ARTICLE OPEN

doi:10.1038/nature12962

Diversity and dynamics of the Drosophila

transcriptome
James B. Brown1,2*, Nathan Boley1*, Robert Eisman3*, Gemma E. May4*, Marcus H. Stoiber1*, Michael O. Duff4, Ben W. Booth2,
Jiayu Wen5, Soo Park2, Ana Maria Suzuki6,7, Kenneth H. Wan2, Charles Yu2, Dayu Zhang8, Joseph W. Carlson2, Lucy Cherbas3,
Brian D. Eads3, David Miller3, Keithanne Mockaitis3, Johnny Roberts8, Carrie A. Davis9, Erwin Frise2, Ann S. Hammonds2,
Sara Olson4, Sol Shenker5, David Sturgill10, Anastasia A. Samsonova11,12, Richard Weiszmann2, Garret Robinson1,
Juan Hernandez1, Justen Andrews3, Peter J. Bickel1, Piero Carninci6,7, Peter Cherbas3,8, Thomas R. Gingeras9, Roger A. Hoskins2,
Thomas C. Kaufman3, Eric C. Lai5, Brian Oliver10, Norbert Perrimon11,12, Brenton R. Graveley4 & Susan E. Celniker2

Animal transcriptomes are dynamic, with each cell type, tissue and organ system expressing an ensemble of transcript
isoforms that give rise to substantial diversity. Here we have identified new genes, transcripts and proteins using
poly(A)1 RNA sequencing from Drosophila melanogaster in cultured cell lines, dissected organ systems and under
environmental perturbations. We found that a small set of mostly neural-specific genes has the potential to encode
thousands of transcripts each through extensive alternative promoter usage and RNA splicing. The magnitudes of splicing
changes are larger between tissues than between developmental stages, and most sex-specific splicing is gonad-specific.
Gonads express hundreds of previously unknown coding and long non-coding RNAs (lncRNAs), some of which are
antisense to protein-coding genes and produce short regulatory RNAs. Furthermore, previously identified pervasive
intergenic transcription occurs primarily within newly identified introns. The fly transcriptome is substantially more
complex than previously recognized, with this complexity arising from combinatorial usage of promoters, splice sites and
polyadenylation sites.

Next-generation RNA sequencing (RNA-seq) has permitted the map- start sites (CAGE and 59 rapid amplification of cDNA ends (RACE)),
ping of transcribed regions of the genomes of a variety of organisms1,2. splice sites and exons (RNA-seq), and polyadenylation sites (39 expressed
These studies demonstrated that large fractions of metazoan genomes sequence tags (ESTs), cDNAs and RNA-seq). We analysed poly(A)1
are transcribed, and they also catalogued individual elements of tran- RNA data from a diverse set of developmental stages6, dissected organ
scriptomes, including transcription start sites3, polyadenylation sites4,5, systems and environmental perturbations; most of this data is new and
exons and introns6. However, the complexity of the transcriptome arises strand-specific. Our data provide higher spatiotemporal resolution and
from the combinatorial incorporation of these elements into mature allow for deeper exploration of the Drosophila transcriptome than was
transcript isoforms. Studies that inferred transcript isoforms from short- previously possible. Our analysis reveals a transcriptome of high com-
read sequence data focused on a small subset of isoforms, filtered using plexity that is expressed in discrete, tissue- and condition-specific mes-
stringent criteria7,8. Studies using complementary DNA (cDNA) or ex- senger RNA and lncRNA transcript isoforms that span most of the
pressed sequence tag (EST) data to infer transcript isoforms have not genome and provides valuable insights into metazoan biology.
had sufficient sampling depth to explore the diversity of RNA products
at most genomic loci9. Although the human genome has been the focus A dense landscape of discrete poly(A)1 transcripts
of intensive manual annotation10, analysis of strand-specific RNA-seq To broadly sample the transcriptome, we performed strand-specific,
data from human cell lines reveals over 100,000 splice junctions not paired-end sequencing of poly(A)1 RNA in biological duplicate from
incorporated into transcript models11. Thus, a large gap exists between 29 dissected tissue samples including the nervous, digestive, reproduct-
genome annotations and the emerging transcriptomes observed in next- ive, endocrine, epidermal and muscle organ systems of larvae, pupae
generation sequence data. In Drosophila, we previously described a non- and adults. To detect RNAs not observed under standard conditions, we
strand-specific poly(A)1 RNA-seq analysis of a developmental time sequenced poly(A)1 RNA in biological duplicate from 21 whole-animal
course through the life cycle6 and cap analysis of gene expression (CAGE) samples treated with environmental perturbations. Adults were chal-
analysis of the embryo12, which discovered thousands of unannotated lenged with heat-shock, cold-shock, exposure to heavy metals (cadmium,
exons, introns and promoters, and expanded coverage of the genome copper and zinc), the drug caffeine or the herbicide paraquat. To deter-
by identified transcribed regions, but not all elements were incorpo- mine whether exposing larvae resulted in RNA expression from prev-
rated into full-length transcript models. Here we describe an expansive iously unidentified genes, we treated them with heavy metals, caffeine,
poly(A)1 transcript set modelled by integrative analysis of transcription ethanol or rotenone. Finally, we sequenced poly(A)1 RNA from 21
1
Department of Statistics, University of California Berkeley, Berkeley, California 94720, USA. 2Department of Genome Dynamics, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA.
3
Department of Biology, Indiana University, 1001 East 3rd Street, Bloomington, Indiana 47405, USA. 4Department of Genetics and Developmental Biology, Institute for Systems Genomics, University of
Connecticut Health Center, 400 Farmington Avenue, Farmington, Connecticut 06030, USA. 5Sloan-Kettering Institute, 1017C Rockefeller Research Labs, 1275 York Avenue, Box 252, New York, New York
10065, USA. 6RIKEN Omics Science Center, Yokohama, Kanagawa 230-0045, Japan. 7RIKEN Center for Life Science Technologies, Division of Genomic Technologies, Yokohama, Kanagawa, 230-0045,
Japan. 8Center for Genomics and Bioinformatics, Indiana University, 1001 East 3rd Street, Bloomington, Indiana 47405, USA. 9Cold Spring Harbor Laboratory, Cold Spring Harbor, New York 11724, USA.
10
Section of Developmental Genomics, Laboratory of Cellular and Developmental Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda,
Maryland 20892, USA. 11Department of Genetics, Harvard Medical School, 77 Avenue Louis Pasteur, Boston, Massachusetts 02115, USA. 12Howard Hughes Medical Institute, Harvard Medical School, 77
Avenue Louis Pasteur, Boston, Massachusetts 02115, USA.
*These authors contributed equally to this work.

2 8 AU G U S T 2 0 1 4 | VO L 5 1 2 | N AT U R E | 3 9 3
©2014 Macmillan Publishers Limited. All rights reserved
 RESEARCH ARTICLE

previously described13 and three ovary-derived cell lines (Supplementary from the second promoter of ovo on the opposite strand and spanning
Methods). In total, we produced 12.4 billion strand-specific read pairs 107 kb (Fig. 1c). Exons of 36 new genes overlap molecularly defined
and over a terabase of sequence data, providing 44,000-fold coverage of mutations with associated phenotypes (genome structure correction
the poly(A)1 transcriptome. (GSC) P value ,0.0002), indicating potential functions (Supplemen-
Reads were aligned to the Drosophila genome as described6, and full- tary Table 2). For example, the lethal P-element insertions l(3)L3051
length transcript models were assembled using our custom pipeline and l(3)L4111 (ref. 17) map to promoters of Mgn095159 and Mgn95009,
termed GRIT14, which uses RNA-seq, poly(A)1seq, CAGE, RACE12, respectively, indicating these may be essential genes. Nearly 60% of the
ESTs15 and full-length cDNAs16 to generate gene and transcript models intergenic transcription we previously reported6 is now incorporated
(Supplementary Methods). We integrated these models with our own into gene models.
and community manual curation data sets to obtain an annotation (Sup-
plementary Information, section 12) consisting of 304,788 transcripts Transcript diversity
and 17,564 genes (Fig. 1a and Supplementary Fig. 1), of which 14,692
are protein-coding (Supplementary Data 1 and updates available at Over half of spliced genes (7,412; 56%) encode two or more transcript
http://fruitfly.org). Ninety per cent of genes produce at most 10 tran- isoforms with alternative first exons. Most of such genes produce alterna-
script and five protein isoforms, whereas 1% of genes have highly com- tive first exons through coordinated alternative splicing and promoter
plex patterns of alternative splicing, promoter usage and polyadenylation, usage (59%, 4,389 genes, hypergeometric P value , 1 3 10216); how-
and may each be processed into hundreds of transcripts (Fig. 1a, b). ever, a substantial number of genes use one, but not both mechanisms
Our gene models span 72% of the euchromatin, an increase from 65% (Fig. 2a). Only 1,058 spliced genes have alternative first exons that alter
in FlyBase 5.12 (FB5.12), the reference annotation at the beginning of protein-encoding capacity and increase the complexity of the predicted
the modENCODE project (Supplementary Table 1 compares annota- proteome. Some genes, such as G protein b-subunit 13F (Gb13F, Fig. 2b
tions in 2008–13). There were 64 euchromatic gene-free regions longer and Supplementary Fig. 3) have exceptionally complex 59 UTRs, but
than 50 kb in FB5.12, and 25 remaining in FB5.45. Our annotation encode a single protein.
includes new gene models in each of these regions. Newly identified We measured splicing efficiency using the ‘per cent spliced in’ (Y)
genes (1,468 total) are expressed in spatially and temporally restricted index—the fraction of isoforms that contain the particular exon6. Introns
patterns (Supplementary Fig. 2), and 536 reside in previously unchar- flanked by coding sequence are retained at an average Y 5 0.7, whereas
acterized gene-free regions. Others map to well-characterized regions, introns flanked by non-coding sequence are retained . fivefold more
including the ovo locus, where we discovered a new ovary-specific, often, with an average Y 5 3.8 (P , 13 10216 subsampling/two-sample
poly(A)1 transcript (Mgn94020, Supplementary Data 1 and 2), extending t-test), and is most frequent in 59 UTRs (mean Y 5 5.1, Fig. 2c).

a 10,000
1 c 4,850K ChrX 4,900K 4,950K
5 14,898 938
Transcripts per gene

Tissue-specific Ovaries
10
1000 RNA-seq +
–
50
–437 –13
100 100
Transcript Mgn94021 CG12680
500 models
10 Ptp4E
1,000 ovo
5,000 SIP3 Mgn94020
1
NC 1 10 100 1,000
Proteins per gene (longest ORF per trans.)

b 15,290K 15,310K 15,330K 15,350K 15,370K 15,390K 15,410K Chr 3R

CNS
Tissue-specific
RNA-seq Imaginal disc

Junctions
Dys
Transcript
models

AAAAA AAAAA AAAAA AAAAA

Transcript
models
expansion
of the 3′ ends

CNS
Tissue-specific
RNA-seq Imaginal disc

Junctions
15,419K 15,423K

Figure 1 | Overview of the annotation of the Drosophila melanogaster Highlighted is alternative splicing and polyadenylation at the 39 end. CAGE
transcriptome. a, Scatterplot showing the per gene correlation between (black), RNA-seq (tan, blue), splice junctions (shaded grey as a function of
number of proteins and number of transcripts. The genes Dscam and para usage). c, An internal promoter of ovo is bidirectional in ovaries and produces a
are omitted as extreme outliers both encoding .10,000 unique proteins. lncRNA (430 bp, red) bridging two gene deserts. CAGE (black), RNA-seq
b, Dystrophin (Dys) produces 72 transcripts and encodes 32 proteins. (pink), counts are read-depth (minus-strand given as negative).

3 9 4 | N AT U R E | VO L 5 1 2 | 2 8 AU G U S T 2 0 1 4
©2014 Macmillan Publishers Limited. All rights reserved
 ARTICLE RESEARCH

a 0 c 1.0

First exon variants

0.8
Multiple TSSs Multiple DS
0.6

Ψ
0.4
75 25 Alternative TSSs and DS

0.2
Alternative TSSs and DS in CDS
0

0 1 2 3 4 5 6
Constitutive first exons
50 Length (kb)

b ChrX 15,755K 15,756K 15,757K 15,758K 15,759K 15,760K

206 8622
Tissue-specific Testes
RNA-seq
361 9672
CNS

Junctions

AAAAA
AAAAA AAAAA AAAAA
Gβ13F AAAAA AAAAA AAAAA AAAAA AAAAA
Transcript AAAAA
models

Transcript
models
expansion
of the 5′ ends

Figure 2 | Splicing complexity across the gene body. a, Alternative first exons addition sites (from adult head poly(A)1seq, red) and generates 235 transcripts.
occur in two main configurations: multiple transcription start sites (TSS, pink) The first exon has two alternative splice acceptors that splice to one of eleven
and multiple donor sites (DS, light blue). A subset of the genes in the multiple different donor sites. Only five donor sites are shown owing to the proximity of
TSS category produce transcripts with different TSSs and shared donor sites splice sites. Four splice donors are represented by the single red line differing
(red), and a subset of the genes in the DS category produce transcripts with a by 12, 5 and 19 bp, respectively. Three splice donors are represented by the
shared TSS and different donor sites (blue). Some genes in the multiple TSS single green line differing by 12 and 11 bp. Two splice donors are represented by
category directly affect the encoded protein (maroon), and similarly for DS the single purple line differing by 7 bp. These splice variants are combined with
(dark blue). The overlap of configurations is radially proportional (units four proximal internal splices (Supplementary Fig. 3a) to generate the full
indicate percentage of all spliced genes). b, Poly(A)1 testes (blue) and central complement of transcripts. c, Intron retention rates (Y) across the gene body.
nervous system (CNS) (orange) stranded RNA-seq of Gb13F showing complex The genome-wide mean lengths of exons and introns are connected by red
processing and splicing of the 59 UTR. An expansion of the 59 UTR showing parabolic arcs, which illustrate the upper and lower quartiles of intron retention
some of the complexity. Transcription of the gene initiates from one of three (across all samples) for introns retained at or above 20 Y in at least one sample.
different promoters (green arrows) terminates at one of ten possible poly(A)1

Despite the depth of our RNA-seq, these data show that 42% of genes 44% are targets of RNA editing (4% genome-wide6, P , 1 3 10216,
encode only a single transcript isoform, and 55% encode a single protein with 18 of 21 validated21), and 21% have 39 UTR extensions and
isoform (Supplementary Methods). In mammals, it has been estimated RNA editing sites (10 of 65 genome-wide, P , 13 102100). The capa-
that 95% of genes produce multiple transcript isoforms18,19, (estimates city to encode thousands of transcripts is largely specific to the nervous
for protein-coding capacity have not been reported). system and coincides with other classes of rare, neural-specific RNA
The majority of transcriptome complexity is attributable to forty- processing.
seven genes that have the capacity to encode .1,000 transcript isoforms
each (Supplementary Table 3), and account for 50% of all transcripts Tissue- and sex-specific splicing
(Fig. 3a). Furthermore, 27% of transcripts encoded by these genes were To examine the dynamics of splicing, we calculated switch scores or
detected exclusively in samples enriched for neuronal tissue, and another DY, for each splicing event by comparing the maximal and minimal
56% only in the embryo (83% total). To determine their tissue specifi- Y values across all samples, and in subsets including just the devel-
cities we conducted embryonic in situ expression assays (Fig. 3b) and opmental and tissue samples. In contrast to the median Y values, the
found that 18 of 35 are detected only in neural tissue (51% compared distribution of DY values is strikingly different between the develop-
with 10% genome-wide, hypergeometric P value , 1 3 10216, Sup- mental and tissue samples. Among the developmental samples, 38%
plementary Table 4). Of these genes, 48% have 39 UTR extensions in of events have a DY $ 50%, whereas between the tissue samples 63%
embryonic neural tissue20 (5% genome-wide, P , 13 10216). Furthermore, of events have a DY $ 50%. This difference is even more pronounced

a b Syp CAP
1%
0.2% <5
4% 8%
15% 8%
79% 50% 5–19
10%
20–99 rdgA GluClα
24%
100–999

Transcripts per gene Transcript isoforms per class >1000

Figure 3 | Complex splicing patterns are mainly limited to neural tissues. GluCla show specific late embryonic neural expression in the ventral midline
a, A small minority of genes (47, 0.2%) encode most transcripts. b, In situ RNA neurons; dorsal/lateral and ventral sensory complexes; Bolwig’s organ or larval
staining of constitutive exons of four genes with highly complex splicing eye; and central nervous system, respectively.
patterns in the embryo. Syncrip (Syp), CAP, retinal degeneration A (rdgA) and

2 8 AU G U S T 2 0 1 4 | VO L 5 1 2 | N AT U R E | 3 9 5
©2014 Macmillan Publishers Limited. All rights reserved
 RESEARCH ARTICLE

at higher DY thresholds—only 6% of events have a DY $ 80% between Long non-coding RNAs

the developmental samples, whereas 31% of events have a DY $ 80% A growing set of candidate long non-coding RNAs (lncRNAs) have
between the tissue samples. Thus, most splicing events are highly tissue- been identified in Drosophila6,26,27. In FB5.45 there were 392 annotated
specific. Of the 17,447 alternative splicing events analysed (Supplemen- lncRNAs, and it has been suggested that as many as 1,119 lncRNAs
tary Information, section 19), we find that 56.6% changed significantly may be transcribed in the fly28. However, this number was based on
(DY . 20%, Bayes factor .20). Clustering revealed groups of splicing transcribed regions, not transcript models, and used non-stranded RNA-
events that are co-ordinately regulated in a tissue-specific manner. For seq data28. We find 3,880 genes produce transcripts with ORFs encod-
example, 1,147 splicing events are specifically included in heads and ing fewer than 100 amino acids. Of these, 795 encode conserved proteins
excluded in testes or ovaries, whereas 797 splicing events are excluded in (Methods) longer than 20 amino acids. For example, a single exon gene
heads but included in testes or ovaries (Fig. 4a). on the opposite strand and in the last intron of the early developmental
We identified hundreds of sex-specific splicing events from adult growth factor spätzle encodes a 42-amino-acid protein that is highly
male and female RNA-seq data6. To further explore sex-specific splic- conserved across all sequenced Drosophila species. We identified 1,875
ing, we compared the splicing patterns in male and female heads enriched candidate lncRNA genes producing 3,085 transcripts, 2,990 of which have
for brain tissues. There were striking differences in gene expression no overlap with protein-coding genes on the same strand (Supplementary
levels, however, only seven splicing events were consistently differenti- Data 2). Some of these putative lncRNAs may encode short polypep-
ally spliced at each time point after eclosion (average DY . 20%), and tides, for example, the gene tarsal-less encodes three 11-amino-acid
these largely corresponded to genes in the known sex-determination ORFs with important developmental functions29. We determined protein
pathway (Supplementary Information, section 19A). We find few exam- conservation scores for each ORF between 20 and 100 amino acids
ples of head sex-specific splicing. This is in contrast to previous studies, (Supplementary Table 6). Of the 1,119 predicted lncRNAs28, we pro-
which have come to conflicting conclusions and used either microar- vide full-length transcript models for 246 transcribed loci; the remain-
rays analysing only a subset of splicing events or single read 36-bp RNA- der were expressed at levels beneath thresholds used in this study.
seq22,23 with an order of magnitude fewer reads24. This is not surprising, the expression patterns of lncRNAs are more
We identified 575 alternative splicing events that are differentially restricted than those of protein-coding genes: the average lncRNA is
spliced in whole male and female animals (DY . 20%) and analysed expressed (bases per kilobase per million mapped bases6 (BPKM) . 1)
the tissue-specific splicing patterns of each event (Fig. 4b). We found that in 1.5 developmental and 3.2 tissue samples, compared to 6.6 and 17 for
186 of the 321 male-biased splicing events were most strongly included protein-coding genes, respectively. Many lncRNAs (563 or 30%) have
in testes or accessory glands, and 157 of 254 female-biased exons were peak expression in testes, and 125 are detectable only in testes. Similarly
ovary-enriched. Consistent with the extensive transcriptional differ- restricted expression patterns have been reported for lncRNAs in humans
ences observed in testes compared to other tissues, the genes containing and other mammals30,31.
male-specific exons are enriched in functions related to transcription. Interestingly, all newly annotated genes overlapping molecularly
In contrast, the female-specific exon containing genes are enriched in defined mutations with phenotypes are lncRNAs (Supplementary Table 2).
functions involved in signalling and splicing ((http://reactome.org)25, For instance, the mutation D114.3 is a regulatory allele of spineless (ss)
Supplementary Table 6). Together, these results indicate that the major- that maps 4 kb upstream of ss32 and within the promoter of Mgn4221.
ity of sex-specific splicing is due to tissue-specific splicing in tissues Similarly, Mgn00541 corresponds to a described, but unannotated 2.0 kb
present only in males or females. transcript overlapping the regulatory mutant allele ci57 of cubitus

a b Figure 4 | Sex-specific splicing is mainly tissue-

specific splicing. a, Clusters of tissue-specific
splicing events. The scale bar indicates z-scores of
Y. Adult mated male (AdMM), adult mated female
(AdMF) and adult virgin female (AdVF) heads are
from 1-, 4- and 20-day-old animals, respectively.
Testes are from 4-day-old adult males, and ovaries
are from mated and virgin 4-day-old adult females.
b, Sex-specific splicing events in whole animals are
primarily testes- or ovary-specific splicing events.
Adult male (AdM) and adult female (AdF) animals
are 1, 5 and 30 days old. Accessory glands were
dissected from 4-day-old adult males. The RNA-
seq columns from heads, testes and ovaries are as
described in a.

2.0
1.5
1.0
0.5
0.0
–0.5
–1.0
–1.5
–2.0
es

s
ds

s
M

O tes

s
rie

Te d
ad

ad

ad

ad

ad

rie
st

Ad
a

an
Ad
va
he

he

he

s
Te

he

he

va
he

gl
O
M

VF

y
F

VF
M

or
M

M
M
Ad
Ad

Ad

ss
Ad

Ad
Ad

ce
Ac

3 9 6 | N AT U R E | VO L 5 1 2 | 2 8 AU G U S T 2 0 1 4
©2014 Macmillan Publishers Limited. All rights reserved
 ARTICLE RESEARCH

interruptus33. It remains to be determined whether these mutations are and across all samples (P 5 3 3 1025). The tissue distribution of these
a result of the loss-of-function of newly annotated transcripts or cis- cis-NAT-siRNAs showed a bias for testis expression (Supplementary
acting regulatory elements (for example, enhancers) or both. Fig. 7b), with fourfold greater number relative to ovaries (P 5 2 3 10217,
binomial test) and sevenfold relative to heads (P 5 4 3 10225) and
Antisense transcription expression levels of siRNAs were substantially higher in testes than
Drosophila antisense transcription has been reported34, but the cata- other tissues (Supplementary Fig. 7c).
logue of antisense transcription has been largely limited to overlapping Over 80% of cis-NAT-siRNAs were derived from 39-convergent gene
mRNAs transcribed on opposite strands. We identify non-coding anti- pairs. Abundant siRNAs emanate from an overlap of the gryzun and
sense transcript models for 402 lncRNA loci that are antisense to mRNA CG14967 39 UTRs (Supplementary Fig. 5). The remainder were dis-
transcripts of 422 protein-coding genes (for example, prd, Fig. 5a), and tributed amongst CDSs, introns and 59 UTRs. We identified abundant
36 lncRNAs form ‘sense-antisense gene-chains’ overlapping more than testis-enriched siRNA production from a 59-divergent overlap of Cyt-
one protein-coding locus, as observed in mammals30,35. In Drosophila, c-d and CG31808 (Fig. 5b) and from the entire CDS of dUTPase and its
21% of lncRNAs are antisense to mRNAs, whereas in human 15% of antisense non-coding transcript Mgn99994.
annotated lncRNAs are antisense to mRNAs (GENCODE v.10). We
assembled antisense transcript models for 5,057 genes (29%, compared Transcriptional effects of environmental stress
to previous estimates of 15%34). For 67% of these loci, antisense expres- Whole-animal perturbations each exhibited condition-specific effects,
sion is observable in at least one cell line, indicating that sense/antisense for example, the metallothionein genes were induced by heavy metals
transcripts may be present in the same cells. LncRNA-mediated anti- (Fig. 6a), but not by other treatments (Supplementary Table 9). The
sense accounts for a small minority of antisense transcription: 94% of genome-wide transcriptional response to cadmium (Cd) exposure involves
antisense loci correspond to overlapping protein-coding mRNAs tran- small changes in expression level in thousands of genes (48 h after
scribed on opposite strands, and of these, 323 loci (667 genes) share exposure), but only a small group of genes change . 20-fold, and this
overlapping CDSs. The majority of antisense is due to overlapping UTRs: group includes six lncRNAs (the third most strongly induced gene is
1,389 genes have overlapping 59 UTRs (divergent transcription), 3,430 CR44138, Fig. 6a, Supplementary Fig. 8a). Four newly modelled lncRNAs
have overlapping 39 UTRs (convergent transcription), and 540 have
are differentially expressed (1% false discovery rate (FDR)) in at least
both, meaning that, as with many lncRNAs, they form gene-chains across
one treatment, and constitute newly described eco-responsive genes.
contiguously transcribed regions. A subset of antisense gene-pairs overlap
Furthermore, 57 genes and 5,259 transcripts (of 811 genes) were detected
almost completely (.90%), which we term reciprocal transcription.
exclusively in these treatment samples. Although no two perturbations
There are 13 such loci (Supplementary Fig. 5) and seven are male-
revealed identical transcriptional landscapes, we find a homogeneous
specific (none are female-specific).
response to environmental stressors (Fig. 6b, Supplementary Fig. 8b).
The mRNA/lncRNA sense-antisense pairs tend to be more posi-
The direction of regulation for most genes is consistent across all treat-
tively correlated in their expression than mRNA/mRNA pairs, (mean
r 5 0.16 compared with 0.13, Kolmogorov–Smirnov (KS) two-sample ments; very few are upregulated in one condition and downregulated
one-sided test P , 1029), and although this mean effect is subtle, the in another. Classes of strongly upregulated genes included those anno-
trend is clearly visible in the quantiles (95th percentile lncRNA/mRNA tated with the GO term ‘‘Response to Stimulus, GO:0050896’’ (most
0.729 versus mRNA/mRNA 0.634, Supplementary Fig. 6a). This effect enriched, P value , 1 3 10216, Supplementary Fig. 8c), and those that
is stronger when the analysis is restricted to cell line samples (Sup- encode lysozymes (. tenfold), cytochrome P450s, and mitochrondrial
plementary Fig. 6b). components mt:ATPase6, mt:CoI, mt:CoIII (. fivefold). Genes encod-
Even in homogenous cell cultures, evidence for sense-antisense tran- ing egg-shell, yolk and seminal fluid proteins are strongly downregu-
scription does not guarantee that both transcripts exist within individual lated in response to every treatment except ‘cold2’ and ‘heat shock’
cells: transcription could originate from exclusive events occurring in (Supplementary Fig. 8d). For these two stressors, samples were collected
different cells. Cis-natural antisense transcripts (cis-NATs) are a sub- 30 min after exposure, corresponding to an ‘early response test’ showing
stantial source of endogenous siRNAs36, and their existence directly suppression of germ cell production is not immediate.
reflects the existence of precursor dsRNA. Cis-NAT-siRNA production
typically involves convergent transcription units that overlap on their Discussion
39 ends, but other documented loci generate siRNAs across internal Most transcriptional complexity in Drosophila occurs in tissues of the
exons, introns or 59 UTRs37–39. Analysis of head, ovary and testis RNAs nervous system, and particularly in the functionally differentiating
showed that 328 unique sense/antisense gene pair regions generated central and peripheral nervous systems. A subset of ultra-complex genes
21-nucleotide RNAs indicative of siRNA production (Supplementary encodes more than half of detected transcript isoforms and these are
Table 8), and these were significantly enriched (Supplementary Fig. 7a, dramatically enriched for RNA editing events and 39 UTR extensions,
Supplementary Methods) for pairs showing positively correlated expres- both phenomena largely specific to the nervous system. Our study indi-
sion between sense and antisense levels across tissues (P 5 2 3 1025), cates that the total information output of an animal transcriptome may
embryo developmental stages (P 5 4 3 1023), conditions (P 5 9 3 1024) be heavily weighted by the needs of the developing nervous system.

a Chr2L 12,084K 12,085K 12,086K 12,087K b 16,705K 16,710K 16,715K Chr2L

72 78
456
+ Imaginal disc
21nt testes
–
Tissue-specific –456 201 1,800 –45
52,202
4,086
RNA-seq + Testes
– - Testes
–1,800 –17

Gene models –397 –5

Mgn00375
Cyt-c-d
Mgn00376 CG31782
prd
CG31808

Figure 5 | Examples of antisense transcription. a, 59/59 overlapping in this region (data not shown). b, A 59/59 antisense region that produces
bidirectional antisense transcription at the prd locus. Short RNA sequencing substantial small RNA signal on both strands. nt, nucleotide.
does not reveal substantial siRNA (that is, 21-nt-dominant small RNA) signal

2 8 AU G U S T 2 0 1 4 | VO L 5 1 2 | N AT U R E | 3 9 7
©2014 Macmillan Publishers Limited. All rights reserved
 RESEARCH ARTICLE

a b

400 MtnB

MtnD
200
CR44138

Genes (FDR 10%)

100 Lcp65Ag2 Fst Hsp70A
Uhg3
CR43482 GstD2 Mgn1457
Gadd45 LysD vsg GstD5
CG7191 MtnA GstD6 mt:ND1
0
Obp56g Cp38
Vm26Ab Vm34Ca Lectin−galC1 IM23 Acp70A Spn77 7SLRNA Vml
Amy−p Jon99C
Jon25 Jon99F 5.0
–100 Acp54A1 CG18180
Acp53 Jon65 Anp
Cp36
CR40959 Mgn570 2.5

Cp19 0.0
–200 Try
–2.5
CG12374
–300 –5.0
0 20 40 60 80 100 120 140 160

C 05 L
0 M
1 M
4. M

C M
H Co 1
ts 2
k
10 M
M
PQ hoc
0. f P
m

d
ea ld
d m
C .1 m
Zn 5 m

PQ 5 m

m
2L 2R 3L 3R X

ol
.5

5
f2

u
C

d
C
Figure 6 | Effects of environmental perturbations on the Drosophila those identified in larvae50. Some genes are omitted for readability, the complete
transcriptome. Adults were treated with caffeine (Cf), Cd, Cu, Zn, cold, heat or figure and list of omitted genes are given in Supplementary Fig. 8a. b, Heat map
paraquat (PQ). a, A genome-wide map of genes that are up- or downregulated showing the fold change of genes with a FDR , 10% (differential expression) in
as a function of Cd treatment. Labelled genes are those that showed a 20-fold at least one sample (log2 scale). PL, pre-lethal.
(,10% FDR) change in response (linear scale). Genes highlighted in red are

The improved depth of sampling and spatiotemporal resolution they are not translated44. Therefore, while we refer to these RNAs as
resulted in the identification of more than 1,200 new genes not dis- ‘non-coding’, additional data are needed to determine if they produce
covered in our previous study of Drosophila development6. A large small polypeptides.
fraction of the new genes are testes-specific, and many of these are The biological consequences of many of the phenomena reported
antisense RNAs, as previously described in mammals30. Some new here, including the observation that many genes encoding RNA bind-
lncRNAs, such as Mgn94020 (Fig. 1), form sense/antisense gene-chains ing proteins exhibit extraordinary splicing complexity, often within
that bring distant protein-coding genes into transcriptional relation- their 59 UTRs, require further study. The splicing factor pUf68 encodes
ships, another phenomenon previously described only in mammals40. more than 100 alternatively spliced 59 UTR variants, but encodes a single
Whenever Mgn94020 is detectably transcribed, the genes on the oppo- protein. The idea that splicing factors may regulate one another to gen-
site strand in its introns are not, indicating that its transcription may erate complex patterns of splicing is consistent with recent computa-
serve a regulatory function independent of the RNA transcribed. The tional models45. More generally, the role of complex splicing in the
presence of short RNAs at many regions of antisense transcription adult and developing nervous system is unclear. To answer the ques-
indicates that sense and antisense transcripts are present in the same tions that come with increasingly complete transcriptomes in higher
cells at the same times. Many of these Drosophila antisense transcripts organisms, it will be necessary to study gene regulation downstream of
correspond to ‘positionally equivalent’30 antisense transcripts in human. transcription initiation, including the regulation of splicing, local-
In the two species we found antisense lncRNAs opposite to ortholo- ization and translation.
gous protein-coding genes. The apparent positional equivalence of fly
and human antisense transcription at genes like Monocarboxylate METHODS SUMMARY
transporter 1 (MCT1), even-skipped (EVX1), CTCF (CTCF), Adenosine Animal staging, collection and RNA extraction. Tissues were dissected from
receptor (ADORA2A), and many others10,31 across 600 million years of Oregon R larval, pupal and adult staged animals synchronized with appropriate
evolution suggests a conserved regulatory mechanism basal to sexual age indicators. Pupal and adult animals were treated with a number of environmental
reproduction in metazoans. stresses. RNA was isolated using TRIzol (Invitrogen), treated with DNase and
Perturbation experiments identified new genes and transcripts, but purified on a RNAeasy column (Qiagen). Poly(A)1 RNA was prepared from an
aliquot of each total RNA sample using an Oligotex kit (Qiagen).
perhaps more importantly, a general response to stress that is broader
RNA-seq. Libraries were generated and sequenced on an Illumina Genome Analyzer
than the heat shock pathway. A similar study conducted on marsh IIx or HiSeq 2000 using paired-end chemistry and 76-bp or 100-bp cycles. The 454
fishes in the wake of the Deepwater Horizon incident in the Gulf of sequencing used poly(A)1 RNA from Oregon R adult males and females and
Mexico41 demonstrated that the killifish response to chronic hydro- mixed-staged y1 cn1 bw1 sp1. embryos. Sequences are available from the Short
carbon exposure included induction of lyzosome genes, P450 cyto- Read Archive (Accession numbers available in Supplementary Table 10) and the
chromes and mitochondrial components, and the downregulation of modENCODE website (http://www.modencode.org/, Supplementary Table 10).
genes encoding eggshell and yolk proteins41. This overlap of expres- CAGE46 was sequenced on an Illumina Genome Analyzer IIx with 36-bp reads.
sional responses by gene families across phyla suggests a conserved Poly(A)1seq was generated using a custom protocol (Supplementary Methods).
metazoan stress response involving enhanced metabolism and the Analysis. RNA-seq, CAGE and poly(A)1 reads were mapped and filtered12. GRIT
suppression of genes involved in reproduction. was used to identify transcript models14. Expression levels for genes and exons
were computed in BPKM6. GSC P values were computed47. Y values were calcu-
We defined an extensive catalogue of putative lncRNAs. However,
lated with MISO48. Differential expression analysis was conducted with a custom
many genes are known to encode poorly conserved, short polypep- method (Supplementary Methods) and with DEseq49. RPS-BLAST was used to
tides, including genes specific to the male gonad and accessory gland. conduct the conserved domain search with version v3.08 of the NCBI Conserved
Analysis of ribosome profiling initially indicated that a number of Domains Database (CDD) (Supplementary Methods). Orthology analysis between
mammalian lncRNAs may be translated42, but this observation has human and fly was conducted using DIOPT (http://www.flyrnai.org/cgi-bin/DRSC_
been difficult to validate by proteomics43, and further analysis has sug- orthologs.pl). Phenotypic alleles were downloaded from FlyBase r5.50, and were
gested that although lncRNAs have signatures of ribosome occupancy, selected as any allele localized to the genome with a disease phenotype.

3 9 8 | N AT U R E | VO L 5 1 2 | 2 8 AU G U S T 2 0 1 4
©2014 Macmillan Publishers Limited. All rights reserved
 ARTICLE RESEARCH

Received 20 April; accepted 18 December 2013. 37. Okamura, K., Balla, S., Martin, R., Liu, N. & Lai, E. C. Two distinct mechanisms
generate endogenous siRNAs from bidirectional transcription in Drosophila
Published online 16 March 2014. melanogaster. Nature Struct. Mol. Biol. 15, 581–590 (2008).
38. Czech, B. et al. An endogenous small interfering RNA pathway in Drosophila. Nature
1. Mortazavi, A., Williams, B. A., McCue, K., Schaeffer, L. & Wold, B. Mapping and 453, 798–802 (2008).
quantifying mammalian transcriptomes by RNA-Seq. Nature Methods 5, 621–628 39. Ghildiyal, M. et al. Endogenous siRNAs derived from transposons and mRNAs in
(2008). Drosophila somatic cells. Science 320, 1077–1081 (2008).
2. Nagalakshmi, U. et al. The transcriptional landscape of the yeast genome defined 40. Engström, P. G. et al. Complex loci in human and mouse genomes. PLoS Genet. 2,
by RNA sequencing. Science 320, 1344–1349 (2008). e47 (2006).
3. Takahashi, H., Kato, S., Murata, M. & Carninci, P. CAGE (cap analysis of gene
41. Whitehead, A. et al. Genomic and physiological footprint of the Deepwater Horizon
expression): a protocol for the detection of promoter and transcriptional networks.
oil spill on resident marsh fishes. Proc. Natl Acad. Sci. USA 109, 20298–20302
Methods Mol. Biol. 786, 181–200 (2012).
(2012).
4. Mangone, M. et al. The landscape of C. elegans 39UTRs. Science 329, 432–435
42. Ingolia, N. T., Ghaemmaghami, S., Newman, J. R. & Weissman, J. S. Genome-wide
(2010).
analysis in vivo of translation with nucleotide resolution using ribosome profiling.
5. Jan, C. H., Friedman, R. C., Ruby, J. G. & Bartel, D. P. Formation, regulation and
Science 324, 218–223 (2009).
evolution of Caenorhabditis elegans 39UTRs. Nature 469, 97–101 (2011).
43. Bánfai, B. et al. Long noncoding RNAs are rarely translated in two human cell lines.
6. Graveley, B. R. et al. The developmental transcriptome of Drosophila melanogaster.
Genome Res. 22, 1646–1657 (2012).
Nature 471, 473–479 (2011).
7. Trapnell, C. et al. Differential gene and transcript expression analysis of RNA-seq 44. Guttman, M., Russell, P., Ingolia, N. T., Weissman, J. S. & Lander, E. S. Ribosome
experiments with TopHat and Cufflinks. Nature Protocols 7, 562–578 (2012). profiling provides evidence that large noncoding RNAs do not encode proteins.
8. Collins, J. E., White, S., Searle, S. M. & Stemple, D. L. Incorporating RNA-seq data Cell 154, 240–251 (2013).
into the zebrafish Ensembl genebuild. Genome Res. 22, 2067–2078 (2012). 45. Huelga, S. C. et al. Integrative genome-wide analysis reveals cooperative
9. Carninci, P. et al. Targeting a complex transcriptome: the construction of the regulation of alternative splicing by hnRNP proteins. Cell Rep. 1, 167–178
mouse full-length cDNA encyclopedia. Genome Res. 13, 1273–1289 (2003). (2012).
10. Harrow, J. et al. GENCODE: the reference human genome annotation for The 46. Takahashi, H., Lassmann, T., Murata, M. & Carninci, P. 59 end-centered expression
ENCODE Project. Genome Res. 22, 1760–1774 (2012). profiling using cap-analysis gene expression and next-generation sequencing.
11. Djebali, S. et al. Landscape of transcription in human cells. Nature 489, 101–108 Nature Protocols 7, 542–561 (2012).
(2012). 47. Bickel, P. J., Boley, N., Brown, J. B., Huang, H. & Zhang, N. R. Subsampling methods
12. Hoskins, R. A. et al. Genome-wide analysis of promoter architecture in Drosophila for genomic inference. Ann. Appl. Stat. 4, 1660–1697 (2010).
melanogaster. Genome Res. 21, 182–192 (2011). 48. Katz, Y., Wang, E. T., Airoldi, E. M. & Burge, C. B. Analysis and design of RNA
13. Cherbas, L. The transcriptional diversity of 25 Drosophila cell lines. Genome Res. sequencing experiments for identifying isoform regulation. Nature Methods 7,
21, 301–314 (2011). 1009–1015 (2010).
14. Boley, N. et al. Genome guided transcript construction from integrative analysis of 49. Anders, S. & Huber, W. Differential expression analysis for sequence count data.
RNA sequence data. Nature Biotechnol. http://dx.doi.org/10.1038/nbt.2850 Genome Biol. 11, R106 (2010).
(2014). 50. Yepiskoposyan, H. et al. Transcriptome response to heavy metal stress in
15. Celniker, S. E. & Rubin, G. M. The Drosophila melanogaster genome. Annu. Rev. Drosophila reveals a new zinc transporter that confers resistance to zinc. Nucleic
Genomics Hum. Genet. 4, 89–117 (2003). Acids Res. 34, 4866–4877 (2006).
16. Stapleton, M. et al. The Drosophila gene collection: identification of putative full-
Supplementary Information is available in the online version of the paper.
length cDNAs for 70% of D. melanogaster genes. Genome Res. 12, 1294–1300
(2002). Acknowledgements We thank the members of the modENCODE transcription
17. Spradling, A. C. et al. The Berkeley Drosophila Genome Project gene disruption consortium, especially J. Landolin and J. Sandler for their early contributions to these
project: single P-element insertions mutating 25% of vital Drosophila genes. studies. We also thank A. Kundaje and H. Huang for helpful discussions. This work was
Genetics 153, 135–177 (1999). funded by a contract from the National Human Genome Research Institute
18. Wang, E. T. et al. Alternative isoform regulation in human tissue transcriptomes. modENCODE Project, contract U01 HG004271 and U54 HG006944, to S.E.C.
Nature 456, 470–476 (2008). (principal investigator) and P.C., T.R.G., R.A.H. and B.R.G. (co-principal investigators)
19. Pan, Q., Shai, O., Lee, L. J., Frey, B. J. & Blencowe, B. J. Deep surveying of alternative with additional support from R01 GM076655 (S.E.C.) both under Department of
splicing complexity in the human transcriptome by high-throughput sequencing. Energy contract no. DE-AC02-05CH11231. J.B.B.’s work was supported by NHGRI K99
Nature Genet. 40, 1413–1415 (2008). HG006698. Work in P.J.B.’s group was supported by the modENCODE DAC sub-award
20. Smibert, P. et al. Global patterns of tissue-specific alternative polyadenylation in 5710003102, 1U01HG007031-01 and the ENCODE DAC 5U01HG004695-04. Work
Drosophila. Cell Rep. 1, 277–289 (2012). in Bloomington was supported in part by the Indiana METACyt Initiative of Indiana
21. St Laurent, G. et al. Genome-wide analysis of A-to-I RNA editing by single-molecule University, funded by an award from the Lilly Endowment. Work in E.C.L.’s group was
sequencing in Drosophila. Nature Struct. Mol. Biol. 20, 1333–1339 (2013). supported by U01-HG004261 and RC2-HG005639.
22. Telonis-Scott, M., Kopp, A., Wayne, M. L., Nuzhdin, S. V. & McIntyre, L. M. Sex-
specific splicing in Drosophila: widespread occurrence, tissue specificity and Author Contributions J.A., T.R.G., B.R.G., R.A.H., T.C.K. and S.E.C. designed the project.
evolutionary conservation. Genetics 181, 421–434 (2009). J.A., P.Ch., T.R.G., B.R.G., R.A.H., J.B.B., B.O. and S.E.C. managed the project. R.E.
23. Hartmann, B. et al. Distinct regulatory programs establish widespread sex-specific designed treatment protocols and prepared biological samples. T.C.K., J.A. and L.C.
alternative splicing in Drosophila melanogaster. RNA 17, 453–468 (2011). oversaw biological sample production. B.D.E., D.M. and J.R. prepared biological
24. Chang, P. L., Dunham, J. P., Nuzhdin, S. V. & Arbeitman, M. N. Somatic sex-specific samples. D.Z. and B.E. prepared RNA samples. J.A. oversaw RNA sample production.
transcriptome differences in Drosophila revealed by whole transcriptome G.E.M., S.O. and L.Y. prepared Illumina RNA-seq libraries. A.M.S. prepared CAGE
sequencing. BMC Genomics 12, 364 (2011). libraries. P.Ca. oversaw production of CAGE libraries. C.A.D., G.E.M., S.O., L.Y., S.P. and
25. Matthews, L. et al. Reactome knowledgebase of human biological pathways and K.H.W. performed Illumina sequencing. B.R.G. and S.E.C. managed Illumina
processes. Nucleic Acids Res. 37, D619–D622 (2009). sequencing production. R.A.H. conceived the poly(A)1seq method. R.W. and R.A.H.
26. Lipshitz, H. D., Peattie, D. A. & Hogness, D. S. Novel transcripts from the developed the poly(A)1seq protocol and produced the libraries. K.M. performed 454
Ultrabithorax domain of the bithorax complex. Genes Dev. 1, 307–322 (1987). sequencing. C.Y., S.P. and K.H.W. performed cDNA library screens and full-insert cDNA
27. Tupy, J. L. et al. Identification of putative noncoding polyadenylated transcripts in sequencing. S.E.C. oversaw cDNA production. E.F. and N.B. installed and administered
Drosophila melanogaster. Proc. Natl Acad. Sci. USA 102, 5495–5500 (2005). computer infrastructure for data storage and analysis. J.B.B., N.B., M.H.S., M.O.D.,
28. Young, R. S. et al. Identification and properties of 1,119 candidate lincRNA loci in B.W.B., D.S., J.W.C., S.S., J.W., A.A.S., N.P., E.C.L., P.J.B. and B.R.G. developed analysis
the Drosophila melanogaster genome. Genome Biol. Evol. 4, 427–442 (2012). methods. J.B.B., N.B., M.H.S., M.O.D., B.W.B., A.S.H., E.F., R.A.H., S.S., D.S., L.C., G.R., J.H.,
29. Kondo, T. et al. Small peptide regulators of actin-based cell morphogenesis J.W., A.A.S., E.C.L., K.H.W., B.R.G. and S.E.C. analysed data. N.B., J.B.B. M.H.S., K.H.W. and
encoded by a polycistronic mRNA. Nature Cell Biol. 9, 660–665 (2007). S.E.C. generated annotations. D.S. and B.O. analysed species validation data. S.S.,
30. Katayama, S. et al. Antisense transcription in the mammalian transcriptome. J.W. and E.C.L. analysed 39 UTR and antisense data. A.S.H., E.F. and S.E.C. analysed
Science 309, 1564–1566 (2005). image data. M.H.S. analysed proteomics data. M.H.S., S.S., D.S., B.O., E.C.L., T.C.K., R.E.,
31. Derrien, T. et al. The GENCODE v7 catalog of human long noncoding RNAs: R.A.H. and P.Ch. contributed to the text. A.S.H. assisted with manuscript preparation.
analysis of their gene structure, evolution, and expression. Genome Res. 22, J.B.B., B.R.G. and S.E.C. wrote the paper with input from all authors. All authors
1775–1789 (2012). discussed the results and commented on the manuscript.
32. Duncan, D. M., Burgess, E. A. & Duncan, I. Control of distal antennal identity and
tarsal development in Drosophila by spineless-aristapedia, a homolog of the Author Information Sequences are available from the Short Read Archive and the
mammalian dioxin receptor. Genes Dev. 12, 1290–1303 (1998). modENCODE website, a list of accession numbers is given in Supplementary Table 10.
33. Schwartz, C., Locke, J., Nishida, C. & Kornberg, T. B. Analysis of cubitus interruptus Reprints and permissions information is available at www.nature.com/reprints.
regulation in Drosophila embryos and imaginal disks. Development 121, The authors declare no competing financial interests. Readers are welcome to
1625–1635 (1995). comment on the online version of the paper. Correspondence and requests for
34. Misra, S. et al. Annotation of the Drosophila melanogaster euchromatic genome: a materials should be addressed to J.B.B. ([email protected]),
systematic review. Genome Biology 3, research0083 (2002). B.R.G. ([email protected]) or S.E.C. ([email protected]).
35. Lipovich, L. et al. Activity-dependent human brain coding/noncoding gene
regulatory networks. Genetics 192, 1133–1148 (2012). This work is licensed under a Creative Commons Attribution-
36. Okamura, K. & Lai, E. C. Endogenous small interfering RNAs in animals. Nature Rev. NonCommercial-Share Alike 3.0 Unported licence. To view a copy of this
Mol. Cell Biol. 9, 673–678 (2008). licence, visit http://creativecommons.org/licenses/by-nc-sa/3.0

2 8 AU G U S T 2 0 1 4 | VO L 5 1 2 | N AT U R E | 3 9 9
©2014 Macmillan Publishers Limited. All rights reserved