Annotation and cluster analysis of spatiotemporal- and sex-related lncRNA expression in rhesus macaque brain

Long noncoding RNAs (lncRNAs) mediate important epigenetic regulation in a wide range of biological processes and diseases. We applied comprehensive analyses of RNA-seq and CAGE-seq (cap analysis of gene expression and sequencing) to characterize the dynamic changes in lncRNA expression in rhesus macaque (Macaca mulatta) brain in four representative age groups. We identified 18 anatomically diverse lncRNA modules and 14 mRNA modules representing spatial, age, and sex specificities. Spatiotemporal- and sex-biased changes in lncRNA expression were generally higher than those observed in mRNA expression. A negative correlation between lncRNA and mRNA expression in cerebral cortex was observed and functionally validated. Our findings offer a fresh insight into spatial-, age-, and sex-biased changes in lncRNA expression in macaque brain and suggest that the changes represent a previously unappreciated regulatory system that potentially contributes to brain development and aging.


Liu et al., LncRNAs link to postnatal brain development and ageing
Transcriptional dynamics has been suggested to be a major contributor to brain architecture, functional evolution as well as the development process and ageing (Belgard et al. 2011;Aprea et al. 2013;Telley et al. 2016).Long noncoding RNAs (lncRNAs) are a subgroup of RNA longer than 200 nucleotides yet have limited protein coding potential.Many lncRNAs are 5'capped, alternatively spliced and polyadenylated like mRNAs (Rinn and Chang 2012;Sun and Kraus 2013).Despite such similarity, lncRNAs are regulated differently and represent the fastest evolving parts of the primate genome (Pollard et al. 2006;Qureshi and Mehler 2012).LncRNA have a broad range of functions in various physiological and pathological contexts (Huarte and Rinn 2010;Guttman et al. 2011;Gutschner and Diederichs 2012;Rinn and Chang 2012;Batista and Chang 2013;Sauvageau et al. 2013;Sun and Kraus 2013;Necsulea et al. 2014;Sun et al. 2015).LncRNAs are epigenetic and transcriptional regulators that serve as scaffolds for the assembly of chromatin-and gene-regulating complexes, and can take part in directing those complexes to specific loci in the genome (Wang and Chang 2011;Rinn and Chang 2012;Vance and Ponting 2014).Alternatively, lncRNAs can act as molecular sponges that buffer various protein factors and thus regulate the processing and post-transcriptional modifications of mRNAs.Also, relying on base pairing mechanisms, they modulate mRNA stability and affect translational control (Fatica and Bozzoni 2014).
The number of identified lncRNAs is close to the number of the protein encoding mRNAs (GENCODE V25, http://www.gencodegenes.org/).While the majority of the Liu et al., LncRNAs link to postnatal brain development and ageing lncRNAs are poorly conserved and expressed at significantly lower levels than mRNAs (Derrien et al. 2012;Briggs et al. 2015;Ulitsky 2016), their expression patterns are tissue-and stage-specific, suggesting their considerable importance in regulating different biological functions, in particular cellular differentiation and development (Mercer et al. 2009;Ponting et al. 2009;Fatica and Bozzoni 2014;Briggs et al. 2015).The brain is an excellent example of this function.Around 40% of mammalian lncRNAs are expressed in the brain in a precise temporal and spatial pattern.This suggests that lncRNAs are part of the machinery needed to regulate specific neuronal functions (Mercer et al. 2008a;Mercer et al. 2008b;Derrien et al. 2012;He et al. 2014;Necsulea et al. 2014;Briggs et al. 2015).Examples of this function include Malat1, MIAT and antisense RNAs to Uchl1 and Kcna2 (Bernard et al. 2010;Carrieri et al. 2012;Zhao et al. 2013;Barry et al. 2014).In addition, reconstruction of an evolutionarily conserved co-expression network suggested that lncRNAs might be involved in synaptic transmission of neurons and other fundamental biological process like spermatogenesis (Necsulea et al. 2014).Despite these new data, the precise mechanism(s) by which lncRNAs play their roles in defining the complexity of brain functions remains unclear.
A recent microarray analysis of the temporal and anatomical expression of protein coding genes but not of lncRNAs in cortical and subcortical regions associated with human neuropsychiatric diseases has yielded wealthy information of the transcriptional regulation in the primate brain development and function, and the transcriptional link with neurological states (Bakken et al. 2016).However, it remains uncertain on the mechanism on how lncRNAs play their roles in defining the complexity of brain functions, especially in primate brain during development and ageing.

LncRNA expression in rhesus macaque brain is highly similar to human
We generated cDNA libraries of polyadenylated RNA extracted from eight macro-dissected brain areas, that included the prefrontal cortex (PFC), posterior cingulate cortex (PCC), temporal cortex (TC), parietal cortex (PC) and occipital cortex (OC), hippocampus CA1 and dentate gyrus (DG), and cerebellar cortex (CB) regions from macaques of four different age groups (1-,4-,10-and 20-year-old) (Fig. 1A, and Supplemental Table S1).We generated RNA-seq data sets (1 library per age-& sex-matched pair samples) at a sequencing depth of 148.1 million reads per sample (Supplemental Table S1).After all, we aligned the filtered reads to the reference sequence (Rhesus Macaque Genome et al. 2007) by TopHat2 (Kim et al. 2013) with 2 mismatches, we were able to detect and characterize the expression patterns of ~96.26% of known annotated genes (Fig. 1B, Supplemental Fig. S1B).
In order to identify lncRNAs from different brain regions, we used Cufflinks (Trapnell et al. 2010) to perform ab initio transcript assembly and reconstructed a Liu et al., LncRNAs link to postnatal brain development and ageing 7 total of 620,089 transcripts.After a series of filtering as described Supplemental Methods, 19,509 multi-exonic lncRNAs encoded by 9904 genomic loci were identified from the remaining transcripts (Supplemental Table S2), among which 2492 (12.77%) were originated from antisense regions.The lengths of these lncRNAs were generally shorter than mRNAs (Supplemental Fig. S1C-D).In addition, lncRNA genes expressed in macaque brains have much lower GC contents in comparison with mRNA-coding genes (Supplemental Fig. S1E).A larger number of lncRNAs (759) were exclusively expressed in the neocortex compared to those in CB (270) and hippocampus (388) (Fig. 1C).In contrast, there were more mRNA specifically expressed in CB (705, p-value = 0.03, Fisher's exact test).Our data also reflected that expression of lncRNA genes was less conserved than mRNA genes among the main brain regions (Fig. 1B-C).
To further explore the conservation of macaque brain lncRNAs, we first downloaded 9,325; 20,785; 141,353; and 117,405 lncRNAs specific for macaque, gorilla, human and mouse respectively from the NONCODE database (Zhao et al. 2016) followed by a comparative analysis.Among them, 19,509 macaque brain lncRNAs were aligned to 4388 of macaque, 4236 of gorilla, 6752 of human and 3036 of mouse lncRNAs, respectively.Note that the homologues identified between macaque brain lncRNAs and the NONCODE primate lncRNAs were not significantly reduced by increasing the BLAST (Altschul et al. 1990) stringency, while those between the macaque brain lncRNA and the mouse lncRNAs were significantly decreased (Fig. 1D).When mammalian brain-related lncRNAs homologous to macaque brain lncRNAs were Liu et al., LncRNAs link to postnatal brain development and ageing 8 aligned with each other, and approximately half of them (2241) were shared by all four species (E-value = 1e-3).The number of lncRNAs shared by macaque, gorilla and human were found to be higher with a stricter threshold (2039, E-value = 1e-10) than with a looser threshold (1093, E-value = 1e-3) (Supplemental Fig. S1F).

The extent of regulation of the expression of brain lncRNAs is higher than that of mRNAs
In order to understand the spatiotemporal expression patterns of all mRNAs and lncRNAs in our data sets, principal component analysis (PCA) was performed.The mRNA expression pattern in CB representing a distinct cluster, those in TC and OC representing another one, and the rest of the five regions represent the third one; whereas for lncRNA expression, only CB cluster was separable from another cluster comprising all other samples (Fig. 2A).Next, Pearson correlation analysis for all pairs of RNA-seq samples was performed, demonstrating similar results (Fig. 2B, and Supplemental Fig. S2A).Expression of mRNAs in each cluster was closer than that of lncRNAs, consistent with a higher expression dynamics of lncRNAs except for the CB cluster (Fig. 2 and Supplemental Fig. S2).The clustering of cerebral lncRNAs showed close similarities in all samples from the 1-year-old age group but a clear divergent expression at later ages (Fig. 2B).To eliminate the influence of expression discrepancy due to any spatiotemporal features of these two classes of RNAs, we performed a similar analysis with filtered lncRNAs and mRNAs having expression

Liu et al., LncRNAs link to postnatal brain development and ageing
9 RPKM values ranging from 0.1 to 20.The results showed the same clustering profiles as those of the unfiltered data sets (Supplemental Fig. S2B-C).
LncRNAs are well-known for their tissue-specific expression patterns than protein-coding genes, and Jensen-Shannon (JS) divergence analysis reveals high tissue specificity scores of lncRNAs expressed from different human tissues (Cabili et al. 2011).The same analysis also suggests that considerable cellular specificity of lncRNAs among different neuronal cell types (Molyneaux et al. 2015).We further performed JS divergence analysis for brain mRNAs and lncRNAs from the brain samples and found that majority of lncRNAs and mRNAs scored lower than 0.25, which were lower than the lncRNA scores corresponding to studies mentioned above.
Interestingly, no significant difference in specificity scores was observed between lncRNAs and mRNAs (Supplemental Fig. S2D, p-value = 0.91, Kolmogorov-Smirnov test, KS test).We also calculated the tissue specificity scores for lncRNAs and mRNAs within similar expression levels (RPKM value ranging from 0.1 to 20).
Profiles of cumulative specificity scores between filtered lncRNAs and mRNAs were similar (Supplemental Fig. S2D, p Next, we determined how lncRNAs were differentially expressed by studying the expression of known lncRNAs (Supplemental Fig. S3).We identified 19 copies of KCNQ1OT1, 3 copies of RMST, one copy of XIST and its antisense non-coding RNA TSIX, SOX21-AS1 and MIAT.As being sex-determined, XIST was exclusively expressed in high levels among all female macaque brain samples without significant changes at different ages (Supplemental Fig. S3A).Such female exclusive expression pattern was further confirmed by qPCR (Supplemental Fig. S3A).Interestingly, expression of TSIX was highly neocortex-specific, and the expression level was the highest in 1-year-old OC samples (Supplemental Fig. S3B).RMST is known to be regulated by the transcription factor REST which then drives the recruitment of the neural transcription factor, SOX2, to turn on key neurogenesis-promoting genes, such as DLX1 and ASCL1 (Ng et al. 2013).We observed that among three copies of RMST, one was expressed at very low level while the other two expressed in an age-and sex-dependent manner.RMST-2 (the second copy of RMST) was more negatively correlated to age in female samples; while RMST-3 (the third copy of RMST) expression was more temporally regulated in both female and male macaques (Supplemental Fig. S3C-D).The temporal regulation of MIAT expression showed more spatial-specific (standard deviation, SD = 12.51) than that of SOX21-AS1 expression (SD = 0.20) (Supplemental Fig. S3E-F).
Lastly, we determined the differential expression of lncRNAs and mRNAs of the same anatomic structure between any two adjacent age groups (1-, 4-, 10-, and 20-year-old).Stages from 1-year-old to 4-year-old showed that expression of lncRNAs changed the most in all the regions except DG.Such changing pattern was also evident in mRNA expression, with exception that both CB and DG failed to show the most significant changes at 1-year.Substantial changes in expression of both lncRNAs and mRNAs were also observed in the period from 4-year-old to

Liu et al., LncRNAs link to postnatal brain development and ageing
10-year-old.However, changes observed between 10-year to 20-year-old period were the least (Fig. 2C-D).

Liu et al., LncRNAs link to postnatal brain development and ageing 12
Tempo-spatial modules demonstrated a more pronounced and patterned regulation by postnatal development and ageing, but were less patterned by structure separation (Fig. 3D).Expression of lncRNA and mRNA genes in sex-temporal modules was specific to both sex and age, but was less associated with specific structure (Fig. 3E, Supplemental Fig. S4E10, and Supplemental Table S3).Surprisingly, reciprocal sex-specific patterns of lncRNAs were observed across the four ages (Fig. 3E1-4).
Such a reciprocal sex-specific regulation was also shown by mRNAs (Supplemental Fig. S4E10).
To validate the spatial-specific lncRNAs in macaque brain, we determined the expression levels of three CB-specific lncRNA RP11-491F9.1, Gm37142 and LINC00670 which were abundant in brain and predicted with potential roles in brain function.Both qPCR and RNA-seq data revealed that RP11-491F9.1, Gm37142 and LINC00670 exclusively preserved in CB across the four ages (Supplemental Fig. S5A and S6A).ISH data from 10-year old CB slices validated that RP11-491F9.1 exclusively expressed in CB (Supplemental Fig. S5B and S6B).This phenotype was also confirmed by the CA1-and DG-enriched NONHSAG047825.1 too (Supplemental Fig. S6C-D).

High dynamics of lncRNA expression in the cerebral cortices
Among the lncRNA co-expression modules, the third largest M3 contains 396 lncRNAs.M3 did not express in CB or DG, but highly expressed in PFC, PCC, TC,

Liu et al., LncRNAs link to postnatal brain development and ageing
13 PC and OC in an age-regulated pattern (Fig. 3D).These lncRNAs were mostly expressed in 1-year-old, and their expression was reduced significantly at other ages especially in regions like PCC, PC and TC.We named this class of cerebral lncRNAs as "Early lncRNAs".In CA1 region, these classes of lncRNAs were expressed in a similar pattern.M9 (magenta, 58) represented another class of lncRNAs.Similar to M3, M9 lncRNAs expressed at highest level in 1-year-old macaques but reduced significantly in other ages, especially in regions of PFC, PCC, TC, PC, OC and CA1.

Sex difference in lncRNA expression
Sex-biased expression of protein coding genes has been reported in the human brain (Kang et al. 2011).Here, we have identified a 4-year-old male mRNA module (81 protein-encoding genes) and 4 sex-temporal lncRNA modules (183 lncRNA genes).
We next used paired t-test module (p-value<0.01) to identify sex-specific lncRNAs regardless to their temporal regulation.A total of 307 sex-biased lncRNAs were identified, including 148 female-biased and 159 male-biased (Fig. 4A).Among these sex-specific lcnRNAs, 5 and 2 sex-biased lncRNAs were encoded from the X Chromosome of female and male macaques, respectively.The same approach identified 90 female-biased and 129 male-biased mRNAs (Supplemental Fig. S9A-D).
Therefore, sex-biased lncRNAs (307/9904, 3.1%) were present at a much higher frequency than mRNAs (219/26654, 0.8%, p-value < 2.2e-16, Fisher's exact test) across all ages and brain structures.The number of overlaps found between sex-biased and sex-temporal mRNA and lncRNAs were very limited, with only 7 mRNAs and 24 lncRNAs that being confirmed as two populations of sex-specific lncRNAs.Finally, we have determined the expression levels of three sex-biased lncRNAs AC027613.1,NONGGOT004660.1 and AC132825.2 in macaque brain which were abundant in brain.Further analyses of RNA-seq, qPCR and ISH data revealed high correlation of the AC027613.1,NONGGOT004660.1 and AC132825.2with sex and age specificities (Fig. 4B-E, Supplemental Fig. S10A-D, and S11).

CAGE-seq analysis of the transcription start sites and 5'-capping dynamics of lncRNAs
Annotation of the transcriptional start sites on lncRNAs is important but this concept is challenged by their diversity in biogenesis and by their low expression and conservation levels.Until recently, serious efforts have been made to gather human lncRNA transcript models with accurate 5' ends by integration of various CAGE-seq data (Hon et al. 2017).In order to identify the more accurate transcription start sites (TSS), we used a modified CAGE-seq technology to analyze lncRNAs identified by RNA sequencing assembly in macaque brains (Fig. 1).This technology selected polyadenylated lncRNAs to identify the 5'-cap structures.As a result, full-length lncRNAs with both 5'-cap and 3'-poly (A) tail were enriched.CAGE-seq was used to generate 5'-cap sequencing reads from the very same 64 macaque brain samples used to generate RNA sequencing reads for assembly.Therefore, TSSs were annotated to the same sets of lncRNAs assembled by RNA reads.Detailed sequencing and alignment results were shown in Supplemental Table S4, with an average of 66.83% mapping efficiency.Note that these reads were significantly enriched at the TSSs of known mRNAs (Supplemental Fig. S12A).
With the CAGE transcript start sites (CTSSs) falling within 20 bp being clustered into transcript clusters (TCs) as previously reported (Nepal et al. 2013), each TC may then represent a potential TSS.A total of 103,766 TCs were identified from all brain samples; 52.49% of them were distributed across 15,592 annotated genes.80% of TCs had a width of no more than 4-nt (Supplemental Fig. S12B), illustrating the strict usage of TSS in macaque brain.Over 82% of CTSS were grouped into TCs, among which 32.93% TCs were detected from more than one sample.TCs were enriched around the TSSs of both known protein encoding genes as well as lncRNAs identified in this study (Fig. 5A).We found that CAGE-seq reads and TCs were strongly enriched at 5'UTR but not at 3'UTR and intronic regions (Fig. 5B).
A total of 6991 of the intergenic TCs fell into 3084 (31.14%) lncRNAs annotated in this study, among which 2324 lncRNAs were homologous to those of human, showing significant enrichment (p-value < 0.001, Fisher's exact test).A total of 13,269 mRNAs (43.87%) had at least one TC support.Different TC frequencies of lncRNAs and mRNAs were well correlated with their differential expression levels (Supplemental Fig. S12C, p-value=4.369e-08,Mann-Whitney U test).TCs of a gene identified within the gene body or at 2-kb upstream of its previously annotated TSS, or genes containing more than one TCs were assigned alternative promoters.About 66% and 46% of mRNAs and lncRNAs contained more than one TC, respectively, showing alternative promoter evidence (Fig. 5C, and Supplemental Fig. S12D).
We further plotted the ratio of multiple TC-containing genes to all TC-containing genes in all samples.Alternative promoter usage in mRNA genes was dynamically regulated by age in a spatial and sex-dependent manner (Supplemental Fig. S12E).
When the same analysis was applied for lncRNA with alternative TCs, similar dynamic patterns were demonstrated for both male and female subjects (Fig. 5D, R = 0.57, Pearson correlation coefficient).The frequency of 5'-capped mRNA and lncRNAs among all lncRNAs and mRNAs demonstrated that the 5' capping efficiency could be regulated spatially in different brain regions, as well as by the degree of brain maturation and age-related degeneration (Fig. 5E, and Supplemental Fig. S12F).We also noticed the presence of sex-dependent regulation of 5' capping efficiency and alternative promoter usage, one example is the lnc-CHRM3-1 gene, shown in Fig. 5F-G.

LncRNA-mRNA co-expressed network
To explore the functions of brain lncRNAs, a correlation matrix between 9904 lncRNAs and 26654 mRNAs was generated by computing the Pearson correlation coefficient for all pairwise combinations based on their expression in our 64 transcriptomes.At a stringency of p-value ≤ 0.01 and absolute Pearson correlation coefficient abs (PCC) ≥ 0.7, a total of 3341261 co-expression pairs were detected between 5084 lncRNAs and 18418 mRNAs (Supplemental Table S5).For example, 237 mRNAs and 93 lncRNAs were co-expressed with MIAT, functional clustering of the interacted mRNAs revealed that this lncRNA is extensively involved in neuroactive ligand-receptor interaction, GABAergic synapse, dopaminergic synapse, glutamatergic synapse and morphine addiction (Fig. 6A).
Enriched Gene Ontology (GO) terms and KEGG pathways were further obtained for all mRNAs interacted with each lncRNA module.Fanconi anemia pathway was the most enriched for M1 lncRNAs (CB module); circadian rhythm pathway was mostly enriched for M2 lncRNAs (CA1/DG-module); allograft rejection and autoimmune disease were found for M13 lncRNAs (1-year male module).For the neocortex lncRNA modules such as M3, M5, M6, M7 and M9, mRNAs interacted were more enriched in conferring synaptic functions (Fig. 6B).
Lastly, we analyzed the number of lncRNA-mRNA pairs between any two module pairs (18 lncRNAs and 14 mRNA modules).We found numbers of strong module-module co-expression pairs, which include exclusive pairs such as L1-M1, L4-M4 as well as multiple pairs such as L2 with M1 and M10, and L6 with M2, 7 and M8 (Fig. 6E, and Supplemental Table S5).

Negatively regulatory networks between mRNA-lncRNA and lncRNA-lncRNA
Co-expression of the gene pairs was then established between all pairs of lncRNA-lncRNA and mRNA-mRNA.We found that positive pairs were the predominant species, as also consistent with the co-expression pattern of most genes involved (D'Haeseleer et al. 2000;Zhang and Horvath 2005).Strikingly, as high as 22.37% lncRNA-mRNA and 25.35% lncRNA-lncRNA were negative pairs, in contrast to the 5.03% observed in mRNA-mRNA pairs.This suggests that the higher population of lncRNA takes part in negative pairs supported the notion that lncRNAs is gene repressing in nature, this includes the repression of both mRNAs and lncRNAs expression (Supplemental Table S5).
After the stringent filtering, 3341261 lncRNA-mRNA pairs remained in our lncRNA-mRNA network containing 5084 lncRNAs and 18418 mRNAs.The network of a million co-expressed pairs contains 92.93% of positive lncRNA-mRNA pairs and 7.07% negative pairs.We determined to illustrate the resulting negative regulatory network with that of Ptbp1 gene, a conservative heterogeneous nuclear ribonucleoprotein (hnRNP) that regulates neuronal gene expression.We found that Ptbp1 level was negatively correlated with 61 mRNA genes.Functions of these genes included neuron differentiation, cell projection organization, neuron and nervous system development (Fig. 7A-B).And genes negatively regulated by Ptbp1 formed extensive co-expression Liu et al., LncRNAs link to postnatal brain development and ageing 20 networks (Fig. 7C).Knockdown of Ptbp1 in mouse cortical neurons significantly increased the levels of several of its targeting genes including Emx2, LhxX2, Nr2e1, Kif3a and Foxg1 expression (Supplemental Fig. S14A-I).
Next, we analyzed the lncRNAs in each distinct module that formed negatively correlates with mRNAs (Fig. 7D).We further identified lncRNAs that target mRNA and lncRNA modules, and then analyzed if there are any overlaps among these lncRNAs.Fig. 7E shows the presence of a large number of lncRNAs that controlled both mRNAs and lncRNAs belonged to the M1 module (high expression in CB).
Alternatively, we also noticed substantial number of lncRNAs could negatively regulate the expression of both M1 CB-mRNAs and M6/M7 neocortex-lncRNAs.
Among such negative regulatory network, lncRNAs of M1 and M5 modules were most extensively regulated by other lncRNAs, the same was also true for mRNAs of M6, M7 and M8 modules.

Discussion
Genomic and transcriptomic profiling of brain tissue data sets of different species reveal alterations in genetic and epigenetic systems underlies the processes brain development, ageing, and even mental disorders (Oldham et al. 2008;Belgard et al. 2011;Qureshi and Mehler 2012;Aprea et al. 2013;Bakken et al. 2016).In this study, using RNA-seq and CAGE-seq we generated complementary data sets that allowed The alternative promoter usages of both mRNAs and lncRNAs are also expected to play a role during brain development and ageing, which has not been systematically studied so far.Using full-length CAGE-seq approach, we identified not only the potential transcription start sites of a large fraction of macaque brain lncRNAs (31.14%), but also extended this finding to understand how spatial, temporal and sex parameters regulate of brain lncRNAs expression.Alternative promoter usage and capping efficiency associated with the transcription of lncRNA and mRNAs could represent an important mechanism in regulating macaque brain development and ageing, this may also take part in regulating the expression of these two classes of RNAs.
While mRNA co-expression networks has been described as important code to understand the brain (Cabili et al. 2011;Fertuzinhos et al. 2014;Molyneaux et al. 2015;Zeisel et al. 2015), very few of them appears to reflect the complexity of brain architecture and function.We demonstrated how the dataset can be used to profile trajectories of genes associated with specific neurobiological categories or disorders, many of which are not likely be evident from transcriptomic profiles of commonly studied model systems.Coupled with analysis of co-expressed genes in the dataset, these provide information on specific timing and tissue localization of various genes expressed in the brain, which will also offer insights regarding to their function.Our data enhances genome-wide associations and linkage studies by narrowing the focus to any candidate genes that are specifically expressed during development or restricted to a specific region known to be afflicted in neurological diseases.
Additional parameter like how lncRNA represses other lncRNA expression is still currently understudied.Therefore, it is likely that our report on the dynamic changes observed in lncRNA co-expression networks may serve as a regulatory system that truly contribute to the complexity of the brain architectures and function, particularly in primates.(C) Pie chart of percentage for lncRNA genes with one and more promoters.">5" indicates lncRNA genes with 6 and more promoters.

Figure legends
Liu et al., LncRNAs link to postnatal brain development and ageing21 the identification and confirmation of full-length orthologous lncRNA sequences, novel transcripts from macaque brain across postnatal development and ageing.We expect that our new resource should contribute to the understanding of the importance of lncRNA-mediated regulation, not only in aspects of brain development and ageing, but also to brain-related disorders during at different periods of a life time.Although the contribution of sex difference in human cognition is well conceived, very limited information were available in the literature which explains their relationship (McCarthy and Arnold 2011).Our analysis of macaque brain lncRNAs is the first to identify hundreds of sex-temporal and sex-biased lncRNAs related to the postnatal development and ageing, indicating that lncRNAs might play significant roles in shaping the cognitive difference observed between male and female subjects.

Figure 1 .
Figure 1.A comprehensive catalog of lncRNA genes in rhesus monkey Brain

Figure 2 .
Figure 2.The discrete expression patterns of lncRNAs and mRNAs

Figure 3 .
Figure 3.The discrete expression modules of lncRNA expression by the WGCNA

Figure 5 .
Figure 5. CAGE-seq characterization of alternative promoter usage and