GENOME-WIDE IDENTIFICATION OF LONG NON-CODING RNAS IN RAT MODELS OF CARDIOVASCULAR AND RENAL DISEASE

Abstract

Long noncoding RNAs (lncRNAs) are an emerging class of genomic regulatory molecules reported in various species. In the rat, which is one of the major mammalian model organisms, discovery of lncRNAs on a genome-wide scale is lagging. Renal LncRNA sequencing and lncRNA transcriptome analysis was conducted in three rat strains that are widely used in cardiovascular and renal research, the Dahl salt-sensitive (S) rat, the Spontaneously Hypertensive Rat (SHR) and the Dahl salt-resistant (R) rat. Through the RNA sequencing approach, 3,273 transcripts were identified as rat lncRNAs. A majority of lncRNAs were without predicted target genes. Differential expression of 273 and 749 lncRNAs was detected between S versus R and S versus SHR comparisons respectively. To couple the observed differential expression of lncRNAs with the status of mRNAs, an mRNA transcriptome analysis was conducted. Several cis mRNA genes were co-regulated with lncRNAs. Of these, the protein expression status of four target genes, Asb3, Chac2, Pex11b and Sp5, were differentially expressed between the relevant strain comparisons thereby suggesting that the differentially expressed lncRNAs associated with these genes are candidate genetic determinants of blood pressure. This study serves as a first-generation catalog of rat lncRNAs and illustrates the prioritization of lncRNAs as positional candidates for complex polygenic traits.

Introduction

Evidence from human genome-wide association studies (GWAS) point to a large number of disease associated variations to exist outside of protein-coding genes^1–4. Mapping studies combined with whole-genome sequencing data using rat models of cardiovascular and renal disease are beginning to provide evidence for the genetic basis for inheritance of traits such as blood pressure and renal disease to be exerted by non-protein coding elements^5–7. These converging lines of evidence across species raise the possibility for non-coding RNA as a potential class of molecules that may causally contribute to cardiovascular disease either through genetic or epigenetic mechanisms^8–14.

Long non-coding RNAs (LncRNAs) are non-protein coding transcripts longer than 200 nucleotides¹⁵, which is an arbitrary cut-off value for LncRNAs to be distinguished from small regulatory RNAs such as microRNAs, short interfering RNAs, Piwi-interacting RNAs, small nucleolar RNAs, and other short RNAs¹⁶. LncRNAs are a relatively new class of RNA molecules that are less well characterized than microRNAs. Although only a small number of functional lncRNAs have been well characterized to date, they have been demonstrated to control every level of the multi-level regulated gene expression pathway¹⁷. For example, they have long been implicated in post-transcriptional gene regulation through controlling processes such as RNA maturation and transport, protein synthesis, and transcriptional gene silencing through regulating the chromatin structure^{18, 19}. Some of the lncRNAs converge on the tertiary structure of chromatin to silence multiple genes located on the overlapping and non-overlapping sides. They also interact with DNA and/or chromatin-modifying proteins and recruit them to their target regions. The exact physical association between these lncRNAs and chromatin modifiers and/or gene promoter chromatin remains to be delineated²⁰. Given the large number of lncRNAs whose functions are still to be elucidated, there is clear potential for widespread regulation of chromatin modifications and gene expression. Thus, lncRNA are dynamic in their mechanisms of action, operating through cis or trans mechanisms to influence transcription, splicing and/or translation^{3, 21, 22}. LncRNAs are emerging as important regulators of a variety of cellular responses, developmental and disease processes^23–31. For a deeper understanding and appreciation of the extent to which LncRNAs influence pathophysiology of cardiovascular and renal diseases in a given mammalian species, comprehensive knowledge regarding their existence on a genome-wide scale is an essential pre-requisite. While such information is increasingly becoming available for human and mouse genomes, there are none reported to date for rats, which are important model organisms for cardiovascular and renal disease traits. To begin addressing this limitation, in this study we sought to catalog rat lncRNAs on the rat genome and to identify novel lncRNAs that may contribute to the differences in cardiovascular and renal phenotypes exhibited by rat models. We employed a dual approach of RNA sequencing and LncRNA transcriptome profiling of three inbred rats strains.

The data presented from the RNA sequencing of three strains collectively map 3273 lncRNAs to the rat genome with evidence for intergenic lncRNAs as the largest class of expressed rat lncRNAs. Given that the rat genome is much less annotated than the human or mouse, our data provide novel information on the location and expression status of novel lncRNAs and thereby insights into this invaluable animal model. Differentially expressed lncRNA associated genes between strains revealed enrichment of genes within novel candidate pathways for cardiovascular and renal diseases, which include several immune processes such as the Jak-STAT signaling pathway and the antigen processing and presentation pathway. Further, several protein-coding genes prioritized for human hypertension by GWAS were discovered to be associated with differentially expressed novel lncRNAs and the protein products of the putative targets of these lncRNAs were also differentially expressed between the various rat models. These data lend support to the cross-species convergence of shared molecular mechanisms that are likely influenced by lncRNAs underlying quantitative trait loci in both species.

Methods

RNA Isolation

All animals were from our colony at the University of Toledo College of Medicine and Life Sciences. The SHR strain was originally obtained from Harlan Sprague Dawley (Indianapolis, IN). All Animal breeding and experiments were conducted as per the University of Toledo IACUC approved protocols. Rats were maintained on a 0.3% NaCl containing diet (TD 7034, Harlan Teklad, Madison, WI, USA). RNA was isolated from the kidneys of concomitantly raised, 53 day old male Dahl S rats, Dahl R rats and SHR (n=3/group) using TRIzol and further purified using RNeasy kits (Qiagen). RNA samples were quantitatively analyzed using a NanoDrop ND-1000 and gel electrophoresis prior to dispatch to ArrayStar Inc. (Rockville, MD, USA. www.arraystar.com). Services provided by ArrayStar were used for both RNA sequencing and lncRNA profiling.

Next-Generation Sequencing (NGS) Library Preparation

NGS library was prepared by ArrayStar, Inc as described in the Illumina’s mRNA-Seq sample preparation guide with minor modifications. The quality of all RNA samples from each strain was determined using a bio-analyzer before preparation of the sequencing library. Briefly, rRNA was depleted from 1μg total RNA samples using rRNA removal kit (Ribo-Zero™ rRNA Removal Kit, Epicentre) following the manufacturer’s directions. The rRNA-depleted RNA was then randomly primed with hexamers and reverse-transcribed using the Super Script II cDNA Synthesis Kit (Invitrogen). The second strand was synthesized using DNA polymerase I (Invitrogen). Double-stranded cDNA was fragmented and then blunt-ended with T4 DNA polymerase, Klenow polymerase and T4 PNK. Klenow (exo minus) polymerase was used to add a dA base to the 3′ end of the fragmented ds cDNA. The cDNA was ligated with Illumina adapter oligo mix using Rapid T4 DNA Ligase (Enzymatics). PCR amplification was performed to enrich the ligated fragments using Phusion High Fidelity PCR Master Mix with HF Buffer (Finnzymes Oy). The final product was purified by polyacrylamide gel electrophoresis and ~200–300 bp fragments were selected at the gel size fractionation step.

Nextgen Sequencing

The DNA fragments in the libraries were denatured with 0.1M NaOH to generate single-stranded DNA molecules, and loaded onto channels of the flow cell at 8 pM concentration, amplified in situ using Standard Cluster Generation Kit v4 (#GD-103-4001, Illumina). Sequencing was carried out by running 36 cycles using the Illumina Genome Analyzer IIx according to the manufacturer’s instructions.

The sequencing coverage for all strains were tabulated in the sequencing data summary as follows:

	Dahl_R	SHR	Dahl_S
Pass Solexa CHASTITY Quality Filter	26,665,681	29,353,693	25,088,878
Trimmed 3′-Adaptor	26,355,960	29,079,904	24,418,891
Aligned Reads (<= 2 Mismatches)	17,487,378 (66.35%)	22,242,967 (76.49%)	13,943,565 (57.10%)
Multi-Aligned Reads	3,514,549 (13.33%)	4,576,266 (15.74%)	2,794,800 (11.45%)
Uniquely Aligned Reads	13,972,829 (53.02%)	17,666,701 (60.75%)	11,148,765 (45.66%)

Nextgen Sequencing data analysis

Image analysis and base calling was performed using Solexa pipeline V1.8 (Off-Line Base Caller software, version 1.8) and short reads (pass Solexa CHASTITY quality filter) were aligned to the rat reference genome (Rat Genome RGSC 3.4). The reads aligned to unique location in the reference genome with no more than 2 mismatches were considered as uniquely aligned reads. The sequencing image analysis and base calling were performed using Solexa pipeline version 1.8 (Off-Line Base Caller software, version 1.8) and trimmed short reads (pass Solexa CHASTITY quality filter, trimmed 3′-adaptor bases) were aligned to Rat reference genome (UCSC build RN5) and Rat transcriptome (Ensembl 70) using Tophat2 software (version 2.0.6). The sequencing length was 36 nt. The gene & transcript expression level (FPKM value) and significant changes in gene & transcript expression were then calculated using Cufflinks2 software (version 2.0.2). Differentially alternative splicing genes were identified using MATS software (version 3.0.6.beta), changes in isoform ratios were detected among different samples. cSNP variants in transcriptome were identified and annotated using VarScan2 software (version 2.3.4) and snpEff software (version 3.1m). mRNA targets of differential LncRNA genes were detected using blast. The thresholds for differentially expressed genes were fold change >= 2.0 for up-regulation or <= 0.5 for down-regulation). One-sided t-tests based on the asymptotics of the Jensen-Shannon metric were calculated as implemented in the software. Type I errors were controlled with the Benjamini-Hochberg correction procedure for multiple t-testing on the genes and reported as q-values. The q-value cutoff was 0.2. Mapped locations of LncRNAs were batch processed by entering into RatMine v1.2013 (http://ratmine.mcw.edu/ratmine/genomicRegionSearch.do) to acquire rat QTL locations corresponding with the rat lncRNA locations.

LncRNA Rat Chip

The lncRNA array (ArrayStar, www.arraystar.com) consisted of approximately 9,000 rat lncRNA candidates that were identified from the most authoritative databases, such as NCBI RefSeq, UCSC all_mrna records and orthologs of mouse lncRNAs. Highly similar sequences and ncRNAs shorter than 200 basepairs were excluded. Probes for housekeeping genes and negative probes were printed multiple times to ensure hybridization quality. ArrayStar designed the lncRNA ChIP (GEO platform accession number: GPL15690), taking into account not only the transcripts already in the database, but also with the aim of predicting new transcripts. ‘Potential lncRNA’ probes were designed to aid discovery of new lncRNAs.

RNA labeling and array hybridization

Sample labeling and array hybridization were performed according to the Agilent One-Color Microarray-Based Gene Expression Analysis protocol (Agilent Technologies, Inc., Santa Clara, CA, USA). Briefly, 1 μg of total RNA from each sample was linearly amplified and labeled with Cy3-dCTP. The labeled cRNAs were purified using the RNeasy mini kit (Qiagen). The concentration and specific activity of the labeled cRNAs (pmol Cy3/μg cRNA) were measured using a NanoDrop ND-1000 spectrophotometer. A total of 1 μg of each labeled cRNA was fragmented by the addition of 11 μl 10X blocking agent and 2.2 μl of 25X fragmentation buffer, then the mixture was heated at 60°C for 30 min; finally, 55 μl 2X GE hybridization buffer were added to dilute the labeled cRNA. Hybridization solution (100 μl) was dispensed into the gasket slide and assembled to the lncRNA expression microarray slide. The slides were incubated for 17 h at 65°C in an Agilent hybridization oven. The hybridized arrays were washed, fixed and scanned using the Agilent DNA Microarray Scanner.

Microarray data analysis

Agilent Feature Extraction software (version 10.7.3.1) was used to analyze the acquired array images. Quantile normalization and subsequent data processing were performed using the GeneSpring GX v11.5.1 software package (Agilent Technologies, Inc.). Following quantile normalization of the raw data, lncRNAs and mRNAs with at least 2 out of 2 present or marginal flags were selected for further data analysis. Differentially expressed lncRNAs and mRNAs were identified through fold change filtering using the Agilent GeneSpring GX software (version 11.5.1). Statistical significance was inferred after correcting for multiple comparisons. Hierarchical clustering and heat maps were generated using CIMminer. Gene Ontology (GO) analysis and pathway analysis were performed using a WEB-based GEne SeT AnaLysis Toolkit (http://bioinfo.vanderbilt.edu/webgestalt/), which is a suite of tools for functional enrichment analysis in various biological contexts (Nucleic Acids Res. 2005 Jul 1;33(Web Server issue):W741-8.

The statistical analysis of differentially expressed protein coding RNAs and lncRNAs (at both gene level and transcript level) were performed using Cufflinks software package (version 1.0.3) [PMID: 22383036]. The thresholds for differentially expressed genes were fold change >= 2.0 for up-regulation or <= 0.5 for down-regulation). One-sided t-tests based on the asymptotics of the Jensen-Shannon metric were calculated as implemented in the software. Type I errors were controlled with the Benjamini-Hochberg correction procedure for multiple t-testing on the genes and reported as q-values. The q-value cutoff was 0.2.

All RNA sequencing and LncRNA profiling data was deposited at the GEO database of the NIH (http://www.ncbi.nlm.nih.gov/geo/). This data can be accessed using the accession number. Microarray-GSE59165 and RNA sequencing – GSE58701 Strain comparisons of data: For the purposes of cataloging RNA sequencing data, all data obtained from individual strains are provided. For the purposes of comparing two strains, the two meaningful strain comparisons of S versus SHR and S versus R are presented.

Immunoblotting

Antibodies were from the following sources: anti-Asb3 (biorbyt), anti-Chac2 (Novus biologicals), anti-Pex11b (Novus biologicals), anti-Sp5 (Abcam), anti-Gapdh (Cell Signaling), anti-mouse IgG-HRP conjugate (Cell Signaling) and anti-rabbit IgG-HRP conjugate (Cell Signaling). Dahl S, Dahl R and SHR rat kidney Samples were homogenized in ice-cold RIPA lysis buffer with protease inhibitor cocktail (Pierce) and Proteins were quantitated using the Pierce BCA protein assay kit. In brief, forty microgram of proteins were boiled with Laemmli sample buffer (Bio-Rad) for 5 minutes at 95° C. Protein samples were resolved using 4–15% Criterion Tris-HCl Gel (Bio-Rad) at room temperature, and transferred on to PVDF membrane (Millipore), blocked with 5% fat free milk powder and incubated with a primary antibody overnight at 4°C, washed and incubated with a secondary antibody conjugated with horseradish peroxidase. Immunoreactivity was detected by autoradiography using the Super Signal West Pico Chemiluminescent Substrate (Pierce).

Results

LncRNA identification through RNA sequencing

Through RNA sequencing of three rat strains, the S, SHR and R, a total of 3273 rat lncRNAs were identified as expressed in rat kidneys. Overall, the extent of lncRNA expression was lower than the expression of protein-coding transcripts in all the 3 strains (Figure 1). LncRNAs unique to any strain was less than 5% of the total LncRNAs reported. A total of 186 lncRNAs are expressed in the S but not in the R rats, whereas 199 LncRNAs are expressed in the R but not in the S rats (Figure 2). Similarly, a total of 199 LncRNAs are expressed in the S but not in the SHR rats, whereas 208 lncRNAs are expressed in the SHR but not in the S rat (Figure 2). The chromosomal locations of lncRNA profiles obtained from all three strains are provided in Figure 2 and Supplementary Table S1.

Figure 1.

LncRNA expression in Dahl S, Dahl R and SHR. Raw fragments per kilobase of exon per million fragments mapped (FPKM) sequence count was used to identify the total number of LncRNAs expressed in each strain and unique LncRNAs in each strain (indicated by + and − symbols for strains that detected a particular lncRNA or not, respectively). Individual LncRNA location data and FPKM values are given in Supplementary Table S1.

Figure 2.

Chromosomal locations of LncRNAs detected in each strain. Read count quantification was done by seqmonk to identify the locations of unique lncRNAs identified in a given strain. The basepair locations of all these locations are given in Supplementary Table S1.

LncRNA Expression Profiling

A larger number of rat lncRNAs was detected through lncRNA expression profiling compared to RNA sequencing. One factor for this observed difference could be because of the inherent differences between the two methods. RNA sequencing employs a stringent cut-off value that has to be applied for detecting reads as Fragments Per Kilobase Of Exon Per Million Fragments Mapped whereas lncRNA expression utilizes multiple biological samples for profiling and thereby allows for detection of hybridization to test probes. The combined transcriptome analysis of all the three strains detected a total of 7,182 lncRNAs as transcribed by the rat kidney. Similar to the observation with RNA sequencing (Figure 3), the extent of lncRNAs transcribed was lower than that of the mRNAs transcribed by the same tissue under the same conditions in all three strains tested (Figure 4). To test whether lncRNAs were differentially expressed between strains, lncRNA profiles from S, SHR and R rats were compared. Only 291 lncRNAs were differentially expressed (123 upregulated and 168 downregulated) between the S and R strains compared with the 747 lncRNAs that were differentially expressed (420 upregulated and 327 downregulated) between the S and SHR comparisons (Figure 5, Supplementary Table S2). This could be because both S and R originated from the Sprague Dawley rats whereas S and SHR originated from genetically distant, Sprague Dawley and Wistar strains, respectively. When classified based on their location with respect to the location of protein-coding genes, a majority of them belonged to the sense exon and intron overlap group followed by the intergenic group (Figure 5). Differentially expressed LncRNAs that were antisense overlaps or bidirectional were also observed, but not as abundant as the sense exon and intron overlaps and intergenic lncRNAs (Figure 5).

Figure 3.

Box plots (whiskers: 10– 90 percentile) show the distribution of mean expression levels as fragments per kilobase of exon per million fragments mapped (FPKM) for each strain. Raw sequence count FPKM data was normalized by dividing each value by the sample mean, and then taking the square root. Median expression is marked with a line. Mean expression is marked with a + symbol. Mean values for protein coding transcripts were 0.5753 (Dahl S), 0.5616 (Dahl R) and 0.5642 (SHR). For lncRNA the mean values were 0.4756 (Dahl S), 0.4801(Dahl R), and 0.4853 (SHR). Outlier peaks show expression as high as 34.09 for protein coding genes and 26.9 for LncRNAs.

Figure 4.

Box plots (whiskers: 1–99 percentile) of normalized intensities indicate the distribution of mean expression levels from the microarray experiments.

Figure 5.

Differentially expressed LncRNAs: The graph represents the total number of lncRNAs differentially expressed in the array, based on the location and direction the LncRNAs were sub-categorized in to Antisense, bidirectional, intergenic and sense exon and intron overlap.

Hierarchical clustering of lncRNA expression patterns differential between S versus R and S versus SHR are shown in panel A of Figures 6 and 7, respectively. This is not particularly useful considering that a large number of lncRNAs represented on this heat map are within intergenic regions with no relationship to target mRNA sequences. Long ncRNAs can epigenetically regulate protein-coding genes by interacting locally in cis³². Panel B in Figures 6 and 7 were therefore developed to exclude such intergenic LncRNAs and only include lncRNAs which are in close proximity or overlap with mRNA transcripts. The mRNA transcripts associated with each lncRNA is given in parenthesis by the name of the lncRNA (Figures 6 and 7, Panel B).

Figure 6.

Color coded heat maps for the differentially expressed lncRNAs and mRNAs (Dahl S v Dahl R). The normalized intensities for each differentially expressed (1.5 fold up/ down, p < 0.05) lncRNA from the array for each animal were used to create the heat map using CIMminer (one matrix) and clustering was done by Euclidean Distance method and average linkage cluster algorithm. Panel A: From the all differentially expressed lncRNA set, a first subset was created by extracting any differentially expressed lncRNA have neighboring genes. Panel B: A second subset created by extracting lncRNA differentially expressed along with mRNA genes in cis (labeled as target genes) that were also differentially expressed (1.5 fold up/ down, P<0.05). Panel C: Expanded view of Panel B with selected lncRNAs and their cis mRNAs; Panel D: Fold change of lncRNAs and their corresponding cis mRNA (labeled as associated gene) expression.

Figure 7.

Color coded heat maps for the differentially expressed lncRNAs and mRNAs (Dahl S v SHR). The normalized intensities for each differentially expressed (1.5 fold up/ down, p < 0.05) lncRNA from the array for each animal were used to create the heat map using CIMminer (one matrix) and clustering was done by Euclidean Distance method and average linkage cluster algorithm. Panel A: From the all differentially expressed lncRNA set, a first subset was created by extracting any differentially expressed lncRNA have neighboring genes. Panel B: A second subset created by extracting lncRNA differentially expressed along with mRNA genes in cis (labeled as target genes) that were also differentially expressed (1.5 fold up/ down, P<0.05). Panel C: Expanded view of Panel B with selected lncRNAs and their cis mRNAs; Panel D: Fold change of lncRNAs and their corresponding cis mRNA (labeled as associated gene) expression.

To assess the potential cis effects (‘cis’ is used to refer to the effects of non-coding RNAs on genes located nearby) of lncRNAs, we profiled mRNA gene expression in the same samples that were used for lncRNA expression analysis. The mRNA profiling data is provided in the Supplementary Table S3. We then generated heat maps of all lncRNAs that were differentially expressed alongside of potential cis mRNAs that were also differentially expressed (Figures 6 and 7, Panel C). In the S versus R comparison, 21 lncRNAs were differentially expressed and were associated with 20 differentially expressed mRNA transcripts (Figure 6, Panel C). The extent of fold changes observed are shown in Figure 6, Panel D. Except for the lncRNAs associated with the mRNA transcripts for Mx2 and Pex11b, which were inversely correlated, a direct correlation in expression was noted for all the other lncRNA-mRNA associations. Between the S and SHR rats, 37 differentially expressed lncRNAs were correlated with differential expression of their closely-associated mRNAs (Figure 7, Panel C). As shown in Figure 7, Panel D, there was a good correlation between the fold changes of each of these differentially expressed lncRNAs and the associated mRNAs. The fold changes of 24 and 12 of the lncRNAs were directly and inversely correlated, respectively, with the fold changes in their associated mRNAs.

To assess the expression status of transcripts with the extent of their protein expression, we prioritized four genes, Asb3, Chac2, Pex11b and Sp5 because collectively they represented genes that are reported for association with cardiovascular traits and encompassed several unique features. The lncRNAs associated with Asb3, Chac2 and Pex11b were >35 fold upregulated in the S compared to R while the lncRNA associated with Sp5 was >6 fold downregulated in the S compared to R. The associated mRNAs were directly correlated with the expression of mRNAs of Asb3 and Sp5 and inversely correlated with the expression of Pex11b mRNA. The mRNA status for Chac2 was unknown as there was no probe representing this gene on the array chip. The protein products of Chac2, Pex11b and Sp5, but not Asb3, were significantly differentially expressed between S and R. Interestingly, all of the protein expression patterns were inversely correlated to the lncRNA expression patterns (Figure 8).

Figure 8.

Protein expression analysis of select cis genes associated with differentially expressed lncRNAs. Whole-kidney lysates from Dahl S (n=3) and Dahl R (n=3) were probed with antibodies against each of the select proteins by immunoblotting. Fold change for each protein is shown on the right along with corresponding lncRNA and the expression level of each gene at the mRNA level from the array (**p<0.01 *** p<0.001 and NS-not significant).

Next, to examine pathways that may be altered due to the downstream effects of the differential expression of lncRNAs between strains, a web-based gene set enrichment analysis was performed with all the genes that are associated exclusively with the differentially expressed lncRNA. The pathways significantly altered between the S and R as well as S and SHR comparisons are shown in Figure 9. As expected due to the differences in the origins of the strains, unique pathways were enriched in each of the strain comparisons. Overall, pathways impacted in the S and R comparisons in order of their hypergeometric p values are as follows-Fc gamma R-mediated phagocytosis, gap junction, salivary secretion, protein digestion and absorption, Alzeimer’s disease, bile secretion, Hedgehog signaling pathway, acute myeloid leukemia, fatty acid metabolism and Jak-STAT signaling pathway (Figure 9). Similarly, the pathways impacted in the S and SHR comparisons are intestinal immune network for IgA production, cell adhesion molecules and a variety of autoimmune, infection and immunity related pathways (Figure 9).

Figure 9.

Web based gene set enrichment analysis of differentially expressed lncRNA associated genes. The LncRNA associated cis mRNA genes from each of the strain comparisons (provided in Supplementary Table S3) were used for differential gene enrichment analysis to detect alterations in physiological/pathological pathways. The top ten pathways with high enrichment value and significant hypergeometric p value are presented.

Discussion

Laboratory rats have been used for decades to model human cardiovascular, renal and metabolic phenotypes associated with disease. Genetic analyses performed using several inbred rat strains has resulted in the mapping of hundreds of quantitative trait loci on the rat genome. While mapping disease loci using this mammalian model has progressed considerably partly due to the availability of genomic sequence data, the non-availability of dense annotations for functional elements on the rat genome remains problematic and thereby poses a significant barrier to the discovery of inherited genetic elements underlying the mapped loci. This is especially a constraint because several of the loci are mapped within genomic segments that lack any known protein-coding genes and/or known functional non-coding elements on the rat genome. In the current study, we focused on a distinct group of non-coding elements, lncRNAs, and profiled renal tissues from three different inbred rat strains. We chose the three strains S, SHR and R for the main purpose of cataloging lncRNA annotations from the most widely used rat models of cardiovascular and renal disease.

We used two approaches, RNA sequencing and lncRNA profiling to further enhance the likelihood of identifying novel rat lncRNAs. The study identified distinct rat lncRNAs that are transcribed from the renal rat tissues. The primary sequences of these lncRNAs suggest that they belong to several classes of being antisense / bidirectional/ intergenic / sense exon overlap / sense intron overlap. However, both approaches suggested that the highest number of lncRNAs are transcribed from intergenic regions, which is interesting from the point of view of mapping studies and human association studies that report a large number of disease-associated variants within intergenic regions.

Long non-coding RNA expression pattern is more variable between tissues compared to the mRNA^{29, 33} and lncRNAs expression median is lower than the median mRNA level^{27, 33–35}. Since our study was conducted exclusively with the kidneys, it remains to be compared for lncRNA expression profiles in other tissues. However, we observed the median lncRNA level is lower than the mRNA of protein coding genes. This lower level of expression of lncRNAs could be either due to the lower abundance of lncRNAs in general or, as reported in other species, due to the less efficient transcription or more efficient degradation³⁶. It could also be due to lower stability because lower stability can be important for regulating lncRNA function³⁷. Among the lncRNAs identified, less than 40% were closely associated with mRNAs. This suggests that a majority of the lncRNAs identified in the rat are ‘orphans’ with no known or predicted target mRNA. Identification of these yet unknown target genes will be an important first step for ascertaining the functional significance of these newly identified rat lncRNA. Nevertheless, because their mapped locations on the rat genome are provided, our study serves as a catalog to query lncRNAs as potential positional candidates for any rat QTL. The strain comparisons presented in our study (that of S and R or S and SHR) are relevant because these strains were previously used to identify putative candidates for physiological QTLs and S and SHR rats represent two of the most widely used inbred strains for studying hypertension and renal disease. While both are genetically hypertensive strains, there are important differences between these two strains. The S and SHR were derived from two very distinct genetic allele pools of the Sprague-Dawley rats and Wistar rats, respectively³⁸. The selection process for developing the S rat, but not the SHR, included an environmental condition of a high salt containing food³⁸. The S rat develops renal disease but the SHR does not^38–40. These important differences point to differences in their pathophysiological mechanisms underlying the etiology of hypertension in these two strains. Both BP and renal disease QTLs have been identified by mapping studies comparing these two strains^39–41. We observed that 238 lncRNAs were differentially expressed between S v R and 598 lncRNAs were differentially expressed between S v SHR and these lncRNAs were located within the previously reported BPQTL regions in rats (cataloged at www.rgd.mcw.edu), suggesting that they are lncRNA expression QTL (eQTL) candidates. Significant differential expressions of their corresponding closely linked mRNA genes were observed and the protein expressions of select genes were also significantly differential. These data indicate that candidate lncRNA within QTLs may act either genetically or epigenetically to alter downstream protein availability. Because of the comparisons presented in this study, which are inbred strain comparisons, we acknowledge that whether any of the observed differences in long noncoding RNA expression patterns are functionally relevant to the strain differences in blood pressure remains to be proven with other experimental designs, such as for example, substitution mapping studies.

Pathway analysis conducted exclusively with the cis genes known to be associated with differentially expressed lncRNAs revealed interesting differences between the strain comparisons. A variety of pathways ranging from secretory mechanisms (salivary and bile secretion), fatty acid metabolism, immune function (Jak-STAT signaling pathway, Fc gamma receptor mediated phagocytosis), neurological processes (Alzheimer’s disease), protein digestion and absorption and other pathways related to cellular mechanisms (gap junction, hedgehog signaling, acute myeloid leukemia) were enriched in the S versus R comparison. The salivary secretory mechanisms are interesting and unique to the S versus R comparison and may indicate a dietary high salt driven selection during generation of the S and R rats. The differences in pathways related to protein digestion and absorption are also notable because of the evidence that epigenetic modifications due to salt-independent dietary effects may alter salt-sensitive blood pressure and renal disease phenotypes in the S rat (http://hyper.ahajournals.org/cgi/content/meeting_abstract/62/3_MeetingAbstracts/A480). The other pathways identified may represent important, yet unknown mechanisms by which non-coding RNA may regulate blood pressure and other cardiovascular and renal physiological differences between the S and R rats.

The pathways altered between the S and SHR rats were different from that of the S and R rat comparison. It is not surprising that given the origins and the major histocompatibility genes of these strains are different, almost all of the major pathways related to mechanisms influencing immunity and infection, including antigen processing and graft-versus host disease are altered between these two strains is (Figure 9).

Having identified a large set differentially expressed lncRNAs, the next challenge is to infer the putative function. Data presented on four of the lncRNA-mRNA associated genes, Ankyrin Repeat And SOCS Box Containing 3 (Asb3), cation transport regulator homolog 2 (Chac2), Peroxisomal membrane protein 11B (Pex11b) and Sp5 transcription factor (Sp5) provide evidence that differential expression of lncRNAs is associated with significant alterations in either mRNA or protein expression of their closely linked genes. Genome-wide association findings from Hypertension Genetic Epidemiology Network reported that Asb3 gene was associated with thickening of walls and increased left ventricular mass⁴². Chac2 gene is significantly associated with age-related deterioration of normal physiological function as reported through the analysis of two independent cohorts, the SardiNIA and the TwinsUK cohort studies⁴³. Similarly, in the context of the S and SHR rats also presenting with features of metabolic syndrome^44–50, it is interesting to note that Pex11b is correlated with fat metabolism, insulin and blood sugar-related traits in rats⁵¹ and Sp5, through regulation of Wnt signaling, is implicated in metabolic diseases^{52, 53}.

The genes Asb3 and Chac2 are both located within the same genome segment. These two genes are therefore potential target genes for a single lncRNA, which was differentially expressed between S and R rats. The corresponding protein expression of Chac2, but not Asb3 was down-regulated significantly in R compared with S rats. These data point to the complexity with which lncRNA molecules can influence multiple downstream molecules differentially. These interpretations are of course based on associated observations and detailed mechanistic studies will be required to confirm the observed cis effects of lncRNAs. Nevertheless, these data combined with observations of associations in human diseases further prioritize lncRNAs as candidate cardiovascular disease causative genetic elements.

Perspectives

The field of study of the genetics and epigenetics of hypertension both in mammalian model organisms and in humans is beginning to focus on genetic variants that exist outside of protein-coding genes as potential drivers of pathogenesis. The current study is one of the first to profile the poorly characterized class of non-coding RNA, called as long non-coding RNA in rat models of hypertension and renal disease. The data presented should set the stage for enquiry into long noncoding RNAs as candidates regulating the etiology of hypertension.

Novelty and Significance.

1. What Is New

   • Detection of long non-coding RNAs in Dahl S, SHR and Dahl R rat kidneys through RNA sequencing.

   • Renal lncRNA expression profiling reveals differential expression of lncRNAs between rat strains

   • Differential cis-mRNA expression with differentially expressed lncRNA molecules point to lncRNAs as candidate genetic determinants of hypertension.

2. What Is Relevant?

   Genome-wide association studies pointing to variants within intronic and intergenic regions of the human genome as being associated with hypertension could be due to variants within elements that are non-coding regulatory molecules that are not yet annotated. Similarly, genetic investigation of rat models of hypertension also point to genomic segments without protein-coding genes as blood pressure quantitative trait loci. Therefore, there is a need to annotate mammalian genomes beyond protein-coding genes.