Evolution of alternative splicing in primate brain transcriptomes.

Abstract

Alternative splicing is a predominant form of gene regulation in higher eukaryotes. The evolution of alternative splicing provides an important mechanism for the acquisition of novel gene functions. In this work, we carried out a genome-wide phylogenetic survey of lineage-specific splicing patterns in the primate brain, via high-density exon junction array profiling of brain transcriptomes of humans, chimpanzees and rhesus macaques. We identified 509 genes showing splicing differences among these species. RT-PCR analysis of 40 exons confirmed the predicted splicing evolution of 33 exons. Of these 33 exons, outgroup analysis using rhesus macaques confirmed 13 exons with human-specific increase or decrease in transcript inclusion levels after humans diverged from chimpanzees. Some of the human-specific brain splicing patterns disrupt domains critical for protein-protein interactions, and some modulate translational efficiency of their host genes. Strikingly, for exons showing splicing differences across species, we observed a significant increase in the rate of silent substitutions within exons, coupled with accelerated sequence divergence in flanking introns. This indicates that evolution of cis-regulatory signals is a major contributor to the emergence of human-specific splicing patterns. In one gene (MAGOH), using minigene reporter assays, we demonstrated that the combination of two human-specific cis-sequence changes created its human-specific splicing pattern. Together, our data reveal widespread human-specific changes of alternative splicing in the brain and suggest an important role of splicing in the evolution of neuronal gene regulation and functions.

Full text

Page 1

Evolution of alternative splicing in primate
brain transcriptomes
LanLin1,{,{, ShihaoShen2,{,{, PengJiang1, SeikoSato1, Beverly L.Davidson1,3,4and YiXing1,2,5,∗
1Department of Internal Medicine,2Department of Biostatistics,3Department of Molecular Physiology and Biophysics,
4Department of Neurology and5Department of Biomedical Engineering, University of Iowa, Iowa City, IA 52242, USA
Received February 27, 2010; Revised April 16, 2010; Accepted May 7, 2010
Alternative splicing is a predominant form of gene regulation in higher eukaryotes. The evolution of alterna-
tive splicing provides an important mechanism for the acquisition of novel gene functions. In this work, we
carried out a genome-wide phylogenetic survey of lineage-specific splicing patterns in the primate brain, via
high-density exon junction array profiling of brain transcriptomes of humans, chimpanzees and rhesus maca-
ques. We identified 509 genes showing splicing differences among these species. RT–PCR analysis of 40
exons confirmed the predicted splicing evolution of 33 exons. Of these 33 exons, outgroup analysis using
rhesus macaques confirmed 13 exons with human-specific increase or decrease in transcript inclusion
levels after humans diverged from chimpanzees. Some of the human-specific brain splicing patterns disrupt
domains critical for protein–protein interactions, and some modulate translational efficiency of their host
genes. Strikingly, for exons showing splicing differences across species, we observed a significant increase
in the rate of silent substitutions within exons, coupled with accelerated sequence divergence in flanking
introns. This indicates that evolution of cis-regulatory signals is a major contributor to the emergence of
human-specific splicing patterns. In one gene (MAGOH), using minigene reporter assays, we demonstrated
that the combination of two human-specific cis-sequence changes created its human-specific splicing pat-
tern. Together, our data reveal widespread human-specific changes of alternative splicing in the brain and
suggest an important role of splicing in the evolution of neuronal gene regulation and functions.
INTRODUCTION
Despite their close evolutionary relationships, humans and
nonhuman primates have notable differences in phenotypic
traits and susceptibility to various diseases (1). A long-
standing challenge in evolutionary biology is to uncover
genomic changes that account for the unique attributes of
the human species. It has been proposed that evolution of
gene regulation is a driving force for phenotypic divergence
between species (2). Consistent with this theory, studies of
human and nonhuman primate transcriptomes revealed wide-
spread changes in steady-state transcript levels during recent
human evolution (3–8).
In addition to transcriptional regulation, alternative splicing
is another predominant mechanism of gene regulation in
higher eukaryotes. Alternative splicing occurs among different
tissues (9) or developmental states (10,11), during cellular
responses to external cues (12,13) and in a wide range of
human diseases (14,15). In humans, the vast majority of
protein-coding genes are alternatively spliced (16–18). For
example, deep RNA sequencing indicates that .90% of multi-
exon human genes undergo alternative splicing (18,19). By
producing multiple mRNA and protein products from a
single gene, alternative splicing is capable of generating tre-
mendous functional and regulatory diversity from a limited
repertoire of protein-coding genes in eukaryotic genomes
(20,21).
There are numerous examples of genes with species-specific
exons or splicing patterns (22,23). Comparative analyses of
cDNA and EST data indicate that many alternative splicing
events are not conserved between species (24–27). In fact,
during evolution, ancient exons undergo creation and loss of
alternative splicing patterns (28–30), and new exons are fre-
quently added to existing functional genes (23,24,31,32).
†These two authors contributed equally to this work.
‡The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.
∗To whom correspondence should be addressed. Tel: +1 3193843099; Fax: +1 3193843150; Email: yi-xing@uiowa.edu
# The Author 2010. Published by Oxford University Press. All rights reserved.
For Permissions, please email: journals.permissions@oxfordjournals.org
Human Molecular Genetics, 2010, Vol. 19, No. 15
doi:10.1093/hmg/ddq201
Advance Access published on May 11, 2010
2958–2973
at University of Iowa Libraries/Serials Acquisitions on July 11, 2010
http://hmg.oxfordjournals.org
Downloaded from

Page 2

Thus, evolution could use splicing to create species-specific
post-transcriptional regulation of gene expression as well as
novel protein function. One well-known example is the
primate-specific exon in a human gene ADAR2 (adenosine
de-aminase, RNA-specific, B1). This exon is created from
an Alu retrotransposon during primate evolution (33). It
inserts a new peptide segment into the catalytic domain of
ADAR2, altering the catalytic activity of the protein product
(34). Even between closely related species, such as humans
and chimpanzees, differences in splicing exist (35,36). For
example, using a custom Agilent microarray targeting
approximately 1700 exon-skipping events and subsequent
RT–PCR assays, we and colleagues identified 30 exons with
different levels of transcript inclusion in corresponding
tissues (frontal cortex or heart) of humans and chimpanzees
(35). These data suggest that the splicing patterns of ortholo-
gous genes can diverge over relatively short evolutionary
timescales [i.e. ?5–7 million years that separate humans
and chimpanzees (37)].
In this work, we conducted a systematic survey of splicing
differences between human and nonhuman primate brains,
particularly lineage-specific changes in splicing after humans
diverged from chimpanzees. We carried out a high-throughput
microarray analysis of alternative splicing in brain transcrip-
tomes of human, chimpanzee and rhesus macaque, followed
by extensive RT–PCR tests in a large panel of tissue
samples from all three species. Our study revealed widespread
changes of alternative splicing in the brain transcriptome
during primate and human evolution.
RESULTS
Comparative analysis of alternative splicing between
humans and nonhuman primates using a high-density
exon junction array
To study alternative splicing in human and primate brains, we
extracted total RNAs from the cerebellum of six chimpanzees,
six rhesus macaques and two pooled human cerebellum
samples each with 10 or more individual donors (see Materials
and Methods). These RNA samples were processed and hybri-
dized to the Affymetrix Human Exon Junction Array (HJAY)
for analysis of alternative splicing patterns (38). The HJAY
array is a next-generation Affymetrix exon array for genome-
wide analysis of alternative splicing. It averages eight probes
per probe set for 315 137 exons in the human genome, and
also includes probe sets for 260 488 exon–exon junctions.
The high probe density, inclusion of exon–exon junction
probes and comprehensive coverage of known alternative spli-
cing events make the HJAY array a powerful tool for analysis
of alternative splicing (38–40). Our recent studies have shown
that this new array detects changes in gene expression and
alternative splicing at a very low false positive rate (38,41).
Although originally designed for analysis of human genes, a
large percentage of HJAY probes perfectly match orthologous
transcripts from closely related nonhuman primates (41).
Thus, we can use this array for comparative analysis of
alternative splicing patterns by directly hybridizing RNA
samples from closely related primate species to the HJAY
array. For this work, we generated six HJAY profiles per
species, with one replicate per sample for chimpanzee and
rhesus cerebellums and three replicates per sample for the
two human RNA pools.
To identify alternative splicing differences of orthologous
genes among these species, we compared HJAY profiles of
human, chimpanzee and rhesus cerebellums. We used our pub-
lished computational procedures for alternative splicing analy-
sis of the HJAY array data (38,41–45), after filtering
microarray probes with mismatches for orthologous chimpan-
zee/rhesus transcripts. The procedures of our HJAY array data
analysis are summarized in Materials and Methods and
described in detail in Supplementary Material, Methods.
Briefly, we identified all HJAY exon probes and exon–exon
junction probes that perfectly matched orthologous transcripts
from chimpanzees or rhesus macaques, using the UCSC pair-
wise genome alignments of the human genome (hg18) to the
genomes of chimpanzee (panTro2) or rhesus macaque
(rheMac2) (46–47). Of the 20 815 mRNA/EST-supported
human alternative splicing events interrogated by the HJAY
array, 15 544 and 10 147 events had sufficient perfect-match
probes for orthologous chimpanzee and rhesus transcript iso-
forms (see definition in Supplementary Material, Methods).
For these events, we used our MADS+ (Microarray Analysis
of Differential Splicing) algorithm (38,45) in pairwise com-
parisons of human versus chimpanzee or human versus
rhesus array profiles to identify differential splicing events.
Specifically, we restricted the MADS+ analysis to HJAY
probes perfectly matching orthologous transcripts from mul-
tiple species. For every alternative splicing event analyzed,
MADS+ calculated P-values of multiple probe sets targeting
exons and exon–exon junctions (see Fig. 1A for the probe
design for exon-skipping events), requiring opposite trends
for probe sets targeting competing isoforms as evidence of
differential alternative splicing (38). For example, as shown
in Figure 1, the exon 2 of GPBP1L1 (GC-rich promoter
binding protein 1-like 1) was predicted to be differentially
spliced between human and rhesus cerebellums. From the
HJAY array data of this exon, we observed significantly
higher intensities of the exon inclusion probe sets (i.e. probe
sets targeting the upstream exon–exon junction, downstream
exon–exon junction and exon) and significantly lower intensi-
ties of the exon-skipping probe set in human samples com-
pared with rhesus samples (Fig. 1B; solid blue lines).
Meanwhile,the estimated
GPBP1L1 was comparable in human and rhesus (Fig. 1B;
dashed red lines). These data indicated that this exon was
included at a higher level in the human cerebellum. Indeed,
our subsequent RT–PCR validation confirmed that this exon
was included in human transcripts at intermediate-to-high
levels, but almost completely skipped in orthologous rhesus
transcripts (Fig. 1C).
Using these computational procedures, our analysis of the
HJAY profilesrevealedwidespread
between humans and closely related nonhuman primates. In
total, we identified 336 alternative splicing differences between
human and chimpanzee cerebellums and 415 differences
between human and rhesus cerebellums. These events covered
all types of alternative splicing patterns, with cassette exons (i.e.
exon skipping) being the most prevalent type (Table 1). It
should be noted that because of our filtering criteria, in particular
geneexpression levelof
splicing differences
Human Molecular Genetics, 2010, Vol. 19, No. 152959
at University of Iowa Libraries/Serials Acquisitions on July 11, 2010
http://hmg.oxfordjournals.org
Downloaded from

Page 3

thenumberofperfect-matchprobesfororthologoustranscripts,a
much smaller number of alternative splicing events can be ana-
lyzedforhuman–rhesusdifferencesthanforhuman–chimpanzee
differences (1501 versus 2646, see Table 1). Despite this, we
identified more alternative splicing events with human–rhesus
differences (415) than with human–chimpanzee differences
(336), representing 27.6% and 12.7% of all events analyzed
between humans and rhesus macaques or between humans and
chimpanzees,respectively.Thiswasconsistentwiththephyloge-
netic relationship among these three species (37).
The percentage of alternative splicing events with predicted
human–chimpanzee differences in the cerebellum (12.7%)
Figure 1. Detection of a differentially spliced cassette exon between human and rhesus cerebellums by the HJAY array. (A) The HJAY probe design for cassette
exons. Each probe set has eight probes. (B) The HJAY array data indicate significantly elevated exon inclusion of GPBP1L1 exon 2 in the human cerebellum
compared with the rhesus cerebellum. In each panel, the dashed red line indicates the average gene expression levels of GPBP1L1 in six replicate samples from
humans or rhesus macaques. The individual blue lines indicate the average background-corrected intensities of individual probes in a probe set. Only probes that
perfectly match orthologous human and rhesus transcripts are shown. MADS+ P-value for differential splicing is shown below the name of the probe set. The
‘+’ and ‘2’ signs indicate the predicted direction of change in exon inclusion levels in the human cerebellum over the rhesus cerebellum. (C) The predicted
splicing difference is confirmed by RT–PCR.
Table 1. Summary of differential splicing events between human, chimpanzee and rhesus cerebellums identified by the HJAY array
Cassette exonsAlternative 5′/3′splice sitesMutually exclusive exon usage
Human versus chimpanzeeDifferential splicing eventsa
Splicing events analyzedb
Differential splicing events
Splicing events analyzed
281
1985
340
1126
2926
338
24
180
323
51
195
Human versus rhesus
aIdentified by MADS+ (see Supplementary Material, Methods).
bThe total set of splicing events that passed the filters for gene expression level in the cerebellum, probe intensity and number of perfect-match probes for
orthologous chimpanzee or rhesus transcripts (see Supplementary Material, Methods). These splicing events were analyzed by MADS+ for the detection of
splicing differences between species.
2960Human Molecular Genetics, 2010, Vol. 19, No. 15
at University of Iowa Libraries/Serials Acquisitions on July 11, 2010
http://hmg.oxfordjournals.org
Downloaded from

Page 4

appeared to be higher than the percentage reported for two
other tissues (frontal cortex and heart) (6–8%) in our earlier
study (35). However, these statistics were not directly compar-
able, as distinct microarray platforms were used in these two
studies. Specifically, in the present work, we used a high-
density Affymetrix short oligonucleotide exon junction
array, with approximately 32 probes for each exon-skipping
event (38). In contrast, the earlier study used a custom
Agilent long oligonucleotide array, with six probes designed
to interrogate each exon-skipping event (35). Thus, the differ-
ence in the percentage of alternative splicing events with pre-
dicted human–chimpanzee differences could be largely
attributed to the sensitivity of these two microarray platforms.
We also noted that in the previous study using the Agilent
custom array, we did not observe any significant difference
in the rate of splicing evolution in frontal cortex and heart
(35).
RT–PCR validation of alternative splicing evolution and
human-specific splicing patterns in the brain
To confirm the identified alternative splicing differences
betweenspecies,weselected40cassetteexonsforexperimental
validation, from 281 cassette exons predicted to have splicing
differences between human and chimpanzee and 340 cassette
exons predicted to have splicing differences between human
and rhesus. All predicted events in the human versus chimpan-
zee comparison or the human versus rhesus comparison were
ranked according to the overall level of statistical significance
summarized from all available exon probes and exon–exon
junction probes (see details in Supplementary Material,
Methods). The 40 selected exons were analyzed by semi-
quantitative RT–PCR in a large panel of cerebellum tissues
from all three species. For a subset of these exons with subtle
changes in inclusion levels, we also used a highly sensitive
fluorescently labeled RT–PCR protocol, which allowed us to
accurately quantify the levels oftranscript inclusion indifferent
species (see Materials and Methods). To distinguish bona fide
inter-species splicing divergence from intra-species splicing
variability, our RT–PCR analysis examined all individual
samples used for generating the HJAY array profiles, as well
as RNAs from additional human/chimpanzee/rhesus tissue
samples (see Supplementary Material, Table S1). Of the 40
exons tested, 33 showed splicing differences between species
as predicted by the HJAY array, yielding a validation rate of
83%. For most of the RT–PCR confirmed events, the splicing
patterns of orthologous exons were consistent within species,
while different between species. For example, the exon 2 in
the 5′-UTR of GPBP1L1 was predicted according to the
HJAY array data to be differentially spliced between humans
and rhesus macaques (Fig. 1B). This exon was completely
skipped in all eight rhesus cerebellum samples from different
animals. In contrast, we observed a consistent pattern of
intermediate-to-highlevelsofexoninclusioninfourhumancer-
ebellum samples from single or multiple donors, and in eight
chimpanzee cerebellum samples from different animals
(Fig. 2A).
It should be emphasized that in selecting candidate exons
for RT–PCR analysis, we included exons with various
levels of statistical significance and did not cherry-pick
candidates from the top of our predicted events to achieve a
high validation rate. In fact, a primary selection criterion for
experimental validation was to facilitate RT–PCR primer
design and data interpretation, i.e. the selected cassette
exons did not have alternative splice site usage at the upstream
and downstream exon–intron boundaries and were flanked by
constitutively spliced exons based on the UCSC Genome
Browser annotation. All cassette exons analyzed by the
HJAY array in the human versus chimpanzee or the human
versus rhesus comparison were ranked by a single combined
P-value for the overall statistical significance of inter-species
splicing difference, which was summarized from individual
P-values of all available perfect-match probes targeting ortho-
logous exons and exon–exon junctions as in Shen et al. (38)
and Xing et al. (45). Among the 281 predicted cassette exon
events in the human versus chimpanzee comparison, the
median rank of RT–PCR confirmed events was 65 with an
inter-quartile range (IQR) of 9 to 152. The lowest rank of vali-
dated events was 239. In the human versus rhesus comparison,
among the 340 predicted cassette exon events, the median rank
of RT–PCR confirmed events was 80 with an IQR of 43 to
169 and the lowest rank of 320. Thus, based on our validation
rate (83%), we expect that the vast majority of our HJAY-
predicted events represent real splicing differences between
humans and closely related nonhuman primates. The list of
validated events and their RT–PCR gel pictures are provided
in Supplementary Material, Table S2. The rankings of RT–
PCR-validated events among all HJAY-predicted events are
provided in Supplementary Material, Table S3. The complete
lists of HJAY-predicted cassette exon events are provided in
Supplementary Material, Tables S4 and S5.
Our RT–PCR analysis also revealed a number of exons
with human-specific changes in splicing compared with both
chimpanzee and rhesus orthologous exons. In this phyloge-
netic analysis, the observed splicing pattern in the rhesus cer-
ebellum was used as the outgroup to infer the direction of
splicing changes of the human exon after humans diverged
from chimpanzees. For example, in DDX42 (DEAD box
protein 42), the exon 2 in its 5′-UTR had different levels of
exon inclusion in human, chimpanzee and rhesus cerebellums
(Fig. 2B). Based on RT–PCR analysis, this exon was comple-
tely skipped in the rhesus transcripts, weakly included in the
chimpanzee transcripts and included at an intermediate level
in the human transcripts. This suggests that the splicing
activity of DDX42 in the cerebellum was gradually strength-
ened in the lineage leading to humans. Together, of these 33
exons with RT–PCR-validated splicing differences among
species, 13 exons (39%) had consistently higher or lower
levels of transcript inclusion in the human cerebellum com-
pared withbothchimpanzee
suggesting recent increases or decreases in splicing activities
after humans diverged from chimpanzees. These genes
covered a broad range of functional categories (Table 2).
Extrapolating from the total number of events identified by
the HJAY analysis (Table 1), the overall validation rate in
our RT–PCR analysis (83%) and the percentage of exons
with human-specific splicing changes among all exons with
validated splicing differences (39%), we estimated that more
than 100 alternative splicing events in our entire data set
underwent human-specific splicing changes. These splicing
and rhesuscerebellums,
Human Molecular Genetics, 2010, Vol. 19, No. 15 2961
at University of Iowa Libraries/Serials Acquisitions on July 11, 2010
http://hmg.oxfordjournals.org
Downloaded from

Page 5

events represent a group of unique transcriptome signatures in
the human brain that can distinguish humans from other
closely related primate species.
Previous studies have suggested a difference in the rate of
protein sequence evolution between the human and chimpan-
zee lineages (48–50). Using our RT–PCR data of 33 exons
(Supplementary Material, Table S2), we compared the
numbers of exons with human-specific or chimpanzee-specific
splicing changes in the brain. To enable an unbiased compari-
son, we focused on 16 exons predicted by the HJAY array to
have splicing differences between humans and chimpanzees
and subsequently validated by RT–PCR. Of these 16 exons,
2 exons did not have RT–PCR data in rhesus macaques due
to the difficulty of primer design. For the remaining 14
exons, using the rhesus macaque as the outgroup, we identified
9 exons with human-specific splicing changes and 5 exons
with chimpanzee-specific splicing changes. However, we
caution that any conclusion on the rate difference of splicing
changes in the human and chimpanzee lineages would
require a much more extensive set of experimentally validated
splicing differences between all three species.
Functional and regulatory implications of alternative
splicing evolution in the brain
On all genes with identified splicing differences between
human, chimpanzee and rhesus cerebellums, Gene Ontology
analysis using the DAVID tool (51) found four strongly
enriched GO terms: ‘cytoskeleton organization and biogen-
esis’ (P ¼ 0.007), ‘RNA processing’ (P ¼ 0.012), ‘cell–cell
adhesion’ (P ¼ 0.025) and ‘neurological system process’
(P ¼ 0.028) (see the DAVID analysis procedure in Materials
and Methods). A previous study of natural selection on
protein-coding regions of human genes found cytoskeletal
proteins to be under strong negative selection pressure
during human evolution (52). Our result thus suggests an intri-
guing scenario that genes encoding cytoskeletal proteins could
experience contrasting modes of selection pressure at the RNA
level and at the protein level during recent human evolution.
Next, we investigated potential evidence of positive
selection for exons with human-specific splicing changes.
Because of the small size of individual exons, methods to
detect positive selection using divergence data between
Figure 2. Examples of exons with splicing differences between human and chimpanzee/rhesus cerebellums. (A) RT–PCR analysis of the exon 2 of GPBP1L1 in
human (Hs), chimpanzee (Pt) and rhesus macaque (Rm) samples. (B) RT–PCR analysis of the exon 2 of DDX42 in human (Hs), chimpanzee (Pt) and rhesus
macaque (Rm) samples. (C) RT–PCR analysis of the exon 2 of HIP14L (also referred to as ZDHHC13) in human (Hs), chimpanzee (Pt) and rhesus macaque
(Rm) samples. ANK, ankyrin repeat; TM, transmembrane domain; DHHC (Asp–His–His–Cys), cysteine-rich domain. (D) RT–PCR analysis of the exon 3 of
MAGOH in human (Hs), chimpanzee (Pt) and rhesus macaque (Rm) samples. In each figure, the exon with differential splicing between species is indicated as
the green box in the gene structure diagram. Translation start site is indicated by the forward arrow. The sizes of exons and introns are not drawn according to the
scale.
2962Human Molecular Genetics, 2010, Vol. 19, No. 15
at University of Iowa Libraries/Serials Acquisitions on July 11, 2010
http://hmg.oxfordjournals.org
Downloaded from