Epigenetic coordination of embryonic hearttranscription by dynamically regulated longnoncoding RNAsScot J. Matkovich1, John R. Edwards, Tiffani C. Grossenheider, Cristina de Guzman Strong, and Gerald W. Dorn II1

Center for Pharmacogenomics, Department of Internal Medicine, Washington University School of Medicine, St. Louis, MO 63110

Edited* by Andrew R.Marks, Columbia University College of Physicians and Surgeons, New York, NY, and approved July 1, 2014 (received for review June 9, 2014)

The vast majority of mammalian DNA does not encode for proteinsbut instead is transcribed into noncoding (nc)RNAs having diverseregulatory functions. The poorly characterized subclass of longncRNAs (lncRNAs) can epigenetically regulate protein-coding genesby interacting locally in cis or distally in trans. A few reports haveimplicated specific lncRNAs in cardiac development or failure, butprecise details of lncRNAs expressed in hearts and how their expres-sion may be altered during embryonic heart development or byadult heart disease is unknown. Using comprehensive quantitativeRNA sequencing data from mouse hearts, livers, and skin cells, weidentified 321 lncRNAs present in the heart, 117 of which exhibita cardiac-enriched pattern of expression. By comparing lncRNA pro-files of normal embryonic (∼E14), normal adult, and hypertrophiedadult hearts, we defined a distinct fetal lncRNA abundance signa-ture that includes 157 lncRNAs differentially expressed comparedwith adults (fold-change ≥ 50%, false discovery rate = 0.02) andthat was only poorly recapitulated in hypertrophied hearts (17 dif-ferentially expressed lncRNAs; 13 of these observed in embryonichearts). Analysis of protein-coding mRNAs from the same samplesidentified 22 concordantly and 11 reciprocally regulated mRNAswithin 10 kb of dynamically expressed lncRNAs, and reciprocal rela-tionships of lncRNA and mRNA levels were validated for the Mccc1and Relb genes using in vitro lncRNA knockdown in C2C12 cells.Network analysis suggested a central role for lncRNAs in modulat-ing NFκB- and CREB1-regulated genes during embryonic heartgrowth and identified multiple mRNAs within these pathways thatare also regulated, but independently of lncRNAs.

fetal heart | pressure overload

One of the revelations from sequencing whole genomes andthe Encyclopedia of DNA Elements project is the smallproportion of the mammalian genome dedicated to protein-coding genes. The majority of genomic DNA encode regulatorynoncoding (nc)RNAs, i.e., transcripts that instead of simply actingas templates for protein translation exert their own intrinsicfunctions. MicroRNAs are the best-studied subclass of ncRNAs,being dynamically regulated small (∼20 nt) single-stranded RNAsthat, in the heart, are recognized as central orchestrators of cardiacdevelopment and stress adaptation. MicroRNAs control entirebiological pathways by targeting multiple mRNAs involved in cellgrowth, differentiation, and apoptosis by suppressing the trans-lation of central protein effectors (1). By contrast, long ncRNAs(lncRNAs) of 200–2,000 nt or larger are distinguished by a diversityof molecular functioning derived from their ability to fold intocomplex structures and act as scaffolds for protein-protein inter-actions and/or chaperones that direct protein complexes to specificRNA or DNA sequences (2). Important roles for some lncRNAsare emerging in heart development (3, 4) and have been suggestedin experimental and human heart failure (5, 6). However, in-terpretation and broader application of these early findings isconstrained by uncertainty as to how lncRNAs are regulated indifferent cardiac developmental and disease states and whetherregulated lncRNAs differ between these states. Indeed, it is not yet

known with certainty which lncRNAs are expressed in mousehearts, nor have the identities of lncRNAs exhibiting “cardiac-enriched” expression been defined. To address this deficit, weapplied comprehensive next-generation sequencing and advancedcomputational approaches to identify cardiac-expressed and car-diac-specific lncRNAs, defining cardiac lncRNA expression sig-natures of late embryonic, normal adult, and hemodynamicallystressed adult hearts. Building on this foundation, we used bio-informatic analysis to integrate expression profiles and genomiclocations of dynamically regulated lncRNAs and mRNAs, identi-fying and biologically validating cardiac mRNAs whose expressionin the developing embryonic heart appears to be directed in part byregulated lncRNAs.

ResultsDelineation of Cardiac-Expressed and Cardiac-Enriched lncRNAs. Asa first step to defining mouse cardiac lncRNAs, we interrogatedarchived raw deepRNA sequencing data from n= 25 normal adultFVB/N mouse hearts (age, 8–16 wk) (7–11) and compared theseresults to RNA sequencing data from n = 7 mouse livers and n = 6independent cultures of primary mouse keratinocytes (skin cells).Noncode v3.0 lists ∼37,000 potential lncRNAs in the mouse ge-nome, but these predictions are based largely on unvalidatedFANTOM3 cDNAs (12). Therefore, we developed a curated list

Significance

The role of noncoding RNAs in mammalian biology is of greatinterest, especially since the Encyclopedia of DNA Elementsresults were published. We and others have studiedmicroRNAs in the heart, but little is known about their largercousins, long noncoding RNAs (lncRNAs). Here, we used ge-nome-wide sequencing and improved bioinformatics to quan-tify lncRNA expression in mouse hearts, define a subset ofcardiac-specific lncRNAs, and measure dynamic lncRNA regu-lation during the transition between embryo and adult, and inthe adult heart after experimental pressure overload (a modelresembling human hypertensive cardiomyopathy). We linkedspecific regulated lncRNAs to cardiac-expressed mRNAs thatthey target and, through network analyses, discovered a broaderrole of regulated cardiac lncRNAs as modulators of key cardiactranscriptional pathways.

Author contributions: S.J.M. and G.W.D. designed research; S.J.M., T.C.G., and C.d.G.S.performed research; J.R.E. contributed new reagents/analytic tools; S.J.M. and G.W.D.analyzed data; and S.J.M. and G.W.D. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: The data reported in this paper have been deposited in the Gene ExpressionOmnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession nos. GSE58453, GSE58455,and GSE56890) and Sequence Read Archive (SRA) database (accession no. SRP026654).1To whom correspondence may be addressed. Email: smatkovi@dom.wustl.edu or gdorn@dom.wustl.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental.

12264–12269 | PNAS | August 19, 2014 | vol. 111 | no. 33 www.pnas.org/cgi/doi/10.1073/pnas.1410622111

Dow

nloa

ded

by g

uest

on

June

7, 2

021

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1410622111&domain=pdfhttp://www.ncbi.nlm.nih.gov/geohttp://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE58453http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE58455http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE56890http://www.ncbi.nlm.nih.gov/sra/SRP026654mailto:smatkovi@dom.wustl.edumailto:gdorn@dom.wustl.edumailto:gdorn@dom.wustl.eduhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1410622111

of 2,997 mouse lncRNAs by combining the annotated lncRNAsfrom Noncode 2.0, lncRNAdb, Scripture, fRNAdb, Ensembl,RefSeq, and the UCSC Genome database, but eliminatedsequences that overlapped with known mRNA exons, leaving2,140 mouse lncRNAs (SI Methods and Dataset S1). Of these, wedetected 736 lncRNAs (∼33% of the annotated list) in at leasthalf of any of the three tissue samples (Dataset S2); lncRNAscomprised 0.3–0.7%of the sequencing readsmapped to transcribedRNAs, which include all defined mRNAs together with the 2,140defined lncRNAs. Both principal components analysis (Fig. 1A)and unsupervised hierarchical clustering of individual lncRNAsequence read abundance (Fig. 1B and Fig. S1) revealed tissue-specific lncRNA expression profiles, consistent with previousobservations that lncRNAs exhibit greater tissue specificity inexpression profile than mRNAs (13). A total of 546 lncRNAs weredetected in the adult heart samples at levels ranging from 210RPKM (reads per kilobase of sequence per million reads mappingto transcribed RNAs) for the most abundant lncRNA, n415312, to0.006 RPKM for lncRNA n411743, one of the longer lncRNAspresent at the specified threshold of detection based on se-quencing read counts (SI Methods). Approximately 200 lncRNAsdetected in hearts were present at very low levels (0.3 in at least one cardiac subgroup (normal embryo,normal adult, adult sham-operated, or pressure overloaded). Ofthe 321 cardiac-expressed lncRNAs (>0.3 RPKM), 117 wereenriched at least threefold in adult hearts compared with liver andskin (Fig. 1 B and C), including the requisite cardiac lncRNABraveheart (n267831; 16.4 RPKM in adult hearts) (3).A total of 152 cardiac lncRNAs were expressed at levels >1

RPKM in normal adult hearts and were designated as abundantcardiac lncRNAs; these comprised at least 90% of the total cardiaclncRNA sequencing reads (Fig. S1 and Dataset S2). Forty-eight ofthese abundant cardiac lncRNAs, including Braveheart (vide su-pra), were defined as cardiac enriched. In total, our studies innormal adult mouse hearts identified 48 abundant cardiac-enriched lncRNAs and 104 abundant lncRNAs that are moregenerally expressed in liver and/or skin, as well as hearts. To de-termine the cellular fraction(s) to which these abundant cardiac

lncRNAs belong, we subjected adult mouse hearts to collagenasedigestion (SI Methods) and performed RNA sequencing assays oncardiomyocytes and fibroblasts (Dataset S3). Of the 48 abundant,cardiac-enriched lncRNAs, 32 (67%) were enriched at leastthreefold in the cardiomyocyte fraction of the heart, whereas only 2(4.2%) were enriched in the fibroblast fraction (Dataset S3).Differences in mouse strain can be a source of genetic and

transcriptional variation and a confounding factor when comparingexperimental results from two or more research groups. We com-pared lncRNA expression detected by RNA-sequencing on 25 FVB/N hearts and 16 C57BL/6 hearts, two lines with known biologicaldistinctiveness (14–17). Similar lncRNA expression signatures wereobserved for 546 detectable lncRNAs in these adult mouse hearts(Fig. S2 A and B) with only a few statistically significant differences,most of which involved lncRNAs present at relatively low levels(Fig. S2 C and D and Dataset S4). Thus, lncRNA expression islargely conserved between normal FVB/N and C57BL/6 adult hearts.

lncRNA Expression Profiles Differ in Embryonic vs. Healthy Adult MouseHearts. Cardiac-specific patterns of expression for some lncRNAssuggest that they may play a role in heart development. Indeed, thecardiac-enriched lncRNA Braveheart/n267831 reportedly playsa central regulatory role in cardiomyocyte differentiation (3). Wedetermined what other cardiac lncRNAs are differentiallyexpressed in developing embryonic hearts at ∼E13.5 (range, E13–E14.5), a stage when progenitor cell commitment/differentiation islargely complete and myocardial growth through cardiomyocyteproliferation is active. Because we were interested in identifyingthose lncRNAs with possible roles in cardiac development, wefocused our analyses on lncRNAs present at >0.3 RPKM.Unsupervised clustering of the raw expression (RPM) values

of 321 cardiac-expressed lncRNAs perfectly segregated embry-onic from adult hearts (Fig. 2A), identifying a distinct lncRNAsignature for developing hearts. A total of 157 lncRNAs (ap-proximately half of cardiac-expressed lncRNAs) exhibited sig-nificant differences [false discovery rate (FDR) = 0.02, >50%increase or decrease] in expression between embryonic and adult

Fig. 1. Tissue selective patterns of mouse lncRNA expression. (A) Principalcomponents analyses of n = 25 adult mouse hearts, n = 7 adult mouse livers,and n = 6 cultured mouse keratinocytes (skin). (B) Heat map display forunsupervised hierarchical clustering of fold-change in expression betweenthe three mouse tissues [reads per million lncRNA aligned reads (RPM)]. (C)Venn diagram revealing patterns of tissue-selective lncRNAs expressed at>0.3 RPKM (reads per kilobase of RNA length per million reads mapped to alltranscribed RNA, where all transcribed RNA comprises reads mapped tomRNAs and 2,140 defined lncRNAs).

Fig. 2. lncRNA expression is regulated during embryonic development butnot after TAC. (A–C) Unsupervised cluster analysis of normal adult and ∼E14C57BL/6 embryonic heart lncRNA expression. (A) Three hundred twenty-onelncRNAs expressed >0.3 RPKM in hearts. (B) Fifty-two cardiac-selective abun-dant (>1 RPKM) lncRNAs. (C) One hundred five non–cardiac-selective abun-dant lncRNAs. (D–F) Same lncRNAs as displayed in A–C, showing data fromsham-operated vs. pressure overloaded (TAC) adult hearts, 1 wk, and 4 wk.

Matkovich et al. PNAS | August 19, 2014 | vol. 111 | no. 33 | 12265

SYST

EMSBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

7, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd02.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd02.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd02.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd02.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd03.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd03.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd04.xlsx

mouse hearts (Fig. 2 B and C and Dataset S5), elucidatinga “fetal gene program” for cardiac lncRNAs analogous to thatpreviously described for cardiac mRNAs (18).

lncRNA and mRNA Regulation in Early and Late Cardiac PressureOverload. We asked if adult cardiac disease induced fetal-likechanges in the lncRNA expression signature. RNA sequencingwas performed on four pairs of sham-operated hearts and heartsof mice 1 and 4 wk after induction of acute pressure overload bymicrosurgical subtotal ligation of the transverse aorta [transverseaortic coarctation (TAC)], which evokes changes in cardiacmRNA and microRNA expression also seen in clinical heartfailure (8). In accordance with our previous studies (8, 19), weobserved hypertrophy without severe deterioration of cardiacfunction at both of these time points (Fig. S3). Eleven lncRNAswere significantly regulated 1 wk after TAC (Fig. 2 D–F), withsimilar trends after 4 wk (Fig. 2 D–F and Dataset S6).Compared with sham-operated controls, only 17 cardiac-

expressed lncRNAs were expressed at different levels after 1 or4 wk of TAC (Fig. 3A and Dataset S6), despite the usual alter-ations in fetal gene mRNAs (Fig. 3B). Thirteen of these 17lncRNAs also showed differences in expression between embry-onic and adult hearts. Thus, dynamic expression of lncRNAs wasmore prominent in cardiac growth when transitioning from em-bryo to adult (157 lncRNAs) than in the hypertrophic response toincreased hemodynamic stress (17 lncRNAs). Comparison of theexpression signatures for 321 cardiac-expressed lncRNAs (re-gardless of statistical categorization) across normal E14.5 embry-onic, normal adult, sham operated adult, and pressure overloadedadult (TAC) hearts supports the conclusion that lncRNA ex-pression markedly differs between fetal and adult hearts but is

similar in healthy and hypertrophied adult hearts (Fig. 3 C and Dand Fig. S4).Because lncRNAs exert their effects in part via epigenetic regu-

lation of gene expression (2), we asked what mRNA changes occurconcomitantly with changes in lncRNA levels. As expected, mRNAsequencing (10) of the same hearts revealed distinct mRNA sig-natures for embryonic, adult, and hemodynamically stressed hearts(Fig. 3 E and F, Fig. S4, and Dataset S7). Partial overlap of regu-lated lncRNAs in embryonic and pressure overloaded hearts (Fig.3D, Venn diagram) resembles the partial recapitulation of embry-onic heart mRNA (Fig. 3F, Venn diagram) and microRNA ex-pression signatures in diseased adult hearts (18, 20–22). Theminimal overlap of regulated lncRNAs in fetal and pressure over-loaded (TAC) hearts suggests a limited role for lncRNAs in geneticreprogramming of hemodynamically stressed adult hearts.

Coregulated cis lncRNA-mRNA Pairs in the Embryonic-to-Adult HeartTransition.Genomic location is critical to the actions of lncRNAsthat regulate neighboring genes in cis and can help uncoverfunctional lncRNA-mRNA relationships (23). To gain insightinto the consequences of lncRNAs on cardiac gene expression,we mapped regulated cardiac lncRNAs to the mouse genome,identified coding mRNAs within 10 kb, and assessed their mu-tual dynamism as a function of cardiac condition.We identified 33 lncRNA-mRNA cis pairs in which both the

lncRNA and mRNA were regulated (FDR < 0.02; >50% in-crease or decrease) in embryonic hearts (Fig. 4A). Importantly,none of these lncRNAs were significantly regulated in pressureoverloaded adult hearts. Twenty-two lncRNA-mRNA pairs (ex-cluding lncRNAs and mRNAs at overlapping genomic loci)exhibited concordant regulation between embryonic and adult

Fig. 3. Comparison of embryonic and pressure over-load heart lncRNA and mRNA signatures. (A) Un-supervised cluster analysis for 17 abundant lncRNAsregulated after 1 or 4 wk TAC. *Opposite regulation byTAC vs. embryonic state; +regulated by TAC only, notby embryonic state. (B) Analysis of fetal gene mRNAexpression in the same embryonic (E), adult (A),sham (S), and 1- and 4-wk TAC hearts (1, 4). (C andD) Combined lncRNA analysis across the four studygroups; pressure overload responses are shown onlyfor the 4-wk time point. (C) Principal componentsanalysis. (D) Standardized lncRNA expression (fold-change). (E and F) Same as C and D for mRNA ex-pression. Venn diagrams to the right show patternsof overlap between fetal- and TAC-regulated lncRNAs(Upper) and mRNAs (Lower).

12266 | www.pnas.org/cgi/doi/10.1073/pnas.1410622111 Matkovich et al.

Dow

nloa

ded

by g

uest

on

June

7, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd05.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd06.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd06.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF4http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd07.xlsxwww.pnas.org/cgi/doi/10.1073/pnas.1410622111

hearts (Fig. S5) (24), whereas 11 lncRNA-mRNA pairs showednonconcordant regulation (Fig. 4A, asterisks), suggesting a sup-pressive function of these lncRNAs (24).The lncRNAs in reciprocally regulated pairs comprised mem-

bers of natural antisense transcript, intronic, and intergenic families(Fig. S6), including Kcnq1ot1/n413804 that is encoded on the op-posite strand of, and has been validated as regulating, the Kcnq1gene in embryonic hearts (Fig. 4B) (25). As the remaining pairsrepresented novel relationships, we used gapmeR (antisense-mediated) knockdown of lncRNAs in the C2C12 mouse skeletalmyoblast line to validate the dependency of mRNA expression on

cis lncRNA levels for two: lncRNA n411949/mRNA Mccc1 andlncRNA n413445/mRNA Relb. Although C2C12 cells are oftenused in differentiation studies, we performed these experimentsin nondifferentiating C2C12 cells to take advantage of muscle-like background gene expression in cells that could nonetheless bereadily transfected. lncRNA n411949 is an example of an antisenselncRNA overlapping protein coding exons, whereas lncRNAn413445 represents an intronic lncRNA. Their cognate mRNAs,Mccc1 and Relb, were increased in response to anti-lncRNAgapmeR transfection (compared with a negative control oligo-nucleotide transfection), using gapmeRs targeted toward two

Fig. 4. Expression of 10-kb lncRNA-mRNA partners.(A) Heat map of standardized (fold-change) ex-pression for 33 coregulated lncRNA-mRNA pairs.lncRNA/mRNA is indicated at the top. Asterisks showinstances of nonconcordant regulation. (B) Quanti-tative expression of the nonconcordantly regulatedlncRNA/mRNA pair n413804/Kcnq1 (black is lncRNA;red is mRNA).

Fig. 5. Nonconcordantly regulated 10-kb lncRNA-mRNA partners in embryonic hearts. (A) (Left) Quantitative expression of the nonconcordantly regulatedlncRNA/mRNA pair n411949/Mccc1 (black is lncRNA; red is mRNA) in embryonic and adult hearts; expression in sham and 4-wk TAC hearts is shown forcomparison (differences with TAC do not meet significance criteria). (Right) Relative positions of lncRNA n411949 and mRNA Mccc1 within the genome; othertranscripts within the locus are colored blue. (B) (Left) Positions targeted by anti-lncRNA gapmers, and primer design for lncRNA (arrows) and mRNA (arrows +internal probe) qPCR detection. (Right) lncRNA and mRNA expression in C2C12 cells transfected with anti-lncRNA gapmers, relative to geometric mean ofActb, Gapdh, and Hmbs (mean ± SEM, n = 6; representative of at least two independent experiments). Non, nontransfected cells; Neg, transfected withfluorescein amidite (FAM)-labeled control gapmer; gapmeR-1 and -2, transfected with one of two different gapmeRs against the chosen lncRNA. *P < 0.05relative to negative control (Neg). (C and D) Same as A and B, but for the nonconcordantly regulated lncRNA/mRNA pair n413445/Relb.

Matkovich et al. PNAS | August 19, 2014 | vol. 111 | no. 33 | 12267

SYST

EMSBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

7, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF5http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=SF6

different positions on each lncRNA (Fig. 5). Thus, these dataconfirm these lncRNA-mRNA regulatory relationships.

lncRNA Involvement in Regulated Expression of Genes Important forCardiac Development. Two-thirds of the lncRNA-associated mRNAsin the 33 coregulated lncRNA-mRNA pairs could be assignedto functional classes relating to tissue growth and development,including DNA replication and transcription, mRNA processingand translation, and protein synthesis and transport (Table S1).Network analysis of these lncRNA-modulated mRNAs uncoveredinvolvement in major cardiac development and metabolic pathwaysand also pointed to central roles for the transcription factorsCREB1 and NFκB (of which RelB is a critical subunit). Otherregulated mRNAs in these pathways are apparently modulatedthrough conventional transcriptional mechanisms rather than bydynamically expressed cardiac lncRNAs (Fig. 6).

DiscussionThere is an explosion of interest in regulatory noncoding RNAs,especially lncRNAs. We and others have been frustrated by theabsence of basic general information on lncRNAs in the fully dif-ferentiated mammalian heart and especially by the lack of rigor-ously annotated, quantitative expression data in adult mousemodels (26–28). Here, we identified hundreds of cardiac-expressedand dozens of cardiac-enriched (not cardiac-specific) lncRNAs,providing a foundation on which we can advance our understandingof lncRNA biology in the heart. Our studies delineate greater than100 dynamically regulated lncRNAs in embryonic hearts but rela-tively few in reactive adult cardiac hypertrophy. In fact, lncRNAsregulated in hypertrophy thus exhibit only limited recapitulation offetal lncRNA expression. This observation is consistent with thenotion, supported by functional classification and network analysisof regulated lncRNAs and their co- or reciprocally regulatedmRNA partners, that a dominant function of cardiac lncRNAs isepigenetic modulation of developmental gene expression.lncRNAs use diverse molecular regulatory mechanisms to

regulate their target genes, including acting as antisense tran-scripts that directly bind mRNA or acting as chaperones thatrecruit macromolecular protein complexes to specific sequence-specified locations in the genome (29). A common theme for thelatter mechanism is chromatin remodeling that evokes long-termchanges in transcriptional activity (27). Such remodeling repre-sents chronic, rather than acute, regulation of gene expression,consistent with our observation that dynamic regulation of

lncRNAs and their partner mRNAs is more evident duringembryonic heart development than in the adult cardiac stressresponse. The abundance of lncRNAs that are an order ofmagnitude less than that of their validated and putative partnermRNAs is also consistent with the notion that at least somelncRNAs act in a nonlinear manner via recruitment of chromatinremodeling protein complexes to regulate mRNA transcription.Little is known about overall lncRNA function in the heart,

but three cardiac-expressed lncRNAs have been studied indetail. Kcnq1ot1 (n413804) mediates transcriptional silencingthrough histone methylation of the overlapping Kcnq1/Kv7.1potassium channel gene and other genes at the same genomiclocus (30–32). In our studies, Kcnq1ot1 is expressed at high levelsin embryonic hearts, but decreases by fivefold in adult hearts. ItsmRNA partner, Kcnq1, encodes the Kv7.1 slow delayed rectify-ing potassium channel that is essential for normal cardiomyocyterepolarization that terminates the action potential and car-diomyocyte contraction. Coinciding with decreased Kcnq1ot1expression, levels of Kcnq1 mRNA increase, consistent with theneed for enhanced pump function during the transition betweendeveloping embryonic and fully functioning adult hearts.Fendrr is expressed specifically in embryonic lateral mesoderm

where it regulates heart development, likely by modifying thechromatin signatures of genes encoding transcription factors thatdirect cardiomyocyte differentiation (4). We did not detect Fendrrin hearts of ∼E14.5 mouse embryos or of adult mice, consistentwith a transient role in the specification and differentiation ofmesoderm to cardiomyocytes, which is largely complete andsupplanted by cardiomyocyte proliferation at E14.5.Braveheart (Bvht) is another cardiac-expressed lncRNA that

epigenetically regulates cardiomyocyte differentiation (3). In ourstudies, Braveheart (annotated as n267831) was cardiac-enrichedapproximately threefold compared with other tissues, butexpressed in the heart at similar levels in E14.5 embryos andadults; Braveheart is also not regulated late after hemodynamicstress. Constitutive cardiac expression of Braveheart suggeststhat it may have “housekeeping” roles in adult hearts in additionto its canonical role upstream of MesP1 to stimulate and main-tain cardiomyocyte fate.Our studies identified possible functional interactions between

10 further lncRNAs and neighboring mRNAs that were re-ciprocally regulated between embryonic and adult hearts; wevalidated two of these using lncRNA knockdown in mouse C2C12skeletal myoblasts. The importance of one of these lncRNA-regulated mRNAs, Relb, was captured in our network analysispointing to a central role for NFκB in modulating altered geneexpression during cardiac growth and transition to the adultheart. lncRNA n413445 expression is high in the embryonic heartbut is quite low in adult hearts. As its levels decline, levels of RelbmRNA increase, providing increased expression of the RelBtranscription factor that is an essential component of the NFκBpathway (33). An increase of the lncRNA n411949-regulatedmRNA Mccc1 during fetal cardiac growth may be of key im-portance for sensing free leucine levels and thus the availability ofbranched-chain amino acids for anabolic signaling in muscle (34,35).Mccc1 is one among several enzymes that metabolize leucine;interestingly, the mRNA levels of at least three others in the samecatabolic pathway (Ivd, Mccc2, and Auh) increase dramaticallyduring the transition to the adult heart (Dataset S7).

Study Limitations. Our studies of embryonic hearts were designed tobe compared with pressure overloaded adult hearts and thereforefocused on a relatively late time during development when car-diomyocyte growthwas dominant andmesodermal specification anddifferentiation were largely complete. A detailed time course oflncRNA expression in earlier stage embryos will be required to un-derstand the role of lncRNA regulation in cardiomyocyte differen-tiation. Likewise, studies in postnatal hearts during the transition

Fig. 6. Signaling networks of mRNAs involved in lncRNA-mRNA relation-ships. MetaCore analysis of functional networks involving regulated cardiacmRNAs linked to regulated cardiac lncRNAs. Green arrows, up-regulation;red arrows, down-regulation; gray lines, context-dependent interaction.Blue circles are mRNAs directly linked to regulated lncRNAs, green circles aremRNAs also regulated, but not connected to lncRNAs described in our study.Note central nodes for NFκB and CREB1.

12268 | www.pnas.org/cgi/doi/10.1073/pnas.1410622111 Matkovich et al.

Dow

nloa

ded

by g

uest

on

June

7, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd07.xlsxwww.pnas.org/cgi/doi/10.1073/pnas.1410622111

from cardiomyocyte proliferation to cardiomyocyte hypertrophy areneeded to define the roles of lncRNAs during that important tran-sition. We studied pressure overloaded hearts at 1 and 4 wk aftersurgical transverse banding, i.e., early cardiomyocyte hypertrophy(1 wk) (8) and late hypertrophy with mild deterioration of heartfunction (4 wk) (19). We selected these time points to capturelncRNA regulation in either cardiac growth or the transition tocardiac failure and found relatively few alterations in lncRNAs.Nonetheless, analyses performed even earlier after surgical model-ing or late in overt heart failure may reveal additional regulatedlncRNAs that our studies did not identify. Finally, we used conven-tional RNA sequencing that does not discriminate between mRNAand lncRNAexonsencodedonopposite strandsat the samegenomiclocation. Future studies could use strand-specific RNA sequencingfrom nuclear and cytoplasmic fractions to expand the universe ofconstitutively expressed and dynamically regulated cardiac lncRNAsdescribed here.On a positive note, our studies of anti-lncRNA knockdown

and assessment of neighboring mRNA expression demonstratedfunctional relationships between cis lncRNA and mRNA pairs.This approach should be useful in future studies designed tofurther elucidate cardiac lncRNA function.

MethodsDefinition of lncRNA and mRNA Sequences. lncRNA sequences were obtainedfrom annotated lncRNA entries from Noncode 2.0, lncRNAdb, Scripture,fRNAdb, Ensembl, RefSeq, and the UCSC Genome database. Annotation detailsand genomic locations (using theNCBIM37/mm9 release of themouse genome)for 2,140 lncRNAs can be found in Dataset S1. mRNA sequences were obtainedfrom the gtf supplied with the Ensembl iGenomes 2011 (NCBIM37) index.

RNA Sequencing Library Preparation, Read Generation, and Mapping. Mouseheart, liver, and primary keratinocyte (skin) RNA sequencing libraries wereprepared from polyA-selected RNA as previously described (19, 36). Align-ment with Tophat before differential expression analysis with DESeq wasperformed as previously described (37).

Heart libraries were sequenced to a depth of 6.7 ± 0.8 × 106 (mean ± SEM)reads mapped to mRNAs and 3.0 ± 0.3 × 104 reads mapped unequivocally tolncRNAs within the same libraries. Similar sequencing depths were obtainedfor liver and skin libraries. Overall, lncRNAs comprised ∼0.4–0.7% of thetotal reads assigned to transcribed RNA.

RNA Quantitation and Differential Expression. RNA count data in heatmaps arepresented as read counts normalized for total read number only, e.g., RPM(lncRNA reads per million reads mapped exclusively to lncRNAs). Alterna-tively, RPM input data were standardized (mean of 0, SD of 1) to betterdisplay the extent of variation among sample groups for a given lncRNA. Useof raw or standardized heatmaps is denoted in figure panels. All unsupervisedclustering was performed with Euclidean distance and average linkage.

RNA expression is presented in the text, tables, and datasets as RPKM.Partek Genomics Suite 6.6 (Partek) was used for principal componentsanalyses, standardization, and heatmap generation.

For differential expression analyses, RNA read count data were not nor-malized or standardized before input into the DESeq package (38). Cutoffswere established a priori at a fold-change ≥ 50% (FDR < 0.02). Output data inthe tables and datasets are presented as RPKM for consistency with the re-mainder of the text, together with fold-change and adjusted P values (FDRs)computed by DESeq.

Further details are given in SI Methods.

ACKNOWLEDGMENTS. This work was supported by National Institutes ofHealth Grants R01 HL108943-02 (to G.W.D.) and UL1 TR000448 (to theWashington University Institute of Clinical and Translational Sciences fromthe National Center for Advancing Translational Sciences).

1. Mendell JT, Olson EN (2012) MicroRNAs in stress signaling and human disease. Cell148(6):1172–1187.

2. Batista PJ, Chang HY (2013) Long noncoding RNAs: Cellular address codes in de-velopment and disease. Cell 152(6):1298–1307.

3. Klattenhoff CA, et al. (2013) Braveheart, a long noncoding RNA required for car-diovascular lineage commitment. Cell 152(3):570–583.

4. Grote P, et al. (2013) The tissue-specific lncRNA Fendrr is an essential regulator ofheart and body wall development in the mouse. Dev Cell 24(2):206–214.

5. Lee JH, et al. (2011) Analysis of transcriptome complexity through RNA sequencing innormal and failing murine hearts. Circ Res 109(12):1332–1341.

6. Yang KC, et al. (2014) Deep RNA sequencing reveals dynamic regulation of myocardialnoncoding RNAs in failing human heart and remodeling with mechanical circulatorysupport. Circulation 129(9):1009–1021.

7. Dorn GW, 2nd (2012) Decoding the cardiac message: The 2011 Thomas W. SmithMemorial Lecture. Circ Res 110(5):755–763.

8. Hu Y, et al. (2012) Epitranscriptional orchestration of genetic reprogramming is anemergent property of stress-regulated cardiac microRNAs. Proc Natl Acad Sci USA109(48):19864–19869.

9. Matkovich SJ, Van BoovenDJ, EschenbacherWH, Dorn GW, 2nd (2011) RISC RNA sequencingfor context-specific identification of in vivo microRNA targets. Circ Res 108(1):18–26.

10. Matkovich SJ, et al. (2010) MicroRNA-133a protects against myocardial fibrosis andmodulates electrical repolarization without affecting hypertrophy in pressure-over-loaded adult hearts. Circ Res 106(1):166–175.

11. Chen Y, et al. (2013) A nucleus-targeted alternately spliced Nix/Bnip3L protein iso-form modifies nuclear factor κB (NFκB)-mediated cardiac transcription. J Biol Chem288(22):15455–15465.

12. Maeda N, et al. (2006) Transcript annotation in FANTOM3: Mouse gene catalog basedon physical cDNAs. PLoS Genet 2(4):e62.

13. Cabili MN, et al. (2011) Integrative annotation of human large intergenic noncodingRNAs reveals global properties and specific subclasses. Genes Dev 25(18):1915–1927.

14. Dansky HM, et al. (1999) Genetic background determines the extent of atherosclerosisin ApoE-deficient mice. Arterioscler Thromb Vasc Biol 19(8):1960–1968.

15. Jo J, et al. (2009) Hypertrophy and/or hyperplasia: Dynamics of adipose tissue growth.PLOS Comput Biol 5(3):e1000324.

16. Li TT, et al. (2004) Genetic variation responsible for mouse strain differences in in-tegrin alpha 2 expression is associated with altered platelet responses to collagen.Blood 103(9):3396–3402.

17. Kim DH, Gutierrez-Aguilar R, Kim HJ, Woods SC, Seeley RJ (2013) Increased adiposetissue hypoxia and capacity for angiogenesis and inflammation in young diet-sensi-tive C57 mice compared with diet-resistant FVB mice. Int J Obes (Lond) 37(6):853–860.

18. Izumo S, Nadal-Ginard B, Mahdavi V (1988) Protooncogene induction and reprogram-ming of cardiac gene expression produced by pressure overload. Proc Natl Acad Sci USA85(2):339–343.

19. Matkovich SJ, Hu Y, Eschenbacher WH, Dorn LE, Dorn GW, 2nd (2012) Direct andindirect involvement of microRNA-499 in clinical and experimental cardiomyopathy.Circ Res 111(5):521–531.

20. Frey N, Olson EN (2003) Cardiac hypertrophy: The good, the bad, and the ugly. Annu

Rev Physiol 65:45–79.21. Rajabi M, Kassiotis C, Razeghi P, Taegtmeyer H (2007) Return to the fetal gene pro-

gram protects the stressed heart: A strong hypothesis. Heart Fail Rev 12(3-4):331–343.22. Thum T, et al. (2007) MicroRNAs in the human heart: A clue to fetal gene re-

programming in heart failure. Circulation 116(3):258–267.23. Derrien T, Guigó R, Johnson R (2011) The long non-coding RNAs: A new (p)layer in the

“dark matter”. Front Genet 2:107.24. Katayama S, et al.; RIKEN Genome Exploration Research Group; Genome Science

Group (Genome Network Project Core Group); FANTOM Consortium (2005) Antisense

transcription in the mammalian transcriptome. Science 309(5740):1564–1566.25. Korostowski L, Sedlak N, Engel N (2012) The Kcnq1ot1 long non-coding RNA affects

chromatin conformation and expression of Kcnq1, but does not regulate its im-

printing in the developing heart. PLoS Genet 8(9):e1002956.26. Ounzain S, Crippa S, Pedrazzini T (2013) Small and long non-coding RNAs in cardiac

homeostasis and regeneration. Biochim Biophys Acta 1833(4):923–933.27. Papait R, Kunderfranco P, Stirparo GG, Latronico MV, Condorelli G (2013) Long

noncoding RNA: A new player of heart failure? J Cardiovasc Transl Res 6(6):876–883.28. Kataoka M, Huang ZP, Wang DZ (2013) Build a braveheart: The missing linc (RNA).

Circ Res 112(12):1532–1534.29. Wang KC, Chang HY (2011) Molecular mechanisms of long noncoding RNAs. Mol Cell

43(6):904–914.30. Mohammad F, Mondal T, Guseva N, Pandey GK, Kanduri C (2010) Kcnq1ot1 non-

coding RNA mediates transcriptional gene silencing by interacting with Dnmt1. De-

velopment 137(15):2493–2499.31. Mohammad F, et al. (2008) Kcnq1ot1/Lit1 noncoding RNA mediates transcriptional

silencing by targeting to the perinucleolar region. Mol Cell Biol 28(11):3713–3728.32. Pandey RR, et al. (2008) Kcnq1ot1 antisense noncoding RNA mediates lineage-specific

transcriptional silencing through chromatin-level regulation. Mol Cell 32(2):232–246.33. Perkins ND (2007) Integrating cell-signalling pathways with NF-kappaB and IKK

function. Nat Rev Mol Cell Biol 8(1):49–62.34. Norton LE, Layman DK (2006) Leucine regulates translation initiation of protein

synthesis in skeletal muscle after exercise. J Nutr 136(2):533S–537S.35. Suryawan A, et al. (2012) Enteral leucine supplementation increases protein synthesis

in skeletal and cardiac muscles and visceral tissues of neonatal pigs through mTORC1-

dependent pathways. Pediatr Res 71(4 Pt 1):324–331.36. Matkovich SJ, Zhang Y, Van Booven DJ, Dorn GW, 2nd (2010) Deep mRNA sequencing

for in vivo functional analysis of cardiac transcriptional regulators: Application to

Galphaq. Circ Res 106(9):1459–1467.37. Matkovich SJ, Hu Y, Dorn GW, 2nd (2013) Regulation of cardiac microRNAs by cardiac

microRNAs. Circ Res 113(1):62–71.38. Anders S, Huber W (2010) Differential expression analysis for sequence count data.

Genome Biol 11(10):R106.

Matkovich et al. PNAS | August 19, 2014 | vol. 111 | no. 33 | 12269

SYST

EMSBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

7, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.1410622111.sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1410622111/-/DCSupplemental/pnas.201410622SI.pdf?targetid=nameddest=STXT