Document Detail

Profiling of T helper cell-derived small RNAs reveals unique antisense transcripts and differential association of miRNAs with argonaute proteins 1 and 2.
Jump to Full Text
MedLine Citation:
PMID:  23185045     Owner:  NLM     Status:  MEDLINE    
RNA interference mediated through antisense transcripts is a fundamentally important mechanism regulating gene expression that remains incompletely understood. Here, we have used next-generation sequencing to determine from mouse CD4+ T cells the functional implications of antisense transcripts binding to argonaute (AGO) proteins that mediate RNA interference and post-transcriptional gene silencing. This effort identified 90 new microRNAs (miRNAs) and six endogenous hairpin RNA-derived small interfering RNAs (siRNAs) mapping to distinct introns. Unexpectedly, 69 miRNAs were expressed as non-canonical isomiRs as the dominant AGO-binding transcript, with extensive 3' terminal nucleotide modifications. Furthermore, differential expression analysis between AGO1- and AGO2-bound miRNAs suggested preferential binding of isomiRs ending with 3' adenine residues to AGO1 and 3' uridine residues to AGO2. Analysis of the putative targets of all miRNAs suggested a striking preference for regulating transcription and transcription factors with additional evidence of a functional division of labor between AGO proteins in this regard. We further provide evidence that multiple mitochondrial genomic loci serve as the source of endogenous cis-natural antisense transcripts. These findings imply diversity in AGO protein function based on differential miRNA binding and indicate that RNA interference-based gene regulation is more complex than previously recognized.
Sumanth Polikepahad; David B Corry
Related Documents :
11752665 - Rna localization meets wingless signaling.
21356375 - Novel roles for hypoxia and prostaglandin e2 in the regulation of il-8 during endometri...
3224555 - Localization of transcripts from the wingless gene in whole drosophila embryos.
11486045 - Herp, a novel heterodimer partner of hes/e(spl) in notch signaling.
20509905 - Retinoblastoma binding protein-1 (rbp1) is a runx2 coactivator and promotes osteoblasti...
12941765 - Peripheral, but not central, administration of adiponectin reduces visceral adiposity a...
Publication Detail:
Type:  Journal Article; Research Support, N.I.H., Extramural; Research Support, Non-U.S. Gov't     Date:  2012-11-26
Journal Detail:
Title:  Nucleic acids research     Volume:  41     ISSN:  1362-4962     ISO Abbreviation:  Nucleic Acids Res.     Publication Date:  2013 Jan 
Date Detail:
Created Date:  2013-01-21     Completed Date:  2013-04-04     Revised Date:  2013-07-11    
Medline Journal Info:
Nlm Unique ID:  0411011     Medline TA:  Nucleic Acids Res     Country:  England    
Other Details:
Languages:  eng     Pagination:  1164-77     Citation Subset:  IM    
Department of Medicine, Baylor College of Medicine, Houston, TX 77030, USA.
Export Citation:
APA/MLA Format     Download EndNote     Download BibTex
MeSH Terms
Argonaute Proteins / metabolism*
Eukaryotic Initiation Factors / metabolism*
Mice, Inbred C57BL
MicroRNAs / chemistry,  metabolism*
RNA / genetics
RNA Precursors / chemistry,  genetics
RNA, Antisense / genetics,  metabolism*
RNA, Small Interfering / chemistry,  genetics
T-Lymphocytes, Helper-Inducer / metabolism*
Grant Support
Reg. No./Substance:
0/Argonaute Proteins; 0/EIF2C2 protein, mouse; 0/Eif2c1 protein, mouse; 0/Eukaryotic Initiation Factors; 0/MicroRNAs; 0/RNA Precursors; 0/RNA, Antisense; 0/RNA, Small Interfering; 0/RNA, mitochondrial; 63231-63-0/RNA

From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine

Full Text
Journal Information
Journal ID (nlm-ta): Nucleic Acids Res
Journal ID (iso-abbrev): Nucleic Acids Res
Journal ID (publisher-id): nar
Journal ID (hwp): nar
ISSN: 0305-1048
ISSN: 1362-4962
Publisher: Oxford University Press
Article Information
Download PDF
Published by Oxford University Press 2012.
Received Day: 9 Month: 6 Year: 2012
Revision Received Day: 15 Month: 10 Year: 2012
Accepted Day: 21 Month: 10 Year: 2012
Print publication date: Month: 1 Year: 2013
Electronic publication date: Day: 25 Month: 11 Year: 2012
pmc-release publication date: Day: 25 Month: 11 Year: 2012
Volume: 41 Issue: 2
First Page: 1164 Last Page: 1177
PubMed Id: 23185045
ID: 3553939
DOI: 10.1093/nar/gks1098
Publisher Id: gks1098

Profiling of T helper cell-derived small RNAs reveals unique antisense transcripts and differential association of miRNAs with argonaute proteins 1 and 2
Sumanth Polikepahad1
David B. Corry12*
1Department of Medicine and 2Department of Pathology and Immunology, Baylor College of Medicine, Houston, TX 77030, USA
Correspondence: *To whom correspondence should be addressed. Tel: +1 713 798 8740; Fax: +1 713 798 5780; Email:


The argonaute proteins comprise an evolutionarily conserved family of RNA-binding proteins that are involved in the regulation of transcribed RNAs through RNA interference. The two major classes of argonaute proteins, AGO and PIWI, both bind to 19–35 nt short RNAs and regulate expression of genes and transposons in somatic and gonadal tissues, respectively. AGO proteins bind to microRNAs (miRNAs) or small interfering RNAs (siRNAs) within the RNA-induced silencing complex (RISC) to suppress target gene expression either by endonucleolytic cleavage or translational repression through a process termed post-transcriptional gene silencing (PTGS). AGO proteins are essential for embryological development and cellular differentiation, and in plants and primitive animals are required for host defense (1). The number of argonaute family proteins varies widely among species, ranging from 27 in Caenorhabditis elegans to eight in mice and humans (2).

AGO proteins further play crucial roles in the biogenesis of mature miRNAs. After Drosha cleavage of precursor miRNAs, the miRNA/miRNA* duplex binds to AGO proteins, a process termed ‘RISC loading’ to form the RISC complex. Numerous factors influence loading of small RNAs into AGO proteins, including Dicer processing, the structure of the small-RNA duplex, terminal (5′ and 3′) nucleotides, thermodynamic properties of the duplex and individual strands and the structure of the AGO protein itself (3). In Drosophila, Dicer 1 processes miRNA precursors, whereas Dicer 2 processes siRNA precursors. Accordingly, miRNAs differentially load into AGO1 and siRNAs into AGO2. However, notable exceptions to this rule in which some Dicer 1 produced miRNAs preferentially bind to AGO2 suggested the presence of a post-dicer processing mechanism that influenced AGO loading (3). The degree of base pairing in duplex small RNAs further appears to influence the loading of small RNAs into AGO proteins. siRNAs, which form fully complementary duplexes, preferentially bind AGO2, whereas miRNA duplexes that contain unpaired bulges preferentially load into AGO1. AGO1 further favors a terminal U and AGO2, 5′C residues (4). Furthermore, strict strand selection of miRNAs is observed with miRNA mature/guide strands binding to AGO1 and miRNA*/passenger strands binding to AGO2 (3,5). This preferential binding is primarily attributed to the structure of the miRNA/miRNA* and siRNA duplexes where the strand with the least 5′ thermodynamic stability is selected as the guide strand (5).

Mice and humans possess four biochemically distinct AGO proteins, AGO1-4. Although functional differences among these proteins remain largely speculative, AGO2 likely possesses a unique slicer, i.e. endonucleolytic, activity required to cleave target mRNAs (6). Differential loading of miRNAs into AGO proteins has been reported from transformed human cells (7), but mammalian AGO proteins are nonetheless believed to be functionally redundant (3). In addition to miRNAs and siRNAs, next-generation sequencing (NGS) technologies have enabled the discovery of many additional classes of non-coding RNAs (ncRNAs) from animals and plants with potential to participate in PTGS, including promoter-associated RNAs (paRNAs), transcription start site-associated RNAs (TSSaRNAs), 5′ capped promoter-associated small RNAs (PASRs), intron-derived miRNAs, i.e. mirtrons, short hairpin RNAs (shRNAs), antisense miRNAs (asmiRNAs), animal endogenous siRNAs (endo-siRNAs), natural antisense transcripts (NATs) and many others (8). Although the molecular mechanisms and AGO-associated functional pathways of most of these ncRNAs remain undefined, their manifold existence confirms the highly complex nature of PTGS.

In this study, we have identified multiple RNA transcripts that bind to AGO1 and AGO2 proteins in mouse splenic CD4+T cells. By combining co-immunoprecipitation of AGO proteins with NGS and advanced bioinformatics approaches, we show the differential AGO binding of several known miRNAs and their isomiRs, in addition to the discovery of many novel miRNAs, hpRNA derived endo-siRNAs and novel NATs deriving from the mitochondrial genome.

Isolation of CD4+T cells and argonaute bound RNA co-immunoprecipitation

Spleens were isolated from eight (two control, three AGO1 and three AGO2) 5 weeks old C57BL/6 mice and CD4+T cells were extracted by using immunomagnetic selection. Twenty million cells per each mouse were used for co-immunoprecipitation with Rabbit IgG (Millipore, Billerica, MA, USA) control, anti-AGO1 (Millipore) and anti-AGO2 (Wako, Richmond, VA, USA) monoclonal antibodies.

Preparation of deep sequencing libraries

Total RNA including small RNAs were extracted by using miRNAEasy kit (Qiagen, Valencia, CA, USA) and quality assessed by 260/230 nm absorbance ratios. Small-RNA transcripts of 15–40 nucleotides were gel purified after running 5 μg of total RNA on 15% [Tris-borate-EDTA (ethylene diamine tetraacetic acid) buffer (TBE)]-urea polyacrylamide gels. A synthetic Illumina-specific 26-residue adapter RNA oligonucleotide (5′-GUU CAG AGU UCU ACA GUC CGA CGA UC-3′) was ligated to the 5′-end of the small RNAs. The ligated small RNA was gel purified to remove free adapter. A synthetic Illumina-specific 22-residue 3′-adapter with inverted dideoxythymidine (idT) added at the 3′-end (5′-p UCG UAU GCC GUC UUC UGC UUG idT-3′) was ligated to the 5′-ligated small RNA and gel purified. The resultant RNA library was reverse transcribed and amplified by PCR for 19 cycles using adapter-specific primers. The PCR products were then sequenced using Illumina-Genome Analyzer IIx, at the Institute for Molecular Design Center—University of Houston.

Read processing, mapping and aligning and post-alignment processing of deep sequenced libraries

Raw sequences were uploaded to NCBI Sequence Read Archives (SRA) and are available through accession number SRA056111. The quality of raw sequences was evaluated using FASTX-Toolkit version 0.6, generating quality statistics and box-whisker plots to summarize the base call quality score range at each position in all eight sequence files (Supplementary Figure S1). We considered base quality as poor if the median score was <28 or lower quartile was <20. Based on these criteria, read quality was very high for positions 1–30 (with median scores ranging between 30 and 40), but declined thereafter in all libraries with each additional nucleotide. As expected, we observed an overrepresentation of adapter sequences at positions 31–40, which were removed from all full-length reads by using a custom ‘perl’ script that searches for and removes a minimum of 8 nt of perfectly matched adapter sequence from the sequenced read. The region that was most frequently deleted was the 31–40 segment, confirming the over-representation of adapter sequences. A tag file for each library was then created by collapsing unique reads and calculating tag counts for each read using the same ‘perl’ script.

The collapsed reads were first mapped to the known mouse miRNA database by using standalone miRanalyzer version 0.2 which uses Bowtie 2.0, an ultrafast memory-efficient short read aligner (9,10), using the following parameters: –best –strata –e –k10 –l17 –n1. For non-microRNA reads, we used Bowtie 2.0 directly for mapping to other transcript classes. This is mainly because miRanalyzer is mainly made for identifying miRNAs and hence has a maximum read length limitation of 28, precluding mapping of longer reads to longer transcripts. We first mapped these reads to mouse genome [NCBI Build 37, University of California at Santa Cruz (UCSC) mm9] using the following parameters: –best –k10 –v1. Reads mapping 10 or more times to the genome were discarded. The mapped reads were aligned to the mouse transcriptome by using BEDTools suite (11). For miRNA and other transcript mapping, we selected a minimum length of 17 nt. Reads aligning to more than one RNA class or more than one type of transcript in each class were considered ambiguous and are not shown.

miRBASE v. 18 indexes were downloaded from the miRanalyzer website; mouse protein coding genes and repeat sequences, UCSC genome browser; ncRNA, ENSEMBL browser and RFam database; PASR, deepBase database. Visualization of the mapped reads was performed on the UCSC genome browser. RNA folding of pre-miRNA and hpRNAs to predict hairpin structures and calculate minimum free energy (MFE) levels was performed on the RNAfold server (12). Multiple alignments of the precursor miRNAs, canonical isomiRs and detected miRNAs, and multiple species seed sequences of novel miRNAs (candidate_21) were performed by the clustalw2 program.

Differential expression analysis

We first checked for the consistency of the data among biological replicates in each group by performing unsupervised clustering analysis that measures Euclidean distances among the samples, which clearly revealed that replicates from each group cluster together (Supplementary Figure S2). Next, differential expression analysis between AGO1 and AGO2 samples was performed by the Empirical analysis of digital gene expression data in R (EdgeR) package version 2.4.6 (13). By using this package, first the tags with >5 counts per million (CPM) were removed in each library, the resultant reads were normalized by the Trimmed Mean of M component method, normalization factors were calculated, effective library sizes were estimated by using normalization factors, fitted according to the Negative Binomial model, estimates of dispersions were made by using the quantile-adjusted conditional maximum likelihood method and differential expression tests were performed. The resultant P-values were adjusted for false discovery rate (FDR) by using Benjamini and Hochberg’s approach and only adjusted P-values with <0.05 were considered statistically significant for the miRNA differential expression analysis. To generate heatmaps, the log count values of each miRNA or isomiR were first moderated by the variance stabilizing transformation (VST) method outlined in the DESeq version 1.6.1 package manual and heatmaps were created by the heatmap2 function of the gplots2.10.1 package (14). The VST data were also used to calculate the Euclidean distances among samples and sample clustering, which showed similarity among different samples (14).

Target prediction and gene ontology enrichment

The targets of differentially expressed miRNAs were predicted by using TargetScanMouse 6.0. Predicted targets that were targeted by at least two miRNAs and with <−0.10 total context score (2582 targets in AGO1 and 1549 targets in AGO2) were targeted for GO enrichment, which was performed by the ‘functional annotation tools’ function of Database for Annotation Visualization, and Integrated Discovery (DAVID) bioinformatics tool (15). The enrichment was performed using the medium classification stringency option against the total mouse gene population background (15).

Novel miRNA discovery

Novel miRNAs were discovered by using miRanalyzer version 0.2 (9,16). This program uses stringent criteria for putative miRNA discovery by employing several prediction models. For accuracy, only those miRNAs were considered novel that were predicted by at least three different prediction models and expressed in at least three of the six samples (three AGO1 and three AGO2).

Statistical analysis

Data are presented as means ± SEM. Quantitative real time PCR data are presented using the ΔΔCt relative expression method. The U6RNA was used as an endogenous control and data were normalized to the control IgG sample. Differential expression analysis of isomiRs and primary-isomiRs was performed by the EdgeR package as described earlier with P-values adjusted using Benjamini and Hochberg’s approach to correct for FDR (13). For determining the significance of 3′ and 5′ terminal nucleotide occurrences in differentially expressed isomiRs, chi-square goodness-of-fit test with expected terminal nucleotide frequency was calculated from all 6141 isomiRs using Graphpad Prism version 5.01 software. Gene ontology (GO) enrichment analysis was performed by the DAVID bioinformatics tool with P-values calculated as previously described (15). For all analyses, a P-value of <0.05 was considered significant.

AGO1 and AGO2 bind distinct miRNAs and other small RNAs

Although the specificity and lack of cross-reactivity between AGO1 and AGO2 antibodies have been reported (17–19; and, we conducted additional studies to confirm the specificity of anti-AGO antibodies. First, we performed qRT-PCR for let-7f using small RNAs derived from AGO1- and AGO2-specific immunoprecipitates of CD4+ T cells as compared with non-specific IgG immunoprecipitates (Figure 1A). This analysis indicated a marked differential binding of let-7f to AGO1 and AGO2 as compared with the relatively weak and non-specific binding in association with rat IgG. Second, the length distribution analysis of the sequenced and trimmed reads selected for mapping revealed predominant binding of both AGO1 and AGO2 to transcripts of between 21 and 23 nucleotides; again, only random, low-grade transcript binding was observed with IgG (Figure 1B). Third, the vast majority of transcripts binding to AGO1 (74.4%) and especially AGO2 (97.2%) were miRNAs, comprising a total of 682 mature miRNAs, with 591 binding to AGO1 and 637 to AGO2 (Figure 1C and D; Supplementary Table S1). In contrast, although substantial numbers of transcripts were found in association with control immunoglobulin, no dominant (i.e. comprising >50% of all transcripts) transcript class was apparent. Repeat elements (29.6%) were the most abundant transcript class found in association with control rat IgG with miRNAs comprising an average of 14.8%. Thus, these data suggest that miRNAs specifically and preferentially associate with both AGO1 and AGO2, in marked contrast to the non-specific transcript binding to control immunoglobulin.

These findings support the specificity of the anti-AGO antibodies in comparison to IgG control, but do not fully address the specificity of the anti-AGO antibodies themselves. Thus, we next analyzed the specificity of transcript binding to AGO1 and AGO2. For this purpose, we performed sample clustering, which calculates the Euclidian sample to sample distance after VST of the count data after removing <5 CPM reads from each sample (the resulting data of which was used for differential expression analysis), by using the dist function of the DESeq package. During this clustering, the dispersion of count data was estimated in the ‘blind’ mode, i.e. the variance-stabilized transformation was not ‘aware’ of the experimental design, to avoid bias. This analysis clearly demonstrated that biological replicates of the AGO1 or AGO2 immunoprecipitations cluster together, strongly indicating that these antibodies lack cross reactivity (Supplementary Figure S2). Notably, the most abundant miRNAs for both groups were of the let-7 family (Figure 1D; Supplementary Table S2), which parallels observations from whole mouse and human lung (20) and D. Corry et al. (unpublished data).

To further analyze the transcript binding specificity of AGO1 and AGO2, we determined the differential association of distinct transcript species with the argonaute proteins. Although AGO1 and AGO2 both predominantly bound miRNAs, AGO2 binding of miRNAs was far more robust (P < 1.407; Supplementary Table S1). In contrast, AGO1 bound a greater diversity of small RNA species, including tRNA, srpRNA, scRNA, snoRNA and snRNA, with greater avidity relative to AGO2 (P < 0.01 for each transcript). Together, these studies suggest that, although similarities in transcript binding are clear, AGO1 and AGO2 are functionally non-redundant.

Dominant expression of non-canonical isomiRs for some miRNAs

Several recent NGS-based studies have suggested that the miRNA-generating arm of the pre-miRNA produces several sequence variants of the ‘canonical’ or ‘reference’ miRNA (miRBASE sequences for each miRNA), known as isomiRs (21,22). We identified 92 019 isomiRs for 682 known miRNAs in at least one of the six T-cell samples. Most were of low copy number and after filtering the data for low copy (<5 CPM) and potentially spurious counts (variable counts among the replicate samples), we revised the isomir count to 6140. There were 214 miRNAs having at least two isomiRs, with let-7 family members occupying 7 of the top 10 most abundant, and let-7f-5p having the highest number of isomiRs (589) followed by miR-140-3p (551) (Figure 2A). The sequences, raw counts and percent abundance of all isomiRs are shown in Supplementary Table S3.

Next, we determined if the highest expressed isomiR for each miRNA was the same as the canonical miRNA. For this analysis, we first isolated the most abundant isomiR for each miRNA in each of the six libraries and then selected only those isomiRs that were expressed in either all three replicates of each of the two groups, i.e. AGO1 and AGO2, or expressed in all six libraries. This yielded 130 miRNAs, of which 69 isomiRs (∼53%) were non-canonical in at least one of the two groups (Supplementary Table S3). Among these 69, the most highly expressed isomiR differed for each AGO sample (Supplementary Table S4). Most of these isomiRs occupied >50% of the fraction, indicating their predominant expression. Some miRNAs, such as let-7b and let-7i were some of the most highly expressed of all miRNAs. Intriguingly, no canonical sequences were found for 17 miRNAs. Even where the canonical miRNA was the primary isomiR, the non-canonical miRNA fraction remained robust. For example, at least 40% of the let-7 miRNA fraction consisted of non-canonical isomiRs. Furthermore, all modifications (i.e. edits) of the highly expressed isomiRs occurred exclusively at the 3′ termini. Isolated 5′ modifications were completely absent from our libraries, and although 5′ modifications in combination with 3′ modifications were detected, they were exceptionally rare (total of three examples found). Additional modifications included single and multiple nucleotide extensions, deletions and substitutions. The most common (∼28%) of the 3′ modifications was a single U extension. Nine isomiRs showed 3′ substitutions of nucleotides, representing 3′ edits that do not map to the mouse genome (Supplementary Table S4). Exemplary alignments are shown in Figure 2B.

Preferential binding of isomiRs with 3′A to AGO1 and 3′U to AGO2

To determine if isomiRs exhibit preferential binding to AGO proteins, we performed differential expression analysis. Comparing isomiRs associating with AGO1 and AGO2, only isomiRs that showed >3 log-fold difference and adjusted P-value <0.05 were considered to be significantly differentially associated. Of 6140 isomiRs, 614 (10%) corresponding to 127 miRNAs, showed differential association, with 392 showing preference to AGO1 and 222 to AGO2 (Figure 3A and B). Fold changes in expression and all P-values are shown in Supplementary Table S5. Of these 392 isomiRs binding to AGO1, 53.4% possessed adenine as the 3′ terminal nucleotide and 58.8% of 222 isomiRs binding to AGO2 had uracil as the 3′ terminal nucleotide (Figure 3C). As for 5′ nucleotides, both AGO1- and AGO2-binding isomiRs had uracil as the predominant nucleotide (Figure 3D). These observations suggest that loading of isomiRs within argonaute proteins in part depends on terminal adenine and uracil expression, with cytosine and guanine expression being of far less importance. The findings further indicate that AGO1 and AGO2 bind distinct isomiRs based in part by terminal nucleotide sequence.

Functional sorting of miRNAs between AGO1 and AGO2

Next, to begin to understand possible functional differences in AGO proteins, we assessed the differential binding of miRNAs to AGO1 and AGO2. For this purpose, we extracted isomiRs that were of highest expression in all six libraries and further filtered the data for low copy number (<5 CPM) and spurious counts. The final dataset yielded 295 miRNAs of 682 total. Differential expression analysis of these miRNAs yielded 74 miRNAs (55 AGO1; 19 AGO2) that were at least 1.5-fold different and with adjusted P-value <0.05 (Figure 4A and B; Supplementary Table S6). Of note, the 3′ terminal nucleotide bias remained.

We then predicted the targets of these 74 miRNAs by TargetScanMouse 6.0. To avoid non-specific targets, we selected targets with total context scores of <−0.10 and targeted by at least two miRNAs, after which GO enrichment by the DAVID algorithm was performed under medium stringency against total mouse gene population background. Targets belonging to both AGO1 and AGO2 were very significantly enriched for the term ‘transcription regulation’ (Table 1) with the majority of targets belonging to the categories of transcription factor, promoter binding gene, DNA binding, transcriptional co-activator and transcriptional co-repressor (Supplementary Table S7). To determine if these miRNAs regulate common genes, we identified the common genes of the top annotation cluster with the lowest P-value between the AGO1 and AGO2 groups and found that ∼129 genes were shared between these two groups (34% of 390 AGO1; 57% of 228 AGO2) (Supplementary Table S7). These findings indicate that distinct sets of miRNAs acting through two different AGO proteins co-regulate transcription by employing more than one-third of the same genes.

Mitochondrial cis-NATs

Cis-NATs are sense–antisense transcripts transcribed from the same genomic locus, but from both directions (23). We found several sense–antisense transcripts from all T-cell libraries that overlap by at least 18 nt and which derive from mitochondrial loci encoding eight distinct mitochondrial genes and which associated with both AGO1 and AGO2 (Supplementary Table S8). Five of eight cis-NAT sequences are apparently transcribed from regions encoding five distinct NADH dehydrogenase subunits (Mt-ND2-6). Two additional cis-NAT sequences derive from tRNA encoding regions (Mt-Tq and Mt-Tp) and one from the cytochrome-b gene (Table 2; Figure 5; Supplementary Figure S3). The reads mapping to the opposite strand of these functionally important genes are complementary to ESTs and/or GenBank mRNAs, the complete sequences and functions of which remain unknown. It should be noted that these reads map only to these locations and nowhere else in the mouse genome.

Hairpin-precursor derived endogenous T-cell siRNAs

We have further discovered sequences that arise from long double-stranded stem loop precursors at six potential locations (Vpas13a, Arhgap17, Cpne5, Spata6, Tcf20 and Aftph) (Figure 6A). All sequences derive from intronic regions and are candidate endogenous-siRNAs based on previously published criteria (Supplementary Table S9) (24). Most of these sequences derive from sample AGO2c, which contained the highest number of usable reads. In agreement with a prior observation, this suggests that endogenous siRNAs are profoundly rare transcripts that are only to be found in the most comprehensively sequenced samples (24). Despite their low copy number, two candidate siRNAs appear in more than one sample at the exact location. The Vps13a-derived siRNA is particularly interesting because of its high read count and appearance in five of the six samples at the same location. All putative siRNAs satisfy the following criteria: (i) mapping only to one genomic locus; (ii) mapping sense and antisense to proximal intronic regions (Figure 6B); (iii) the read length ranges from 18 to 24 nt. Moreover, these sequences fail to meet criteria for miRNAs as the pre-miRNA stem loop structures do not form perfectly complementary double stranded stems. Also unlike miRNAs, endogenous siRNAs exhibit no species conservation, i.e. the sequences are only found in the mouse (Figure 6B). Endogenous siRNAs further failed to map to spliced ESTs, suggesting that they are distinct from NATs (data not shown).

Many novel miRNAs discovered

Finally, by using standalone miRanalyzer software, we have discovered 90 novel miRNAs that are expressed in at least three of the six samples. The complete list is shown in Supplementary Table S10. A representative novel miRNA having species conservation within the seed sequence is shown in Figure 7.


More than 90% of the eukaryotic genome is believed to be transcribed to RNA, but <2% of such transcripts encode functional proteins (8). Recent discoveries of the important roles played by several types of ncRNAs including miRNAs, piRNAs and siRNAs have clearly suggested pivotal roles in multiple biological processes. Moreover, advanced bioinformatics approaches combined with increasing power of deep sequencing technologies have made it possible to discover novel types of ncRNAs that markedly enrich the ncRNA databases. In this study of mouse T helper cells, we have identified three types of antisense ncRNAs, including miRNAs, cis-NATs and endogenous siRNAs, including many novel sequences. Moreover, we have identified isomiRs that demonstrate differential binding to AGO1 and AGO2, a property that appears in part to be due to editing of the 3′-end. Together, our findings point to an extraordinarily complex role played by ncRNAs in regulating T-cell function, especially regarding transcription.

Tissue based differential expression and function of primary-isomiRs was first suggested by Morin et al. (25). Wyman et al., further reported that 3′ modification of miRNAs is an evolutionarily conserved process that is functionally relevant in diverse tissues and disease states (22). They also suggested that the post-transcriptional addition of adenine to generate 3′NTAs (non-templated additions), here termed 3′ substitutions, is a physiologically regulated process and several members of the nucleotidyl transferase family, including PAPD4, PAPD5, TUT1, MTPAP, etc., might regulate this process (22). In this study, the majority of the 3′ modifications were extensions, i.e. the modified isomiR aligns to the genome perfectly, yet still differ from the canonical sequence. Such modifications can be caused by either differential processing by DICER during miRNA synthesis or nucleotidyl transferases post-maturation or both (26). We also identified nine cases of 3′ substitutions, in which the sequences align imperfectly with the genome. These modifications are most likely caused by the post-maturational 3′ addition of nucleotides by nucleotidyl transferases (26). Regarding their function, isomiRs are differentially expressed among distinct human cancers including gastric cancers and metastatic prostate cancer, murine cardiomyocytes and in pre-eclampsia and psoriasis (27–31). Furthermore, Cloonan et al., reported that isomiRs function cooperatively with their corresponding canonical miRNAs during target regulation in various human tissues, confirming their functional importance (32). Our studies indicate that a large number of miRNAs from CD4 T cells are expressed primarily as non-canonical isomiRs. This observation may be particularly relevant because of the need of these T cells to differentiate into specialized effector subtypes (Th1, Th2, Th17, etc.) that are required to defend against a wide variety of pathogens (33).

The differential binding affinity of isomiRs to AGO proteins was first shown for miR-142-5p in human Jurkat T cells (34). Burroughs et al. also showed differential binding of isomiRs to AGO proteins in human THP-1 cell lines for the most abundant isomiRs having 3′ nucleotide variations and less 5′ variation (7). However, in this study we showed that isomiRs deriving from 127 distinct miRNAs associate differentially between AGO1 and AGO2, suggesting that differential AGO binding might be more prevalent in normal somatic cells than previously thought. Moreover, we have shown that isomiRs ending with 3′adenine preferentially bind to AGO1 and those ending with 3′ uracil, to AGO2 (Figure 3C). Such 3′ nucleotide bias for AGO proteins has previously been observed in C. elegans (35,36), whereas this study reports such bias in mammalian cells for the first time. Although further study is required, terminal nucleotide modifications potentially explain the differential sorting of miRNA isomiRs between AGO proteins and further support the notion of distinct functions of AGO proteins. Our laboratory and others have further shown that post-transcriptional editing of miRNAs can significantly alter the target repertoire of miRNAs, again suggesting distinct functions for isomiRs (20,26). Thus, analysis of AGO-associated transcripts confirms that miRNA isomiRs are extensively edited post-transcriptionally and imply differential functions for distinct isomiRs deriving from the same canonical sequences and AGO proteins.

Little is currently known of the biochemical properties of mammalian AGO proteins. miRNAs of plants and lower organisms are known to sort differentially to AGO proteins, but this is currently thought be a primitive trait that is largely lost in higher eukaryotes (3). Our findings from T cells indicate, in contrast, that many miRNAs in fact differentially bind to AGO1 and AGO2. Deeper analysis of the putative targets of these distinct miRNA classes indicates that only approximately a third of these targets overlap despite having a predominant focus on the regulation of transcription factors. Together, these findings indicate that isomiRs work cooperatively with AGO1 and 2 to regulate extremely fundamental cellular processes, especially transcription, but imply the existence of a division of labor with regard to the regulation of distinct genes.

In addition to miRNAs, our findings demonstrate the existence of other classes of antisense transcripts in T cells, including NATs and endogenous siRNAs. The existence and differential expression of antisense transcripts deriving from mitochondrially encoded genes has previously been demonstrated from porcine and murine brain (37,38). Similarly, Burzio et al., showed the existence of long antisense transcripts deriving from mitochondrially encoded rRNA 16 in tumors and normal proliferating human cells (800 and 2000 bp, respectively). These mtncRNAs have inverted repeats and also showed differential expression between tumor and normal cells (39). NATs are critically important regulators of genomic imprinting, RNA interference, alternative splicing and X-chromosome inactivation (23), functions that are no less important in T cells. However, T-cell NATs all derive from unique loci including five distinct subunits of the NADH dehydrogenase complex, which is a crucial enzyme regulating cellular respiration. Several NADH dehydrogenase subunits have been critically linked to several neurological disorders (40–43). A critical extension of the current studies is to determine if the unique cis-NATs discovered here may regulate T cell and therefore immune function.

The lack of RNA dependent RNA polymerase (RdRP) that is essential for siRNA mediated gene regulation and the marked sensitivity of mammalian cells to the long dsRNA precursors of siRNAs that resemble viral genomes have suggested that endogenous siRNAs are not functional in higher eukaryotes (24). However, large numbers of hairpinRNA (hpRNA)-derived endo-siRNAs have been demonstrated in mouse oocytes (44). Interferon-based immune responses to long dsRNA are suppressed in oocytes, providing a means for the physiological expression of these transcripts in haploid cells. However, Smalheiser et al., reported finding eight endogenous siRNA candidates in mouse hippocampus (24). Our studies confirm the existence of endogenous, AGO-binding siRNAs in mouse somatic T cells, suggesting that such transcripts are expressed physiologically, albeit at extremely low levels.

Much further work is required to establish the functional significance of distinct isomiRs, NATs and siRNAs, both generally in mammals and specifically in T cells. We note further that the complete lack of species conservation of siRNAs and NATs implies that these transcripts regulate morphology and function in a species-specific manner. Additional studies are further required to determine the redundant and unique functions of the other known mammalian AGO proteins. More complete sequencing of AGO-related samples from diverse species and tissues combined with specialized techniques for manipulating AGO proteins in vitro and in vivo will provide insight into these critical questions.


Supplementary Data are available at NAR Online: Supplementary Tables 1–10 and Supplementary Figures 1–3.


Clayton Foundation for Research and the National Institutes of Health (NIH) [HL095382 to D.B.C.]. Funding for open access charge: NIH [HL095382 to D.B.C.].

Conflict of interest statement. None declared.

1. Joshua-Tor L,Hannon GJ. Ancestral roles of small RNAs: an Ago-centric perspectiveCold Spring Harb. Perspect. Biol.Year: 20113a00377220810548
2. Ender C,Meister G. Argonaute proteins at a glanceJ. Cell Sci.Year: 20101231819182320484662
3. Czech B,Hannon GJ. Small RNA sorting: matchmaking for ArgonautesNat. Rev. Genet.Year: 201112193121116305
4. Czech B,Malone CD,Zhou R,Stark A,Schlingeheyde C,Dus M,Perrimon N,Kellis M,Wohlschlegel JA,Sachidanandam R,et al. An endogenous small interfering RNA pathway in DrosophilaNatureYear: 200845379880218463631
5. Cenik ES,Zamore PD. Argonaute proteinsCurr. Biol.Year: 20112121
6. Liu J,Carmell MA,Rivas FV,Marsden CG,Thomson JM,Song JJ,Hammond SM,Joshua-Tor L,Hannon GJ. Argonaute2 is the catalytic engine of mammalian RNAiScienceYear: 20043051437144115284456
7. Burroughs AM,Ando Y,de Hoon MJ,Tomaru Y,Suzuki H,Hayashizaki Y,Daub CO. Deep-sequencing of human Argonaute-associated small RNAs provides insight into miRNA sorting and reveals Argonaute association with RNA fragments of diverse originRNA Biol.Year: 2011815817721282978
8. Rother S,Meister G. Small RNAs derived from longer non-coding RNAsBiochimieYear: 2011931905191521843590
9. Hackenberg M,Sturm M,Langenberger D,Falcon-Perez JM,Aransay AM. miRanalyzer: a microRNA detection and analysis tool for next-generation sequencing experimentsNucleic Acids Res.Year: 200937W68W7619433510
10. Langmead B. Aligning short sequencing reads with BowtieCurr. Protoc. BioinformaticsYear: 201011124
11. Quinlan AR,Hall IM. BEDTools: a flexible suite of utilities for comparing genomic featuresBioinformaticsYear: 20102684184220110278
12. Hofacker IL. RNA secondary structure analysis using the Vienna RNA packageCurr. Protoc. Bioinform.Year: 200912116
13. Robinson MD,McCarthy DJ,Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression dataBioinformaticsYear: 20102613914019910308
14. Anders S,Huber W. Differential expression analysis for sequence count dataGenome Biol.Year: 201011R10620979621
15. Dennis G Jr,Sherman BT,Hosack DA,Yang J,Gao W,Lane HC,Lempicki RA. DAVID: database for annotation, visualization, and integrated discoveryGenome Biol.Year: 200343
16. Hackenberg M,Rodriguez-Ezpeleta N,Aransay AM. miRanalyzer: an update on the detection and analysis of microRNAs in high-throughput sequencing experimentsNucleic Acids Res.Year: 201139W132W13821515631
17. Janowski BA,Huffman KE,Schwartz JC,Ram R,Nordsell R,Shames DS,Minna JD,Corey DR. Involvement of AGO1 and AGO2 in mammalian transcriptional silencingNat. Struct. Mol. Biol.Year: 20061378779216936728
18. Leung AK,Young AG,Bhutkar A,Zheng GX,Bosson AD,Nielsen CB,Sharp PA. Genome-wide identification of Ago2 binding sites from mouse embryonic stem cells with and without mature microRNAs. [Erratum appears in Nat. Struct. Mol. Biol. 2011 Sep;18(9):1084]Nat. Struct. Mol. Biol.Year: 20111823724421258322
19. Wu C,So J,Davis-Dusenbery BN,Qi HH,Bloch DB,Shi Y,Lagna G,Hata A. Hypoxia potentiates microRNA-mediated gene silencing through posttranslational modification of Argonaute2Mol. Cell. Biol.Year: 2011314760477421969601
20. Polikepahad S,Knight JM,Naghavi AO,Oplt T,Creighton CJ,Shaw C,Benham AL,Kim J,Soibam B,Harris RA,et al. Proinflammatory role for let-7 microRNAS in experimental asthmaJ. Biol. Chem.Year: 2010285301393014920630862
21. Morin RD,O'Connor MD,Griffith M,Kuchenbauer F,Delaney A,Prabhu A-L,Zhao Y,McDonald H,Zeng T,Hirst M,et al. Application of massively parallel sequencing to microRNA profiling and discovery in human embryonic stem cellsGenome Res.Year: 20081861062118285502
22. Wyman SK,Knouf EC,Parkin RK,Fritz BR,Lin DW,Dennis LM,Krouse MA,Webster PJ,Tewari M. Post-transcriptional generation of miRNA variants by multiple nucleotidyl transferases contributes to miRNA transcriptome complexityGenome Res.Year: 2011211450146121813625
23. Zhang Y,Liu XS,Liu QR,Wei L. Genome-wide in silico identification and analysis of cis natural antisense transcripts (cis-NATs) in ten speciesNucleic Acids Res.Year: 2006343465347516849434
24. Smalheiser NR,Lugli G,Thimmapuram J,Cook EH,Larson J. Endogenous siRNAs and noncoding RNA-derived small RNAs are expressed in adult mouse hippocampus and are up-regulated in olfactory discrimination trainingRNAYear: 20111716618121045079
25. Morin RD,O'Connor MD,Griffith M,Kuchenbauer F,Delaney A,Prabhu AL,Zhao Y,McDonald H,Zeng T,Hirst M,et al. Application of massively parallel sequencing to microRNA profiling and discovery in human embryonic stem cells. [Erratum appears in Genome Res. 2009 May;19(5):958]Genome Res.Year: 20081861062118285502
26. Burroughs AM,Ando Y,de Hoon MJ,Tomaru Y,Nishibu T,Ukekawa R,Funakoshi T,Kurokawa T,Suzuki H,Hayashizaki Y,et al. A comprehensive survey of 3′ animal miRNA modification events and a possible role for 3′ adenylation in modulating miRNA targeting effectivenessGenome Res.Year: 2010201398141020719920
27. Guo L,Li H,Liang T,Lu J,Yang Q,Ge Q,Lu Z. Consistent isomiR expression patterns and 3' addition events in miRNA gene clusters and families implicate functional and evolutionary relationshipsMol. Biol. Rep.Year: 2012396699670622391651
28. Humphreys DT,Hynes CJ,Patel HR,Wei GH,Cannon L,Fatkin D,Suter CM,Clancy JL,Preiss T. Complexity of murine cardiomyocyte miRNA biogenesis, sequence variant expression and functionPLoS OneYear: 20127e3093322319597
29. Joyce CE,Zhou X,Xia J,Ryan C,Thrash B,Menter A,Zhang W,Bowcock AM. Deep sequencing of small RNAs from human skin reveals major alterations in the psoriasis miRNAomeHum. Mol. Genet.Year: 2011204025404021807764
30. Li SC,Liao YL,Ho MR,Tsai KW,Lai CH,Lin WC. miRNA arm selection and isomiR distribution in gastric cancerBMC GenomicsYear: 201213Suppl 1S1322369582
31. Watahiki A,Wang Y,Morris J,Dennis K,O'Dwyer HM,Gleave M,Gout PW. MicroRNAs associated with metastatic prostate cancerPLoS OneYear: 20116e2495021980368
32. Cloonan N,Wani S,Xu Q,Gu J,Lea K,Heater S,Barbacioru C,Steptoe AL,Martin HC,Nourbakhsh E,et al. MicroRNAs and their isomiRs function cooperatively to target common biological pathwaysGenome Biol.Year: 201112R12622208850
33. Dong C. Diversification of T-helper-cell lineages: finding the family root of IL-17-producing cellsNat. Rev. Immunol.Year: 2006632933316557264
34. Siomi MC,Siomi H. Characterization of endogenous human Argonautes and their miRNA partners in RNA silencingNucleic Acids Symp. Ser.Year: 2004525960
35. Frank F,Sonenberg N,Nagar B. Structural basis for 5′-nucleotide base-specific recognition of guide RNA by human AGO2NatureYear: 201046581882220505670
36. Lau NC,Lim LP,Weinstein EG,Bartel DP. An abundant class of tiny RNAs with probable regulatory roles in Caenorhabditis elegansScienceYear: 200129485886211679671
37. Michel U,Stringaris AK,Nau R,Rieckmann P. Differential expression of sense and antisense transcripts of the mitochondrial DNA region coding for ATPase 6 in fetal and adult porcine brain: identification of novel unusually assembled mitochondrial RNAsBiochem. Biophys. Res. Commun.Year: 200027117018010777698
38. Smalheiser NR,Lugli G,Thimmapuram J,Cook EH,Larson J. Mitochondrial small RNAs that are up-regulated in hippocampus during olfactory discrimination training in miceMitochondrionYear: 20111199499521925291
39. Burzio VA,Villota C,Villegas J,Landerer E,Boccardo E,Villa LL,Martinez R,Lopez C,Gaete F,Toro V,et al. Expression of a family of noncoding mitochondrial RNAs distinguishes normal from cancer cellsProc. Natl Acad. Sci. USAYear: 20091069430943419470459
40. Crimi M,Papadimitriou A,Galbiati S,Palamidou P,Fortunato F,Bordoni A,Papandreou U,Papadimitriou D,Hadjigeorgiou GM,Drogari E,et al. A new mitochondrial DNA mutation in ND3 gene causing severe Leigh syndrome with early lethalityPediatr. Res.Year: 20045584284614764913
41. Johns DR,Neufeld MJ,Park RD. An ND-6 mitochondrial DNA mutation associated with Leber hereditary optic neuropathyBiochem. Biophys. Res. Commun.Year: 1992187155115571417830
42. Lertrit P,Noer AS,Jean-Francois MJ,Kapsa R,Dennett X,Thyagarajan D,Lethlean K,Byrne E,Marzuki S. A new disease-related mutation for mitochondrial encephalopathy lactic acidosis and strokelike episodes (MELAS) syndrome affects the ND4 subunit of the respiratory complex IAm. J. Hum. Genet.Year: 1992514574681323207
43. Lin FH,Lin R,Wisniewski HM,Hwang YW,Grundke-Iqbal I,Healy-Louie G,Iqbal K. Detection of point mutations in codon 331 of mitochondrial NADH dehydrogenase subunit 2 in Alzheimer's brainsBiochem. Biophys. Res. Commun.Year: 19921822382461370613
44. Okamura K,Chung WJ,Lai EC. The long and short of inverted repeat genes in animals: microRNAs, mirtrons and hairpin RNAsCell CycleYear: 200872840284518769156


[Figure ID: gks1098-F1]
Figure 1. 

AGO1 and AGO2 bind predominantly to miRNAs from T helper cells. (A) Quantitative RT-PCR of miRNA let-7f from the co-immunoprecipitated small-RNA fraction of mouse splenic CD4+T cells with data normalized first to U6RNA within each group and then to IgG controls. The data were analyzed by ΔΔCt method and −ΔΔCt values are shown. n = 3, *P < 0.05. (B) Size distribution of RNA sequences binding to AGO1, AGO2 and control IgG plotted as a percent of total reads. (C) Percentage of mapped reads plotted as a function of transcript types. (D) Piecharts showing the distribution of average percentages of various types of miRNAs in AGO samples. Legend includes the top 10 most highly expressed miRNAs.

[Figure ID: gks1098-F2]
Figure 2. 

Distribution of isomiR types and 3′ modifications of primary isomiRs. (A) Relative abundance of isomiRs after filtering for low copy number and spurious reads. The legend indicates the top 10 miRNAs with the most abundant isomiRs. (B) Representative 3′-end modifications of the most abundant isomiRs, including extensions, deletions and substitutions.

[Figure ID: gks1098-F3]
Figure 3. 

Differential expression of isomiRs and associated 3′ terminal nucleotides between AGO1 and AGO2. (A) Smearplot showing the differential expression of isomiRs between AGO1 and AGO2 samples. Reads that are at least 1.5 log-fold different with adjusted P-value < 0.05 are indicated in red. (B) Heatmap and hierarchial clustering of the differentially expressed isomiRs indicating clearly distinct AGO1 and AGO2 groups. Cyan colored lines depict the deviation of the absolute count value for each isomiR from the mean count. The color key index plots the normalized value against the isomiRs counts having the same value. (C) 3′-end nucleotides of differentially expressed isomiRs between AGO1 and AGO2 samples. The four types of nucleotides on the x-axis are plotted against the percentage of isomiRs ending with the corresponding nucleotide in all the differentially expressed isomiRs with at least 1.5 log-fold change. Chi-squared goodness-of-fit test was performed with expected frequencies of 3′-end nucleotides calculated from all the 6140 isomiRs. Chi-squared test, AGO1 P < 0.0001, DF = 3 and AGO2 P = 0.01, DF = 3. (D) 5′-end nucleotides of differentially expressed isomiRs between AGO1 and AGO2 samples. The four types of nucleotides on the x-axis are plotted against the percentage of isomiRs ending with the corresponding nucleotide in all the differentially expressed isomiRs with at least 1.5 log-fold change. Chi-squared test, not significant.

[Figure ID: gks1098-F4]
Figure 4. 

Differential expression of the most abundant isomiRs. (A) Log-fold change in expression comparing AGO2- and AGO1-associated isomiRs. A positive fold change indicates that the isomiR is binding to the AGO2; negative to AGO1. Only reads with at least 1.5-fold difference and adjusted P-values <0.05 were considered statistically significant. (B) Heatmap and hierarchial clustering analysis of the same isomiRs. Cyan lines depict the deviation of the count value for each isomiR from the mean count. The color key index plots the normalized value against the isomiR counts having the same value.

[Figure ID: gks1098-F5]
Figure 5. 

Representative AGO-associated cis-NATs. (A) Representative reads mapping sense and antisense to the same genomic locus encoding the Mt-Tp tRNA gene. Reads shown in dark shading align sense with respect to the genomic locus; light shading, antisense. Multiple alignments with various vertebrate species are also shown. The annotation of the reads follows length_copy number format. (B) Representative reads mapping sense and antisense to the mitochondrial ‘NADH-Dehydrogenase 6’ locus.

[Figure ID: gks1098-F6]
Figure 6. 

Hairpin precursors of candidate endogenous siRNAs and the representative genome alignment. (A) Folding diagrams of candidate endogenous siRNA precursors. Except for Spata6, the junction of stem and loop is constituted by A = U base pairing suitable for Dicer cleavage. The color key index depicts 0 as lowest base pair probability and 1 highest. Underscored regions indicate most likely mature sequences. (B) Alignment of the candidate endo-siRNAs derived from Vps13a introns. The same reads that map in sense orientation to the genome map antisense to a very proximal locus. Alignments for various species are also shown.

[Figure ID: gks1098-F7]
Figure 7. 

Features of a candidate novel miRNA. miRNA Candidate_21 genomic coordinates, mature sequence, seed region, species conservation of the seed region, MFE value after folding, multiple alignments from other species and the precursor RNA structure are shown. The gray highlighted region of the precursor RNA is the seed sequence.

[TableWrap ID: gks1098-T1] Table 1. 

GO enrichment of the targets of the same isomiRs

gks1098-TF1‘Functional annotation tools’ feature of DAVID bioinformatics program was used to enrich the targets. The top clusters with highest enrichment score in both groups are shown. The fourth column depicts the count percentage of the total number of targets in each category. SP_PIR_KEYWORDS: keywords of the Swiss-Prot Protein Information Resource.

[TableWrap ID: gks1098-T2] Table 2. 

Genomic loci of mitochondrial cis-NATs

Chromosome locus Possible source of sense Possible source of antisense Figure
chrM:3772-3842 AK138995 Mt-Tq-201 (tRNA) Supplementary Figure S3A
chrM:3914-4951 NADH Dehydrogenase 2 AK501316 Supplementary Figure S3B
chrM:9459–9806 NADH Dehydrogenase 3 AK141612 Supplementary Figure S3C
chrM:10 167–11 544 NADH Dehydrogenase 4 AA467310 or AA433514 or CB597336 Supplementary Figure S3D
chrM:11 742–13 565 NADH Dehydrogenase 5 AK138272 or CB597336 Supplementary Figure S3E
chrM:14 145–15 288 Cytochrome b BF785953 Supplementary Figure S3F
chrM:15 356–15 422 AK154943 or AK131592 Mt-Tp-201 (tRNA) Figure 5A
chrM:13 552–14 070 AK140457 or BC055066 or BC081549 or AK142712 or AK154943 or AK159184 NADH Dehydrogenase 6 Figure 5B

gks1098-TF2Sense and antisense pairs located in eight mitochondrial genomic loci and the coordinates and possible source of the reads are shown.

Article Categories:
  • RNA

Previous Document:  OikoBase: a genomics and developmental transcriptomics resource for the urochordate Oikopleura dioic...
Next Document:  Extensive white matter hyperintensities may increase brain volume in cerebral autosomal-dominant art...