The rat has been used extensively as a model for evaluating chemical toxicities and for understanding drug mechanisms. However, its transcriptome across multiple organs, or developmental stages, has not yet been reported. Here we show, as part of the SEQC consortium efforts, a comprehensive rat transcriptomic BodyMap created by performing RNASeq on 320 samples from 11 organs of both sexes of juvenile, adolescent, adult and aged Fischer 344 rats. We catalogue the expression pro?les of 40,064 genes, 65,167 transcripts, 31,909 alternatively spliced transcript variants and 2,367 non-coding genes/non-coding RNAs (ncRNAs) annotated in AceView. We ?nd that organ-enriched, differentially expressed genes re?ect the known organ-speci?c biological activities. A large number of transcripts show organ-speci?c, age-dependent or sex-speci?c differential expression patterns. We create a web-based, open-access rat BodyMap database of expression pro?les with crosslinks to other widely used databases, anticipating that it will serve as a primary resource for biomedical research using the rat model.
The rat is used extensively by the pharmaceutical, regulatory and academic communities to test drug and chemical toxicities, to evaluate the mechanisms underlying drug effects and to model human diseases. Although several community-wide efforts are preparing a catalogue of genes expressed during normal development of mice1,2 and humans3,4, such efforts are less advanced for the rat. Furthermore, the rat genome is still incomplete, containing many gaps and missing genes, and the rat transcriptome is not well annotated. Next-generation sequencing technologies have revolutionized genomic research and allow the genome and transcriptome of any organism to be explored without a priori assumptions and with unprecedented throughput5–11. RNA-Seq is able to provide single-nucleotide resolution, strand speci?city and short-range connectivity through paired-end sequencing5,8,9,12–14. Using RNA-Seq to catalogue the variations in the transcriptome between sexes and over the lifespan of the rat, from birth to old age, can provide insights into disease susceptibility, drug ef?cacy and safety, and toxicity mechanisms, and could ultimately improve the translation of preclinical ?ndings to humans. Several transcriptomic BodyMap studies have been reported in Drosophila melanogaster12,15, mouse and human16–18, and these studies show large age-dependent variations in gene expression in various organs19. In rat, the liver has been examined in detail because of its central role in the metabolism of drugs and xenobiotics20–23. Kwekel et al.23 found that nearly 3,800 genes in the Fisher 344 rat liver were differentially expressed when evaluated by either age or sex over the life cycle. Such large differences in the transcriptome at various life stages may contribute to age- and/or sex-speci?c susceptibilities to disease or to adverse reactions to drugs or environmental pollutants. Accounting for these differences may help in developing mechanism-based drug safety assessment and prediction24,25, as well as in re?ning environmental risk assessments. Through the US Food and Drug Administration’s sequencing quality control (SEQC) consortium, we use RNA-Seq to comprehensively catalogue transcriptomic pro?les across 11 organs and 4 developmental stages (juvenile, adolescence, adult and aged) in both sexes of Fischer 344 rats. To assess inter-animal biological variations, four individual rats are tested per condition. We validate many transcripts that were previously only annotated in AceView26 based on cDNAs in GenBank and dbEST, including 31,909 alternatively spliced (AS) transcripts and 2,367 spliced non-coding genes/non-coding RNAs (ncRNAs) that were not annotated in RefSeq. This represents the ?rst usage of large amounts of next-generation deep sequence data in rat crossvalidated against AceView annotation. We then construct a webbased, open-access rat BodyMap database
ratbodymap/index.html) to catalogue the expression pro?les for 40,064 AceView-annotated genes and 65,167 transcripts measured in 320 RNA-Seq libraries, with crosslinks to other widely used databases, including AceView, GenBank, Entrez, Ensembl, RGD, UniProt, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes. Our study, accompanied by the online database searching capabilities, can serve as a useful resource for both academic biologists and pharmaceutical companies that utilize rats for assessing chemical safety pro?les and for studying human diseases.
Results Study design. To study the rat transcriptome at single-base resolution, we constructed and sequenced 320 RNA-Seq libraries from 320 RNA samples derived from 16 female and 16 male rats from the Fischer 344 strain. Ten organs were evaluated per rat (adrenal gland, brain, heart, kidney, liver, lung, muscle, spleen,
thymus and testis or uterus) at four developmental stages—that is, juvenile (2-weeks old), adolescence (6-weeks old), adult (21-weeks old) and aged (104-weeks old); eight rats (four female and four male rats) were evaluated per developmental stage (Supplementary Fig. 1). To monitor the quality of the RNA-Seq, we added External RNA Control Consortium (ERCC) spike-in controls in an amount equivalent to about 1% of the mRNA in each sample before library construction27. RNA-Seq libraries were constructed starting with total RNA using Ribo-Zero kit (Epicentre) for rRNA depletion, combined with Illumina’s TruSeq RNA kit (skipping the Poly(A)t selection step) for each single biological sample, which allowed us to detect both polyadenylated and non-polyadenylated transcripts, including ncRNAs. We generated B13.2 billion reads of 50-bp single-end RNA-Seq data for this study, corresponding to an average of 40 million sequence reads per sample.
Overview of the landscape of the rat transcriptome. We mapped the reads to the rat AceView transcriptome26, UCSC rn4 genome and ERCC transcripts. On average, 88.5% of the reads were mapped to genomic regions, 41.7% to AceView exons, 8.2% to rRNA and 0.92% to the ERCC transcripts (Supplementary Fig. 2). The pair-wise Pearson correlation coef?cient (R) between any two of the four biological replicates within each sample group was calculated based on the 40,064 genes, yielding six pair-wise R values per sample group. The mean R value and the s.e. were calculated per group (n?6), yielding 80 mean R values and 80 s.e. values with a grand mean of 0.9679 and 0.0014 (n?80), respectively, indicating a high level of measurement consistency among biological replicates (Supplementary Fig. 3). Scatterplots of ERCC log2(FPKM) versus log2(spike-in concentration) showed an overall linear relationship between RNA-Seq-detected signal and the true concentration of the ERCC spike-in controls, in particular for controls with higher concentrations (Supplementary Fig. 4a,b and Supplementary Data 1). In addition, the average detected expression values of the 92 ERCC controls were similar (log2FPKM B7.2) in 318 of the 320 samples (Supplementary Fig. 4c). In general, the expression values of ERCC spike-in controls measured in this study, where an rRNA-depletion protocol (Ribo-Zero) was used for mRNA enrichment, were much closer to the expected values than what was observed using a poly(A)-selection protocol for mRNA enrichment. It was the poly(A)-selection process that introduced the ERCC transcript-speci?c biases in mRNA enrichment. A combination of quality-control assessment of the sequence data (Supplementary Figs 3 and 4) demonstrated a high level of reproducibility of biological replicates, and the expected behaviour of external spike-in controls ensured that our data are of high quality for follow-up analyses. Consequently, a ?nal data matrix consisting of 40,064 AceView-annotated genes and 65,167 transcripts across all 320 biological samples was generated and used for further analyses as described in the following sections. The mapping pipelines are outlined in Supplementary Fig. 5.
On average, 25,523 (63.7%) of the 40,064 AceView-annotated genes were de?ned as expressed (FPKM Z1) per organ. Differences in the numbers of genes and transcripts expressed were observed among organs, in particular for those only annotated in AceView (Fig. 1a,b). For example, 22,995 genes were expressed in the liver, whereas 27,521 were expressed in the lung. Liver and muscle had the lowest numbers of expressed genes in comparison to the other nine organs (Fig. 1a). Large numbers of genes (15,894 or 39.7%, Fig. 1a) and transcripts (27,795 or 42.7%, Fig. 1b) were expressed in all the 11 organs at all developmental stages and in both sexes, including ‘novel’ genes.
Figure 1 | Landscape of the rat RNA-Seq transcriptome. Number of expressed genes (a) and transcripts (b) detected per organ across four developmental stages in both males and females (4 biological replicates each). For each panel (a,b), the x axis indicates organs and developmental stages in either sex, whereas the y axes (left and right) indicate the numbers of genes or transcripts expressed (1,000; left) or the percentages of all annotated genes or transcripts (right) in each organ across four developmental stages in either sex. Red bars represent the number of expressed genes or transcripts (mean±s.e., N?8; for Tst and Utr N?4) annotated in both RefSeq and AceView, while blue bars represent the additional expressed genes or transcripts (mean±s.e., N?8; for Tst and Utr N?4) annotated only in AceView. The green lines (unions) represent the number of genes or transcripts expressed per organ and per age in at least one biological replica, male or female (N?8). A gene or transcript was considered expressed if its average expression level in FPKM is Z1. (c) Hierarchical clustering analysis of gene expression pro?les from 320 rat samples with 40,064 genes. (d) Principal variance component analysis (PVCA) of the relative contribution of main effects (organ, age, sex and replicate) and their combinations (asterisk) to total model variance. RefSeq genes are characterized by a gene ID; for RefSeq transcripts, only those well annotated (that is, with ‘NM_’ accessions) were counted. Organs tested are: Adr, adrenal; Brn, brain; Hrt, heart; Kdn, kidney; Lng, lung; Lvr, liver; Msc, skeletal muscle; Spl, spleen; Thm, thymus; Tst, testis; and Utr, uterus; Inter: intersection of genes (a) and transcripts (b) commonly expressed across all 11 organs sampled in this study.
or transcripts that were annotated only in AceView, but not in RefSeq. The majority of these 15,894 commonly expressed genes (FPKM Z1) appear to be primarily involved in basic biological functions—for example, oxidative phosphorylation, GTP–XTP metabolism, cytoskeletal remodelling and the cell cycle (Supplementary Table 1); these genes are referred to as ‘commonly expressed genes’.
To obtain an overview of gene expression pro?les of the 320 rat samples, we performed a hierarchical cluster analysis (Fig. 1c). This analysis showed a clear separation of the organs by gene expression except for testes and thymus, which are further separated by age group. Testis 2- and 104-week-old differ from testis 6- and 21-week-old, re?ecting adolescence and sexual maturity. In contrast, the uterus, even though it mainly contains smooth muscle tissue, was clustered far from both heart and skeletal muscle. The distinct cluster seen in the thymus at 104 weeks re?ects the known thymus atrophy in aging animals.
Analysis of the sources of variance in our data set by principal variance component analysis showed that organ accounted for 70.47% of the total variance (Fig. 1d). All other effects and interactions were less than the residual variance of the model (17.66%). We observed that the sex difference was subtle and only accounted for 0.22% of overall variance in expression pro?les. It should be noted that the Y chromosome of the rat has not been sequenced and annotated, explaining the relatively small between-sex differences observed in our data. We, therefore, combined the data from female and male rats for most of our analyses, including those of differentially expressed organenriched genes across the four developmental stages.
Organ-dependent differentially expressed genes. We used a t-test (P-value r0.05, fold change (FC) Z2 orr0.5) to identify genes that were differentially expressed between any two organs. The number of differentially expressed genes (DEGs) was signi?cantly different depending on the pair of organs compared. The overall DEGs between any organ and the other 10 organs over the 4 developmental stages are shown in Fig. 2a. DEGs in the liver and muscle were generally underexpressed compared with the other organs, while DEGs in the brain, testes and lung were generally overexpressed compared with other organs.
We looked to identify organ-enriched genes that were highly expressed and relatively speci?c to each organ. To identify organenriched genes during development (Supplementary Fig. 6), we used a t-test with a Bonferroni-corrected P-value r0.05 to generate a list of organ-enriched genes at increasing FC cutoff values (that is, 2, 4, 8, 16, 32, 64, 128 and 256). When a FC of 2 was used, we identi?ed 3,413 organ-enriched AceView genes (Supplementary Table 2), 2,052 (60.1%) of which were annotated in RefSeq and the remaining 1,361 (39.9%) annotated only in AceView. Of these organ-enriched genes, 1,401 (41.0%) were detected speci?cally in the brain, 454 (13.3%) speci?cally in the liver and 386 (11.3%) speci?cally in the kidney (Supplementary Table 2). The numbers of organ-enriched genes identi?ed in the other eight organs ranged from 25 to 306.
Organ-enriched genes re?ect organ biological functions. We conducted GO enrichment analysis of organ-enriched genes in each organ type that were annotated in RefSeq (n?2,052; Fig. 2b). Supplementary Data 2 contain a list of all pathways that were signi?cantly enriched (Pr0.05, hypergeometric test, Benjamini–Hochberg FDR-adjusted P-value) and unique to each organ. In general, the GO enrichment and the selection and ranking of the pathways based on organ-enriched genes were highly consistent with the biological functional activities of the organ for which the genes were enriched. For example, brain-enriched genes were associated with neurophysiological processes including dopamine and GABA signalling, whereas
heart-enriched genes were associated with muscle contraction, signal transduction and regulation of cardiac hypertrophy (Supplementary Table 3). Examples of pathways de?ned by the organ-enriched genes are shown for the brain (role of CDK5 in presynaptic signalling, Supplementary Fig. 7), liver (bile acid biosynthesis, Supplementary Fig. 8) and kidney (renal secretion of organic electrolytes, Supplementary Fig. 9).
Development-dependent genes. To evaluate developmentdependent differential gene expression in various organs, we used an analysis of variance (ANOVA) model and applied FC Z2 (or r0.5) plus a Bonferroni-adjusted P-value r0.05. Overall, we identi?ed 18,640 genes differentially expressed during development in at least one of the 11 organs, of which 10,572 were annotated in both RefSeq and AceView; the remaining 8,068 were only annotated in AceView. The number of developmentdependent genes varied by organ, from 2,211 in the brain to 16,186 in the testis (Supplementary Table 4). As expected, the greatest differential gene expression was observed when juvenile 2-week-old rats were compared with older rats. Moreover, a large number of genes were differentially expressed in the testis across ages, as seen in a comparison with sexually mature 6- and 21-week-old rats with 2- and 104-week-old rats (Supplementary Table 4), which have young and atrophying testes, respectively. Major known functions of the 10,572 development-dependent DEGs annotated in RefSeq included protein folding and maturation, cell cycle, cell adhesion, immune response, glutathione metabolism and transcription (Fig. 2c and Supplementary Data 3).
Development-dependent gene expression patterns. To evaluate the time course and development-dependent transcriptomic activities across the life cycle of the rat, we performed a time course differential gene expression analysis by comparing any two adjacent developmental stages, using the younger developmental stage group as the denominator (see Online Methods) for each of the 11 organs. There were 27 possible patterns (three change points during development, or 33 possibilities), including those that increased across all developmental stage boundaries, termed ‘up-up-up’ (UUU); those that were similar across all boundaries, termed ‘maintain-maintain-maintain’ (MMM); and those that decreased across all boundaries, termed ‘decrease-decreasedecrease’ (DDD). Genes were non-randomly represented across all patterns. The overall development-dependent patterns across all organs are shown in Fig. 3a. Relatively few genes continuously increased (UUU) or decreased (DDD) in expression during aging,
the vast majority of genes remained unchanged over the lifespan (over 74% for any organ except for the testis) (Fig. 3b). However, the onset of adolescence and adulthood triggers signi?cant changes in genes (UMM or DMM) as well as the onset of old age (MMD or MMU) (Fig. 3).
Sex-speci?c DEGs. Even though they represented a small proportion of overall variation in gene expression, differential gene expression pro?les between female and male rats for all nine nonsex organs were examined at all four developmental stages (Figs 4a,b, and Supplementary Fig. 10). A number of genes were signi?cantly different between male and female rats, particularly in the liver, muscle and kidney, and to a lesser extent in the spleen and brain. Most DEGs were found at 21 weeks, in adults (Supplementary Table 5). More notable were, at 6 weeks, the 2,230 female-dominant genes (sexually dimorphic expression with higher expression in female) compared with 1,668 maledominant genes (sexually dimorphic expression with higher expression in male). Female-dominant genes were outnumbered by male-dominant genes at all other ages (female versus male: 1,921 versus 3,409 for week 2; 2,769 versus 2,945 for week 21; and 2,116 versus 2,571 for week 104). More genes showed sex-speci?c expression in the liver and kidney in week 21 with large FCs (Fig. 4a,b, and Supplementary Table 5). Genes involved in metabolism, particularly cytochrome P450s, are also known to be differentially expressed between the sexes. We found P450 differences to be variable; however, expression levels of Cyp1a1, Cyp1a2, Cyp2c7, Cyp3a9 and Cyp26a1 were higher in the female liver, predominantly at sexual maturity, whereas Cyp2a2, Cyp2c, Cyp3a2 and Cyp3a18 were expressed higher in the male liver (data not shown). Major known functions of the 6,677 sexspeci?c DEGs annotated in RefSeq included cell cycle, blood coagulation and CREM signalling in the testis and GABA-B receptor signalling in presynaptic nerve terminals (Fig. 2d and Supplementary Data 4).
Using a gene with many different alternatively spliced variants as an illustration, we also explored the organ-dependent and sex-speci?c differential isoform expression of Ugt1a1 (UDP glucuronosyltransferase 1 family, polypeptide A1), an enzyme playing an essential role in the detoxi?cation of xenobiotics and endogenous compounds by conjugating bilirubin with glucuronic acid28–30. Twelve Ugt1a1 isoforms were annotated for rat in AceView, two of which (Ugt1a1 g and h) showed organdependent differential expression between female and male rats. Ugt1a1 h was expressed signi?cantly higher in female liver, while Ugt1a1 g was more highly expressed in the male adrenal gland
Figure 2 | Organ- and development-dependent and sex-speci?c genes. (a) Comparison of relative gene expression between organs. Circos plot illustrating the relative number of DEGs between any two organs (orange, overexpressed; green, underexpressed). Concentric circles from inside to outside represent; A, organ under comparison (colour-coded as described in the outer-most ring); B, the 11 organs being compared with organ A (note no change in DEGs for organ compared with itself); C–F, total number of DEGs, more (orange) or fewer (green) in organ Aversus the other organs at weeks 2 (C), 6 (D), 21 (E) or 104 (F). Each bar represents the combination of either four or eight biological replicates for a given organ at the same developmental stage. A gene was considered differentially expressed between two organs if the fold change was Z2 orr0.5 (t-test, P-value r0.05). (b) Expression pro?les of organ-enriched genes with corresponding signi?cantly and uniquely enriched GeneGo canonical pathway maps. Expression data for 3,413 organ-enriched genes across 320 samples were arranged by organ type (in decreasing order in terms of the number of organ-enriched genes), sex and developmental stage. (c) Development-dependent clusters with signi?cantly enriched GeneGo canonical pathway maps. Development-dependent genes in each organ were identi?ed using a combination of ANOVA with Bonferroni-corrected Pr0.05 plus a FCZ2. Hierarchical clustering analysis grouped the developmentdependent genes into 10 clusters. (d) Sex-speci?c clusters with signi?cantly enriched GeneGo canonical pathway maps. The 288 samples (except uterus and testis samples) were separated into 36 groups based on four developmental stages and nine organ types. For any organ at any development stage, genes with a FCZ2 (or r0.5) and Pr0.05 between female and male were considered sex-speci?c. Hierarchical clustering analysis grouped the sex-speci?c genes into six clusters. Expression data were Z-score standardized (mean zero and s.d. of one) per gene. Ontology terms of enriched GeneGo canonical pathway maps were listed next to each organ type (b) or cluster (c,d) on the right along with –log10(FDR q-value) in the parentheses. Organs tested are: Adr, adrenal; Brn, brain; Hrt, heart; Kdn, kidney; Lng, lung; Lvr, liver; Msc, skeletal muscle; Spl, spleen; Thm, thymus; Tst, testis; and Utr, uterus.
Figure 3 | Development-dependent patterns of rat gene expression. (a) Differentially expressed genes were determined based on a combination of ANOVA with the Bonferroni-corrected P-value r0.05 and FCZ2 (or r0.5) between the four developmental stages. For each organ, data from two sequential developmental stages were compared, with the younger developmental stage used as the denominator. Genes were grouped into Up (U; ‘upregulated’ based on FCZ2), Down (D; ‘downregulated’ based on FCr0.5), or Maintain (M; ‘no change’ based on 0.5oFCo2). Shown here are the 27 possible combinatorial patterns. The x axis depicts time point (in weeks) and the y axis depicts fold change. The number shown in each box (for example, 129 genes for pattern DDD) was derived based on the number of genes, across all 11 organs—where each gene was counted only once regardless of how many organs shared that same pattern. (b) The percentage of genes within each pattern per organ exhibiting speci?c development-dependent expression patterns. The numbers in the table are colour-coded; red indicates a relatively large percentage of genes with that expression pattern and blue represents a relatively small percentage of genes with that pattern. Organs tested are: Adr, adrenal; Brn, brain; Hrt, heart; Kdn, kidney; Lng, lung; Lvr, liver; Msc, skeletal muscle; Spl, spleen; Thm, thymus; Tst, testis; and Utr, uterus.
and lung (Supplementary Fig. 11). The Ugt1a1 gene itself, as well as its other 10 isoforms (data not shown), did not show any sexspeci?c differential expression.
Alternative splicing and organ-speci?c isoform expression. On the basis of the cDNA sequences deposited in NCBI GenBank and dbEST databases, 2,430 novel spliced non-coding genes have been annotated in AceView. Among them, 2,367 non-coding genes were cross-validated with the data set from the current study (Supplementary Data 5). We also cross-validated 31,909 alternatively spliced transcripts (Supplementary Data 6) only annotated in AceView. Both of these tables are linked to AceView. We further measured and mapped the expression level of these alternatively spliced transcripts and non-coding genes/ ncRNAs across the 11 organs in our rat BodyMap database. Of the 2,367 spliced non-coding genes, 326 were expressed in all organs across the four developmental stages (Supplementary Fig. 12a), whereas 139 displayed organ-enriched expression, with 44 speci?cally expressed in the brain (Supplementary Fig. 12b). We found that soyshee, one of the spliced non-coding genes/ ncRNAs in AceView, was most highly expressed in the liver with somewhat lesser expression in the testis (6 and 21 weeks, Supplementary Fig. 13).
We also examined the expression of all alternatively spliced transcripts (including those annotated in RefSeq). The brain contained the vast majority of organ-enriched transcript variants (1,902) followed by the liver (774), kidney (598) and muscle (452) (Fig. 5a). The number of organ-enriched transcript variants per gene varied from 1 to 10, with 2,956 genes having one variant, 23 genes having ?ve variants and one gene having 10 variants de?ned as organ-enriched (Fig. 5b). Most of the organ-enriched transcript variants showed the same expression pattern as the gene itself. However, some organ-enriched transcript variants showed a different, organ-dependent expression pattern, such as Dlg2 (disks large homologue 2, Fig. 5c). In addition to Dlg2 variant a, which is annotated in RefSeq, ?ve additional Dlg2 variants (named Dlg2.b, c, d, e and f) were annotated in AceView. Dlg2.b was highly enriched in the adrenal gland, whereas Dlg2.e. was enriched in the brain. Another gene that showed organ-speci?c differential variant expression was Pecr (peroxisomal trans-2-enoyl-CoA reductase), coding for an enzyme involved in fatty acid elongation31,32. Four transcript variants (named a, b, c and d) were annotated in AceView. Our data demonstrated that, while the Pecr gene, as well as its variants a and d, showed a similar expression pro?le and were highly enriched in the liver, Pecr.c was expressed almost exclusively in the kidney (Supplementary Fig. 14).
Transcriptional expression pro?les can also serve as an important resource for developing a functional understanding of regulation of splicing events and selection of alternative promoters and polyadenylation sites33–35. For example, troponin Tnni1.c and Tnni1.d variants were both annotated in AceView as encoding the same isoform of troponin 1, skeletal, slow 1, but differ by AS affecting the 30 untranslated region (UTR) and alternative polyadenylation (APA) site selection. Illustration of
Figure 4 | Sex differences of rat transcriptomic pro?les. Gene expression values, expressed as log2FPKM, for male (y axis) and female (x axis) rats are depicted in the scatter plots. These plots show gene expression pro?les between female and male rats in nine organs (a) and four developmental stages (b) using pooled data from either all four stages (a) or all organs for a given stage (b). Non-DEGs form the centre line (in grey), while DEGs are coloured and occur above and below the centre line. 002?2 weeks; 006?6 weeks; 021?21 weeks; 104?104 weeks. Organs tested are: Adr, adrenal; Brn, brain; Hrt, heart; Kdn, kidney; Lng, lung; Lvr, liver; Msc, skeletal muscle; Spl, spleen; and Thm, thymus.
the AS/APA events and expression patterns in three organs of 6-week-old female rats are shown (Fig. 6a). As expected for troponin protein-coding transcripts, neither Tnni1.c nor Tnni1.d were expressed in the brain, but both were highly expressed in the muscle, where expression of Tnni1.d was 94% higher than that of Tnni1.c. Interestingly, only Tnni1.d, which is an AceView-only transcript, was detectable in the thymus (proportion value ?0.994, see Online Methods). The underlying biological mechanism of the organ-dependent expression of Tnni1.c and Tnni1.d warrants further investigation.
For genes without any clear function description in AceView, their co-expression patterns across the 320 RNA samples with genes of known functions under the same GO term has the potential to provide an indication of their functions based on the ‘guilty by association’ principle (see Online Methods). Two examples of functional inference are shown in Fig. 6b,c. The gene muwey was annotated in AceView with one potential non-coding transcript. Our RNA-Seq data showed that the trend of expression pro?le of muwey across the 320 rat RNA samples was highly similar to that of the gene Nat1/Nat2, which is a member of the GO:0007507 (heart development) group; thus, the function of muwey may also be associated with heart development. However, we note that the absolute expression level of muwey was much higher than that of Nat1/Nat2. Similarly, the AceView-only gene ga?o may be related to the glutathione biosynthetic process because its expression pro?le was similar to that of Avpr1a, a member of the GO:0006750 group (glutathione biosynthetic process).
We investigated the transcriptome of the Fischer 344 rat by constructing a rat RNA-Seq transcriptomic BodyMap including 11 organs, from both sexes, at 4 developmental stages from juvenile to old age. Although many genes showed organ-speci?c differential expression across the lifespan, thousands of genes were commonly expressed across all organs and 4 developmental stages. Interestingly, genes that were commonly expressed in all organs were more likely to be annotated in RefSeq than in AceView, while genes that were enriched in organs were increasingly represented in AceView. RefSeq contains wellstudied genes expressed at high levels, mostly the conserved coding genes. AceView26 is a more comprehensive annotation based on cDNAs and is more likely to contain novel or uncommon genes. We found that organ-enriched genes are well correlated with the biological functions of each organ. For example, brain-enriched genes were active in pathways related to a variety of neurophysiological processes, including dopamine signalling, CDK5 signalling and GABA signalling. The pathway enrichment pattern for liver-enriched genes was very different from that for the brain and included various metabolic processes such as fatty acid oxidation and bile acid biosynthesis, whereas thymus-enriched genes were associated with various immunerelated processes and signalling pathways. In contrast, genes de?ned as commonly expressed across all organs tended to be enriched in non-organ-speci?c pathways such as oxidative phosphorylation, GTP–XTP metabolism and cytoskeleton remodelling. Sex-speci?c and organ-dependent differential gene