and memory between diginorm and raw reads was not
as large when using the Trinity assembler. Diginorm completed its assemblies several hours
faster than assembling raw reads, ⇠15 hours compared to ⇠26 hours for M. oculata and
⇠24

hours compared to ⇠39 hours for M. occulta. M. oculata unnormalized reads did not

require much more memory than the normalized reads—16.8 GB and 15.65 GB, respectively.
Diginorm had a larger e↵ect on M. occulta, assembling M. occulta normalized reads with 23.17

20

A to
Walltime
to complete
assemblies RAM needed
RAMtoneeded
to assemblies
complete assemblies
A Walltime
complete
assemblies
complete
occ
oases ocu
occoases
trinity
occ trinity

ocu trinity
oases
ocu

ocu trinity

300

50

Total memory used (GB)

75

occtrinity
oases ocu
occoases
trinity ocu
ocutrinity
oases
occ

ocu trinity

DN RAW
RAW DN
DN RAW
RAW DN
DNRAW
RAW
DN

DN RAW

300

100

Total walltime (hrs)

Total walltime (hrs)

100

occ oases

75

200

50

100

25

25

0

0
DN RAW

0

DN RAW
RAW
DN

DN RAW
RAW
DN

DN RAW
RAW
DN

DN RAW

Sample Sample

Total memory used (GB)

125

occ oases
125

200

100

0
DN RAW

Sample Sample

(a) Time to complete assemblies

(b) Memory to complete assemblies

Figure 3.1: Wall time and memory requirements for assemblies. Wall time (a) in
hours to complete the diginorm (DN) and raw read (RAW) assemblies for both species and
assemblers. Oases assembled multiple k’s, 21  k  35 opposed to Trinity that uses only
a single k. This is one reason the assembly times di↵ered. (b) Shows the memory used
to assemble each of the transcriptomes. M. oculata (ocu) transcriptomes assemble in less
time than M. occulta (occ) because they have fewer lanes of reads to assemble. In all cases
diginorm required less time and memory to complete the assembly.
GB of RAM versus 34.14 GB of RAM for the unnormalized reads (Figure 3.1b).

3.3.2

Assembly statistics varied by preprocessing approach and
assembler

Oases run with the diginormed reads yielded fewer total transcripts than Oases run with
the unnormalized reads. The M. oculata diginorm assembly produced 300 fewer transcripts,
and the M. occulta diginorm assembly produced 227 fewer transcripts (Table 3.2). Digital
21

Species
M. occulta
M. occulta
M. occulta
M. occulta
M. oculata
M. oculata
M. oculata
M. oculata

Method
DN Oases
Oases
DN Trinity
Trinity
DN Oases
Oases
DN Trinity
Trinity

N50
14,606
14,492
14,738
12,300
7,274
7,158
10,141
8,018

Table 2: Assembly Statistics
Mean transcripts length Total number of transcripts
888
89,465
912
89,692
978
96,287
914
87,090
1,478
39,438
1,380
39,738
1,450
57,105
1,275
49,265

Total number of base pairs
79,447,700
81,824,388
94,200,549
79,672,435
58,291,461
54,869,493
82,856,337
62,817,433

Table 3.2: Transcriptome metrics. Several metrics used to assess the assembled transcriptomes. The N50, mean transcript length, total number of transcripts and total number
of base pairs are listed for each transcriptomes.
normalization had the opposite a↵ect when using Trinity for assembly, increasing the total
number of assembled transcripts by 7,840 for M. oculata and 9,197 for M. occulta.
Trinity produces 6.8k (7.6%) more transcripts than Oases for M. occulta using the digitally normalized reads, and a 2.6k (2.9%) decrease in the number of transcripts using the
unnormalized reads. Trinity assembled more transcripts for both M. oculata assemblies, a
17.6k (44.8%) increase for diginorm and a 9.5k (24%) increase for the raw reads.

3.3.3

Trinity assemblies include more low-abundance k-mers than
Oases assemblies

We next examined the k-mer spectrum of the assembled transcripts using k-mer abundances
from the digitally normalized reads. The k-mer
spectrum is an account of the information
1
content of the reads and can be used to evaluate the ability of the assemblers to recover
low-abundance transcripts [92]. We first used digital normalization to reduce the reads to
a median k-mer coverage of 20, so that the k-mer frequency spectrum peaked at a coverage
of 20, and then plotted a cumulative abundance plot of those k-mers shared between the
normalized reads and the assemblies. The results, displayed in Figure 2, show that Trinity
recovers more low-abundance k-mers. Also note that between assemblies done with the same
22

assemblers, the k-mer distributions were very similar, suggesting that the k-mer spectrum
is reflective of the underlying graph traversal algorithm used by the assembler. In addition
the Trinity assemblies included more unique k-mers (Figure 3.3)

23

M. oculata distribution

0.4

0.6

0.8

DN Oases
Oases
DN Trinity
Trinity

0.0

0.0

0.2

Cumulative fraction

0.4

0.6

DN Oases
Oases
DN Trinity
Trinity

0.2

abundance

0.8

1.0

1.0

M. occulta distribution

0

5

10

15

20

25

0

30

5

10

15

20

25

30

k−mer count

kmer count

(a) M. occulta

(b) M. oculata

Figure 3.2: K-mer distribution. The k-mer distribution is shown for each assembler and assembly condition, diginorm (DN)
and unnormalized reads. The k-mer distribution is the coverage of a given k-mer verses how many k-mers of that coverage
are incorporated in the respective assemblies. Both Oases and Trinity assemblies are shown for (3.2a) the M. occulta k-mer
distribution and (3.2b) the M. oculata k-mer distribution. Trinity had a higher k-mer distribution for both species, reflective of
the inclusion of more low abundance reads into the Trinity assemblies.

24

Species
M. occulta
M. occulta
M. occulta
M. occulta
M. oculata
M. oculata
M. oculata
M. oculata

Table 3: K-mer multiplicity
Method
n=1 n=2
DN Oases
60.7
18.4
Oases
60.3
17.4
DN Trinity
68.5
17.5
Trinity
73.5
16
DN Oases
65
17.7
Oases
67.1
16.4
DN Trinity
66.1
17.3
Trinity
74.2
15

n 3
20.9
22.3
14
10.5
17.3
16.5
16.6
10.8

Table 3.3: Multiplicity. The k-mer multiplicity shows uniqueness of each assembly. All
k-mers with a multiplicity of one are unique. Trinity has a higher percentage of unique
k-mers when comparing assemblers. The unnormalized Trinity had the highest number of
unique k-mers overall.

3.3.4

Read mapping shows high inclusion of reads in the assembled transcriptomes

We mapped the quality-filtered reads to the assembled transcriptomes to evaluate their
inclusiveness. The F+3 stage of reads from M. occulta had the lowest percentage of mapped
reads, with the Oases unnormalized assembly mapping only 49% of the reads, and the
Trinity unnormalized assembly mapping 67% (Figure 3.3a). This was an isolated case: all
other Oases assemblies contained at least 75% of the reads for each time point and the
Trinity assemblies contained at least 93% of the reads for each time point. Trinity raw read
assemblies tended to contain slightly more reads than the diginorm assemblies, while the
opposite was true for Oases; however, in no case did the mapping of raw-reads assembly
di↵er from the diginorm assemblies in more than 3% of their read content.

1

25

Percentage of reads mapped back to transcriptome
100

f+3

f+3.2

f+4

f+5

Percentage of reads mapped back to transcriptome

f+6

100

f+4

f+6

90

Reads Mapped (%)

Reads Mapped (%)

90

f+3

80
Assembly method
Trinity
DN Trinity
Oases
DN Oases

70

80

70

60

60

50

50

M. occulta hours post fertilization

Assembly method
Trinity
DN Trinity
Oases
DN Oases

M. oculata hours post fertilization

(a) M. occulta

(b) M. oculata

Figure 3.3: Read mapping. Unnormalized reads were mapped back to each of the assemblies to determine the inclusion of
reads in the assembly. M. occulta first round of gastrulation reads (f+3), showed the lowest mapping quality for all assemblies,
with the lowest being raw Oases at 48.57%. M. occulta f+3 is the only case where mapping is less than 74% and the only case
where DN Trinity mapped more reads than raw Trinity. M. oculata unnormalized Oases performed the worst, with the Trinity
assembly having the best mappings. Trinity assemblies have more mapped reads than Oases for all conditions, with at least
93% read mapping for both species. Raw Trinity typically mapped slightly more reads than DN, and the opposite occurs for
Oases, with DN having more reads mapped to its assembly. Note that the Y axis starts at 45%.

26

3.3.5

All assemblies recovered transcripts with high accuracy but
varied completeness

mRNAseq assembly accuracy can be calculated based on known transcripts generated from
longer reads or reference genomes [133, 72]. We use Molgulid nucleotide sequences from
NCBI to measure accuracy, and we define accuracy as the average BLAST identity score
for the best match for each gene recovered [61]. There are 178 sequences from within the
Molgula clade in the NCBI database. With the exception of M. occulta unnormalized Oases
assembly, all assemblies have hits to at least 113 out of these Molgula sequences (Figure
4). The Trinity assemblies for both species have hits to all 178 sequences. Oases assemblies
have hits for more sequences using digital normalized reads, two additional hits for M.
oculata and 40 additional hits for M. occulta. M. oculata assemblies hits have high average
accuracy in the 90 and 99 percentile for Oases and Trinity, respectively. Completeness is
the percentage of a gene, transcript or protein that is recovered. Within the M. oculata
assemblies, the unnormalized Oases assembly has the lowest average completeness at 36%,
the Trinity assemblies round out at 60% and the digital normalized Oases assembly has
the highest average completeness at 72%. (Note that many of the Molgula sequences are
genomic, which includes intronic regions, so we would expect this to lower the completeness
scores.)

27

Assembly metrics for M. occulta vs known Molgula nucleotides
accuracy

completeness

Assembly metrics for M. oculata vs known Molgula nucleotides
accuracy

recovery

100

100

75

Assembly method
Trinity
DN Trinity
Oases
DN Oases

50

Percent

75

Percent

completeness

recovery

50

25

25

0

0

(a) M. occulta

Assembly method
Trinity
DN Trinity
Oases
DN Oases

(b) M. oculata

Figure 3.4: Accuracy, completeness and recovery rate against known Molgula sequences. The NCBI has 178 Molgula
sequence in its database. Transcripts were searched against these sequences using BLASTN with a cut-o↵ of 1e-12. Trinity
assemblies performed the best, recovering all known sequences. The M. occulta unnormalized assembly performed the worst,
only recovering 79 (44%) of the transcripts. M. occulta tended to recover fewer of the known transcripts as well.

28

Of these 178 nucleotide sequences, 8 of them are M. occulta sequences and 15 of them
are M. oculata sequences. All M. occulta assemblies recovered all 8 of the NCBI M. occulta
sequences with a 94% or greater accuracy. M. oculata assemblies recovered M. oculata transcripts at a 93% accuracy as well. M. occulta assemblies produced the lowest completeness of
the two species, 41% and 43% for unnormalized Oases and diginorm Oases respectively, and
75% for both Trinity assemblies. M. oculata assemblies produced more complete transcripts
66, 75, 86, and 83 percent for unnormalized Oases, Diginorm Oases, unnormalized Trinity
and Diginorm Trinity respectively.

3.3.6

Both unnormalized and normalized assemblies recovered many
of the same transcripts

We evaluated the two diginorm and unnormalized assemblies against one another to test
whether either method missed significant portions of the transcriptome assembled by the
other. We used BLAT to compare unnormalized and diginorm assemblies in both directions.
In M. occulta, both methods recovered at least 93% of the transcripts, with Trinity diginorm
recovering ⇠99% of Trinity’s unnormalized assembly. M. oculata assemblies showed high
overlap as well, all recovering greater than 98% of each other with the exception of diginorm
Oases recovering 94% of unnormalized Oases assembly.

3.3.7

Homology search against the Ciona proteome shows similar
recovery of ascidian genes across assemblies

We used Ciona intestinalis to evaluate the completeness of our transcriptomes. C. intestinalis has an assembled genome that is well annotated and is the closest available genome
29

to the Molgulids. C. intestinalis has a genome of 160 Mb and contains ⇠16,000 genes [99].
A total of 13, 835 (86%) of the C. intestinalis proteins found in NCBI had hits in the M.
occulta transcriptomes (Figure 5), with 2,288 genes (14%) having no hits due presumably to
either lack of expression, high divergence, or loss M. occulta. When comparing transcripts
excluded by either diginorm or unnormalized reads for all assemblies, the unnormalized read
assemblies produced an additional 0.04% hits to C. intestinalis and there was additional
0.03% for the diginorm assemblies. There was little di↵erence between the assemblies when
compared to C. intestinalis, with 99% of the C. intestinalis genes being found in all M. occulta assemblies (Figure 4a). Eighty-six percent of the C. intestinalis proteins had matches
in the M. occulta and M. oculata assemblies with less than 1% di↵erence in presence between
the several assemblies (Figure 4b).

30

Diginorm Trinity

Trinity

2

1

Diginorm Oases

68

0
2

Oases

37

Trinity

0

2

Diginorm Oases

1

32

Diginorm Trinity

61

0
4

5

10

1

56

13618
2

Oases

11

8

13641
4

0

1

13

1

54

8

(a) M. occulta

(b) M. oculata

0

Figure 3.5: Gene recovery, raw reads versus normalized. Gene similarity with C. intestinalis via BLAST for M. occulta
(left) and M. oculata (right). Each oval represent the total number of putative homologous sequences recovered. In both
species the Trinity assembler assembled more homologous sequences. There was almost complete overlap in homology for both
assemblers and both assembly conditions.

31

We next examined the di↵erence between the unnormalized and digitally normalized assemblies. Transcripts in the unnormalized assembly with BLAST hits to C. intestinalis but
without hits in diginorm assemblies were extracted, and searched using BLASTN against
the diginorm assemblies; we found fragmented versions of these transcripts, suggesting that
they were partially assembled. We then mapped the diginorm reads to the extracted unnormalized transcripts and found that some portions of the transcripts were not covered by the
normalized reads. This demonstrates that these transcripts were lost due to a loss of information from the diginorm process. However, the overall loss was minimal and complemented
by an increase in the recovery of other conserved transcripts; this is clearly a direction for
further study.

3.3.8

CEGMA analysis shows high recovery of genes

CEGMA uses a list of highly conserved eukaryotic proteins to evaluate genome and transcriptome completeness [89]. We used CEGMA to analyze the number of protein families that are
present in each assembly. The default CEGMA parameters were used for analysis. CEGMA
reports recovery as “complete” or “partial”, where a match is marked as “complete” if 70%
or more of the amino acid sequence is recovered. More than 90% of the CEGMA genes were
recovered completely in each of the transcriptome assemblies, while greater than 98% of the
CEGMA genes were recovered at least partially.

32

3.4
3.4.1

Discussion
Transcriptome assembly accurately recovers known transcripts and many genes

All of the transcriptome assemblies yielded homologs for an almost identical subset of
the Ciona intestinalis proteome. While the evolutionary distance between the Molgulids
and C. intestinalis may be large – the Molgulids are stolidobranch ascidians and are believed to be very divergent from C. intestinalis, which is a phlebobranch ascidian [34,
109]—approximately 84% of Ciona proteins were found in all assemblies via BLAST, and
more than 44% of Ciona proteins had putative orthologs in each of our assemblies via reciprocal best hit. Since both transcriptomes are from a limited set of embryonic tissues that do
not express all genes, these are surprisingly high numbers! We infer that we have recovered
almost all embryonic genes and the majority of genes present in the Molgula genomes.
Read mapping and CEGMA analyses further confirm that the transcriptome assemblies
are of high quality and inclusiveness. The assemblies represent 75% or more of the reads
from all but one time point, contain complete matches to 90% or more of the conserved
eukaryotic gene families in CEGMA, and contain partial matches to 98% or more of the
CEGMA families. It is important to note that the CEGMA results are almost certainly
biased upwards by the nature of the CEGMA families, which represent many more metabolic
and cellular function genes than e.g. animal-specific transcription factors; thus the CEGMA
numbers do not directly demonstrate the inclusiveness of the transcriptome families, as they
would for a genome assembly [89].

33

3.4.2

Digital normalization eases assembly without strongly affecting assembly content

One of our goals in this study was explore the impact of digital normalization on the biological
interpretation of transcriptome assemblies; while previous studies have shown that digital
normalization can make assembly faster and less memory intensive, gene recovery has been
less well studied [23, 5]. Here we confirm the computational results: diginorm dramatically
reduces the computational cost of Oases assemblies, and also decreases the time and memory
requirements for Trinity assemblies.
While digital normalization does alter the number of transcripts significantly, it does
not strongly a↵ect either read inclusion or the conserved gene content of the assemblies.
Read inclusion by mapping never decreased more than 3% after digital normalization, and
in many cases increased. The conserved gene content, measured by a proteome comparison,
showed that we recover essentially the same set of proteins with all four treatments on both
transcriptomes.
Combined, these results suggest that the varying number of transcripts largely reflect
di↵erences in the splice variants reported by di↵erent assemblers under di↵erent conditions.
These results also strongly support the idea that preprocessing with digital normalization
does not strongly a↵ect assembly content. We note, however, that the few transcripts not
recovered in assemblies of the digitally normalized reads were probably not recovered because
the underlying reads were eliminated during digital normalization. This is an area where
digital normalization can be improved.
Only a small number (well below 1%) of di↵erent homology matches were reported between the various assemblies. Because of this we decided not to merge or otherwise com-

34

bine the di↵erent assemblies: the likely benefits were outweighed by the risk of introducing
chimeric transcripts or combining isoforms.
We also note that the variation in number of assembled transcripts due to read preprocessing and choice of assembler despite the similar gene content suggests that traditional
genome assembly metrics such as number of transcripts, total bp assembled, and N50 are
not useful for transcriptome evaluation as previously suggested [87]. For example, the same
exon may be included in multiple splice variants, inflating the total bp assembled; some assemblers may choose to report more isoforms than others even with the same read support;
and N50 makes little sense for transcriptomes.

3.4.3

Trinity assemblies are more sensitive to low-abundance kmers but contain no new conserved genes

The di↵erence in transcript numbers between Trinity and Oases assemblies is stark: for the
same data set, with the same treatment, Trinity always produces thousands more transcripts
than Oases. Moreover, many more reads can be mapped to the Trinity assemblies —an
additional 10% or more, for every stage. Despite this greater inclusion of reads, we see
no substantial gain in either CEGMA matches or Ciona proteome matches for the Trinity
assemblies.
This conundrum can be resolved by examining the k-mer spectra, which show that the
Trinity assemblies include many more low-abundance k-mers from the read data set. This
demonstrates that Trinity is more sensitive to low-abundance sequences, and may include
more isoforms in its assemblies—by design, Trinity attempts to be more sensitive to isoforms
than Oases, and focuses particularly on low-coverage isoforms [133, 21, 131]. Those tran-

35

scripts were indeed the results of Trinity assembling low coverage reads, having an average
coverage of 5x compared to 75x.

3.5

Conclusions

We show that transcriptome assembly on two closely related species of Molgulid ascidians
produced accurate and high-quality transcriptomes, as determined by several di↵erent metrics. Importantly, four di↵erent assembly protocols produced transcriptomes that contained
nearly identical complements of homologs to the nearest model organism, Ciona intestinalis.
While variations in isoform content were observed, these variations had little apparent impact on sensitivity of homologous gene recovery. We provide detailed assembly protocols
that should enable others to easily achieve de novo transcriptome assemblies.

3.6

Acknowledgments

EKL and this research were supported by the National Science Foundation under Cooperative Agreement No. DBI-0939454 (BEACON). Any opinions, findings, and conclusions or
recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. CTB was supported in part by
Agriculture and Food Research Initiative Competitive Grant no. 2010-65205-20361 from the
United States Department of Agriculture, National Institute of Food and Agriculture.

36

Chapter 4
Genome assembly and
characterization1

4.1

Introduction

Ascidians are marine invertebrates that spend their adult life filter feeding through an incurrent siphon and an outcurrent siphon. Ascidians are evolutionarily interesting because of
the phylogenetic position – they are tunicates, the sister group to vertebrates and cephalochordates, with whom they form the chordate phylum. Although ascidians share little morphological resemblance to vertebrates in their adult stage, they do share several features in
their larval stage: a notochord, dorsal hollow neural tube, and gill slits during development
[136, 7].
The development of ascidians is well documented, and the cell lineage from fertilization
to gastrulation has been described thoroughly in Ciona intestinalis [84, 85, 82]. Studies of
other ascidian species have shown that the majority of the phyla members have an invariant
cell lineage and typical development [2]. However, a few solitary ascidians have deviated
from the typical developmental program and undergone tail-loss [120, 129]. M. occulta and
M. oculata are two species that are found in the shallow waters for Rosco↵, France that
closely resemble each other– in their adult stage, they di↵er only by a white pigment spot
1 Portions

of this section were published in Stolfi et al., [116].

37

found between the siphons of M. oculata (Figure 4.1). These two Molgula species, however,
have di↵erent methods of development—M. oculata develops as a typical tadpole larvae and
M. occulta develops without a tail. The underlying molecular reasons for divergence are
unknown.
Many ascidian genes have been studied across a number of ascidians, showing that gene
function tends to be orthologous within the phyla [98]. Although genes tend to be expressed
in homologous patterns and tissues, the presence of genes are not the same across species.
There are a number of cases where a gene that has been shown to be necessary for a phenotype in one species is completely absent in other ascidian species with the same phenotype
[60]. Ascidian species are far more divergent than they appear phenotypically. It has been
shown that in ascidians with the same phenotype and gene expression, regulatory modules
are not necessarily the same [35, 53, 115]. This is often attributed to the conservation of
gene regulatory networks (GRNs) and the flexibility of TF binding site distribution in a
given enhancer, which contribute to conservation of enhancer function [25]. This regulatory
turnover is termed “developmental system/systems drift” (DSD) [128]. This term broadly
applies to the divergence in the molecular or morphogenetic basis for the development of
identical homologous characters.
Tunicates have even deviated from hox patterning and function [37], and here genomics
has shed some light on the area. Ascidians are broadcast spawners, which leads to them being
highly polymorphic and having rapid rates of evolution [14]. This drives rapid divergence in
genomes, as well as change of gene function when compared to other chordates [60]. Through
whole genome sequencing and assembly of two closely related species M. occulta and M.
oculata, and the more divergent M. occidentalis, we have demonstrated that M. occidentalis,
M. occulta, and M. oculata all have di↵erent hox configurations, and while having invariant
38

A

B

C

D

Figure 4.1: Adult ascidians. M. occulta (A) and M. oculata (B) are nearly identical in their
adult stage with the white pigment spot (red arrow). Their tunic is covered in sand, since
they are found on the sandy sea bottoms. Under their sand covered tunic, the two species
di↵er by the color of their eggs—purple in M. oculata, pictured, and an orange-yellowish
color in M. occulta—which are found just above the kidney complex (C). C. intestinalis (D)
is the best studied ascidian and has a well-assembled and annotated genome [14, 102].
development, the cis-regulatory elements behind the development have diverged between
Ciona and Molgula.

4.2
4.2.1

Materials and methods
Genomic DNA library preparation and sequencing

Genomic DNA was phenol/chloroform extracted from dissected gonads of Molgula occulta
(Kup↵er) and Molgula oculata (Forbes) adults from Rosco↵, France, and a Molgula occidentalis (Traustedt) adult from Panacea, Florida, USA (Gulf Specimen Marine Lab). Genomic
DNA was sheared using an M220 Focused-ultrasonicator (Covaris, Woburn, MA). Sequencing libraries were prepared using KAPA HiFi Library Preparation Kit (KAPA Biosystems,
Wilmington, MA) indexed with DNA barcoded adapters (BioO, Austin, TX). Size selection was performed using Agencourt (Beckman-Coulter, Brea, CA) AMPure XP purification
39

beads (300-400 bp fragments), or Sage Science (Beverly, MA) Pippin Prep (650-750 bp and
875-975 bp fragments). For M. occulta and M. occidentalis libraries, 6 PCR cycles were
used. For M. oculata libraries, 8 cycles were used for the 300-400 bp library, and 10 cycles
were used for the 650-750 and 875-975 bp libraries. Libraries of di↵erent species but same
insert size ranges were multiplexed for sequencing in three 2x100 PE lanes on a HiSeq 2000
sequencing system (Illumina, San Diego, CA) at the Genomics Sequencing Core Facility,
Center for Genomics and Systems Biology at New York University (New York, NY). Thus,
each lane was dedicated to a mix of species, specifically barcoded libraries of a given insert
size range. Raw sequencing reads were deposited as a BioProject at NCBI under the ID#
PRJNA253689.

4.2.2

Genome sequence assembly

All genomes were assembled on Michigan State University High Performance Computing
Cluster (http://contact.icer.msu.edu). Prior to assembly, read quality was examined
using FastQC v0.10.1. Reads were then quality trimmed on both the 5’ and 3’ end using seqtk
trimfq (https://github.com/lh3/seqtk) which uses the Phred algorithm to determine the
quality of a given base pair. Seqtk trimfq only trims bases, so no reads were discarded. Each
library per species was then abundance filtered using 3-pass digital normalization to remove
repetitive and erroneous reads [5, 104, 33]. Genome assembly was done using velvet v1.2.08
[141] with k-mer overlap length (‘k’) ranging from 19 to 69 and sca↵olding was done by Velvet,
by default. Velvet does not produce separate files for contigs and sca↵olds; because Velvet
sca↵olded conservatively, contigs dominated the assemblies so we refer to both contigs and
sca↵olds as contigs. CEGMA scores were then computed to evaluate genome completeness
[89]. The latest versions of three species’ genome assemblies have been deposited on the
40

ANISEED (Ascidian Network for In Situ Expression and Embryological Data) database
for browsing and BLAST searching at http://www.aniseed.cnrs.fr/ [126]. Scripts for
genome assembly and CEGMA analysis can be found in the following github repository:
https://github.com/ged-lab/2014-molgula-genome-assemblies

4.2.3

Gene identification and alignments

Thirty-nine hox genes were identified in human and downloaded from the NCBI database.
These sequences were then BLASTed against each of the three assembled Molgula genomes.
The alignments were then extracted and BLASTed against the NCBI non-redundant database.
Molgula aligning sequences were extracted, annotated and placed in the following files,
mocc hox aa.fa, mocu hox aa.fa, and moxi hox aa.fa, which are located at https://github.
com/ged-lab/2014-elijahlowe-thesis in the directory data/. Hox1-13 sequences for human, fruit fly, and Amphioxus were download from ‘Homeobox Database’ http://homeodb.
zoo.ox.ac.uk/. These sequences were then joined in a multifasta file with the identified
Molgula hox genes and used to produce a phylogenetic trees using MAFFT version 7 online rough tree program at http://mafft.cbrc.jp/alignment/server/clustering.html
[47, 46]. Additional alignments between the three species were conducted using mVista
[73, 18, 135] with M. oculata as the anchoring sequence because it shows the most similarity between the three Molgula species. The LAGAN alignment algorithm was used with
translated anchoring to improve alignment because of evolutionary distances[6].

41

1

Species
M. occidentalis
M. occulta
M. oculata

N50
26,298
13,011
34,042

Table 1: Assembly Statistics
Mean contig length Total
Total number of base pairs
5,072
51,761
262,547,660
3,233
58,489
189,110,562
6,270
25,497
159,886,716

CEGMA C1
81.45
77.42
89.92

CEGMA P2
96.77
98.79
99.19

Table 4.1: Genome assembly statistics. The contig N50 length, mean contig length,
total number of contigs, total number of base pairs and CEGMA scores were collected for
each draft assembly. The CEGMA score is a metric of completeness measured against highly
conserved eukaryotic genes. Alignments of 70% or greater of the protein length are called
complete (C) and all other statistically significant alignments are called partial (P).

4.3
4.3.1

Results
Genome assemblies assessment

Genomes of three Molgula species (M. occidentalis, M. oculata, and M. occulta) were sequenced using next-generation sequencing technology and assembled. A common metric for
judging the quality of a genome assembly is the contig N50 length, which is determined
such that 50% of the assembly is contained in contigs of this length or greater. We used the
contig N50 length to select the best assembly for each species given the varying ‘k’ parameter
(length of k-mer overlap). A ‘k’ of 39 yields the best assembly for both M. occidentalis and
M. occulta. The best ‘k’ for M. oculata was 61. M. occidentalis, M. occulta, and M. oculata
N50 lengths were approximately 26.3 kb, 13 kb, and 34 kb, respectively (Table 4.1).
In addition to N50 lengths, we also used CEGMA (Core Eukaryotic Genes Mapping
Approach) scores, in order to evaluate the assemblies’ representative completeness [89].
CEGMA reports scores for complete and partial alignments to a subset of core eukaryotic
genes. An alignment is considered “complete” if at least 70% of a given protein model aligns
to a contig in the assembly, while a partial alignment indicates that a statistically significant
portion of the protein model aligns. The partial alignment scores are ⇠97% or higher for all
assemblies. M. oculata has the best complete alignment score at ⇠90%. M. occidentalis and
42

M. occulta have complete alignment scores of 81% and 77% respectively (Table 4.1). These
scores indicate that our assemblies contain at least partial sequences for the vast majority of
protein-coding genes in the genomes of these species. Various factors make it unreliable to
predict genome size and gene density based on assembly metrics alone [3]. Of the handful
of sequences we isolated and analyzed, we found that the sizes of introns and upstream regulatory regions were roughly comparable to those from their Ciona orthologs. This suggests
that the Molgula genomes may be as compact as the C. intestinalis genome (i.e., ⇠150-170
Mb, ⇠16,000 genes) [55, 107, 102].

4.3.2

Gene complexes

Hox genes are a subset of the homeobox genes and are known to be involved with the
establishment of morphological identities along the anteroposterior axis of bilaterians and
cnidarians [17]. All hox genes have a highly conserved 60 amino acid (aa) homeobox sequence
[74, 19] and hox genes are distinguished by variation of the homeobox domain. There are 4
HOX clusters in humans totaling in 39 genes. Within tunicates, C. intestinalis, Halocythia
roretzi and Oikopleura dioica hox genes have been characterized. C. intestinalis has 9 hox
genes, Hox1 through 6, Hox10, and Hox12-13 [14]. The hox gene of C. intestinalis were
initially found on 5 sca↵olds spanning ⇠980 kb using the draft assembly, with hox2-4, hox5-6
and hox12-13 being found on the same sca↵old and later identified to be two clusters of hox
genes across two chromosomes [38]. O. dioica also has 9 hox genes, hox1-2, hox4, a duplicate
hox9, and hox10-13, however, none of the genes have been found on the same sca↵old, even
using a 250 kb window [108].
Eight hox genes have be found in M. occulta and M. oculata, while nine have been found
in M. occidentalis. The eight found were Hox1, hox2, hox3-4, hox5, hox10 and hox12-13, with
43

hox3 and hox 4 being found on the same contig in all three species (Figure 4.2). Additionally
hox10, and hox12-13 are found on the same contig in M. oculata with only hox12-13 being
found on the same contig in M. occidentalis. However, it appears that the hox genes have
been rearranged in M. oculata, since hox10 is downstream of hox12-13. M. occidentalis had
one additional hox gene compared to M. occulta and M. oculata, a duplicate hox10 gene
⇠12kb

apart found on the same contig. Also, the M. occidentalis hox2 has a stop codon

located in the 3-4 helix, although this may be specific to the animal examined and not the
total population. The second hox10 sequence was not fully sequenced, missing 14 aa of
the homeobox domain, and the identity of the two sequences is 52.1% at a nucleotide level,
53.4% at a protein level, and 91.3% identical within the homeobox domain (Figure 1.4). M.
occulta, M. oculata, and M. occidentalis hox genes span across 7, 5 and 6 contigs respectively
and are 197 kb, 311.7 kb and 279 kb in length, respectively. This is more compact than the
Ciona hox cluster which exhibits longer than usual intergenic regions, averaging in the 5Mb
range, when typically the hox genes have 100-120 kb intergenic regions separating them [75].
The hox gene in the Molgula had far smaller intergenic regions, 10-25 kb in length for the
hox genes that were found on the same contigs.

4.3.3

Divergence of Gene regulatory networks

Our sequencing e↵orts revealed extreme genetic divergence not only between Ciona and
Molgula, as expected, but even within the Molgulids. For example, in Stolfi et al., [115] we
used BLAST to identify the Molgula orthologs of C. intestinalis Mesp. C. intestinalis Mesp
is the sole ortholog of vertebrate genes coding for MesP and Mesogenin bHLH transcription
factor family members [102]. VISTA alignment shows high sequence similarity between
sequences 5’ upstream of the Mesp genes from the closely related M. oculata and M. occulta.
44

!1!!2!!3!!4!!5!!6!!!7!!8!!9!10!11!12!13!#!!

Vertebrata!
!!(Human)!
Chordata!
!1!!!2!!!3!!!4!!!5!!!6!!!7!!8!!!9!!10!11!12!13!

Ascidians!
Tunicata!

C.#intes)nalis#
M.#occidentalis#
M.#oculata#
M.#occulta#

+!
*!

!!Larvacean!
(Oikopleura)!

Figure 4.2: Hox clusters for M. occulta, M. oculata and M. occidentalis Eight hox
genes were found in M. occulta and M. oculata, while nine were found in M. occidentalis.
Hox1, hox2, hox3-4, hox5, hox10 and hox12-13 were found in all three Molgula species.
Hox3-4 was found on the same contig in all species, with hox12-13 found on the same contig
in M. occidentalis and M. oculata. *M. occulta hox12-13 is not found on the same contig, but
when aligned using mVista, there is high sequence similarity, showing the possible placement
of hox12-13 in M. occulta. + M. occidentalis hox2 gene had a stop codon found in the 3-4
helix. # numbers correspond to gene color, and rearrangements have been found in Ciona
and Molgula.

45

However, there is no conservation of Mesp DNA sequences, coding or non-coding, between
M. oculata/occulta and M. occidentalis, nor between C. intestinalis and any of the three
Molgula species. In previous phylogenetic surveys, M. occidentalis has been placed as an
early-branching Molgula species, often grouped together in a subfamily with species ascribed
to the genera Eugyra and Bostrichobranchus instead [24, 34, 129]. Our sequencing results
support the view that M. occidentalis is highly diverged from other Molgula species.
This sequence divergence is also evident when analyzing the hox genes. When comparing
sequence similarity of the hox genes that were found on the same contigs (hox3-4 and hox1213 ), only regions clustered around the coding region for M. occulta when compared to M.
oculata showed similarity, and only highly conserved coding domains showed similarity in
M. occidentalis when compared M. oculata (Figure 1.1). This sequence similarity was a lot
less obvious when comparing M. occulta to M. occidentalis. However, because of a lack of
synteny outside of coding regions between M. occidentalis and M. oculata we were able to
identify distal-less, downstream of Hox13, which is expressed in endodermal strand cells in
Ciona.

4.4

Expression of the M. occidentalis Mesp gene marks
the B7.5 cells

Ciinte.Mesp specifies the B7.5 cells as the sole progenitors of the cardiopharyngeal lineage
[101, 13, 29, 113]. We performed RNA in situ hybridization (ISH) for M. occidentalis Mesp
(Moocci.Mesp) and found that this gene is also expressed only in the B7.5 cells of M. occidentalis embryos (Figure 4.3C). These cells are unequivocally identified due to the perfect
conservation of early embryonic cell cleavage patterns in all ascidians. ISH for the M. oc46

Figure 4.3: The B7.5 lineage in M. occidentalis (A) Diagram comparing M. occidentalis (top) and C. intestinalis (bottom) embryogenesis at 24C. Embryos were stained with
Alexa Fluor dye-conjugated phalloidin to visualize cell outlines and DAPI to visualize cell
nuclei. (B) Diagram of mVISTA ([18]; genome.lbl.gov/vista/) alignment of M. oculata Mesp
(Moocul.Mesp) locus to orthologs in M. occulta, M. occidentalis, and C. intestinalis. Shaded
peaks indicate sequence conservation above 70% over 100-bp windows (blue = protein-coding,
pink = non-coding). Arrows indicate direction of transcription of protein-coding genes. Noncoding sequences upstream of Mesp are only conserved between M. oculata and M. occulta.
M. occidentalis and C. intestinalis show considerable divergence even in protein-coding sequences. Note that microsynteny with SLC5A-related gene supports the orthology of these
sequences among the Molgulids.

47

Figure 4.3 (cont’d): (C) In situ hybridization (ISH) for Moocci.Mesp in 110-cell stage embryo
(vegetal view), showing mRNA detection (green) in B7.5 blastomeres. Nuclei were counterstained with DAPI (blue). Staging is given by hours post-fertilization (hpf). (D) Vegetal
view of a 110-cell stage embryo electroporated with Moocci.Mesp>GFP reporter construct.
Reporter gene expression was detected by ISH for GFP transcripts (green). Nuclei were
stained with DAPI (blue). (E) Lateral view of a mid-tailbud stage embryo electroporated
with Moocci. Mesp>Histone2B::GFP reporter construct. GFP fluorescence reveals B7.5
descendants on left side of embryo: two trunk ventral cells (TVCs) and two anterior tail
muscles (ATMs). (F) Diagram of B7.5 lineage divisions from 110-cell stage to mid-tailbud
stage, inferred from previous C. intestinalis studies. Cells are named according to Conklin’s
method (Conklin, 1905). The lineage is bilaterally symmetric, but only cells on the left
side are indicated and named. Relative staging given for M. occidentalis (Mo.occi) and C.
intestinalis (Ci.inte). 110-cell and late gastrula: vegetal view. Initial tailbud: dorsal view.
Mid-tailbud: lateral view. Anterior pole is on the left in all images and illustrations. DOI:
10.7554/eLife.03728.003

culta Mesp gene (Mooccu.Mesp) also revealed conserved expression in this tail-less species
(Figure 4.3). We successfully adapted the Ciona electroporation protocol for simultaneous
transfection of reporter gene plasmids into hundreds of synchronized M. occidentalis embryos
(Figure 4.3D,E). We were also able to electroporate M. occulta embryos (Figure 4.3). However, only M. occidentalis was routinely available to us for in vivo studies, so we focused our
experiments on this species. Development of M. occidentalis embryos was optimal at 24C
and faster than that of C. intestinalis (Figure 4.3A). Using electroporation-based transfection, we determined that an ⇠1.1 kb genomic DNA fragment upstream of Moocci.Mesp is
able to drive expression of fused reporter genes specifically in the B7.5 cells with no ‘leaky’
expression in other cells as is commonly observed in C. intestinalis (Figure 4.3D; [112]).
This faithful recapitulation of Moocci.Mesp expression and the persistence of GFP allows for
visualization of the descendants of B7.5 long after endogenous Moocci.Mesp transcription
has ceased (Figure 4.3; [13]). At the tailbud stage, we find that each B7.5 blastomere gives
rise to four grand-daughter cells (Figure 4.3E). The two anterior B7.5 grand-daughter cells
on either side of the bilaterally symmetric embryo migrate anteriorly and are termed the
48

trunk ventral cells (TVCs) due to their final position in the C. intestinalis and H. roretzi
embryos [82]. Their posterior sister cells remain in the tail and become anterior tail muscles
(ATMs). As far as we can tell, B7.5 lineage ontogeny is perfectly conserved between M.
occidentalis and C. intestinalis (Figure 4.3F).

4.5

Divergence of Mesp cis-regulatory sequence function between M. occidentalis and C. intestinalis

Given the obvious parallels between C. intestinalis and M. occidentalis cardiopharyngeal
development, we expected transcriptional regulatory mechanisms to also be highly conserved between the two species. We tested this assumption by electroporating C. intestinalis reporter constructs into M. occidentalis embryos, and vice-versa. We observed that
a Ciinte.Mesp reporter construct [13], when electroporated into M. occidentalis embryos,
drives relatively weak reporter gene expression in B7.5 with substantial leaky expression in
other tissues (Figure 4.4 (cont’d)A). Conversely, the Moocci.Mesp enhancer fails to drive
any reporter gene expression when electroporated into C. intestinalis embryos (Figure 4.4
(cont’d)B), despite recapitulating robust B7.5-specific expression in M. occidentalis embryos
(Figure 4.3D,E). These data suggest acute DSD of transcriptional regulatory mechanisms underlying otherwise identical Mesp expression patterns. More specifically, the trans-regulatory
environment of the B7.5 blastomeres has diverged between Molgula and Ciona, and compensatory changes in the respective Mesp cis-regulatory sequences must have rendered these
unable to function adequately outside of that milieu.
The observation of Mesp incompatibility or unintelligibility, lead to the examination of
Foxf cis-elements, who is also involved in TVC specification, both revealed divergent cis49

regulatiory logic for underlying identical gene expression patterns. This sparked an interest
to see if there was a general trend of cis-regulatory unintelligibility between C. intestinalis
and M. occidentalis. This revealed that M. occidentalis Tbx6 we found that the Ciinte.Hand-r
reporter can drive reporter gene expression in M. occidentalis TVCs. Thus, unlike the case of
Foxf, there is an asymmetric intelligibility of Hand-r TVC enhancers between M. occidentalis
and C. intestinalis. Moreover, a M. oculata Hand-r TVC enhancer is functional in M.
occidentalis but not in C. intestinalis. Taken together, these data suggest that di↵erences
in enhancer logic may have accumulated over the course of the deep evolutionary history
between Molgula and Ciona but not between M. occidentalis and M. oculata, and that
some enhancers may have evolved asymmetrically in the two branches, retaining greater
pan-ascidian “intelligibility” in one or the other.

4.6

Discussion

Three Molgula—M. occulta, M. oculata and M. occidentalis—species have been sequenced
and assembled; these are the first of any molgulids to have assembled genomes. Developmentally the three species are very similar up to the gastrula stage, where M. occulta diverge
from the typical solitary ascidian body plan and develop without a tail [2, 82].
In vertebrate and other bilaterians the hox genes has been shown to be important for
patterning along the anterior-posterior axis [17, 70]. The same has not been shown in ascidians, where hox has more of a tissue specific role [37]. Ciona has 10 hox genes and is missing
hox7-9 and 11, with hox10 and 12 being the only two genes to show morphological e↵ects
when knocked down. Hox10 is involved in the regulation of the motor neuron di↵erentiation
and hox12 is involved in tail development, through the elongation of the posterior most

50

Figure 4.4: The B7.5 lineage in M. occidentalis Cross-species reporter construct assays
reveal multiple cases of cis-regulatory unintelligibility. (A) C. intestinalis embryo electroporated with Ciinte.Tbx6-r.b>GFP reporter construct [9], which drives GFP expression in tail
muscles and the B7.5 lineage cells (arrowheads), recapitulating endogenous Ciinte.Tbx6-r.b
expression. (B) M. occidentalis embryo electroporated with Moocci.Tbx6-r.b>GFP reporter
construct, which recapitulates expression in tail muscle cells including B7.5 lineage cells (arrowheads). (C) C. intestinalis embryo electroporated with Moocci.Tbx6-r.b>GFP, which
drives expression in B-line tail muscle and mesenchyme cells but is excluded from the B7.5
lineage. (D) C. intestinalis embryo electroporated with Ciinte.Hand-r>H2B::GFP reporter
(Davidson and Levine, 2003), which drives H2B::GFP expression in anterior endoderm, A7.6
lineage, and TVCs (arrowheads), recapitulating endogenous Ciinte.Hand-r expression. (E)
M. occidentalis embryo electroporated with Moccci.Hand-r>H2B::GFP construct, which recapitulates the same expression pattern. (F) C. intestinalis embryo electroporated with
Moocci.Hand-r>H2B::GFP, which drives expression in endoderm and A7.6 lineage, but not
in B7.5. (G) M. occidentalis embryo electroporated with Ciinte.Ebf neuron-specific YFP
(green) and H2B::mCherry (red) reporter constructs electroporated [114], which drive very
weak expression in a limited subset of motor ganglion neurons (green and red).

51

Figure 4.4 (cont’d): (H) C. intestinalis embryo electroporated with a Moocci.Ebf>YFP
reporter, which drives robust reporter gene expression in several brain, motor ganglion, and
nerve cord neurons. (I) C. intestinalis embryos electroporated with Ciinte.Sox1/2/3 (left)
and Moocci.Sox1/2/3 (right) H2B::mCherry reporter constructs, both of which recapitulate
Sox1/2/3 expression in ectoderm. Panels A-F are lateral views of tailbud embryos, panels G
is a dorsal view of a tailbud embryo, panel H is a dorso-lateral view of a hatched larva, and
panel I shows vegetal views of mid-gastrula stage embryos. Anterior is to the right, except
in Panel I, in which anterior is to the top. DOI: 10.7554/eLife.03728.023

section of the tail and of the epidermal cells at the tail tip [37]. We observed the absence of
hox6 in all three Molgula species, which is not surprising, seeing that hox6 is also missing
in O. dioica and no expression is detected in C. intestinalis through whole mount in situ
hybridization at any stage of development [38, 105]. No two of the Molgula species show the
same hox cluster and all show a strong divergence outside of coding regions, even more so in
M. occidentalis. There is a duplication of M. occidentalis hox10 which could lead to a split in
function since Ciona Hox10 is expressed in two region during the mid-tailbud stage—a small
region of the anterior nerve cord, and a small area of the posterior ventral endoderm and
adjacent tissue [38]. It is proposed that ascidians evolved their simple body plans and rapid
embryogenesis through extensive genomic rearrangement and gene loss, with specific loss of
several hox genes [36]. Hox12-13 are not found on the same contig in M. occulta, however
when aligned with mVista there appears to be a strong case for synteny (Figure 1.1), so it
is possible they are clustered, but the contigs are not joined because the M. occulta genome
assembly is too fragmented. Hox function does not have the same level and so far their has
not been a consistence between hox clusters in no Tunicate studied, showing that hox is
tends do be divergent within the tunicates.
In Ciona dll stains in the endodermal strand cell during early-tailbud embryos, showing
positive signals in two cells of the endodermal strand [8] which derives from the primary

52

muscle or mesenchyme lineage [96], while in Drosophila embryos dll is required for the gene
pathway for limb formation in the thoracic segment [130]. Because of the lack of homology
outside of coding regions, we were able to identify distal-less (dll ) downstream of hox13
in both M. occidentalis and M. oculata, but we did not find it in M. occulta. Further
investigation is needed, as this could potentially be tied to failed muscle cell di↵erentiation
in M. occulta.
From our initial survey of a handful of enhancers from C. intestinalis, M. occidentalis, and
M. oculata, we encountered several instances of either mutual unintelligibility, or asymmetric
intelligibility of enhancers. These results add to the mounting evidence suggesting that acute
and pervasive DSD may have occurred over the course of ascidian evolution, obfuscated by
the identical cell lineages and highly conserved gene expression patterns of ascidian embryos
[118]. The multiple examples of cis-regulatory unintelligibility we identified were rather
unexpected given (a) the extremely conserved pattern of expression of orthologous genes from
Molgula and Ciona [124] and (b) previous observations of mutual intelligibility of enhancers
between C. intestinalis and H. roretzi (e.g., Otx), and between C. intestinalis and the more
closely related C. savignyi [44, 4]. Large-scale, quantitative cross-species assays and detailed
GRN studies will illuminate factors that may contribute to conservation or divergence in
regulatory mechanisms.

4.7

Conclusion

Hox genes function is not conserved within the tunicates, which have also undergone substantial hox gene loss. All central hox genes are missing in all three Molgula species. Studying
these three molgula species have compounded the evidence that Tunicate has lost both hox7

53

and hox8 after diverging from the vertebrate and cephalochordate lineages, while ascidians
have lost hox9 and hox11 after diverging from the larvaceans. It appears because hox does
not act as an anterior-posterior organizer in ascidians, hox genes are more easily lost in
ascidians and other genes have taken over this role.
Although tunicates have a well defined and invariant cleavage, the mechanism behind development is not always the same from species to species. We speculate that a high frequency
of compensatory changes, required for the rapidly evolving ascidians to accommodate the
constraints imposed by their invariant embryonic cell lineages and highly compact genomes,
has given rise to a preponderance of cross-species cis-regulatory unintelligibility, following
the DSD/SPC model. This perfect storm of intrinsic factors may be the key to explaining the dichotomy observed between highly conserved embryos and divergent cis-regulatory
structure/function in ascidians.

54

Chapter 5
Di↵erential expression analysis of tail
loss in an invertebrate chordate

5.1

Introduction

Chordates are composed of three subphyla—vertebra, tunicata and cephalochordata—that
all share several characteristics, with the notochord being the key characteristic (from which
the phylum name comes). The tail development of larvaceans such as Oikopleura dioica
and several species of ascidians and tunicates has been well studied [39, 79, 53]. Larvaceans
form tailed larvae with a hollow dorsal notochord and keep their tail throughout their adult
life stage. Ascidians generally form their tail in a similar manner before undergoing the
process of metamorphosis, in which the larval tail is absorbed in to the trunk region [88].
A typical ascidian larvae tail forms through the convergence, intercalation and extension
of the notochord and the di↵erentiation of the posterior muscle cells [117]. When fully
formed the ascidian notochord contains 40 cells, flanked by three rows of muscle cells. The
ancestral notochord or notochord-like structure is believed to have been muscle based, which
is perhaps the reason that the tail formation is often contingent upon proper development of
both notochord and muscles [57]. This idea is supported by the observation that the primary
and secondary notochord and muscle lineage are derived from the same blastomere, and the
ascidian tail needs both the notochord and di↵erentiated muscle to form a larval tail [82, 16].
55

Of the ⇠3000 species of ascidians fewer than 20 do not form a tail, with the majority being Molgulidae species [2, 34]. This likely represents several independent instances
of evolutionary loss of the tail, and introduces the question of why are the Molgulidae so
susceptible to tail-loss. Although the mechanism behind tail-loss di↵ers by species, a common characteristic is the lack of a notochord that intercalates and extends, as well as a
less di↵erentiated central nervous system (CNS) structures and tail muscles [117]. Molgula
bleizi notochord cells converge to the midline, and began to extend, but cells never properly
intercalate and the tail formation stops before it is fully formed [42]. In M. bleizi there is
an early down-regulation of brachyury (bra)—a key notochord inducer—and larva-specific
muscle actin genes have become pseudo genes [42]. A similar situation is observed in Molgula
occulta: muscle actins have become pseudogenes, independently of their conversion in M.
bleizi.
M. occulta and Molgula oculata are two closely related species, who in their adult form
are virtually identical, with the exception of a white pigment spot between the two siphons
of the tailed species, M. oculata. During development the species are indistinguishable up
to the gastrula stage. In late gastrula when the notochord and muscle progenitor begin
to move posteriorly [123] and the notochord begins to form, the morphological divergence
becomes evident. Through subtractive screening Swalla and Je↵ery [121] identified manx,
a zinc finger transcription factor (TF) and the cytoskeletal protein p58 as down-regulated
in M. occulta relative to M. oculata [119], and this was shown to be one of the causes of
the tail loss in M. occulta. There are several steps that take place to form the notochord
and tail: first, the notochord cells move mediolaterally to the midline; and next, the cells
polarize and intercalate, changing their shape and extending posteriorly [49, 43, 111]. This
process is known as convergence and extension. Although the two species have di↵erent
56

developmental programs, crossing the tail-less M. occulta eggs with sperm of the tailed M.
oculata produces a hybrid with 20 notochord cells like the tail-less species, but in which the
cells intercalate and extend like the tailed species. In the hybrids the expression of manx and
p58 are restored, and antisense phosphorothiated oligodeoxynucleotide manx in the hybrids
have shown that zygotic manx is necessary for tail formation [121].
Several key tail development genes have been identified as present in the M. occulta
genome, but expressed at low levels during embryogenesis, and when expressed, some of these
genes were shown to restore features in the hybrid [120, 39, 123, 121]. With advances in high
throughput sequencing technologies, gene expression of M. occulta, M. oculata, and hybrid
species can be analyzed on a whole transcriptome level [22, 91]. mRNA of three di↵erent
developmental stages for M. occulta, M. oculata, and their hybrid has been sequenced and
assembled at Michigan State University (MSU). These three transcriptomes were used to
assess the expression levels of known notochord genes downstream of bra using C. intestinalis
data from the NCBI database (ncbi.nlm.nih.gov). BLAST searches were done against
known notochord genes, and several of them were selected for further analysis, including
FGF9/16/20, prickle (pk), and several other downstream genes. These genes were then
used to construct a putative brachyury gene regulatory network for both M. occulta and M.
oculata.

5.2
5.2.1

Methods
Sample collection, sequencing and assembly

RNA was extracted from all three Molgula species using the methods discussed in Lowe et
al. [66]. RNAs for the gastrula (3hpf), neurula (4hpf) and mid-tailbud (6hpf) stages were
57

extracted for both M. occulta and M. oculata, with a replicate for the gastrula and a sample
for early-tailbud stage sequenced from M. occulta. DNA was extracted from the gonads
of an individual adult specimen for M. occulta, and M. oculata. Two paired-end jumping
libraries were collected for each species ranging from ⇠300bp to ⇠950bp. Further details
about extraction methods and libraries can be found in Stolfi et al., [115]. Sequencing for
M. occulta and M. oculata RNA were conducted at the Michigan State University, while
all other sequencing was done at New York University. All libraries were paired-end, with
75 base pair (bp) reads for the sequencing done at MSU and 100 bp reads for the NYU
sequencing.
Genome assemblies were conducted using 3-pass digital normalization [5] and assembled
using Velvet[141]. Other assemblers were tested, however, Velvet produced the best results
with the least fragmented assemblies. Assemblies were initially done with 21

k

71, for

intervals of 10. We selected the ’k’ value with the highest N50, and then the genomes were
reassembled for a k±10 with a step size of 2 for the selected assembly, and the best N50 was
chosen. A k of 31, and 49 were select for M. occulta, and M. oculata, respectively.
Both de novo and reference based assemblies were used to create gene models. Reads
were mapped to their respective genomes using bowtie2 and tophat to identify genes and
alternative splicing variants [56, 127]. The accepted.bam files were then sorted and indexed using samtools [61]. The sorted bam files were then processed using cu✏inks and
cu↵merge to generated consensus gtf annotation files. The digitally normalized trinity de
novo assembled transcripts from Lowe et al. [66], were aligned to their respective genomes
using BLAT [23]. The cu✏inks/cu↵merged gtf files were then converted into bed files
and merged with the annotation files from the mapped de novo assembly aligned using
gimme (https://github.com/likit/gimme).
58

Gimme joins gene models using a graph

based method to develop more complete transcripts. The gimme gene models were then
converted to g↵ format using the script bed2g↵ in the gimme utils folder in order to
extract the transcripts from the genome in a multi fast file. Transcripts were then extracted using “g↵read -w transcripts.fa -g /path/to/genome.fa transcripts.gtf” which is included in the cu✏inks package. The extracted transcripts were partitioned into transcript
families and annotated using the khmer suite and steps found in the eel-pond protocol
(https://khmer-protocols.readthedocs.org/). Ciona intestinalis was used as an annotation reference, and the sequences were retrieved as discussed in Chapter 3.

5.2.2

Gene counts and di↵erential expression analysis

Reads were mapped to transcripts from the gimme gene models for their respective species.
Reads from the hybrids were mapped onto both the M. occulta and M. oculata transcripts,
since the hybrids are F1 hybrids and should contain an allele from each parent. Read counts
were generated using eXpress [94]. eXpress gives the option of “Total counts” and “E↵ective
counts”, which reports the number of reads mapped per transcript and the normalized counts
based on transcript length, respectively. Because EdgeR uses unnormalized reads, “Total
counts” were used. Counts for hybrid reads mapped to M. occulta and M. oculata were
combined to calculate total expression at a given stage. A replicate was only available for
one of the samples, 3hpf, and because of this 5hpf was treated as a replicate for 6hpf. These
time points correspond to early and mid-tail bud stages in the tailed ascidian.
Di↵erential expression was calculated using the Bioconductor EdgeR package because of
its ability to work with minimal replicates [95]. We ran RBH for M. occulta and M. oculata
in order to identify orthologous transcripts to conduct allele specific di↵erential expression
analysis on the hybrid. We used both estimateGLMCommonDisp and estimateGLMTag59

wiseDisp to calculate dispersions. There is only a replicate for M. occulta gastrula stage
(F+3), because of this we used M. occulta early tailbud stage as a replicate for the mid
tailbud stage. These replicates were used to calculate dispersions for all samples. Exacttest with p = 0.05 was used to determine di↵erential expression. Transcripts with a false
discovery rate (FDR) of 0.05 were called as di↵erentially expressed.

5.3
5.3.1

Results
M. occulta and M. oculata have strong overlap in gene expression

C. intestinalis is the closest ascidian species with a well annotated genome, and so we used
Ciona proteins obtained from the NCBI to annotate the genomes of both M. occulta and
M. oculata. Reciprocal best hit (RBH) blast with an e-value of 1e-3 was done with the M.
occulta and M. oculata transcriptomes against C. intestinalis for the annotation of both
Molgula species. We are aware this is a low threshold for homology, however, information
for these species are not known and we wanted to gain as much insight as possible for genes
and gene function. Moreover, the reciprocal best hit criterion is extremely stringent.
The gene models for M. occulta and M. oculata produced 42,365 and 40,775 sequences
total, respectively. Precisely 8,627 M. occulta transcripts were annotated as orthologs and
22,700 transcripts were annotated as showing homology to C. intestinalis. Similar annotation numbers were produced with M. oculata: 8,677 showed orthology, and 22,583 showed
homology. M. occulta and M. oculata have a high overlap in number of translated transcripts
that showed any level of homology with C. intestinalis proteins from the NCBI database.

60

Figure 5.1: Brachyury gene regulatory network. Bra is a key notochord inducer,
and without its expression neither the notochord nor the chordate tail forms. Downstream
genes have been identified in various studies and of these 67 genes 11 were missing from
the transcriptomes of both species. The missing genes were cofilin, entactin (nidogen-2like), fibrinogen-like protein (FGL), fibronectin, multidom, noto1, noto5, noto14, noto16,
and tropomyosin. netrin, and noto9 are missing in M. occulta and Klf15 is missing in M.
oculata. (Blue) missing in M. occulta, (Red) missing in M. oculata, and (Purple) missing
in both species. This GRN was built from previous studies [31, 30, 32, 54, 53], using the
BioTapestry software [65].
Of the 16,414 Ciona proteins, M. occulta had BLAST hits for 83.6% and 86.5% had hits in
M. oculata. M. occulta had hits for 453 proteins that were not found in M. oculata and M.
oculata had hits for 921 transcripts that did not have hits in M. occulta, yielding an overlap
of 97% in assembled homologs.

61

5.3.2

Notochord gene network

Next, we examined genes associated with notochord development in C. intestinalis to investigate the molecular development of the tail. Gene candidates were identified as being
involved in tail formation and notochord development through previous analysis of genes
downstream of bra [31, 30, 32, 54, 53]. From these studies many potential notochord genes
were identified; we compiled a list of 67 genes and identified those expressed during the gastrula, neurula or tailbud stages within the transcriptomes of both M. occulta and M. oculata
(Figure 5.1). Of the 67 genes, 11 were not expressed in the transcriptomes of either species:
cofilin, entactin (nidogen-2-like), fibrinogen-like protein (FGL), fibronectin, multidom, noto1,
noto5, noto14, noto16, and tropomyosin. The remaining genes without orthologous sequences
were netrin, and noto9 in M. occulta and Klf15 in M. oculata. In Ciona, netrin is expressed
in the notochord and the central nervous system and is associated with axon guidance [30].
Klf15 was detected in the notochord, but there is currently no known information regarding
its function [90]. Taken together, this demonstrates a strong overlap in the presence of the
number of genes associated in notochord formation in both M. occulta and M. oculata.

5.3.3

Di↵erential expression between neurula and tailbud appears
to be key factor in tail development

62

Species
M. occulta
M. oculata
Hybrid

Condition
from Gastrula to Neurula
from Neurula to Tailbud
from Gastrula to Neurula
from Neurula to Tailbud
from Gastrula to Neurula
from Neurula to Tailbud

Number of transcripts that show...
Up-regulation Down-regulation No di↵erential expression
260
8
20197
1
4
20460
119
66
20280
1170
626
18669
21
99
20345
1270
129
19066

Table 5.1: Di↵erential expression: Species vs time

63

10

10

10

●
●

●

●●

●
●●●

●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●

●
●●
●●

●
●●
●

●●●

●●●
●

●●
●

●
●●
●
●●
● ●
●●

●●●

●●
●● ●

●
●●
●
● ●
●
●
●●
●●●●

●

●

●

●

●
●
●
●

●
●
●
●

●
●
●
●
●

●● ●● ●
●
●

●
●

●

●

●

●
●

●

●
●
●
●

●
●
●

●
●

●

●
●
●

●
●
●

●
●
●
●

●
●
●
●
●
●

−5

●
●●●●
●●●
●

●
●●
●
●
●

●

●
●

●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●● ●
●
●●
●
●
●
●
●
●
● ●
●
●
●●●● ●
●
●
●
●
●
● ●
●●
● ●●
●
●
●● ● ● ● ●●
●●
●
●●
●● ●● ●
●
●
●
●
●
●
●
●●
●
●
●
● ● ●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● ●● ●
●
●●
●● ● ● ●
●
● ●● ●
●
●
●
●
●
● ●
●
●
●
●
●● ●● ●● ● ●
● ●●
● ●● ● ●
●
●
●
●
●
●
●
●● ●●● ● ●
●
●● ●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
● ●● ● ●
●
●
●
● ● ● ● ●●
●
●
●
●
●●
●
●●●● ●● ●
●
●
●
●
●
●
●
●
● ●● ●
●
●● ●
● ●
● ●● ● ● ●
●
●
●●●
●
●
●●
●
●●
●
●
●● ●
●
●
●
●● ●
●
●
●
● ● ●
●
●
●
●
●● ●● ●
●●●● ●
●●●●●●
● ●●●● ●●●●
●
●
●● ●●● ●● ●●
●●● ●●
●
●●
●
●
●
●
●
●
●●
● ●●● ● ● ●●●●
●
●
●
●
●
● ●●
●●
● ●
●
●
●
●
●
● ● ●●
● ●
●
●
●
●
●
●
●
●
●
●●
● ● ●● ● ●
●
●
●
● ● ● ●● ●●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● ● ●● ●● ●●●● ●
●
●
●●●● ● ● ●●● ●
●
●
● ●
●
●
●
● ●●
●
● ●●
● ●
●
● ● ●● ●
●●●● ●
●
●●●●●● ●
●
●
● ●
●
●●
●●● ●
●●
● ●
●
● ●
●
●● ● ● ●●● ● ●
●
●
●
●
●
●
●
●
●
●
● ●●
●●
●
●
●
●
●●●●● ●
● ●
●
●●
●
●●
●●● ● ●● ● ●●● ●●●
●● ● ●
●
●●
● ●●
●
● ●
●
●
●
●
● ●●
●
● ●●
● ● ● ● ●● ● ●
●
●
●
●●●
●
●
●
●
● ●
●● ●
●
●●●●●
●
● ●
●
●
● ●
●
●●
●●● ●
●● ●
●●
● ●
●
● ●● ● ●●● ● ●● ●
● ●●
●●●●● ●●●●●● ● ●
●●
●
● ●●
● ●
●
●
●
●● ●
● ●●
●
●
●● ●
●
●
●●● ●●● ●● ● ●●●
●●● ●
● ●● ●●●●●●
● ●●● ● ●●
●
●
●
●
●●●
●●
●
● ●
● ●
●
●
●
●
● ●● ●● ●
●●
●
● ●●
●●
●
● ● ●●●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●● ● ●● ●●
●● ●●
●●
●● ●● ●
●● ● ●●
●●●
●
● ●● ● ●●
●
● ●● ●●
●
●●●●●●●
●●
●● ●
●
●●● ●
●●●● ● ●
●
●
●● ●●●● ● ● ●●●
●
●
● ●●
●●●
●
●
●
●●●●
●
●
●
●● ●
●
●
●
●
●
● ●●●●
● ●● ●
● ●●
●
●
●
●
●
●
●
●
●
●
●●● ●
● ●
●● ●●
●●
● ●●●●
●●
●●
●●● ●
●
●
●
●●
● ● ● ●● ●●
●
●●
●
●
●
●
●●
●●● ●
●
●●
●
●
●
●● ●●
●
●
●
●●●● ●●●● ●●●●●●●
●
●
●●
●
● ●● ●
●● ●● ●
●●● ●
●
●● ●●
● ●●
●●●●●
●
●
●
●
●
● ●●●●
●●
●●
●●●
●●
● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●●
●●●●● ●●
●●
●
● ●●●●
●● ●● ● ●
●●
●
●
●●●
●
●
●
●
●
●
●●
●●
●
● ● ● ●●
●●
●
●
●●
●
● ●● ● ●●●● ●●
●
●
● ●●●
●
●
●
● ●
●●
● ●
●●●●
●
● ●●
●
●
●
●
●
●
●
●
● ●●
●●
●
●●
●
●
●
●
●
●
●
●●
●●
● ●●
●
●
●
●
●●● ●●● ●●
●
●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ● ●● ●●● ●
●
●
●
●
●
●
●
●
●
●
●● ● ●
●
●
● ● ●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●●●●●●
●
● ●●
●●●● ●
● ●
●●●● ●●●
●
●●
●●●●●
●●●
●●
●
●● ●
●
●
● ●
●●
● ● ●● ● ● ●●
●
●
●● ●
●
● ●●
● ●●
●●●
●
●
● ●●●
●●●●
●● ● ● ● ●
●●●
●
●
●
● ●
●
●●●●●●●
●●● ●
●●
●
●
●●●
●
●
●
●
●
●●●●●● ●
●●
●●●●●● ●
●●
●
●●●
●
●●
●
●● ●●
●●
●
● ●●
●
●●
●●
● ●
●●
●
●
●●
●●●
●● ●
●
●
●
●
●
●
●●
●
●
● ●●
●
●
●●
●
●
● ●
●
●
●
● ●
●●
●●●
●
●
●
●
●
●
●●
●●
●●●●●
●
●
●●
●●
●●
●
●●
●
●
●
●
●●
●
●
● ●● ●
●●
●
●●●●●●●
●●
●
●● ●
●
●●
●
● ●●●
●●
●
●
●●
●
●
●
●
●●●
●●●
●
● ●
●●
● ●●
●●
●
●
●
●
●
●
●●
●●
●●
●●
●●
●●●●
●
●
●
●●
●●
●
●
●●
● ●●
●●
●
●● ●●●●
●
●●
●
●
●●
●
●
● ●● ● ●●●
●
●
●●
●
●
●
●
●
●● ●
●
●
●
●
●
●●●
●●
●
●
●●●
●
●●●
●●●●
●●●
●
●●
●
●●
●
●●
●●
●●
●●
●●●
●
● ●●
●●
●●●●
●●
●●●
● ●
●
●
●●
●
●●●
●
●
●●
●●
●
●
●
●●
●●
●●
●●●
●●
●
●
●● ●
●●
●●
● ●●●●●●
●
●●●
●●
●●
●
●●
●●●
●
●
●●
● ●● ● ●
●
●
●
●
●●●
●●●
●
●●
●
●●
●
●
●
●●
●
●
●
●●● ● ●
●●
●●●●
●
●
●●
●
●
● ●
●●
●
●
●
●
●● ●●●●●●●●●●●●●
●
●
●
●●
●
●●
●●
●
●
● ●
●●
●
●●
●
●
●●
●●●
●
●●
●
●
●
●
● ●
●
●●
●●
●● ●●
● ●
●
●●●
●
●●●
●
●●
●
●
●
●
●●
●
● ● ● ●●●
●
●●●
●
●
●●
●●●
●
●
●
●●
●●
●
●●
●●
●
●●
●
●
●
●●●
●
●●
●●●●
●
● ●●
●
●
●
●
●
●● ●
●
●
●●
●●
●●
●
●
●●
● ●●●●
●
●
●
●
●
●
●
●●
●
●●●
●●
●
●
●
●
●
● ●
●●●●●●●
●
●
●●
●
●
●
●●●●
●
●
●
●
●●
●
●
●
●●
●
● ●●
●●
●●
●● ●
●
●
●
●
●
●
●
●●●
●●
●●
●●●●●
●●●
●
●●
●
●
●●
●●●
●●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●● ●● ●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●●
●●
●●●
●●
● ●●
●
● ●
●●
●●
●
● ● ● ●●●●
●
● ● ●
●
●●●
●●
●
●
● ● ● ●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●●
●
●
●
●
●
●●●
●
●
●●●
●
●●
●
●
●
●●
●
●●●
●
●
●
●●●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●●
●
●●●●
●
●
●●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●
● ●
●
●
●
●●
●
● ●●●●●●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●●
●●●
●●
●
●
●
●●
●
●
●●●
●
●
●●
● ●
●●
●
●
●
●
●
●●
●
●
●
● ● ●●●●●●●●●
●
●
●●
●
●
●●
●●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
● ● ●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●●●
●●
●●
●
●
●
●
●
●
●
●
●●
●
● ●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●●
●
● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●●
●●
●
●
●●●
●
● ●
●●
●
●●
●
●●●
●
●●
●
●●●
●
●
●
● ●●●●●●●●
●●
●●
●
●●●
●
●
● ●● ●●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ● ● ● ● ●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
● ● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●●●● ●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●●●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
● ● ● ● ●●
●
●
●
●●●●●● ●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●● ● ● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●● ●● ●●●
●
●●●
●
● ●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●●●
●
●
●
●
●● ●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
● ● ● ● ● ●●●●●●●●●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●●●
●
●
●
●●
●●
●
●●●
●
●●
●
●
●
●●●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●● ● ●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●● ● ● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
● ● ● ● ●● ●● ●●●● ●●●●●●●●●●●●
●
●●
●
●
●
●●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●●●●●●●●
●
●
●
●●
●
●
● ●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
● ●●●●●●●
●
●
●
●●● ● ●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●● ●●
●● ●●●●●●
●
●
●
●
●
●
●●
●
●
●●●
● ●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●●●
●
●
●
●
●
●
● ●● ●● ●● ●● ●●●● ●●●●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●●●●●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●●●●●
●●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● ● ●●●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
● ●●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●●
●●
●● ●
●●
●
●
●
●
● ● ●● ●● ●●●●●●●●●● ●●●●●●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ● ●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
● ● ● ● ● ● ● ●● ●●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
● ●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
● ●● ●●
●
●●
●●●
●●●●●●●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
● ●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●● ●●●●●●●●●●●●●●●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●●
●
● ●●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●●●●
●
●
●●●●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●
● ●●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
● ●● ●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●●
●
● ●● ●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ● ● ● ●● ●●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●●
●●
●
●
●●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●● ●
●
●
●
●
● ● ●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●●
●●
●
●●●
●
●●●
●
●● ●● ●●●●●●●●●●●●●●●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
● ●● ● ●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●● ● ●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●●
●●
●
●
●
●
●
●
●● ●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●● ●● ●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●●●●
●
●
●●
● ●● ●
●●
●
●●
●
●●
●●
●
●
●●
●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
● ●●
●●●●●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●●
●
●●
●
●
●
●
●●
●
● ●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●●
●
●
●
●●●●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
● ●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●●
●●●●●
●
●
●●
●●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
● ●●
●●
●
●●
●
●
●●●
●
●●●●●●
●
●●
●●
●
●
●
●
● ●
●
●●
● ●●●●●●
●● ●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●●
●
●
●●● ●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●●●
●●●●●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●
●●
● ●●
●●●
●
●
●
●●
●
●●●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●●
● ● ●●
●
●
●●
●
●●●
●
●
●●●
● ●
●
●
●
●
●●
●
●●
●
●
●
● ●
●
● ●●●
●
●
●●●
●●●
●
●● ●●
●
●
●
● ●● ●● ●
●
● ●
● ●●
● ●● ●●●
●
●●
●
●●●
●●
●●
●●
●●
●
●
●● ● ●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
● ● ●●
●●
●
●
●
●●
●
●●
●
●●
●●●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●●
● ● ● ●●●● ●
●●
●
●●
●
●●
●
●
●
●
●
● ●●
●●
●
●
●●
●●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●●
●●●
●
●
●●
●
●●
●
●●
●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●●●
●
●
●●
●
●
●
●● ●●● ●●● ●● ●●●●●●●
●
●●●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
● ●
●●
●
●
●●●
● ● ● ●●
●
●
●
●●●
●
●
● ●●
●● ●●●●●
●●●●●
●
●
●
●
●●
●●
●
●
●●
●
●
● ●●
●
●
●
●●
●
●
●
●
●
●
●
●
●● ●●●● ●●●
●
●●
●
●
●
● ●●
●● ●
●
●
●
●●
●
●
●
●●●●●
●●●
●
● ●
●●
●● ●●●●
●
●●
●
●
●●●
●
● ●
●
●
●● ●
●
●
●
●
●
●
●●
●●●
●
●
●
● ●
●
●●
●●●●
●●
●
●●
●
●●●●
●● ●
●
●
●
●●
● ●●●
●●●
●
●●
●●
●●● ●
●
●
●●
●
● ●●
●●
●● ●●● ●●●●●●●●●●
●
●●
●●
●●
●●●
●
●
● ●● ●
●
●●
●●
●
●●
●●
●
●
●
●● ●●●
●
●● ●●●
●
●●
● ●●
● ●●
●
●●
●●●●
●
●
●
●
●
●
●
●●●●
●
●
●
●●
●●
●
●●● ●
●
●● ●
●●
●● ●
●
●
●
●●
●
● ● ● ●●●●
● ●
● ●●
●●●●
●●
●
●
●
●
●●
●●
● ●● ●
●
●
● ●
●
●●● ●
●●
●●●● ● ●
●
●●●●●
●● ●●
●
●●
●●●
●●
●
●●●
●●
●● ●●
●
●
●
●
●●
●●●
●
●●
● ●●●●●
●●
●●●
●●●●●●
●
●
●
●
●
●
●● ●
●● ●●●
●●●
● ●●
●
●●
●
●● ●●
●
●● ●
●
●● ●
●●
●
● ●● ●●●
●
●
●● ●●●
●● ●
●●●●
●●
●●
●
●●
● ●●
● ●
●●●
●●●
●●●●●
●●
●●
● ● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
● ●●
●
●
● ●●● ●●
● ●
●
●●
● ●
●● ●● ●
●● ● ●
●
●●●
●
●
● ●● ●
●
●
●
●
●
● ●●
●
● ●●● ● ●●
●
●
●
● ●●
●
●
●
●●
●● ● ●●●●●● ● ●●
● ●
●● ●●● ●● ●● ● ●●
●
● ●●● ●● ●●
●● ● ●
●
●
●
●●● ●●●
●
●
● ●●
●
●●
●●
●
● ●●
● ●
●●
●●
●
●
●
●
●● ●●●●●●●● ●● ● ●● ●● ● ●●●●
●
●●
●
●
●●●
●● ●● ●●
●
● ● ● ●● ●●
●●
●●
●
●
●●●● ●●●● ●
●
●● ●● ● ●
● ●
●● ●●
●
●●
● ● ●●
●
●
●●●
● ●
●
●
●
●
●
● ●●
●● ●
●
●
●● ● ●
●
●●●
●
●
●
●
●●
●●
●
●
● ●
●● ●
●
●
● ●● ● ● ●
●● ● ●
●● ●●
●● ●●
●
●●●
● ● ●●●
●● ●●
●
●●
●
●
●●
●
●
●●
●●● ●
●
●
● ● ●
●● ●
●● ●●●
●
● ●●●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●●
● ● ● ●●●
●●
●
●
●
●●
●●
●●
●
●
●●
●
● ●
●●
●
●
●
●
●●
● ●
●
● ●● ●
●
● ● ● ● ●● ●●
●
●
●
●
●●
●
● ●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
● ●●
●●
●
●●
●
●
●● ●
●
●
●
●●
●
●
● ● ● ●
●● ●
●
●
● ● ●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
● ● ●● ● ●
●
●
●●
●●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●●
●
●● ●
●
●●
●
●
●●
●●
●
●
●● ●●
● ●●
●
●●
●
●
●
●
●
●
●
●● ● ●
●
●●
●●
●●
●●
●●
●
●
●
●● ● ●
●
●
●
●
●
●
●
●
●
●
●
● ●● ●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●

●
●
●
●
●

●
●
●

●
●
●

●

●

●
●
●

●
●●

●
●

●

●
●

●
●

●

●
●

●
●

●

●
●●
● ●●

●
●
●●

●
●

●

●
●●

●

●
●
●●

●
●
●

●
●
●
●
●

●

●
●●●

●

●
●

●
●
●
●
●

●

●

●

●

●●

●

●
●
●
●

●
●
●
●
●
●

●
●
●
●
●●●
●

●
●●● ●

−5

●

●
●
●
●
●

●
●
●●
●
●
●
●
●
●

●●
●● ●
●●●
●●
●

0

●
●
●●

logFC

●

●●

●
●●
●

●
●●●

●●
●●

●
●
●

5

●
●
●
● ●
●
● ●
●
●●
●●
●
●
●
●
●
●●
●
● ● ●
●●
●
●●
● ● ●●
● ●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
● ●
● ●
●
● ● ● ● ●
●
●
●
●
●●
●
● ● ●● ● ● ●●● ●
●●
● ●
●●● ●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●●
● ●
●
●
●
●● ● ●
● ●●
●
● ●
●
●
●
●
●
●
● ●
● ● ●●
●
●
● ● ●●● ● ● ●
●
●
● ●
●
●
●●
●●
●
●
●●
● ●●
●
●● ●
●
● ●
●
●
●●
●
●
●
●● ●●
●
● ●●● ● ●
●
●
●
●
● ●● ●●●
●
●
●
●●
●●● ●
●
●
●
●●
●● ● ●● ● ● ●●
●
●
●
●
●●
●
●
● ●●
● ●
●
● ●
●
●● ●
●
●●●
●●
●●
●●
●
●
● ●● ●
●
●
● ●● ●
●●
● ● ●● ●●
● ● ●
●●●
●●
●
●●
●●
● ● ● ●● ●
●
● ● ● ●● ●
● ● ●●● ●●
●●
●
●
●
●●
●
●●
● ●●
●
●
●● ●
●
●●
●
●●
●●
●
●
●●●
●● ● ●●
●●
● ● ●● ● ● ●●● ●
●●
●●● ●
●
●●
●
●●
●
●
●●
●
●●●● ● ●
●
●
●●
●●
●●●●
●
●
●
●●● ●●● ●
●
●
●●●
●●●
●●
●
●
● ● ●●
●
● ●●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
● ● ●● ●● ●
●●● ●
●
●
●● ●
●●
●
● ●
● ●
●● ●●
●●●
●●
●
●
● ●
●
●
●● ● ●●● ● ●
● ● ● ●●●
●●●
●●
●
●
●
● ●●●●●
● ●
●
●●●
●
●●
●●
●
● ●●●●●●
●●
●
●●● ●● ● ●
●
●●
●●
●
●● ●
●
●●●●●● ●
●● ● ●
●
●●●
●
● ●●● ●● ● ● ●●●
●● ●●
●
●
●●●
●
●●
●
●● ● ● ●
● ●
● ● ● ●● ●
●●
● ● ●
●
●
● ● ●
● ●● ●
●● ●
●
●
●
●
●●●
●●
● ●●●●
●● ●●
●
●●●
●
● ●● ●
●
● ● ●
●●●● ●
●● ●
●
●●●
●
● ●● ●● ●
●
●
●●
●
●●
●●
● ● ●
●●
● ●●
●●
●●
● ●●●●●●
●●
● ●
● ●
● ●● ● ●
●●
● ●
●● ●
● ●● ●● ●
●●●● ●●●●●●
●
●●●
●●
● ● ●●
●● ● ● ● ●
●
● ●●● ● ●●
●
● ●●●
●●
●●●
● ●●●
●●
●●●●
●
●
● ●●●●
●● ●●●●●●●
● ●
●● ●
●
● ●●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●● ●
●
● ● ●●●
●● ●
● ●● ●●
●●● ●●●●●
●
● ●
● ●● ●
●●
●
●●● ● ● ●
●● ●●●
●●●●
●
●●
●●●
●● ● ●●
●●●
●● ● ●●●
●●● ●●
● ● ●●●
● ● ●
●●●●
● ●
● ●●●
●●● ●●
●
● ●
●●● ●
●
●
●
●●●
●●
●●●●
●● ● ●●
●●●
●●
●● ● ●●●● ●●●● ●●●
●●●●
● ●●
●●●● ●
●●●●●
●
● ●●●
●
●●● ●
●
●●
●●
●●●●
●●●● ●●
●●●●
●
● ●
●●●● ●● ●●●
●
● ● ●● ●
●●
●●●●
●●●
●
●
●●
●
●
●●●●
●
●●●● ●
●
●
●
●●
●●
●
●●● ●
●
●● ●
● ● ●●●
●
●●
●●
●●
●
● ●●●
● ● ●●
●●●
●
●
● ●●
●
●●
●
●●●●
●●
●●
●
●
●
● ●●
●
●
●●●
●●●
●
●
●●
●●
●● ●●●●
●
●
●●●
●
●
●
●●●●
●●
●●
●●●
●●
●●
● ●●●
● ●● ●
●
●
●
●●
●
●●
●
●
●
●●●
●●
●●
●
●●●●●●
●●
●
●
●●
●
●●
●
●●
●
●
●●●●●●
●●
●
●
● ●●
●
●● ●●
●●
●●●●●●
●
●●
●●
●
●
●●
● ●● ● ●● ●
●
●
●●
●●
● ● ●●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●●
●●●●●
●
●●●
●
●●●●
●
●
●
●●●●
● ●
● ●
●●●
●●
●
●
●
●●
●●●●● ●●
●
●
●
●
●
●●
●●
●
●
●●●
●
●●
●●●●●
●
●
●
●●
●●●
●
●●
●
●
●
●●
●●●●●
●●
●
●
●●
●●
●
●
●●
●
● ●
●
●
●●●
●
●●●● ●●●●
●
●●
●
● ●
●
●●●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●●
●
●
● ●● ●
●
●●●
●
●
●
●●●
●
●
●●
●●●●●●●●●
●
●
●
●●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●●
●
●
●●
●
●
● ●●
●
●●
●
●●●
● ● ●● ●●●● ● ●●●
●● ●●
●
●
●●
●●
●
●
●●●●
●●
●
●
●
●●●
●
●
● ●●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●●●●
●●●●●●●●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
● ●●
●●●●
●●●●
●
●
● ●●●
●
●
● ●●● ●
●
●●
●
●●
●
●
●
●●●●
●
●
●
●
●●
●
●
●
●
●●●●● ●
●
●
●
●
●
●●● ●
●
●
●●
●
●
●●
●
●
●
●
●●
●● ●●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●●
●●●●●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●●●
●
●●
●
●
●●
●
●
●
●
●
●
●
● ●
●
●●
●
●
●●●●
●
●
●
●●
●●
●●●
●
●
●
●
●
● ●
●
●
●
●●
●
●●
●
●
●●
●
●
●●●●●●
●
●
●
●
●
●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●●●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●●
●●●●● ●●●●
●● ●●●●●● ●
●
●
●●
●●●
●● ●
●●
●
●
●
●●●●
●
●●
●●
● ●
●● ●
●●
●
●
●●●
●●●●●
●●●●●
●
●●
●● ● ●
●
●
●
●●●●●●●●● ●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●●
●●
●● ●
●
●
●●
●●●
●
●● ●●●●
●
●
●
●●●
●
●
●●
●●
●●
●
● ●●●●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● ●●●
●●
●
●
●●
●
●●
●
●
●
●●●
●●
●
●
●
●●
●
●●
●
●
●●
●● ●
●
●
●
●
●
●
●●
●●
●●
●
●●
●
●●
●●
●●
●
●
●●
●
●
●●
●●
●●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●●
●
●
●●●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●●●●●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●●
●
●●●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●●●●●
●●
●
●
●
●
●● ●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●●●●●
●●●
●●
●●
●
●
●
●
●●
●
●
●
●●●
●
●●●
●
●
●
●
●●
●●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●●●●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●●●●●
●● ●● ●
●●●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●●
●
●● ●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●● ● ●● ● ●
●●●
●
●
●
●●
●
●
●
●
●
●
●●●
●● ●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●●
●
●
●●
●●
●●●●●
●
●●●
●●●●
●●●●
●
●●●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●● ●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●●●●
●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
● ●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●●● ●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●● ● ●●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●●
●
●
●
●
●●
●
●
●
●●●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
● ●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●●●
●●
●●●
●
●●
●
●●●
●●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●● ●●
●
●●
●
●●
●●
●
●
●
●
●
●●●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●●
●●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●●●
●
●
●●●●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●●
●
●
●
●
●
●●●●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●● ●●
●●●
●
●
●
●●●●
●●●
●●
●●●
●
●●
●
●●
●
●●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●●
●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
● ●●●●
●●
●
● ●●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●●●●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●
●
●
●
●
●●
●●●
●
●
●●
●
●
●●
●●
●
●●●
●●●
●
●
● ●
●
●
●
●
●●
●
●●●
●
●●
●●
●
●●
●●●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●●
●●
●
●
●
●●
●
●●
●
●●
●●●●
●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●●●●●●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
● ●● ● ● ●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●●
●●●
●
●
●●
●
●
●●●●
●
●
●
●●●●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●●
●
●
●●
●
●
●
●●●
●
●
●●
●
●●
●
●
●
●
●
●●●
●
● ●
●
●
●
●
●
●
●
●
●● ●
●
●
●●
●
●●
●
●●
●●●
●
●
●●
●
●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●●●●●●●●●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●●
●● ●●●●●●●
●
●●
●●
●●
●
●●
●
●●●●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
● ●●●
●
●
●
●
●
●●
●
●●
●
●
●●
●●
●
●
●●
●●●
●
●
●
●● ●●●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●●
●●
●
●●
●
●
●● ●●
●●
●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●●
●
●●
●●●
●
●● ●●
●
●
●●
●
●●
●
● ●●
●
●
● ●
●
●●●
●●
●
●
●
●
●
●
●
●
●●
●
●●
● ● ●●
●●
●
●
●
●●
●●
●●
●●
●●●
●●
●
●
●
●●
●
●
●
●
●●
●● ● ●● ●
●
●●
●
●
●●
●
●
●
●●
●● ●●
● ●
●
●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●● ●
●●
●
●
●
●
●
●
●
●●
●
●●
●● ●●
●●●●
●
●●
●
●
●
●
●
●
●
●
●●
●●●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●
●●
●
●
●
●
● ●
●●
●●●●●
●
●
●●
●
● ●●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●●
●●
●
●●
●●
●
●
● ●●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●●●● ●
●
●●
●
●
●
●● ●
●●
●
●●
●
●
●
●●
●
●
●●
●●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●●●●
●
●●
●
●●●
●●
●
●
●
●
● ●●
●
●
●● ●
●
●●●●
● ●
●
●●●
●●
●
●●●●
● ●●
●●●●●
●●
●
●
●
●
●
●● ●●●●●●●●
●●
●
●●●
●
●
●●
●
●
●
●●●
●●
●
●
●
●
●●
●
●
●
●●
●●●●
●
●●
●●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●●●●
●●
●
●●
●
●
●
●
●
●
●●
● ●
●
●
●
●●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●●●
●●
●
●●
●
●●
● ● ● ●●
●
●
●
●
●
●
●
●●
●
●
● ●
●●●
●
●
●●
●
●●
●● ●●
●
●●
●
●
●●
●
●●●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●
●
●●●
●●● ●●
●
●●●●
●●
●
●●
●
●
●●
●
●
●●●●
●
●
●●●
●●●
●●
●
●
●
●
●
●● ●
●
●
● ● ● ●●
●
● ●●
●
●
●
●●●●
●
●● ●●
●
●
●●
●●
●
●●●●●●●●
●
●
●●
●●
●
●●
●●
●
●
●●●
●●
●
●●●
●
●
●●
●●
●
●
●●●●●
●
●
●
●●
●●
●● ● ●
●● ●
● ●●
●
●
●
● ●●
●●●●
●●
●
●
●
●
●
●
●●●●
● ●●
●●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●●●
●
●●●
●
●●
●
●
●●●●
●
●●● ●●
● ●●●●
●
●● ●●
●●
●
●●●
●●
● ●
●
●●●●●
●
●●●
●●
●●●● ●
●●
●●
●
●
●●
●
●●●
●●●●
●
●●●
●
● ●
●
●
●
● ●
●●● ●
●●
●● ● ●
●
●●●
●●
●● ●
●●
●
●●●
●
● ● ●●●
●●●
●
●
●
●●
●●●● ●
●
●
●●
●●●
●
●
●
●●
●●
● ●●
●
●
●● ●
●●
●
●
●●
● ●●
●
●●● ●
●●
●●●
●
●●
●●
●
●●
●●●●●●
●●●
●
●
●●
●●
●
●●
●
● ● ●●
●● ●
● ●
●
●
●● ●
●
●●●
●
●●●
●●
●●●●●●
●●●●
●
●
●●
●●
●
●●●●●
● ● ●● ● ● ● ●
●
●
● ●●
●●
●●● ●
●
●●●● ●
●● ●●●●●●●
●● ●
●
●●●
●●
●●
●●●● ●●
●
●●
● ●● ●
●●●
●●●
●
●●●● ●
●
●
●
●● ●
●●
●●●●
●●●
● ●
● ●●
●●● ● ●●
●●
●●● ● ●
●●
●●●●●●● ● ●
●
● ●●
● ● ●●●
●
● ●● ●
●●●
●● ● ●
● ●
●●●●
● ●
●● ●
● ●●
●
●●●●
●●●●● ●●
●
●●●●● ●● ●●●●
●
● ● ●● ● ●●
● ●●●
● ● ●
●
●●
●
●●●●● ●
● ●●●
● ●
●● ●
●●●●● ●●●●
●●
●
●●
●
●
●●
●● ●
●●●●
●●
●●●● ●
●
●
●●● ●
●● ● ● ●
●
●●●● ●
●●
● ●● ●●
●●● ●●●●●●
●● ●
●● ● ●●●●
● ●
●
●
●●●
●●●●●● ●
●
●●● ●
●
●
●
● ●
● ●●● ●●●●●● ●
● ● ● ●● ● ● ●
● ●
●●
●
● ●● ●● ●
●●●● ●
●
●●●
●● ●
●●●
●
● ●
●
●●
●●●
● ●●
●
●●●● ●
●
●●
●
● ●
●
● ●●
●
●
●●●
●
●●● ●
●●●
●
●
● ●●
●
●
●●● ●
● ●
●●●●
●●●●●
●
●
●
●
●
●●●
●●
●
●
●●●●
●
● ●●● ●
●●
●
●
●
●
●●
● ●
●
●
●
●
●
●●●
●●●●
●
●
●
●
●
●
●●
●
● ●
●
●
●
●
●
●●●●
●
●
● ●
● ●●
●●
●
●●
●
●
●●●●
●
●
●
●● ● ●
●
●
●
●
●●● ●
●
●
●
●
●●
●●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●

●

●

●
●
●

●
●

●

5

●

5

●
●
●

●
●

0

●

●
●
● ●

●
●

●
●
●

●

logFC

●
●

●

●
●

●

●

●
●

●

●

●
●

●●

●

●

●
●●

●
●

●

●

●

●
●

●

●
●

●
●
●

●

●
●

●
●●

●
●

●

●

●
●

●

●

●

●

●
●

●

●

●
●

●
●

●
●

●
●

●

●

●
●

●

●
●

●

●

●
●
●

●
●

●

●

●

●

●

●

●

●

●
●

●

●

●
●

●

●

●

●

●
●●
●

●

●

●

●
●
●

●
● ●
●
●
●
●
● ● ●● ●

●
●●

●
● ●
● ●

●

●
●

●

●

●
●
●
●
●
● ●

●● ●
●

● ●

●

●●

●
●●

●

●
●
●

●

●

●
●

●
● ●●●● ●
●
●
● ●
●●
●
●● ●
●●
●
●● ●
●
●● ● ●
●
● ●
● ● ●
●
●
● ●●●
●● ●
●
●
● ●
●
●
●
●
● ● ●●
●
●● ●
●●●
●
● ● ● ●● ●● ●
●●
● ●
●
●
●● ● ●
● ●●● ●●●
●
●
●
●● ● ● ●●
●●
●
●
● ● ● ●● ●●● ●● ●
●
●
●
● ●
●
●●
●
●
● ●
●●
●
●
● ●●
●
●
● ● ● ● ●●
●
● ●●
●
●
●
●●
●● ● ● ● ●
● ● ●
● ●●
●
●
● ●●●● ● ● ● ●●
● ●
●
● ●
●●
●●
● ● ●●●●● ● ●●● ●● ●●●
● ●
●● ●
●
●
●
●
●●
●●● ● ●
●
●
●
● ● ●
● ● ● ●●● ●● ● ●
● ●●
●●●
●
●●
●● ●
●●
● ●
● ● ●● ● ●
●
●
●● ● ● ●
●
●●
●●●
●
●●
● ●●
● ●
●
●●
●●
●
●
● ●●●
●
● ● ● ● ● ● ●●● ●●●
●
●● ●● ● ●
●
●
●
●
●
●
● ●●
● ●●
●
●
●
●
● ●
● ●●●●
●
●
●● ●
● ●● ●●●
● ●● ●●
● ●●● ●
●●
●●
●
●
●
●
●
●
●●● ●●
●●●
● ●●●
●
●
●●
● ● ● ●
●●
●
●
●●● ●●●
●
●
●●
●
●
●
●●
●●● ●
●●●
●● ●●●● ● ●
●●●●
●
● ●●● ●●
●
●
●●●● ●
● ●●●
● ●● ●●●
●
●
●● ●
●
●●●●
●●●
●
●●
●
● ● ● ●●●●●●●● ●●●●
●●● ●●
●
●
●
●
● ●
●●●●
●
●
●
●● ●●●
●●
●
● ●●●●
●
●
●
●
●
●
●
●●
●●●
● ●
● ● ●● ●●
●●●●●● ●●
● ●●
●
●●●
●●
●
●
● ●
●●
●
●
●
●● ●
●●
●●
●
●
●
●●●
●
●
●
● ● ●● ●
●
●●
● ●
●
●
● ●●
●
●
●
●
●
● ● ●●
●
●●
● ●
●
●●
● ● ●●
●● ●
●●
●●●
●
●●
●●●●
●
●
● ●●● ●●●
●
●●●
●
●
●
●
●●●●● ●● ●●●●
●
●
●●
● ● ●
●●●
●
●
●
●●
●
● ● ●●
●
●
●
●
●●
● ●● ● ●
●
●
●
●
●●
●●●●
●●
●
●
●
●●
● ● ● ●●
●
●
● ●●● ●●
●●
●
●
●●
●
●
●●●
●
●
● ●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●●
●
●●
●●●
●
●
●●●
●●
●
●
●
●
●●
●●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●●
●●
●
●●
●
●●
●
●
●●
●
●● ●●
●●
●
●
●
●●
●
●●
●
●
●●
●
●
●●●●●
●●●
●
●
●●
●●● ●
●
●●●●●
●
●
●
●●● ● ●●●●
●
●
●
●
●
●
●●●
●●●
●
● ● ●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●●●●●
●
●
●
●
● ● ●● ● ● ●●●●● ●●●
●
●●●●● ● ●● ●
●
●
●●
●
●
●●●
●
●●●
●●
●
●
●
●
●
●●●
●● ● ●
●●● ●●● ●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●● ●
●●
●
●
●●
●
●
●●●●●
●
●
●
●
●●
●●●
●
●●
●
●●●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●●●●
●
● ●●
●●
●
●●
●
●●
●
●●
●
●
●
●
●
● ●● ●
● ● ● ● ● ● ● ● ● ●● ● ● ●●●●●● ●●●
●
●
●●●●
●●
●
●
●
●
●
●
●●●
●●
●
●●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●●
●●
● ●●
●●
●
● ●● ●●●●
●
●
● ●
●
●
●●●●●●
●
● ●●●● ● ●●
●
● ●●
●
●●
●
●
●
●
●
●●●
●
●
●● ●
●
●
●
●● ● ●●
● ●
●
●●●
●
●●
●
●
●●
●●
●
●●
●●
●
●
●
●●
●
●●●
●
● ●● ● ●● ● ●
●
●
●
●
●
●●●
●
●●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●●●
●
●●
●●●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●●
●
●
●●
●
●
●●
●
●
●
●●
●
●●
●
●●●
●●●●●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●●●
●
●●
●●●
●
●
●
●
●
●
●●●
●
● ●●●●● ●
●
●
●
●●
●
●
●●
●
●
●●●
●
●●
●
●●●
●●
●
● ●●●
●●●● ●
●
●●
●●
●●●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●●
●●
●
●●
●
●
● ●●
●
●
●
●●
● ●
●
●
●
●●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●●● ●●●●
●
●●
●●
●●
●●●●●
●
●
●●
●●●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
● ●● ●●●
●
●
●
●●
●
●
●
●
●
●● ●●●
●●
●
●
●
●●
●
●
● ●● ●
●●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●●
●●●
●
●
● ● ● ● ● ● ●●●● ● ●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●
●●
●●
●
●
●
● ●●●
●●
●●●
●● ●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
● ● ●●
●●●
●
●
●●
●
●●●●●
●
●
●●
●
●
●
●
●
●
●●
●
●●
●
●
●●●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
● ● ●●
●
●
●
●
●
●
●
●
●
●
●
● ●●●●●
●
●
●
●
●
●
●
●
●
●
●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
● ●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
● ●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●
●●
●
●●●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●
●●
●●
● ●
●
●
●
●
●
● ● ● ●●● ●●●●●
●
●
●●●●●●● ●●●● ●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
● ● ● ● ●●● ●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
● ● ●●
●
●
●
●
●●
●
●●
●
●●●
●
●●
●
●
●
●
●
●
●
● ●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
● ●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
● ●●
●
●
●
●
●
●
●
●
●
● ● ●●
●
●
●●
●●● ●●●●● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
● ●●●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●●●●
●●
●●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
● ● ● ● ●●● ●●●●●●
●
● ● ●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●●●
●
●
●●
●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●●●
●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
● ●
●●
●
●
●●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●
●●
●
●●●●
●
●
●●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
● ● ●●●●●
● ● ●●●● ● ● ●●●●●●●
●
●
●
●
●●
●
●● ●
●●
●●
●
●
●
●
●
●
● ●
●
●●
●
●●●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●
●●●
●
●●●●●●● ●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●●●
●
●
●
●
●●
●●
●●●
●●
●●
●
●
●
●
● ●
●
●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●●●●
●
●
● ●●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●
●
●●
●
●
●
●●
●
●
● ● ● ●●●●●●●
●●
● ● ● ● ● ● ●●●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●●●●●
●●
●
●
●
●●● ●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●
●
●●
●●●●●●
●
●
●
●
●
●
●●
●
●
●●●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●●
●
●
●●
● ●● ●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● ●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●●●
●●
●
●
●●
●●
●●
●
●
●●
●
●●
●●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●
● ●
●
●
●
●
● ●●●
●
●●●
●
●
●●
●
●●
●
●
●●
●●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●● ●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●●●●
●●●
●
●
●●●●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●●
●
●●
●
●
●●
●
●●
●
● ● ●●
●●
●
●
●
●
●
●
●
●
●
● ● ●●●● ● ●●● ● ●●●●●●● ●
●
●
●●
●●●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●●
●
●●
●
●
●●
●
●
●
● ●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●●
●
●●
●
●● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●●●
●
●●
●
●
●
●
●●
●
●●●
●
●
●
●
●●
●●
●
●●●
●
●●
●●●
●
●
●
●●
●
●●●
●●
●
●
●
●
●
●
●●
●
●●
●
●●
●●
●
●●
●
●
●
●●
●●
●●
●
●
●●●
●●
●●
●
●
●
●
●●
● ●●●
●●
●●
●●●●
●
●
●
●
●
●
●●
●● ●●● ●●
●
●●●●●
●●
●●
●
●
●
●
●
●
●● ●
●
●
● ● ●● ●
●●●
●●
●
●
●●
●
●●
●●
●
●
●
●●
●
●●●
●
●
●
●
●
●●●
●
●●
●
●
● ●● ●●● ●●
●
●
●
●
●●
●●● ●
●
●
●
●
●
●
● ●●●● ●● ●●● ●●●● ●
●
●
●●●
●●
●
●●●●●
●
●
●
●●●
●●●●●●
●
●
●●●
●
●● ● ● ● ●●
●
●
●●
●●● ●
●
●●
●
●●
●●
●● ●●
●●●
●
●●●●
●
●
●
●
●
●●
●●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●●●
●
●
●●
●●●●●
●●●
●
●
●●
●
●●
●● ●
●
●
●
●
●
●●
●
●
●
●●
●●
●
● ●● ●
●
●●
●
●●●
●●●
●
●●
●●
●
●● ●
●
●
●
●●
●
●
●
●●
●● ●●
●
●●●
●
●
●
●●
●
●
●
●● ●●
●
●
●●●●●● ●●
●
●●
●
●●●
●●
●
●
●●
●
●
●●
●
●
●
●●
●● ●
●
●
●
● ●●
●●
●●
●
●
●
● ●●
●
●●
●●
●●●
● ●
●●
● ●
●
●
●●
●
●
●●
●
●
●●
●●●● ● ●●●●●
●
●●
●
●
●
●●
●●
●
●
●●
● ●●●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●●
●
●
●●
●
● ●
●
●
●●●
●
● ●●
●
● ●●
●●
●
●●
●
●●●
● ● ●
●
●
●
●
●●●
●
●
●
●●●●
●
●
●
●
●
●
●
● ● ●● ● ●
●●
●●●●● ● ●
●
●●
●
●●
●
●
● ● ● ●●
●
●
● ●●
●●
● ●
●
●
●●●
●
● ●
●●
● ● ●
●
●
● ●●
●●●
●
●
●
●●
●
●●●●
●
●
●●
●
●●●
●●●●●●
●
●
● ●
●
● ●● ●
●●
●
●
●
●
●
●●●
●●
●
●●
●
●
●
●●
●
●●
●
●● ●
●●●
●
●
●
●● ●●● ●●● ● ●
● ●● ●●●● ●● ● ● ● ● ● ●●●●● ●● ●●●●
●●
●●●●
●● ●
●
●
●
●●●●●●●●●
●●
●
●●
● ●●
●● ●
●
● ● ● ●● ●●
●
●
●
●
●
●
●
● ●●●●●●
●● ●
●●●
● ●●● ●
●
●
●
●●●●
●●
●●
●●●
●●
●●
●● ●
●
●●●
● ●●
●
●
●●
●● ●
●
●●●
●●●●
● ●●
● ●●
●
●
●
●●
●● ●
●●●
● ●●
● ● ●●●
●●
●
●
● ●●
●
●●
●
●● ● ●●●
●
●
●●
●
●
●●●
●●
●
●●
●●
●
●● ● ● ●●● ●● ●
●
●
●
●
●
●
●
● ●●●●
●●●
●●
●●● ●
●●● ● ● ●● ● ●
● ●●● ●
●●●
●
●
●
●
●
●
● ● ●● ● ●
●
●●
●
●
● ●
●●
●●
●●
●●●●●●●● ●● ●● ● ● ●
●●
● ●●
●● ●
●●●
●●●
● ● ●●
● ●● ●
●
●●
●
● ●
●
●
●●
●●
● ●
●● ● ● ●
●
●●
●
●
●
● ●● ●●●●
●● ●
●●
●●
●
●●
● ●
● ●● ●
● ●●
● ●●
●● ● ●
●
●
●
●
● ● ●●● ●
●●●
●
●● ●●
●
● ● ●● ● ●● ●● ●●● ● ●●●●●●●● ●●●
●● ● ●
●
●
● ●●
●●
●●
●● ●
●● ●●
●
●●
●●
● ● ●●● ●
● ●● ● ●●
●●
●
● ●● ● ●
● ● ●●
●
●
●
● ●
●●
●
● ●● ● ●
●
●
●●●
●
● ●
●
●
●
●
●
● ●●●
●
● ● ● ●
●
●● ●●● ●
●
● ●●● ●●
● ● ● ●● ●●
●
●
●
● ●● ●●
●
●
●● ● ● ●● ●
●
●●●●
● ● ● ● ● ● ●●●●
●● ●●
●
●●●
●
●●
● ●●
●
●
● ●● ●● ●● ●● ●● ●● ●●● ●
● ●
● ●● ●
●●
●
● ● ● ● ●●
●
●
● ●
● ●●● ● ●
●
●
●
●
●
●
● ●
●
●
● ● ●● ● ●
● ●
●●
●
●●
●
●
●
●●● ●
●
●● ●● ●
●● ●
● ●
● ●● ● ●● ● ●● ● ● ● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
● ● ●● ●
●
●
●
● ●
● ●●
●
● ●● ●●
●●
●
● ●●● ●
●●
● ●●●
●●
●
●● ●● ●
●●
●
● ●●● ●●● ●
●
●
●
●
●
● ● ● ●
●●
● ●
●
●●
●
● ●●●● ●●
●
●
●● ●
● ● ●
● ●
● ● ● ●● ● ●
●
●●● ●
●
● ●
●
●● ●
● ●
●
●
●
●●
●
●
●
●
●●
●●
●
● ● ● ●●
●●
●●
●
●●
●
●● ●
●
●
●
●
● ●
●●
●●
●●●
●
●
●● ● ●●
● ●● ●
●
● ●
●
●●
●●
●
● ● ●
●
●● ●●
●
●●
●
●
●
●
●
●
●
● ●●●
●
●
●●●
●
● ●●
● ●
● ●
●● ● ● ●
●
●
●
●
●
● ●
●
●
● ●
●
●●
●
●
●
●●●
●
● ●●
●
●
●
●
●●
●●
●
●
●●
● ●● ●
●
●●
●
●●
●
●
●
●●
●
●●
●
●●
●
●
●
●
●
●
●●
●
●●
●●
●● ●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
● ●
●●
●
●
●●
●
● ●
●
●
●
●
●
●
●
●
●
●
●

●

●

●

●

●
●

●
●

●

●

●

●

●
●

●
●
●

●
●

●

●
●

●

●

●
●
●

●

●

●
●
●
●
●

●

●

●

●
●
●

●

●

●
●

●

●

●●
●●

●
●

●

●

●

●

●

●

●

0

●

−5

●
●

ogFC

●

●
●

●
●

●

●

●
●
●

●

●

●

●

●

●

●

●

●

●
●

●

●

●
●
●
●

●

●

●

●

●
●

●

●

●

●

●
●

●

●
●
●

●

●

●

●

5

10

15

0

5

logCPM

15

0

15

10

15

10
ogFC

5

10

−10

−5

0

ogFC

−5
−10
10

15

(c)

5

10
5
0
−5

5

10
ogCPM

(b)

−10

0

5

logCPM

(a)

ogFC

10

●
●●
●●●

0

0

●
●●
●

−10

−10

−10

●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●

0

5

ogCPM

10
ogCPM

(d)

(e)

15

0

5
ogCPM

(f)

Figure 5.2: Di↵erential expression of homologous transcripts. Di↵erential expression in M. occu ta for (a) gastrula vs
neurula, and (d) neurula vs tailbud, M. ocu ata for (b) gastrula vs neurula, and (e) neurula vs tailbud, and the hybrid for (c)
gastrula vs neurula, and (f) neurula vs tailbud. Genes in red are di↵erentially expressed with a FDR <0.05

64

Only M. occulta embryos that expressed manx and p58 produced hybrids with urodele
features [39, 121]. To identify other candidate genes whose di↵erential expression in M.
occulta may contribute to the tail-less condition, we sequenced and assembled three developmental stages—gastrula, neurula, and tailbud—across M. occulta, M. oculata and their
interspecies hybrid. The tail-less species, M. occulta, showed the highest level of di↵erential
expression (FDR <0.05), with 260 (97%) of the identified di↵erentially expressed transcripts
up-regulated in the neurula stage relative to the gastrula stage. (Table 5.1). M. oculata also
had more transcripts up-regulated (65%) than down-regulated at the neurula stage relative
to the gastrula. Hybrids did not follow this trend; the majority of di↵erentially-expressed
hybrid genes are down-regulated at this stage (82%).
When comparing the tailbud stage to the neurula stage there was essentially no significant
di↵erential expression observed in M. occulta. A total of 5 genes were said to be di↵erentially
expressed (FDR <0.05). However a major shift in di↵erential expression was seen in both
M. oculata and the hybrid at the tailbud stage relative to the neurula stage (Figure 5.2).
There were 1170 and 1270 transcripts up-regulated, and transcripts and 129 transcripts
down-regulated in M. oculata and the hybrid respectively. That equates to a 10x increase in
di↵erentially expressed transcripts in both hybrid and M. oculata tailbud embryos relative
to the respective neurula stage embryos.

5.3.4

Overlap between hybrid and M. oculata alleles

When comparing transcript expression from neurula to tailbud in M. oculata and the hybrid
there is a total of 2,440 transcripts up-regulated collectively. Of these 2,440 transcripts, 328
(13%) overlap (Figure 5.3a). There were no transcripts in M. occulta identified as di↵eren-

65

tially expressed that overlapped with either M. oculata or their hybrid. Of the transcripts
that were down regulated, there was only an overlap of 5 transcripts (0.06%) between M.
oculata and the hybrid, and again none with M. occulta. We further examined their allele
specific expression to determine if the M. oculata alleles represented the majority of this gene
expression in the hybrids, or if there was some rescue of expression of the M. occulta alleles.
When analyzing the 328 up-regulated transcripts that overlapped between M. oculata and
the hybrid, there was large skew for expression from the M. oculata allele: 91.7% of expression came from M. oculata (Figure 5.3b). A similar but far less dramatic trend was observed
for the transcripts up-regulated in hybrid but not overlapping with M. oculata, with more
expression coming from the M. oculata allele. The allele specific di↵erential expression at
tailbud was 32% M. oculata, 36% M. occulta and 32% hybrid.

5.4

Discussion

Several genes identified as expressed in the notochord of C. intestinalis [30, 32] were observed
to be missing in both M. occulta and M. oculata; of these genes all were also identified as
being missing in the O. dioica [53]. The lack of expression of these gene is the Molgulids
and O. dioica implies these genes are not necessary for the development of a fully functional
notochord, CNS, or muscles in the ancestral chordate.
Here we present a di↵erential expression analysis of two closely related Molgula species,
M. occulta and M. oculata, using high throughput sequencing technology. We were able to
create gene models using both a mapping based approach and de novo assembly approach,
and then combined the assemblies for better transcript models. We showed that from neurula
to the tailbud stage there is a 10 fold increase in transcripts that are identified as di↵eren-

66

Hybrid only

Total

75%

Percent

Hybrid

M. oculata

M. ocu/hybrid overlap

50%

842

328

942

25%

cu

l

M

.o

cc
.o
M

cu
.o
M

N
eu
tra

cc

l

.o
M

cu

N
eu
tra

.o
M

.o

cc

l

0%
M

M. occulta

N
eu
tra

1

Allele

(a)

(b)

Figure 5.3: Upregulated transcripts overlap between hybrid and M. oculata. When
comparing the gastrula and neurula time point for M. occulta, M. oculata, and their hybrid,
both M. oculata and the hybrid showed di↵erential expression in at least 7% of their transcripts with the majority being up-regulated. (5.3a) There is a 15% overlap in overexpressed
genes M. oculata and the hybrid, for the transcripts overexpressed when comparing gastrula
to neural expression, and there is no overlap with M. occulta. (5.3b) When looking at the
allelic express for the same condition, there is a strong skew in expression coming from the
tailed allele for the unregulated transcripts that overlap between M. oculata and the hybrid.
The highest percentage of transcripts of allelic expression in the hybrid also comes from M.
oculata for up-regulated transcripts that do not overlap with M. oculata, but not to as great
an e↵ect. For genes’ overall allelic expression, the majority comes from M. occulta, 36%,
with 32% coming from M. oculata.
tially expressed (p=0.05) in both M. oculata and the hybrid. Using this same condition
we were able to show that there is almost no di↵erential expression in the the tail-less M.
occulta. This result correlate with the embryo morphologies during this transition; in M.
oculata, the processes of tail formation are occurring, which may explain the increased transcriptional activity. In contrast in M. occulta there are no noticeable morphological changes
occurring. In hybrid embryos, this dynamic regulation of gene expression at the tailbud
stage is rescued. Because hybrids show partially rescued tails and CNS structures, I propose
that these transcripts may have an important role in the formation of the tail. Furthermore,
of those transcripts upregulated at the tailbud stage in the hybrid but not in M. occulta, it
appears that expression is being restored from the M. oculata allele. This suggests that the
67

relative lack of di↵erential gene expression in the neurula-to-tailbud transition in M. occulta
may be due to loss-of-function of cis-regulatory elements controlling the expression of key
genes involved in tail and CNS formation.

5.5

Conclusion

We have used two closely related Molgulids with the ability to hybridize to study three
key time points in tail development: gastrulation, neurulation and tailbud. This study has
shown that from neurula to tailbud a number of transcripts are up-regulated, in both M.
oculata and the hybrid. Not only are transcripts up-regulated in this condition, with no
significant di↵erential expression in the tail-less species, but there is also an overlap between
the di↵erentially expressed genes in the M. oculata and the hybrid. Since both M. occulta
and M. oculata have a strong overlap in genes implicated in notochord development, this
di↵erential expression could be the possible cause of loss of the tail and notochord and CNS
in M. occulta.
The hybrids shed light on the fact that cis-regulatory elements are one of the key causes
for the lack of urodele features in the tail-less M. occulta. Previously, it has been shown that
Molgula have a high turnover in sequences of cis-regulatory elements, diverging in specific
binding sites, while conserving orthologous gene expression [115]. Perhaps, this cis-regulatory
turnover has led to the Molgulidae becoming more susceptible to the deactivation of binding
sites, which in turn contributes to the loss of tails in several species.

68

Chapter 6
Conclusions and Discussion
The chordate body plan is conserved throughout the phyla with only a few species deviating from the conserved body plan; at some point in their life, most chordates develop a
tadpole larvae containing a hollow dorsal neural tube, and a postanal tail containing a notochord flanked by muscle cells. However, it has been documented that 16 species out of the
approximately 3000 species of tunicates—one of the three subphyla of chordates—have independently undergone tail loss, with the majority being within the molgulids [2, 34]. Because
they have both tailed and tail-less species, molgulids are useful models for studying changes
in the development of body plans. Having closely related species with altered development
allows us to look at the short-term modifications that occur during the evolution of alternate
body plans, showing us that larval development can evolve rapidly. Of particular interest,
two molgulids have the ability to hybridize, which allows us to examine the mechanisms of
evolutionary change at an allele-specific level.

6.1

Evaluating a lightweight transcriptome assembly
pipeline on two closely related ascidian species

The drop in sequencing price has aided in our e↵orts to understand body plan development
by allowing us to quickly and inexpensively obtain transcriptome and genome sequences,

69

along with temporal expression profiles. However, for non-model systems with no existing
genomic sequence, we must first assemble the transcriptomes and genomes. The methodology
for assembly is not unambiguous: there are a number of di↵erent steps—quality trimming,
filtering, and choice of assembler(s), with several programs at each step and no clear choice.
Studies have been done to compare assemblers, but no one assembler is clearly the best
choice [10]. Many times assembly methods are chosen on the usability of software and the
availability of resources.
Factors that are most limiting for assemblies are memory, and in some cases time. The redundant nature of sequencing data allows for the removal of redundant reads with approaches
like digital normalization, but at what cost? Here in e↵orts to assemble the first two molgula
transcriptomes, we evaluate the cost of filtering sequencing reads for redundancy. We show
that abundance filtering quality trimmed assembly reads enables transcriptome assembly
with a reduction in both memory requirement and assembly time, while retaining essentially
the same information content (e.g. number of genes, average gene length, and homology to
the closest sequence species). We have demonstrated that the Oases and Trinity assemblers
return similar results, both suitable for downstream analysis using the full or redundancy
filtered dataset. Our pipelines are available so we also provide methodology to be used by
future researchers. In addition assembly metrics to evaluate assemblies are also an important step in the protocol. One of the standard metrics for evaluating assemblies is the N50,
but this is designed for genomic evaluation and does not clearly translate to transcriptomes
because of isoforms. For example, the same exon may be included in multiple splice variants,
inflating the total bp assembled; some assemblers may choose to report more isoforms than
others even with the same read support; and “total length” makes little sense for transcriptomes. In contrast, homology and programs such as CEGMA are more informative and
70

useful for downstream analysis, because they measure the information in the transcriptome
and cover the annotation step as well. When measuring recovered homologies and CEGMA
measurements we find that both assemblers and both treatments compare well.

6.2

Change in gene function and cis-regulatory binding
sites

In tunicates hox genes are not important for patterning along the anterior-posterior axis,
as they are in vertebrate and other bilaterians [17, 70, 37]. This limited function in the
hox genes is evident from the lack of noticeable phenotypic defects in the knockdown of C.
intestinalis hox genes, with the exception of hox10 and hox12. Hox10 is involved in neuronal
development and hox12 is involved in the formation of the posterior most portion of the the
tunicate larval tail [37]. In addition, tunicates have undergone loss and rearrangement within
the hox clusters, unlike most animals that have been studied [36]. There have yet to be two
tunicate species found with the same hox cluster configuration - there are typically changes
in ordering, duplications, and which hox genes are present (figure 4.2). Only C. intestinalis
has enough sca↵olding to completely identify the structure and intergenic spacing between
the hox genes. However, of the hox gene we found on the same sca↵olds in M. occulta,
M. oculata and M. occidentalis the average intergenic spacing is far smaller than in Ciona,
ranging from 10-25 kb in length. Without additionally sca↵olding we cannot fully examine
the rearrangements within the hox clusters of ascidians based on the molgulid sequences.
Perhaps soon after diverging from vertebrae and cephalochordates, the hox cluster genes were
co-opted into other GRNs, leading to higher lability than in species where they retained their
stereotyped roles.
71

We have also identified a number of enhancers between C. intestinalis, M. occidentalis,
and M. oculata that were either mutually unintelligible, or asymmetrically intelligible. We
propose that a high frequency of compensatory changes are required for the rapidly evolving
ascidians to maintain their invariant embryonic cell lineages in the presence of their compact
genomes. These compensatory changes have caused the enhancers to diverge within the
ascidians while allowing them to develop with their typical body plan.

6.3

Di↵erential expression analysis of tail loss in an
invertebrate chordate

Studying closely related organisms gives us insight into the underlying evolutionary mechanisms of divergence and underlying development. Indirect and direct developing sea urchins
have been studied showing that change in axial cleavage patterns and early cell fate leads to
divergent larval body plans that can exhibit similar adult body plans [138, 27]. In contrast,
ascidians retain their typical cell division and invariant cell fate in direct, indirect, tailed
and tail-less ascidians [41, 69].
We have shown that evolutionary drift has occurred in the cis-regulatory modules of
developing ascidian embryos between C. intestinalis and M. occidentalis, and these enhancers
are not always capable of driving expression in other ascidian species. It appears that the
same has happened between the tailed M. oculata and tail-less M. occulta, however those
cis-mechanisms are restored in the (M. oculata x M. occulta) interspecific hybrids, probably
allowing transcription factors to bind their targets at a higher affinity and restoring the
necessary level of expression to develop the urodele features. Further research is needed
to identify and test cis-elements, but as of now we have a better understanding of possible
72

mechanisms of the restoration of the larval tail in hybrids, which we can at least hypothesize
is driven by recovery of cis-regulatory modules.

6.4

Conclusion

The identification of regulatory genes expressed di↵erently in the urodele and anural species
is needed to understand the molecular mechanism underlying the evolutionary transition
from urodele to anural development. When examining the tailed and tail-less Molgula—M.
oculata and M. occulta—there is a strong overlap (91%) in the expression of genes that can be
annotated using the C. intestinalis proteome. There is also a strong overlap in the expression
of genes associated with notochord development, which is the key structure-defining element
not present in M. occulta. We observed that it was not the lack of genes in the transcriptome,
but the expression of said genes at a sufficient level, which led to the tail-less phenotype.
One of the shortcomings of our study is the lack of replicates from all stages; ideally we
would like at least three per developmental stage for each organism and because of this there
are likely to be more false positives and false negatives in our analysis than there would
be with more replicates. However, we were able to identify di↵erentially expressed genes
using the available replicates to estimate the dispersion for each of the samples, and able
to identify a subset of genes involved the development of the urodele features by examining
the correlation between two independent datasets, the tailed M. oculata and the hybrid. For
further examination replicates are needed, along with addition sequencing of other tailed and
tail-less molgula genomes and transcriptomes. This would give us stronger insights into why
the molgula are able to easily undergo tail-loss and the toolkit for molgula tail development.
Here we were able to see in the hybrids that the restored expression came from the

73

tailed allele and we proposed that cis-regulatory modules are the cause of this. The restored
expression of genes involved with the development of urodele features are acting in an allele
specific manner, while expression for other genes (total) has a more balanced expression from
each allele. The tailed allele restored the enhancer binding sites; presumably TF were able to
bind at a higher affinity and restore necessary level of expression to generate urodele features,
including the formation of the notochord and hence the larval tail. To confirm this hypothesis
we would first perform a ChIP-seq analysis to identify direct binding targets, and correlate
them with the di↵erentially expressed genes. A clear starting point is the gene bra which is
known to be important for notochord development, and Tbx2/3 which is downstream of bra,
and involved in convergence and extension [45]. Tbx2/3 was also identified in our di↵erential
expression analysis. From this gene-targeted analysis we can examine the divergence between
M. occulta and M. oculata enhancers, examine their binding affinities, and test the found
sites using transgenesis to see if we can restore the M. occulta’s tail.
Our hypothesis depends heavily on the assumption that the gene regulatory network
(GRN) is conserved within the tunicates. Comparing the GRN of distantly related organisms
can identify the genes necessary to the development of similar phenotypes and body plans.
Currently we plan to expand our analysis by comparing the molgula to the ten tunicate
genomes assemblies found on the aniseed website (http://www.aniseed.cnrs.fr/), and
available chordate genomes found on NCBI (http://www.ncbi.nlm.nih.gov/) and ensembl
(ensembl.org). With the information we currently have we will look more deeply into the
sequence divergence of both coding and non coding regions. To examine the noncoding
regions additional genomic sequencing is needed for gap filling and sca↵olding. However, it is
also necessary to determine if the GRNs have conserve structure by e.g. confirming the spatial
expression patterns for the network. Next steps on this project would be to identify relevant
74

transcription factors by annotating the genes and then selecting di↵erentially expressed genes
associated with DNA binding functions. Once we identify our candidate TFs we would
proceed with whole mount in situ hybridization to confirm conserved spatial patterning.

75

APPENDICES

76

Appendix A
Supplemental Figures
Mocu-hox12_13_10 NODE_17568_length_70313_cov_11.979449:1-70361

HOX12

Alignment 1
Mocc-hox12_13
2 alignments
Criteria: 70%, 100 bp
Regions: 29
Alignment 2
Moxi_12_13
NODE_379254_length_78679_cov_15.131484 (-)
1-78715
Criteria: 70%, 100 bp
Regions: 10

100%

50%
100%

0k

2k

4k

6k

8k

10k

12k

14k

16k

18k

50%
20k

HOX13

X-axis: Mocu-hox12_13_10
Resolution: 31
Window size: 100 bp

100%

contig
gene
exon
UTR
CNS
mRNA

50%
100%

20k

22k

24k

26k

28k

30k

32k

34k

36k

38k

50%
40k

HOX10

Figure A.1: Alignment for hox12-13 in M. occulta, M. oculata and M. occidentalis
100%
The contig containing hox12-13 for M. occidentalis and M. oculata, along with the two contigs
50%
containing hox12 and hox13 for M. occulta. M. oculata was used as the anchor sequence
100%
because it showed the most similarity between the three species. Outside of the coding
50%
regions and its flanking area,40kthere42k is very
little
sequence
the
species,
44k
46k
48k
50k similarity
52k
54kbetween
56k
58k
60k
and M. occidentalis exclusively shows similarity in coding regions. Grey arrows show the
direction of the contig.
100%
50%
100%

50%
60k

62k

64k

66k

77

68k

70k

Helix&1&

Helix&3&

Helix&3/4&

Figure A.2: Alignment of M. occidentalis hox2 genes with Ciona show premature
stop codon. The M. occidentalis hox2 gene has a stop codon in the 3’ region, inside of
the 3/4 helix. hox2 knockdowns in Ciona did not show any phenotypic di↵erence, so the
function of hox2 may not be important in M. occidentalis

78

1_Fruitfly_abd-A_Hox6-8_ANTP
64_Fruitfly_Ubx_Hox6-8_ANTP
3_Fruitfly_Antp_Hox6-8_ANTP
19_Amphioxus_Hox7_Hox6-8_ANTP
31_Human_HOXA7_Hox6-8_ANTP
40_Human_HOXB7_Hox6-8_ANTP
30_Human_HOXA6_Hox6-8_ANTP
39_Human_HOXB6_Hox6-8_ANTP
49_Human_HOXC6_Hox6-8_ANTP
72_gi_27525475_emb_CAD59670_1__putative_homeobox_protein_Hox6/7__partial__Ciona_intestinalis_
18_Amphioxus_Hox6_Hox6-8_ANTP
17_Amphioxus_Hox5_Hox5_ANTP
70_gi_74096097_ref_NP_001027665_1__Hox_5__Ciona_intestinalis_
29_Human_HOXA5_Hox5_ANTP
38_Human_HOXB5_Hox5_ANTP
63_Fruitfly_Scr_Hox5_ANTP
79_Moxi-hox5
88_Mocu-Hox5
96_Mocc-Hox5
20_Amphioxus_Hox8_Hox6-8_ANTP
48_Human_HOXC5_Hox5_ANTP
5_Fruitfly_Dfd_Hox4_ANTP
16_Amphioxus_Hox4_Hox4_ANTP
28_Human_HOXA4_Hox4_ANTP
58_Human_HOXD4_Hox4_ANTP
37_Human_HOXB4_Hox4_ANTP
47_Human_HOXC4_Hox4_ANTP
68_gi_148229589_ref_NP_001027781_2__hox4_protein__Ciona_intestinalis_
78_Moxi-hox4
87_Mocu-Hox4
95_Mocc-Hox4
41_Human_HOXB8_Hox6-8_ANTP
50_Human_HOXC8_Hox6-8_ANTP
59_Human_HOXD8_Hox6-8_ANTP
6_Fruitfly_ftz_Hox6-8_ANTP
14_Amphioxus_Hox2_Hox2_ANTP
15_Amphioxus_Hox3_Hox3_ANTP
27_Human_HOXA3_Hox3_ANTP
36_Human_HOXB3_Hox3_ANTP
57_Human_HOXD3_Hox3_ANTP
26_Human_HOXA2_Hox2_ANTP
35_Human_HOXB2_Hox2_ANTP
62_Fruitfly_pb_Hox2_ANTP
74_gi_27525471_emb_CAD59668_1__putative_homeobox_protein_hox2__partial__Ciona_intestinalis_
76_Moxi-hox2
85_Mocu-Hox2
93_Mocc-Hox2
77_Moxi-hox3
86_Mocu-Hox3
94_Mocc-Hox3
65_Fruitfly_zen_Hox3_ANTP
66_Fruitfly_zen2_Hox3_ANTP
8_Amphioxus_Hox10_Hox9-13_15__ANTP
9_Amphioxus_Hox11_Hox9-13_15__ANTP
21_Amphioxus_Hox9_Hox9-13_15__ANTP
32_Human_HOXA9_Hox9-13_15__ANTP
51_Human_HOXC9_Hox9-13_15__ANTP
60_Human_HOXD9_Hox9-13_15__ANTP
42_Human_HOXB9_Hox9-13_15__ANTP
23_Human_HOXA10_Hox9-13_15__ANTP
43_Human_HOXC10_Hox9-13_15__ANTP
53_Human_HOXD10_Hox9-13_15__ANTP
10_Amphioxus_Hox12_Hox9-13_15__ANTP
69_gi_74096477_ref_NP_001027696_1__hox10_protein__Ciona_intestinalis_
81_Moxi-hox10b
89_Mocu-Hox10
97_Mocc-Hox10
24_Human_HOXA11_Hox9-13_15__ANTP
54_Human_HOXD11_Hox9-13_15__ANTP
44_Human_HOXC11_Hox9-13_15__ANTP
2_Fruitfly_Abd-B_Hox9-13_15__ANTP
11_Amphioxus_Hox13_Hox9-13_15__ANTP
12_Amphioxus_Hox14_Hox9-13_15__ANTP
82_Moxi-hox12
90_Mocu-Hox12
98_Mocc-Hox12
7_Amphioxus_Hox1_Hox1_ANTP
67_gi_193083003_ref_NP_001122333_1__homeobox_transcription_factor_Hox1__Ciona_intestinalis_
73_gi_27525469_emb_CAD59667_1__putative_homeobox_protein_hox1__partial__Ciona_intestinalis_
22_Human_HOXA1_Hox1_ANTP
33_Human_HOXB1_Hox1_ANTP
52_Human_HOXD1_Hox1_ANTP
61_Fruitfly_lab_Hox1_ANTP
75_Moxi-hox1
84_Mocu-Hox1
92_Mocc-Hox1
45_Human_HOXC12_Hox9-13_15__ANTP
55_Human_HOXD12_Hox9-13_15__ANTP
71_gi_193083021_ref_NP_001122344_1__homeobox_transcription_factor_Hox13__Ciona_intestinalis_
83_Moxi-hox13
91_Mocu-Hox13
99_Mocc-Hox13
13_Amphioxus_Hox15_Hox9-13_15__ANTP
25_Human_HOXA13_Hox9-13_15__ANTP
46_Human_HOXC13_Hox9-13_15__ANTP
56_Human_HOXD13_Hox9-13_15__ANTP
34_Human_HOXB13_Hox9-13_15__ANTP
80_Moxi-hox10a
4_Fruitfly_bcd_Hox3_ANTP

Figure A.3: Alignment of hox genes. Alignments of the aa homeobox sequences from all
the Molgula species, show that they group with their respective orthologs. All but one of
the M. occidentalis cluster properly, but M. occidentalis hox10 full homeobox sequence was
not fully assembled, so this is possibly the reason for poor clustering.

79

Formatted Alignments
10

20

30

Occi.hox10a
Occi.hox10b

F S E S D P T R H W L T A N G R K K R V P Y T K F Q L L E L
- - - - - - - - - - - - - - - - - - - - - - - - - - - - E L
F S E S D P T R H W L T A N G R K K R V P Y T K F Q L L E L

Occi.hox10a
Occi.hox10b

EK E F H Y N Q Y L T R E R R L E V A K S V S L S D R Q V K
E K E F H Y N Q Y L T R E R R L E V A K S V H L S D R Q I K
E K E F H Y N Q Y L T R E R R L E V A K S V
L S D R Q . K

40

50

70

80

60

90

Occi.hox10a
Occi.hox10b

I W F QN R RM K WK K E R K E E K M R D GM T I P P P P H
I W F QN R RM K WK K E K K E D S MK S M L D I A S P - N
I W F QN R RM K WK K E . K E .
M .
I
P P

Occi.hox10a
Occi.hox10b

L I S S H L R P Q F P P A S H Y P A A L A A TM Q Q S Y P L
F L S P Q T L P P I A T G S Q Y S G - - - F E F Q Q P Y P F
. S
P
. S
Y
. A L A
Q Q
Y P

100

110

130

140

120

150

Occi.hox10a
Occi.hox10b

H N P F T S P T QA QG F S QH S V G S P P G V S S G T P H
H S A I T A H V Q S - - - - - H Y I G S Q S F I N N D V S Q
H
T
Q
Q G F S QH
. G S
.

Occi.hox10a
Occi.hox10b

H F Q P H Y Q S H S S N T G Y H D N V N QM A A A A A A D F
S Y Q A C QN L Q R TK T E Y D E T P - - - P N Q L A T D F
. Q
.
.
T
Y
.
N QM
. A
D F

Occi.hox10a
Occi.hox10b

F T S FH H - V P Y S M S R E P T L S L GM Y N
F N P FH H Q L P Y QM S R DH A L A L GM Y N
F
F H H Q . P Y
M S R .
L
L GM Y N

160

170

190

200

180

210

Figure A.4: Alignment of M. occidentalis duplicate hox10 genes Two copies of hox10
were found in M. occidentalis ⇠12 kb apart on the same contig.

80

Figure A.5: Gel electrophoresis of cdc45, netrin and controls Netrin was not found
in M. occulta’s transcriptome but was recovered via PCR from a cDNA library. However,
this library has a wider range of developmental stages, so it is still possible netrin is not
expressed at the right stage for tail development.

Figure A.6: Whole Mount In Situ Hybridization of prickle in M. occulta The PCP
gene pk was shown to be expressed in the notochord in a similar pattern to C. intestinalis

81

Appendix B
Contributions

B.1

Chapeter 3: Evaluating a lightweight transcriptome assembly pipeline on two closely related ascidian species

The experimental design for the RNA sequence analysis was decided by C. Titus Brown
(CTB) and Billie J. Swalla (BJS). The fertilization and collection of RNA samples were
done by BJS and myself. Library preparation for Illumina sequencing was done by Kanchan
Pavangadkar. All downstream analysis was done by myself. The idea for evaluating the de
novo assembly pipeline came about through conversations with CTB, as well as methods for
evaluating the transcriptome assemblies. Writing was done by me with edits from BJS and
CTB.

B.2

Chapter 4: Genome assembly and characterization

The experimental design for the DNA sequencing was decided by CTB, BJS, and Lionel
Christiaen (LC). Embryonic samples were collected by BJS, LC, Alberto Stolfi (AS), Claudia

82

Racioppi (CR) and myself. Genome assembly was conducted by me. Ideas for examining
divegand were developed by AS, orthologous sequences were identified through the use of
RBH blast and done so by me because the sequences were not yet annotated on the Aniseed
database. The Hox analysis was done by me and was initiated by a question from David
Arnosti. Writing was done by AS, myself with edits from BJS, CTB, CR, and LC.

B.3

Chapter 5: Di↵erential expression analysis of tail
loss in an invertebrate chordate

The experimental design is the same for the “Evaluating a lightweight transcriptome assembly pipeline on two closely related ascidian species” chapter (3). Transcript models were
assembled by me. Read mapping for di↵erential expression analysis was conducted by me,
and the analysis of hybrid expression counts was the idea of CTB. The di↵erential expression
analysis was conducted by me, including the idea to focus on the overlapping upregulated
gene in M. oculata and the interspecific hybrid. Annotation of the overlapping upregulated
genes was done by AS, Anna Di Gregorio and myself. Writing was done by me with edits
from AS, CR, and CTB.

83

BIBLIOGRAPHY

84

BIBLIOGRAPHY
Bibliography

[1]
Postlethwait.
(t)
in
of
S. Bassham Brachyury
and J. Postlethwait.
Brachyury
(t) expression
in embryos of a la
[1] S.
S. Bassham
Bassham and
and J.
J.[1]
Postlethwait.
Brachyury
(t) expression
expression
in embryos
embryos
of aa larvacean
larvacean
Bibliography
urochordate,
dioica,
the
role
of
urochordate,
dioica,
ancestral role Biology,
of t. Developmental
urochordate, oikopleura
oikopleura
dioica, and
andoikopleura
the ancestral
ancestral
roleand
of t.
t.theDevelopmental
Developmental
Biology,
220(2):322–332,
220(2):322–332, Apr. 2000.
220(2):322–332, Apr.
Apr. 2000.
2000.

[2]
[2]
[1]
[3]
[3]

[2]
[3]

[4]
[4]

[5]
[5]
[4]
[6]
[6]
[5]

N.
in
tunicate
Society,
[2] N. J.
Studies in tunicate
developnent. 1931.
Society, 219:281–346, 1931.
N. J.
J. BERRILL.
BERRILL. Studies
Studies
in BERRILL.
tunicate developnent.
developnent.
Society, 219:281–346,
219:281–346,
1931.
S. Bassham and J. Postlethwait. Brachyury (t) expression in embryos of a larvacean
K.
J.
Fass,
A.
P.
Baranay,
M.
I.
Birol,
[3]
K. dioica,
R. Bradnam,
N. Fass,
A.of Alexandrov,
P. Baranay,
urochordate,
oikopleura
andAlexandrov,
the J.
ancestral
t. Developmental
K. R.
R. Bradnam,
Bradnam,
J. N.
N.
Fass,
A.
Alexandrov,
P.role
Baranay,
M. Bechner,
Bechner,
I.Biology,
Birol,M. Bechner, I
S.
Chapman,
G.
R.
Chitsaz,
W.-C.
Chou,
CorS. Boisvert,
J. A. Chapman,
Chikhi,
H. J.
Chitsaz,
220(2):322–332,
2000.
S. Boisvert,
Boisvert, J.
J. A.
A.Apr.
Chapman,
G. Chapuis,
Chapuis,
R. Chikhi,
Chikhi,G.H.
H.Chapuis,
Chitsaz,R.
W.-C.
Chou,
J.
Cor- W.-C. Chou,
beil,
C.
D.
Fabbro,
T.
R.
Docking,
R.
Durbin,
D.
Earl,
S.
Emrich,
P.
Fedotov,
N.
A.
beil,
C.
D.
Fabbro,
T.
R.
Docking,
R.
Durbin,
D.
Earl,
S.
P. Fedotov
beil, C. D. Fabbro, T. R. Docking, R. Durbin, D. Earl, S. Emrich, P. Fedotov, N.Emrich,
A.
N.
J.
BERRILL.
Studies
in
tunicate
developnent.
Society,
219:281–346,
1931.
Fonseca,
R.
S.
Goldstein,
M.
Fonseca,
G. Ganapathy,
A. Gibbs, S.S.
. Godzaridis,
S. Goldstein, M.
Fonseca, G.
G. Ganapathy,
Ganapathy,
R. A.
A. Gibbs,
Gibbs,
S. Gnerre,
Gnerre,R... Godzaridis,
Godzaridis,
S.Gnerre,
Goldstein,
M. Haimel,
Haimel,
G.
Hall,
D.
Haussler,
J.
B.
Hiatt,
I.
Y.
Ho,
J.
Howard,
M.
Hunt,
S.
D.
Jackman,
D.
B.
G. Hall,
D. I.Haussler,
B. Hiatt,M.
I. Hunt,
Y. Ho,S.J.D.
Howard,
M. D.
Hunt,
G. Hall,
D. Haussler,
Hiatt,
Y. Ho, J.J.Howard,
Jackman,
B. S. D. Jackma
K.
R. Bradnam,
J. J.
N.B.Fass,
A. Alexandrov,
P. Baranay,
M. Bechner,
I. Birol,
Ja↵e,
E.
D.
Jarvis,
H.
Jiang,
S.
Kazakov,
P.
J.
Kersey,
J.
O.
Kitzman,
J.
R.
Knight,
Ja↵e,
E.
D.
Jarvis,
H.
Jiang,
S.
Kazakov,
P.
J.
Kersey,
J.
O.
Kitzman, J. R.
Ja↵e,
E. D. J.
Jarvis,
H. Jiang,G.
S. Chapuis,
Kazakov,R.
P. Chikhi,
J. Kersey,
J. O. Kitzman,
J. R. Knight,
S.
Boisvert,
A. Chapman,
H. Chitsaz,
W.-C. Chou,
J. CorS.
T.-W. Lam,
D.Koren,
Lavenier,
F. Laviolette,
Y. Li,
B.
Y.Y.
Liu,
S.
T.-W.
Lam, D. Lavenier,
F.Li,Laviolette,
Li,R.
Z. Luo,
Li, B. Liu, Y. Liu,
S. Koren,
Koren,
Lam,
Li, Z.
Z.
B. Liu,
Liu,
Liu,
R.
Luo,
beil,
C. D.T.-W.
Fabbro,
T. D.
R. Lavenier,
Docking, F.
R. Laviolette,
Durbin, D.Y.
Earl,
S. Li,
Emrich,
P. Y.
Fedotov,
N.
A.
I.
MacCallum,
M.
D.
MacManes,
N.
Maillet,
S.
Melnikov,
D.
Naquin,
Z.
Ning,
T.
D.
I.
MacCallum,
M.
D.
MacManes,
N.
Maillet,
S.
Melnikov,
D.
Naquin, Z. Ning
I.
MacCallum,
M.
D.
MacManes,
N.
Maillet,
S.
Melnikov,
D.
Naquin,
Z.
Ning,
T.
D.
Fonseca, G. Ganapathy, R. A. Gibbs, S. Gnerre, . Godzaridis, S. Goldstein, M. Haimel,
Otto,
B.
Paten,
O.
Paulo,
A.
M.
Phillippy,
F.
M. Place,
D.
B.
O. S. Paulo,
A. M. Phillippy,
F. Pina-Martins,
M. Place, D. P
Otto,
B. D.
Paten,
O. S.
S.J.Otto,
Paulo,
A.Paten,
M.
Phillippy,
F. Pina-Martins,
Pina-Martins,
Place,
D. PrzybylPrzybylG.
Hall,
Haussler,
B. Hiatt,
I. Y.
Ho, J. Howard,
M. Hunt,M.
S. D.
Jackman,
D. B.
ski,
X.
Qin,
C.
Qu,
F.
J.
Ribeiro,
S.
Richards,
D.
S.
Rokhsar,
J.
G.
Ruby,
S.
Scalabrin,
ski,Ribeiro,
X. Qin,S.C.Richards,
Qu, F. J.D.
Ribeiro,
S. Richards,
D. S.S.Rokhsar,
J. G. Ruby, S. Sc
ski, X.E.Qin,
Qu, F.
S. Rokhsar,
G. Ruby,
Scalabrin,
Ja↵e,
D. C.
Jarvis,
H.J.Jiang,
S. Kazakov, P. J.
Kersey,
J. O.J.Kitzman,
J. R.
Knight,
M.
C.
Schatz,
D.
C.
Schwartz,
A.
Sergushichev,
T.
Sharpe,
T.
I.
Shaw,
J.
Shendure,
M.
C.
Schatz,
D.
C.
Schwartz,
A.
Sergushichev,
T.
Sharpe,
T.
I. Shaw, J. Sh
M.Koren,
C. Schatz,
D.Lam,
C. Schwartz,
A. Sergushichev,
Shaw,
Shendure,
S.
T.-W.
D. Lavenier,
F. Laviolette, T.
Y. Sharpe,
Li, Z. Li,T.B.I. Liu,
Y. J.
Liu,
R. Luo,
Y.
Shi,
Song,
F.
Tsarev,
B. M.
Vieira,
Y.H.
J. T.
H.Vezzi,
Song, R.
F. Vicedomini,
Tsarev, F. Vezzi,
Vicedomini, B. M.
Y.MacCallum,
Shi, J.
J. T.
T. Simpson,
Simpson,
H.Shi,
Song,
F. Simpson,
Tsarev, F.
F.
Vezzi,
R.
Vicedomini,
M. R.
Vieira,
D. Naquin, Z.B.Ning,
T. D.
I.
M. D. MacManes,
N.
Maillet,
S. Melnikov,
J.
Wang,
K.
C.
Worley,
S.
Yin,
S.-M.
Yiu,
J.
Yuan,
G.
Zhang,
H.
Zhang,
S.
Zhou,
and
J.
Wang,
K.
C.
Worley,
S.
Yin,
S.-M.
Yiu,
J.
Yuan,
G.
Zhang,
J. Wang,
K. C. Worley,
S. Yin,
Yiu, J. Yuan,
G. Zhang, H.M.Zhang,
andH. Zhang, S. Zh
Place,S.D.Zhou,
PrzybylOtto,
B. Paten,
O. S. Paulo,
A.S.-M.
M. Phillippy,
F. Pina-Martins,
I.
Korf.
Assemblathon
evaluating
de
methods
of
assembly
in
three
I. F.2:
Assemblathon
2: evaluating
de novo
methods
genome assembly
I. F.
F.X.
Korf.
2:Korf.
evaluating
de novo
novo
of genome
genome
assembly
in of
three
J. G. Ruby,
S. Scalabrin,
ski,
Qin,Assemblathon
C. Qu, F. J. Ribeiro,
S. Richards,
D.methods
S. Rokhsar,
vertebrate
species.
GigaScience,
2(1):10,
July
2013.
vertebrate species. July
GigaScience,
2(1):10, July 2013.
vertebrate
species.
2013.
M.
C. Schatz,
D. C.GigaScience,
Schwartz, A.2(1):10,
Sergushichev,
T. Sharpe, T. I. Shaw, J. Shendure,
Vicedomini,
B. M. Vieira,
Y.
Shi,Brown,
J. T. Simpson,
F. D.
Tsarev,
F.Functional
Vezzi,
R.A.
C.
and
architecture
and
[4] Johnson,
C.H.D.Song,
Brown,
S. Johnson,
and
Sidow. Functional
architecture and ev
C. D.
D. Brown, D.
D. S.
S.
Johnson,
and A.
A. Sidow.
Sidow.
Functional
architecture
and evolution
evolution
H.
Zhang,
S. Zhou,
and
J.
Wang,
K.
C.
Worley,
S.
Yin,
S.-M.
Yiu,
J.
Yuan,
G.
Zhang,
of
transcriptional
elements
that
drive
gene
coexpression.
Science
(New
York,
N.Y.),
of
transcriptional
elements
that
drive
gene
coexpression.
Science
(New York,
of transcriptional elements that drive gene coexpression. Science (New York, N.Y.),
genome
assembly
in
three
I.
F. Korf. Assemblathon
2:
evaluating
de
novo
methods
of
317(5844):1557–1560,
Sept.
2007.
317(5844):1557–1560,
Sept. 2007.
317(5844):1557–1560, Sept.
2007.
vertebrate species. GigaScience, 2(1):10, July 2013.
C.
Pyrkosz,
and
H.
Brom.
A
[5] C.Q.
T. Zhang,
Brown,A.
A.B.
Howe,
Q. Zhang,
Pyrkosz,
and T. H. Brom. A refere
C. T.
T. Brown,
Brown, A.
A. Howe,
Howe,
Q.
Zhang,
A.
B.
Pyrkosz,
and T.
T.A.
H.B.
Brom.
A reference-free
reference-free
architecture
and
evolution
C.
D.
Brown,
D.
S.
Johnson,
and
A.
Sidow.
Functional
algorithm
for
computational
normalization
of
shotgun
sequencing
data.
arXiv
e-print
algorithm
for
computational
normalization
of
shotgun
algorithm for computational normalization of shotgun sequencing data. arXivsequencing
e-print data. arXiv
Science
(New
York,
N.Y.),
of
transcriptional
elements
that
drive
gene
coexpression.
1203.4802,
Mar.
2012.
1203.4802, Mar. 2012. 1203.4802, Mar. 2012.
317(5844):1557–1560, Sept. 2007.
M.
Do,
M.
F.
Kim,
Davydov,
NISC
Comparative
C. B.M.
M. E.
Cooper,
M. F.
Kim,
E. Davydov, NISC Comp
M. Brudno,
Brudno, C.
C. B.
B. [6]
Do,M.G.
G.Brudno,
M. Cooper,
Cooper,
M.Do,
F. G.
Kim,
E.
Davydov,
NISC
Comparative
C.
T.
Brown,
A.
Howe,
Q.
Zhang,
A.
B.
Pyrkosz,
and
T.
H.
Brom.
A
reference-free
Sequencing
Program,
E.
D.
Green,
A.
Sidow,
and
S.
Batzoglou.
LAGAN
and
multiSequencing
Program,
E.
D.
Green,
A.
Sidow,
and
S.
Batzoglou.
Sequencing Program, E. D. Green, A. Sidow, and S. Batzoglou. LAGAN and
multi- LAGAN and
data.
arXiv
e-print
algorithm
for
computational
normalization
of
shotgun
sequencing
LAGAN:
efficient
tools
for
large-scale
multiple
alignment
of
genomic
DNA.
Genome
efficientmultiple
tools for
large-scale
alignment
of genomic DNA. G
LAGAN: efficient toolsLAGAN:
for large-scale
alignment
of multiple
genomic DNA.
Genome
1203.4802,
Mar.
2012.
Research,
13(4):721–731,
Apr.
2003.
Research,
13(4):721–731, Apr. 2003.
Research, 13(4):721–731,
Apr. 2003.

[6]
M. B.
Brudno, C. B.
Do,Garey,
G.B.M.
Cooper,
M.R.F.Garey,
Kim, and
E. Davydov,
NISCEvolution
Comparative
[7]
R.
and
chordate
body
Cameron,
B.of
Swalla.
of the chordate bod
[7] C.
C. B. Cameron,
Cameron, J.
J. [7]
R. C.
Garey,
and B.
B. J.
J.J.Swalla.
Swalla. Evolution
Evolution
ofJ.the
the
chordate
body plan:
plan:
LAGAN
and
multiSequencing
Program,
E.
D.
Green,
A.
Sidow,
and
S.
Batzoglou.
New
insights
from
phylogenetic
analyses
of
deuterostome
phyla.
Proceedings
of
the
New insights
from phylogenetic
analyses
deuterostome
Proceeding
New insights from phylogenetic
analyses
of deuterostome
phyla.of Proceedings
of phyla.
the
of
genomic
DNA.
Genome
LAGAN:
efficient
tools
for
large-scale
multiple
alignment
National
Academy
of
Sciences,
97(9):4469–4474,
Apr.
2000.
National 97(9):4469–4474,
Academy of Sciences,
Apr. 2000.
National Academy of Sciences,
Apr. 97(9):4469–4474,
2000.
Research, 13(4):721–731, Apr. 2003.
81
the chordate body plan:
[7] C. B. Cameron, J. R. Garey, and B. J.81Swalla. Evolution of 85
New insights from phylogenetic analyses of deuterostome phyla. Proceedings of the
National Academy of Sciences, 97(9):4469–4474, Apr. 2000.
81
85

[8] A. Caracciolo, A. Di Gregorio, F. Aniello, R. Di Lauro, and M. Branno. Identification
and developmental expression of three distal-less homeobox containing genes in the
ascidian ciona intestinalis. Mechanisms of Development, 99(12):173–176, Dec. 2000.
[9] L. Christiaen, A. Stolfi, B. Davidson, and M. Levine. Spatio-temporal intersection of
lhx3 and tbx6 defines the cardiac field through synergistic activation of mesp. Developmental Biology, 328(2):552–560, Apr. 2009.
[10] K. Clarke, Y. Yang, R. Marsh, L. Xie, and K. K. Zhang. Comparative analysis of de
novo transcriptome assembly. Science China Life Sciences, 56(2):156–162, Feb. 2013.
[11] E. G. Conklin. The organization and cell-lineage of the ascidian egg. Philadelphia :
[Academy of Natural Sciences], 1905.
[12] J. C. Corbo, M. Levine, and R. W. Zeller. Characterization of a notochord-specific
enhancer from the brachyury promoter region of the ascidian, ciona intestinalis. Development (Cambridge, England), 124(3):589–602, Feb. 1997.
[13] B. Davidson, W. Shi, and M. Levine. Uncoupling heart cell specification and migration in the simple chordate ciona intestinalis. Development (Cambridge, England),
132(21):4811–4818, Nov. 2005.
[14] P. Dehal, Y. Satou, R. K. Campbell, J. Chapman, B. Degnan, A. D. Tomaso, B. Davidson, A. D. Gregorio, M. Gelpke, D. M. Goodstein, N. Harafuji, K. E. M. Hastings,
I. Ho, K. Hotta, W. Huang, T. Kawashima, P. Lemaire, D. Martinez, I. A. Meinertzhagen, S. Necula, M. Nonaka, N. Putnam, S. Rash, H. Saiga, M. Satake, A. Terry,
L. Yamada, H.-G. Wang, S. Awazu, K. Azumi, J. Boore, M. Branno, S. Chin-bow,
R. DeSantis, S. Doyle, P. Francino, D. N. Keys, S. Haga, H. Hayashi, K. Hino,
K. S. Imai, K. Inaba, S. Kano, K. Kobayashi, M. Kobayashi, B.-I. Lee, K. W. Makabe, C. Manohar, G. Matassi, M. Medina, Y. Mochizuki, S. Mount, T. Morishita,
S. Miura, A. Nakayama, S. Nishizaka, H. Nomoto, F. Ohta, K. Oishi, I. Rigoutsos,
M. Sano, A. Sasaki, Y. Sasakura, E. Shoguchi, T. Shin-i, A. Spagnuolo, D. Stainier,
M. M. Suzuki, O. Tassy, N. Takatori, M. Tokuoka, K. Yagi, F. Yoshizaki, S. Wada,
C. Zhang, P. D. Hyatt, F. Larimer, C. Detter, N. Doggett, T. Glavina, T. Hawkins,
P. Richardson, S. Lucas, Y. Kohara, M. Levine, N. Satoh, and D. S. Rokhsar. The draft
genome of ciona intestinalis: Insights into chordate and vertebrate origins. Science,
298(5601):2157–2167, Dec. 2002.
[15] F. Denoeud, S. Henriet, S. Mungpakdee, J.-M. Aury, C. D. Silva, H. Brinkmann,
J. Mikhaleva, L. C. Olsen, C. Jubin, C. Caestro, J.-M. Bouquet, G. Danks, J. Poulain,
C. Campsteijn, M. Adamski, I. Cross, F. Yadetie, M. Mu↵ato, A. Louis, S. Butcher,
G. Tsagkogeorga, A. Konrad, S. Singh, M. F. Jensen, E. H. Cong, H. EikesethOtteraa, B. Noel, V. Anthouard, B. M. Porcel, R. Kachouri-Lafond, A. Nishino,
M. Ugolini, P. Chourrout, H. Nishida, R. Aasland, S. Huzurbazar, E. Westhof, F. Delsuc, H. Lehrach, R. Reinhardt, J. Weissenbach, S. W. Roy, F. Artiguenave, J. H.
Postlethwait, J. R. Manak, E. M. Thompson, O. Jaillon, L. D. Pasquier, P. Boudinot,
D. A. Liberles, J.-N. Vol↵, H. Philippe, B. Lenhard, H. R. Crollius, P. Wincker, and
86

D. Chourrout. Plasticity of animal genome architecture unmasked by rapid evolution
of a pelagic tunicate. Science, 330(6009):1381–1385, Dec. 2010.
[16] A. Di Gregorio, R. M. Harland, M. Levine, and E. S. Casey. Tail morphogenesis in
the ascidian, ciona intestinalis, requires cooperation between notochord and muscle.
Developmental biology, 244(2):385–95, Apr. 2002.
[17] J. R. Finnerty. The origins of axial patterning in the metazoa: how old is bilateral
symmetry? The International Journal of Developmental Biology, 47(7-8):523–529,
2003.
[18] K. A. Frazer, L. Pachter, A. Poliakov, E. M. Rubin, and I. Dubchak. VISTA: computational tools for comparative genomics. Nucleic Acids Research, 32(Web Server
issue):W273–279, July 2004.
[19] W. J. Gehring, Y. Q. Qian, M. Billeter, K. Furukubo-Tokunaga, A. F. Schier,
D. Resendez-Perez, M. A↵olter, G. Otting, and K. Wthrich. Homeodomain-DNA
recognition. Cell, 78(2):211–223, July 1994.
[20] T. C. Glenn. Field guide to next-generation DNA sequencers. Molecular Ecology
Resources, 11(5):759–769, Sept. 2011.
[21] M. G. Grabherr, B. J. Haas, M. Yassour, J. Z. Levin, D. a. Thompson, I. Amit,
X. Adiconis, L. Fan, R. Raychowdhury, Q. Zeng, Z. Chen, E. Mauceli, N. Hacohen,
A. Gnirke, N. Rhind, F. di Palma, B. W. Birren, C. Nusbaum, K. Lindblad-Toh,
N. Friedman, and A. Regev. Full-length transcriptome assembly from RNA-seq data
without a reference genome. Nature biotechnology, 29(7):644–52, July 2011.
[22] F. Gyoja, Y. Satou, T. Shin-i, Y. Kohara, B. J. Swalla, and N. Satoh. Analysis of large
scale expression sequenced tags (ESTs) from the anural ascidian, molgula tectiformis.
Developmental biology, 307(2):460–82, July 2007.
[23] B. J. Haas, A. Papanicolaou, M. Yassour, M. Grabherr, P. D. Blood, J. Bowden, M. B.
Couger, D. Eccles, B. Li, M. Lieber, M. D. MacManes, M. Ott, J. Orvis, N. Pochet,
F. Strozzi, N. Weeks, R. Westerman, T. William, C. N. Dewey, R. Henschel, R. D.
LeDuc, N. Friedman, and A. Regev. De novo transcript sequence reconstruction from
RNA-seq using the trinity platform for reference generation and analysis. Nat. Protocols, 8(8):1494–1512, Aug. 2013.
[24] K. A. Hadfield, B. J. Swalla, and W. R. Je↵ery. Multiple origins of anural development
in ascidians inferred from rDNA sequences. Journal of Molecular Evolution, 40(4):413–
427, Apr. 1995.
[25] E. E. Hare, B. K. Peterson, and M. B. Eisen. A careful look at binding site reorganization in the even-skipped enhancers of drosophila and sepsids. PLoS genetics,
4(11):e1000268, Nov. 2008.

87

[26] H. Hashimoto, T. Enomoto, G. Kumano, and H. Nishida. The transcription factor
FoxB mediates temporal loss of cellular competence for notochord induction in ascidian
embryos. Development, 138(14):3091–3091, June 2011.
[27] J. J. Henry and R. A. Ra↵. Evolutionary change in the process of dorsoventral axis
determination in the direct developing sea urchin, heliocidaris erythrogramma. Developmental Biology, 141(1):55–69, Sept. 1990.
[28] B. G. Herrmann and A. Kispert. The t genes in embryogenesis. Trends in Genetics,
10(8):280–286, Aug. 1994.
[29] T. Hirano and H. Nishida. Developmental fates of larval tissues after metamorphosis
in ascidian halocynthia roretzi. i. origin of mesodermal tissues of the juvenile. Developmental Biology, 192(2):199–210, Dec. 1997.
[30] K. Hotta, H. Takahashi, T. Asakura, B. Saitoh, N. Takatori, Y. Satou, and N. Satoh.
Characterization of brachyury-downstream notochord genes in the ciona intestinalis
embryo. Developmental biology, 224(1):69–80, Aug. 2000.
[31] K. Hotta, H. Takahashi, a. Erives, M. Levine, and N. Satoh. Temporal expression
patterns of 39 brachyury-downstream genes associated with notochord formation in
the ciona intestinalis embryo. Development, growth & di↵erentiation, 41(6):657–64,
Dec. 1999.
[32] K. Hotta, S. Yamada, N. Ueno, N. Satoh, and H. Takahashi. Brachyury-downstream
notochord genes and convergent extension in ciona intestinalis embryos. Development,
Growth & Di↵erentiation, 49(5):373–382, June 2007.
[33] A. C. Howe, J. K. Jansson, S. A. Malfatti, S. G. Tringe, J. M. Tiedje, and C. T. Brown.
Tackling soil diversity with the assembly of large, complex metagenomes. Proceedings
of the National Academy of Sciences, 111(13):4904–4909, Apr. 2014.
[34] J. L. Huber, K. B. da Silva, W. R. Bates, and B. J. Swalla. The evolution of anural
larvae in molgulid ascidians. Seminars in cell & developmental biology, 11(6):419–26,
Dec. 2000.
[35] C. Hudson, M. Ba, C. Rouvire, and H. Yasuo. Divergent mechanisms specify chordate motoneurons: evidence from ascidians. Development (Cambridge, England),
138(8):1643–52, Apr. 2011.
[36] T. Ikuta and H. Saiga. Organization of hox genes in ascidians: present, past, and
future. Developmental Dynamics: An Official Publication of the American Association
of Anatomists, 233(2):382–389, June 2005.
[37] T. Ikuta, N. Satoh, and H. Saiga. Limited functions of hox genes in the larval
development of the ascidian ciona intestinalis. Development (Cambridge, England),
137(9):1505–1513, May 2010.

88

[38] T. Ikuta, N. Yoshida, N. Satoh, and H. Saiga. Ciona intestinalis hox gene cluster: Its
dispersed structure and residual colinear expression in development. Proceedings of the
National Academy of Sciences of the United States of America, 101(42):15118–15123,
Oct. 2004.
[39] Je↵ery, R. billie, and J. Swalla. Factors necessary for restoring an evolutionary change
in an anural ascidian embryo. Developmental biology, 153:194–205, 1992.
[40] W. R. Je↵ery. Minireview ascidian gene-expression profiles. Genome biology, 3(10):1–4,
2002.
[41] W. R. Je↵ery and B. J. Swalla. An evolutionary change in the muscle lineage of
an anural ascidian embryo is restored by interspecific hybridization with a urodele
ascidian. Developmental Biology, 337:328–337, 1991.
[42] W. R. Je↵ery, B. J. Swalla, N. Ewing, and T. Kusakabe. Evolution of the ascidian
anural larva: evidence from embryos and molecules. Molecular biology and evolution,
16(5):646–54, May 1999.
[43] D. Jiang, E. M. Munro, W. C. Smith, S. Barbara, and F. Harbor. Ascidian prickle regulates both mediolateral and anterior-posterior cell polarity of notochord cells. Current
biology, 15:79–85, 2005.
[44] D. S. Johnson, B. Davidson, C. D. Brown, W. C. Smith, and A. Sidow. Noncoding regulatory sequences of ciona exhibit strong correspondence between evolutionary
constraint and functional importance. Genome Research, 14(12):2448–2456, Dec. 2004.
[45] L. Katikala, H. Aihara, Y. J. Passamaneck, S. Gazdoiu, D. S. Jos-Edwards, J. E.
Kugler, I. Oda-Ishii, J. H. Imai, Y. Nibu, and A. Di Gregorio. Functional brachyury
binding sites establish a temporal read-out of gene expression in the ciona notochord.
PLoS Biol, 11(10):e1001697, Oct. 2013.
[46] K. Katoh and D. M. Standley. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Molecular Biology and Evolution,
30(4):772–780, Apr. 2013.
[47] K. Katoh and H. Toh. PartTree: an algorithm to build an approximate tree from a
large number of unaligned sequences. Bioinformatics (Oxford, England), 23(3):372–
374, Feb. 2007.
[48] A. I. Kavka and J. B. A. Green. Tales of tails: Brachyury and the t-box genes.
Biochimica et Biophysica Acta (BBA) - Reviews on Cancer, 1333(2):F73–F84, Oct.
1997.
[49] R. Keller, L. Davidson, a. Edlund, T. Elul, M. Ezin, D. Shook, and P. Skoglund. Mechanisms of convergence and extension by cell intercalation. Philosophical transactions
of the Royal Society of London. Series B, Biological sciences, 355(1399):897–922, July
2000.
89

[50] K. Kobayashi, L. Yamada, and N. Satoh. Di↵erential gene expression in notochord and
nerve cord fate segregation in the ciona intestinalis embryo. Genesis, 51(9):647–659,
2013.
[51] M. J. Kourakis, W. Reeves, E. Newman-Smith, B. Maury, S. Abdul-Wajid, and W. C.
Smith. A one-dimensional model of PCP signaling: Polarized cell behavior in the
notochord of the ascidian ciona. Developmental Biology, 395(1):120–130, Nov. 2014.
[52] A. Kubo, N. Suzuki, X. Yuan, K. Nakai, N. Satoh, K. S. Imai, and Y. Satou. Genomic cis-regulatory networks in the early ciona intestinalis embryo. Development
(Cambridge, England), 137(10):1613–1623, May 2010.
[53] J. E. Kugler, P. Kerner, J.-M. Bouquet, D. Jiang, and A. Di Gregorio. Evolutionary
changes in the notochord genetic toolkit: a comparative analysis of notochord genes in
the ascidian ciona and the larvacean oikopleura. BMC evolutionary biology, 11(1):21,
Jan. 2011.
[54] J. E. Kugler, Y. J. Passamaneck, T. G. Feldman, J. Beh, T. W. Regnier, and A. Di Gregorio. Evolutionary conservation of vertebrate notochord genes in the ascidian ciona
intestinalis. Genesis (New York, N.Y. : 2000), 46(11):697–710, Nov. 2008.
[55] C. D. Laird. Chromatid structure: relationship between DNA content and nucleotide
sequence diversity. Chromosoma, 32(4):378–406, Dec. 1971.
[56] B. Langmead and S. L. Salzberg. Fast gapped-read alignment with bowtie 2. Nature
Methods, 9(4):357–359, Apr. 2012.
[57] A. Lauri, T. Brunet, M. Handberg-Thorsager, A. H. L. Fischer, O. Simakov, P. R. H.
Steinmetz, R. Tomer, P. J. Keller, and D. Arendt. Development of the annelid axochord: Insights into notochord evolution. Science, 345(6202):1365–1368, Sept. 2014.
[58] P. Lemaire. Unfolding a chordate developmental program, one cell at a time: invariant cell lineages, short-range inductions and evolutionary plasticity in ascidians.
Developmental biology, 332(1):48–60, Aug. 2009.
[59] P. Lemaire. Evolutionary crossroads in developmental biology: the tunicates. Development, 138(11):2143–2152, May 2011.
[60] P. Lemaire, W. C. Smith, and H. Nishida. Ascidians and the plasticity of the chordate
developmental program. Current biology : CB, 18(14):R620–31, July 2008.
[61] H. Li, B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis, R. Durbin, and 1000 Genome Project Data Processing Subgroup. The sequence
alignment/map format and SAMtools. Bioinformatics (Oxford, England), 25(16):2078–
2079, Aug. 2009.
[62] W. Li and A. Godzik. Cd-hit: a fast program for clustering and comparing large sets
of protein or nucleotide sequences. Bioinformatics, 22(13):1658–1659, July 2006.
90

[63] L. Liu, Y. Li, S. Li, N. Hu, Y. He, R. Pong, D. Lin, L. Lu, and M. Law. Comparison
of next-generation sequencing systems. BioMed Research International, 2012:e251364,
July 2012.
[64] M. Lohse, A. M. Bolger, A. Nagel, A. R. Fernie, J. E. Lunn, M. Stitt, and B. Usadel.
RobiNA: a user-friendly, integrated software solution for RNA-seq-based transcriptomics. Nucleic Acids Research, 40(W1):W622–W627, June 2012.
[65] W. J. R. Longabaugh, E. H. Davidson, and H. Bolouri. Visualization, documentation,
analysis, and communication of large-scale gene regulatory networks. Biochimica et
Biophysica Acta (BBA) - Gene Regulatory Mechanisms, 1789(4):363–374, Apr. 2009.
[66] E. K. Lowe, B. J. Swalla, and C. T. Brown. Evaluating a lightweight transcriptome
assembly pipeline on two closely related ascidian species. PeerJ PrePrints, 2:e505v1,
2014.
[67] M. D. Macmanes. On the optimal trimming of high-throughput mRNA sequence data.
Frontiers in Genetics, 5:13, 2014.
[68] M. E. Maliska, M. W. Pennell, and B. J. Swalla. Developmental mode influences diversification in ascidians. Biology Letters, 9(3):20130068, June 2013. Ascidian species
(Tunicata: Ascidiacea) usually have tailed, hatching tadpole larvae. In several lineages,
species have evolved larvae that completely lack any tail tissues and are unable to disperse actively. Some tailless species hatch, but some do not hatch before going through
metamorphosis. We show here that ascidian species with the highest speciation rates
are those with the largest range sizes and tailed hatching larval development. We use
methods for examining diversification in binary characters across a posterior distribution of trees, and show that mode of larval development predicts geographical range
sizes. Conversely, we find that species with the least dispersive larval development (tailless, non-hatching) have the lowest speciation rates and smallest geographical ranges.
Our speciation rate results are contrary to findings from sea urchins and snails examined in the fossil record, and further work is necessary to reconcile these disparate
results.
[69] M. E. Maliska and B. J. Swalla. Molgula pugetiensis is a pacific tailless ascidian within
the roscovita clade of molgulids. The Biological Bulletin, 219(3):277–282, Dec. 2010.
[70] M. Mallo and C. R. Alonso. The regulation of hox gene expression during animal.
Development, 140(19):3951–3963, Oct. 2013.
[71] J. C. Marioni, C. E. Mason, S. M. Mane, M. Stephens, and Y. Gilad. RNA-seq: An
assessment of technical reproducibility and comparison with gene expression arrays.
Genome Research, 18(9):1509–1517, Sept. 2008.
[72] J. a. Martin and Z. Wang. Next-generation transcriptome assembly. Nature reviews.
Genetics, 12(10):671–82, Oct. 2011.

91

[73] C. Mayor, M. Brudno, J. R. Schwartz, A. Poliakov, E. M. Rubin, K. A. Frazer, L. S.
Pachter, and I. Dubchak. VISTA : visualizing global DNA sequence alignments of
arbitrary length. Bioinformatics, 16(11):1046–1047, Nov. 2000.
[74] W. McGinnis, R. L. Garber, J. Wirz, A. Kuroiwa, and W. J. Gehring. A homologous
protein-coding sequence in drosophila homeotic genes and its conservation in other
metazoans. Cell, 37(2):403–408, June 1984.
[75] W. McGinnis and R. Krumlauf. Homeobox genes and axial patterning. Cell, 68(2):283–
302, Jan. 1992.
[76] M. L. Metzker. Emerging technologies in DNA sequencing.
15(12):1767–1776, Dec. 2005.

Genome Research,

[77] T. Minokawa, K. Yagi, K. W. Makabe, and H. Nishida. Binary specification of nerve
cord and notochord cell fates in ascidian embryos. Development, 128(11):2007–2017,
June 2001.
[78] D. Miyamoto and R. Crowther. Formation of the notochord in living ascidian embryos.
Journal of Embryology and Experimental Morphology, VOL. 86:1–17, 1985.
[79] Y. Nakatani, R. Moody, and W. C. Smith. Mutations a↵ecting tail and notochord
development in the ascidian ciona savignyi. Development, 126(15):3293–3301, Aug.
1999.
[80] Y. Nakatani and H. Nishida. Duration of competence and inducing capacity of blastomeres in notochord induction during ascidian embryogenesis. Development, Growth
& Di↵erentiation, 41(4):449–453, Aug. 1999.
[81] Y. Nakatani, H. Yasuo, N. Satoh, and H. Nishida. Basic fibroblast growth factor
induces notochord formation and the expression of as-t, a brachyury homolog, during
ascidian embryogenesis. Development (Cambridge, England), 122(7):2023–2031, July
1996.
[82] H. Nishida. Cell lineage analysis in ascidian embryos by intracellular injection of
a tracer enzyme: III. up to the tissue restricted stage. Developmental Biology,
121(2):526–541, June 1987.
[83] H. Nishida. Development of the appendicularian oikopleura dioica: culture, genome,
and cell lineages. Development, Growth & Di↵erentiation, 50 Suppl 1:S239–256, June
2008.
[84] H. Nishida and N. Satoh. Cell lineage analysis in ascidian embryos by intracellular
injection of a tracer enzyme: I. up to the eight-cell stage. Developmental Biology,
99(2):382–394, Oct. 1983.
[85] H. Nishida and N. Satoh. Cell lineage analysis in ascidian embryos by intracellular
injection of a tracer enzyme: II. the 16- and 32-cell stages. Developmental Biology,
110(2):440–454, Aug. 1985.
92

[86] H. Nishida and T. Stach. Cell lineages and fate maps in tunicates: Conservation and
modification. Zoological Science, 31(10):645–652, Oct. 2014.
[87] S. T. ONeil and S. J. Emrich. Assessing de novo transcriptome assembly metrics for
consistency and utility. BMC Genomics, 14(1):465, July 2013.
[88] M. Paris and V. Laudet. The history of a developmental stage: metamorphosis in
chordates. Genesis (New York, N.Y. : 2000), 46(11):657–72, Nov. 2008.
[89] G. Parra, K. Bradnam, and I. Korf. CEGMA: a pipeline to accurately annotate core
genes in eukaryotic genomes. Bioinformatics (Oxford, England), 23(9):1061–1067, May
2007.
[90] Y. J. Passamaneck, L. Katikala, L. Perrone, M. P. Dunn, I. Oda-Ishii, and A. Di Gregorio. Direct activation of a notochord cis-regulatory module by brachyury and FoxA in
the ascidian ciona intestinalis. Development (Cambridge, England), 136(21):3679–89,
Nov. 2009.
[91] J. K. Pickrell, J. C. Marioni, A. A. Pai, J. F. Degner, B. E. Engelhardt, E. Nkadori,
J.-B. Veyrieras, M. Stephens, Y. Gilad, and J. K. Pritchard. variation with RNA
sequencing. Nature, 464(7289):768–772, 2010.
[92] M. Pop. Genome assembly reborn: recent computational challenges. Briefings in
Bioinformatics, 10(4):354–366, July 2009.
[93] M. A. Quail, M. Smith, P. Coupland, T. D. Otto, S. R. Harris, T. R. Connor,
A. Bertoni, H. P. Swerdlow, and Y. Gu. A tale of three next generation sequencing platforms: comparison of ion torrent, pacific biosciences and illumina MiSeq sequencers.
BMC genomics, 13:341, 2012.
[94] A. Roberts and L. Pachter. Streaming fragment assignment for real-time analysis of
sequencing experiments. Nature Methods, 10(1):71–73, Jan. 2013.
[95] M. D. Robinson, D. J. McCarthy, and G. K. Smyth. edgeR: a bioconductor package
for di↵erential expression analysis of digital gene expression data. Bioinformatics,
26(1):139–140, Jan. 2010.
[96] E. F. Rossomando and S. Alexander. Morphogenesis: An Analysis of the Development
of Biological Form: An Analysis of the Development of Biological Form. CRC Press,
Apr. 1992.
[97] N. Satoh. Ascidian embryos as a model system to analyze expression and function
of developmental genes. Di↵erentiation; Research in Biological Diversity, 68(1):1–12,
Aug. 2001.
[98] N. Satoh. The ascidian tadpole larva: comparative molecular development and genomics. Nature reviews. Genetics, 4(4):285–95, Apr. 2003.

93

[99] N. Satoh and M. Levine. Surfing with the tunicates into the post-genome era. Genes
& development, 19(20):2407–11, Oct. 2005.
[100] N. Satoh, Y. Satou, B. Davidson, and M. Levine. Ciona intestinalis: an emerging
model for whole-genome analyses. Trends in genetics : TIG, 19(7):376–81, July 2003.
[101] Y. Satou, K. S. Imai, and N. Satoh. The ascidian mesp gene specifies heart precursor
cells. Development (Cambridge, England), 131(11):2533–2541, June 2004.
[102] Y. Satou, K. Mineta, M. Ogasawara, Y. Sasakura, E. Shoguchi, K. Ueno, L. Yamada,
J. Matsumoto, J. Wasserscheid, K. Dewar, G. B. Wiley, S. L. Macmil, B. A. Roe, R. W.
Zeller, K. E. M. Hastings, P. Lemaire, E. Lindquist, T. Endo, K. Hotta, and K. Inaba.
Improved genome assembly and evidence-based global gene model set for the chordate
ciona intestinalis: new insight into intron and operon populations. Genome Biology,
9(10):R152, 2008.
[103] M. H. Schulz, D. R. Zerbino, M. Vingron, and E. Birney. Oases: robust de novo RNAseq assembly across the dynamic range of expression levels. Bioinformatics (Oxford,
England), 28(8):1086–92, Apr. 2012.
[104] E. M. Schwarz, P. K. Korhonen, B. E. Campbell, N. D. Young, A. R. Jex, A. Jabbar,
R. S. Hall, A. Mondal, A. C. Howe, J. Pell, A. Hofmann, P. R. Boag, X.-Q. Zhu, T. R.
Gregory, A. Loukas, B. A. Williams, I. Antoshechkin, C. T. Brown, P. W. Sternberg,
and R. B. Gasser. The genome and developmental transcriptome of the strongylid
nematode haemonchus contortus. Genome Biology, 14(8):R89, Aug. 2013.
[105] H.-C. Seo, R. B. Edvardsen, A. D. Maeland, M. Bjordal, M. F. Jensen, A. Hansen,
M. Flaat, J. Weissenbach, H. Lehrach, P. Wincker, R. Reinhardt, and D. Chourrout.
Hox cluster disintegration with persistent anteroposterior order of expression in oikopleura dioica. Nature, 431(7004):67–71, Sept. 2004.
[106] H.-C. Seo, M. Kube, R. B. Edvardsen, M. F. Jensen, A. Beck, E. Spriet, G. Gorsky,
E. M. Thompson, H. Lehrach, R. Reinhardt, and D. Chourrout. Miniature genome in
the marine chordate oikopleura dioica. Science, 294(5551):2506–2506, Dec. 2001.
[107] M. W. Simmen, S. Leitgeb, V. H. Clark, S. J. Jones, and A. Bird. Gene number in
an invertebrate chordate, ciona intestinalis. Proceedings of the National Academy of
Sciences of the United States of America, 95(8):4437–4440, Apr. 1998.
[108] W. T. Siok, C. A. Perfetti, Z. Jin, and L. H. Tan. Biological abnormality of impaired
reading is constrained by culture. Nature, 431(7004):71–76, Sept. 2004.
[109] T. Stach and J. M. Turbeville. Phylogeny of tunicata inferred from molecular and
morphological characters. Molecular Phylogenetics and Evolution, 25(3):408–428, Dec.
2002.
[110] J. Stapley, J. Reger, P. G. D. Feulner, C. Smadja, J. Galindo, R. Ekblom, C. Bennison,
A. D. Ball, A. P. Beckerman, and J. Slate. Adaptation genomics: the next generation.
Trends in ecology & evolution, 25(12):705–12, Dec. 2010.
94

[111] D. L. Stemple. Structure and function of the notochord: an essential organ for chordate
development. Development (Cambridge, England), 132(11):2503–12, June 2005.
[112] A. Stolfi and L. Christiaen. Genetic and genomic toolbox of the chordate ciona intestinalis. Genetics, 192(1):55–66, Sept. 2012.
[113] A. Stolfi, T. B. Gainous, J. J. Young, A. Mori, M. Levine, and L. Christiaen. Early
chordate origins of the vertebrate second heart field. Science (New York, N.Y.),
329(5991):565–568, July 2010.
[114] A. Stolfi and M. Levine. Neuronal subtype specification in the spinal cord of a protovertebrate. Development (Cambridge, England), 138(5):995–1004, Mar. 2011.
[115] A. Stolfi, E. K. Lowe, C. Racioppi, F. Ristoratore, C. T. Brown, B. J. Swalla, and
L. Christiaen. Divergent mechanisms regulate conserved cardiopharyngeal development
and gene expression in distantly related ascidians. eLife, 3:e03728, 2014.
[116] A. Stolfi, Y. Sasakura, D. Chalopin, Y. Satou, L. Christiaen, C. Dantec, T. Endo,
M. Naville, H. Nishida, B. J. Swalla, J.-N. Vol↵, A. Voskoboynik, D. Dauga, and
P. Lemaire. Guidelines for the nomenclature of genetic elements in tunicate genomes.
Genesis (New York, N.Y.: 2000), Sept. 2014.
[117] B. J. Swalla. Mechanisms of gastrulation and tail formation in ascidians. Microscopy
research and technique, 26(4):274–84, 1993.
[118] B. J. Swalla. Protochordate gastrulation: lancelets and ascidians. gastrulation. Cold
Spring Harbor: Cold Spring Harbor Press, 49:139, 2004.
[119] B. J. Swalla, M. R. Badgett, and W. R. Je↵ery. Identification of a cytoskeletal protein
localized in the myoplasm of ascidian eggs: localization is modified during anural
development. Development (Cambridge, England), 111(2):425–436, Feb. 1991.
[120] B. J. Swalla and W. R. Je↵ery. Interspecific hybridization between an anural and
urodele ascidian: di↵erential expression of urodele features suggests multiple mechanisms control anural development. Developmental biology, 142(2):319–34, Dec. 1990.
[121] B. J. Swalla and W. R. Je↵ery. Requirement of the manx gene for expression of chordate
features in a tailless ascidian larva. Science (New York, N.Y.), 274(5290):1205–8, Nov.
1996.
[122] B. J. Swalla, M. a. Just, E. L. Pederson, and W. R. Je↵ery. A multigene locus containing the manx and bobcat genes is required for development of chordate features in
the ascidian tadpole larva. Development (Cambridge, England), 126(8):1643–53, Apr.
1999.
[123] B. J. Swalla, K. W. Makabe, N. Satoh, and W. R. Je↵ery. Novel genes expressed
di↵erentially in ascidians with alternate modes of development. Development, 318:307–
318, 1993.
95

[124] N. Takada, N. Satoh, and B. J. Swalla. Expression of tbx6, a muscle lineage t-box
gene, in the tailless embryo of the ascidian molgula tectiformis. Development Genes
and Evolution, 212(7):354–356, Aug. 2002.
[125] N. Takada, J. York, J. M. Davis, B. Schumpert, H. Yasuo, N. Satoh, and B. J. Swalla.
Brachyury expression in tailless molgulid ascidian embryos. Evolution & development,
4(3):205–11, 2002.
[126] O. Tassy, D. Dauga, F. Daian, D. Sobral, F. Robin, P. Khoueiry, D. Salgado, V. Fox,
D. Caillol, R. Schiappa, B. Laporte, A. Rios, G. Luxardi, T. Kusakabe, J.-S. Joly,
S. Darras, L. Christiaen, M. Contensin, H. Auger, C. Lamy, C. Hudson, U. Rothbaecher, M. Gilchrist, K. Makabe, K. Hotta, S. Fujiwara, N. Satoh, Y. Satou, and
P. Lemaire. The ANISEED database: Digital representation, formalization and elucidation of a chordate developmental program. Genome research, pages 1459–1468, July
2010.
[127] C. Trapnell, A. Roberts, L. Go↵, G. Pertea, D. Kim, D. R. Kelley, H. Pimentel, S. L.
Salzberg, J. L. Rinn, and L. Pachter. Di↵erential gene and transcript expression analysis of RNA-seq experiments with TopHat and cu✏inks. Nature Protocols, 7(3):562–578,
Mar. 2012.
[128] J. R. True and E. S. Haag. Developmental system drift and flexibility in evolutionary
trajectories. Evolution & Development, 3(2):109–119, Apr. 2001.
[129] G. Tsagkogeorga, X. Turon, R. R. Hopcroft, M.-K. Tilak, T. Feldstein, N. Shenkar,
Y. Loya, D. Huchon, E. J. Douzery, and F. Delsuc. An updated 18s rRNA phylogeny
of tunicates based on mixture and secondary structure models. BMC Evolutionary
Biology, 9(1):187, Aug. 2009.
[130] G. Vachon, B. Cohen, C. Pfeifle, M. E. McGuffin, J. Botas, and S. M. Cohen.
Homeotic genes of the bithorax complex repress limb development in the abdomen
of the drosophila embryo through the target gene distal-less. Cell, 71(3):437–450, Oct.
1992.
[131] S. M. Van Belleghem, D. Roelofs, J. Van Houdt, and F. Hendrickx. De novo transcriptome assembly and SNP discovery in the wing polymorphic salt marsh beetle pogonus
chalceus (coleoptera, carabidae). PLoS ONE, 7(8):e42605, Aug. 2012.
[132] M. T. Veeman, Y. Nakatani, C. Hendrickson, V. Ericson, C. Lin, and W. C. Smith.
Chongmague reveals an essential role for laminin-mediated boundary formation in
chordate convergence and extension movements. Development (Cambridge, England),
135(1):33–41, Jan. 2008.
[133] N. Vijay, J. W. Poelstra, A. Knstner, and J. B. W. Wolf. Challenges and strategies in
transcriptome assembly and di↵erential gene expression quantification. a comprehensive in silico assessment of RNA-seq experiments. Molecular ecology, 46:620–634, Sept.
2012.
96

[134] J. P. Vinson, D. B. Ja↵e, K. O’Neill, E. K. Karlsson, N. Stange-Thomann, S. Anderson, J. P. Mesirov, N. Satoh, Y. Satou, C. Nusbaum, B. Birren, J. E. Galagan, and
E. S. Lander. Assembly of polymorphic genomes: algorithms and application to ciona
savignyi. Genome research, 15(8):1127–1135, Aug. 2005.
[135] A. Visel, S. Minovitsky, I. Dubchak, and L. a. Pennacchio. VISTA enhancer browsera
database of tissue-specific human enhancers. Nucleic acids research, 35(Database
issue):D88–92, Jan. 2007.
[136] H. Wada and N. Satoh. Details of the evolutionary history from invertebrates to
vertebrates, as deduced from the sequences of 18s rDNA. Proceedings of the National
Academy of Sciences of the United States of America, 91(5):1801–1804, Mar. 1994.
[137] Z. Wang, M. Gerstein, and M. Snyder. RNA-seq: a revolutionary tool for transcriptomics. Nature reviews. Genetics, 10(1):57–63, Jan. 2009.
[138] G. A. Wray and R. A. Ra↵. Evolutionary modification of cell lineage in the directdeveloping sea urchin heliocidaris erythrogramma. Developmental Biology, 132(2):458–
470, Apr. 1989.
[139] H. Yasuo and N. Satoh. An ascidian homolog of the mouse brachyury (t) gene is
expressed exclusively in notochord cells at the fate restricted stage. Development,
Growth & Di↵erentiation, 36(1):9–18, Feb. 1994.
[140] H. Yasuo and N. Satoh. Conservation of the developmental role of brachyury in notochord formation in a urochordate, the ascidian halocynthia roretzi. Developmental
Biology, 200(2):158–170, Aug. 1998.
[141] D. R. Zerbino and E. Birney. Velvet: algorithms for de novo short read assembly using
de bruijn graphs. Genome research, 18(5):821–9, May 2008.
[142] J. Zhang, R. Chiodini, A. Badr, and G. Zhang. The impact of next-generation sequencing on genomics. Journal of genetics and genomics, 38(3):95–109, Mar. 2011.

97