790

Cardiac fibrosis because of collagen deposition increases with age and in disease. It is a prominent feature of sev-

eral cardiomyopathies, such as hypertensive cardiac hyper-trophy, the inherited cardiomyopathies, and cardiomyopathy following myocardial infarction. Cardiac fibrosis has also been associated with primary cardiac electric disease, namely sodium channelopathy caused by mutation in the SCN5A gene (encoding the major sodium channel isoform in heart), both in patient case reports1,2 as well as in Scn5a mouse models.3,4

Clinical Perspective on p 798Cardiac fibrosis reduces cardiac contractility and function.

It also disturbs cell–cell coupling of cardiomyocytes, which in turn provides a substrate for arrhythmia. The development of drug therapies to prevent the development of cardiac fibro-sis relies on knowledge of the molecular pathways involved. Uncovering genetic factors that play a role in the development of fibrosis is likely to provide insight into such pathways. Integrative genomic approaches, such as those that integrate genome-wide genotypic data with phenotypic and gene

transcript abundance data, facilitate the identification of such genetic factors.5,6 Genetically segregating populations of mice or rats, such as filial generation 2 (F2) progeny, recombinant inbred lines, and outbred populations, offer powerful opportu-nities for such approaches because these combine phenotypic diversity with a known genetic background.7

We set out to identify genetic modifiers of collagen depo-sition in the mouse heart. We used a disease-sensitized mouse F2 progeny (129.FVB.F2-Scn5atm1Care/+) harboring the Scn5atm1Care/+ mutation (previously referred to as Scn5a1798insD/+), generated from a 129P2-Scn5atm1Care/+ × FVBN/J- Scn5atm1Care/+ cross.8 The 129P2-Scn5atm1Care/+ and FVBN/J-Scn5atm1Care/+ mouse lines display different severity of cardiac sodium chan-nelopathy9 and have been previously exploited to map quan-titative trait loci (QTLs) for electrocardiographic traits and arrhythmia susceptibility.8 Tnni3k was identified as the gene underlying one of these loci in a subsequent study.5 We here used an integrative genomics approach harnessing the genetic and phenotypic variability in the 129.FVB.F2-Scn5atm1Care/ pop-ulation to map genetic factors underlying collagen deposition

Background—With aging and in cardiac disease, fibrosis caused by collagen deposition is increased, impairing contractility and providing a substrate for arrhythmia. In this study, we set out to identify genetic modifiers of collagen deposition in the heart by exploiting the genetic variability among F2 progeny of 129P2 and FVBN/J mice carrying the Scn5atm1Care/+ mutation.

Methods and Results—Relative amounts of collagen were determined in left ventricular myocardium of 65 F2-mice and combined with genome-wide genotypic and cardiac expression data to identify collagen quantitative trait loci (QTLs) and overlapping expression QTLs (eQTLs). A significant collagen QTL was identified on mouse Chr8; an additional collagen QTL was identified on mouse Chr2 after correction for a genetic covariate uncovered on Chr18 using multiple QTL mapping. Of the 24 eQTLs colocalizing with the Chr8-collagen QTL, 6 transcripts were correlated with relative collagen amount. Similarly, of the 7 eQTLs colocalizing with the Chr2-collagen QTL, 1 transcript, Gpr158, correlated with relative collagen. Of the 12 transcripts with an eQTLs in the Chr18-covariate region, only Fgf1 correlated with relative collagen amount. Furthermore, 2 of the transcripts, Pdlim3 (Chr8) and Itga6 (Chr2), with eQTLs overlapping with collagen QTLs, had a genetic covariate on Chr18 that coincided with the Chr18 collagen covariate locus. Application of recombinant human FGF1 to isolated cardiac fibroblasts in culture affected the level of expression of Pdlim3, Itga6, and Gpr158.

Conclusions—We here mapped a possible novel genetic network modulating collagen deposition in mouse left ventricular myocardium. (Circ Cardiovasc Genet. 2014;7:790-798.)

Key Words: collagen ◼ Fgf1 ◼ fibroblast growth factor

© 2014 American Heart Association, Inc.

Circ Cardiovasc Genet is available at http://circgenetics.ahajournals.org DOI: 10.1161/CIRCGENETICS.114.000537

Received January 29, 2014; accepted July 31, 2014.From the Department of Experimental Cardiology (E.M.L., B.P.S., L.B., M.E.A., C.R.B.), and Department of Clinical Epidemiology, Biostatistics and

Bioinformatics (P.D.M., M.W.T.T.), University of Amsterdam, Academic Medical Center, Amsterdam, Netherlands; and Groningen Bioinformatics Centre, University of Groningen, Groningen, Netherlands (D.A.).

The Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.114.000537/-/DC1.Correspondence to Connie R. Bezzina, PhD, Department of Experimental Cardiology, Room L2-108, AMC Meibergdreef 9, 1105 AZ Amsterdam, The

Netherlands. E-mail c.r.bezzina@amc.uva.nl

Integrative Genomic Approach Identifies Multiple Genes Involved in Cardiac Collagen Deposition

Elisabeth M. Lodder, PhD; Brendon P. Scicluna, PhD; Leander Beekman, BSc; Danny Arends, MSc, PhD; Perry D. Moerland, PhD; Michael W.T. Tanck, PhD;

Michiel E. Adriaens, PhD; Connie R. Bezzina, PhD

Original Article

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

Lodder et al Genetics of Cardiac Collagen Deposition 791

in the heart. Using QTL and multiple QTL mapping (MQM) mapping,10 we identified QTLs for collagen deposition (colla-gen QTLs) on mouse chromosome 2 and 8. We also identified a genetic covariate for cardiac collagen deposition on mouse chromosome 18. As genetic variation underlying a QTL may affect the trait through effects on gene expression, we overlaid cardiac expression QTL (eQTL) data from the same 129.FVB.F2-Scn5atm1Care/+ cross with the 3 loci, identifying overlapping eQTLs. Cognate genes for the eQTLs at the 3 loci, represent-ing candidate causal genes, were then investigated in cultured neonatal cardiac fibroblasts in vitro.

MethodsMouse Breeding and HusbandryThe transgenic 129P2-Scn5atm1Care/+ and FVBN/J- Scn5atm1Care/+ mice were generated as previously described.11 129.FVB.F1-Scn5atm1Care/+ mice were reared from these mice and subsequently intercrossed to produce 129.FVB.F2-Scn5atm1Care/+ progeny. All mice were supplied with the same SDS diet (SDS CRM (E) PL; Special Diets Services, UK) and water ad libitum and maintained on a 12-hour light/dark cycle in a temperature- and humidity-controlled environment. All experiments were performed in accordance with governmental and institutional guidelines for animal use in research.

GenotypingAll 129.FVB.F2-Scn5atm1Care/+mice were genotyped for the Scn5atm1Care mutation as previously described8; only mice heterozygous for the mutation were used. For the genome-wide scan, the 129.FVB.F2-Scn5atm1Care/+and 3 mice from each parental strain were genotyped across the 19 autosomes and X chromosome by means of an Illumina Golden Gate mouse medium density (768 SNP) panel as described previously8; 587 SNPs of this panel are informative in the studied F2 mice.

Harvest and Dissection of Heart SamplesMice were euthanized at 12 to 14 weeks by CO

2–O

2 asphyxiation

followed by cervical dislocation. Hearts were excised, washed in 1× PBS, and dissected left ventricular (LV) free-wall flash-frozen in liquid N

2.

Protein Isolation and AnalysisProtein was isolated from left ventricular tissue using standard pro-cedures. In short, cells/tissue were homogenized in ice-cold radio-immunoprecipitation assay (RIPA) buffer (50 mmol/L Tris HCl pH 7.6, 150 mmol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with protease inhibitors (Complete Mini; Roche) and sodium orthovanadate (final concentration 0.5 mmol/L). The unsoluable parts were spun down (30 sec at 4°C, 13 000 rpm); the supernatant was transferred to a fresh tube. Protein concentra-tions were determined using a BCA protein assay kit (Thermo Fisher Scientific). Total levels of soluble collagen were determined using the Sircol Assay (Biocolor) according to the manufacturer’s protocol and normalized to the total protein input. All experiments were performed in duplicate.

(Multiple) QTL MappingGenome-wide (Multiple) QTL mapping for the amount of collagen was performed using the R/qtl package in the statistical programming language R.12,13 Briefly, for all 587 informative markers in the 129.FVB.F2-Scn5atm1Care/+ mice, association with the amount of colla-gen was tested, correcting for sex and assuming either an additive or dominant genetic model in 2 separate analyses. To detect possible in-teracting loci (MQM), this model (Y≈X+sex+sex:X+ε) was further ex-panded by including all markers on the chromosomes with suggestive

association (logarithm of the odds>2.2, ie, Chr1, Chr2, Chr8, and Chr18) as model covariates, 1 chromosome at a time, followed by backwards elimination of markers that did not add significantly to the model. Markers were retained as covariates in the model at a nominal significance level of FDR <0.05 for the difference between the basic model and including the genetic covariate (referred to as cofactor in the R/qtl documentation). This resulted in the identification of a ge-netic covariate at Chr18 rs6361186. Significance levels for the LOD scores were determined based on a permutation FDR <0.05, resulting in an LOD threshold of 2.9 for significant collagen QTLs.

Fully processed (variance-stabilization transformation, quantile-normalized, log-transformed) gene expression microarray data were used as previously published5 and deposited in the gene expression omnibus database (GSE27236).5 Genome-wide (Multiple) eQTL mapping was performed using the R/eqtl package in the statistical programming language R. Briefly, for all 587 informative markers in the 129.FVB.F2-Scn5atm1Care/+ mice, association with each tran-script level (n=26 529) was tested, correcting for sex and assuming an additive genetic model. The model was further expanded by in-cluding evenly spaced, randomly placed genomic positions as model covariates to detect possible interacting loci. A multiple-transcript genome-wide significance threshold (P<1.88×10−6; Bonferroni cor-rection for 26 529 transcript traits) was applied; the corresponding empirical LOD threshold of 6.6 was determined using 100 000 per-mutations. The expression levels of transcripts for which we iden-tified eQTLs colocalizing with the collagen QTLs were tested for nominal correlation (P<0.05) with collagen levels (Spearman cor-relation coefficient, ρ).

Cardiac Fibroblast Isolation and CultureNeonatal mouse cardiac fibroblasts were isolated, in 2 independent experiments, according to a modified protocol described previously for rats.14 In short, hearts were excised aseptically and large vessels and both atria were removed. Ventricular tissue of all littermates (11 and 8 pups, respectively) was pooled and cut into 2 to 3 mm small pieces and remaining blood was washed out. Tissue fragments were washed in medium and incubated on at 4°C in HBSS (Sigma, H4641) containing 0.1% d-glucose (Merck) with antibiotics and 0.41 μ/mL Trypsin 1:300 (USB22715). Next, the tissue fragments were incubat-ed at 37°C in prewarmed DMEM41965 (Gibco) supplemented with antibiotics and 140 μ/mL collagenase II (Worthington, 4176) for 45 minutes. Gentle trituration with a plastic 5 mL pipet was then used to dissociate the cells, which were filtered through a 70 μm filter to re-move remaining debris. Cells were plated on 1% gelatin-coated 6 well plates in DMEM41965, 10% FBS, antibiotics for 1 hour, after which the supernatant containing the cardiomyocytes was removed, and the attached fibroblasts were cultured for 48 hours in DMEM41965, 2% BSA, supplemented with antibiotics, insulin (SigmaI6634), l-car-nitin (SigmaC0283), and l-glutamin (Gibco25030) in the presence or absence of 20 ng/mL recombinant human FGF1 (peprotech 100-17A). Cells were washed with PBS and scraped in 200 μl Trizol (Life technologies) per well. mRNA was isolated from these cell lysates according to the manufacturers’ protocol.

qRT-PCRmRNA expression levels of the candidate genes was determined in by quantification using the LightCycler system for real-time RT-PCR (Roche Applied Science). Quantitative PCR data were analyzed with the LinRegPCR program.15,16 All samples were processed in tripli-cate and expression levels were normalized to HPRT. A pairwise nonparametric Kruskal–Wallis test was used to assess differences between control and human recombinant FGF1 groups, significance was called at P<0.05.

Data AccessExpression data were deposited previously in the public gene ex-pression omnibus database of NCBI GEO database: GSE27236. Genotype and collagen phenotypic data have been submitted to the mouse phenome database: http://qtlarchive.org/.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

792 Circ Cardiovasc Genet December 2014

ResultsCollagen QTL MappingThe F2-cross used in this study to identify genetic modifiers of collagen deposition in heart has been described earlier.5,8 In brief, F2 progeny were generated from 129P2 and FVBN/J mice both carrying heterozygously the p.1795insD mutation in the cardiac sodium channel Scn5a (Scn5atm1Care/+). Amounts of collagen (relative to total protein content) were determined in LV tissue of 65 129.FVB.F2-Scn5atm1Care/+ mice and in mice from both parental strains with and without the mutation. No differences in the levels of collagen were observed between the 2 founder strains (Figure 1). However, the 129.FVB.F2-Scn5atm1Care/+ mice showed a large spread in relative col-lagen, suggesting the existence of segregating alleles affecting the trait in the founder strains. The relative amount of collagen was then used together with previously generated genotypic data from the same 129.FVB.F2-Scn5atm1Care/+ animals8 as input for QTL mapping. We thus detected a significant domi-nant QTL for cardiac collagen deposition on mouse chromo-some 8 between rs3725286 and rs13479811 (Figure 2A and 2B and Table 1). The FVBN/J allele at this locus was associ-ated with a lower level of cardiac collagen (Figure 3A).

Multiple QTL MappingAs we expected the involvement of multiple loci in the depo-sition of collagen, we investigated the possible existence of additional QTLs and possible genetic covariates using MQM. MQM provides a statistical approach for identifying genetic covariates in a data set based on a systematic screen of possible covariates as additional parameters in the multiple regression model of the trait. The approach protects against overfitting as the residual variance of the model is fixed.10 All chromosomal regions with an LOD score >2.2 were systematically analyzed for the presence of possible covariate loci by MQM using an additive model. We thus identified a significant covari-ate (FDR<0.05) for collagen deposition at chromosome 18 (rs6361186; Figure 3A and Table 1). Furthermore, adding the Chr18 locus as a covariate in QTL mapping uncovered that an additional collagen QTL on Chr2 (rs3713397-c2.loc20; Fig-ures 2A, 2C, and 3A and Table 1) was found. Table 1 provides an overview of the 3 identified loci, their respective borders, and LOD scores.

Overlay of the Identified QTLs With eQTL DataAs genetic variation underlying a QTL may act through effects on gene expression, we assessed the 1.5 LOD drop region around the identified collagen QTLs and the Chr18 covariate

locus for overlapping eQTLs. These eQTLs were identified based on determination of gene transcript abundance in LV tissue from 109 129.FVB.F2-Scn5atm1Care/+ mice and geno-typic information in the same population.5 We thus identified 24 significant eQTLs mapping within the 1.5 LOD-interval of the Chr8 collagen QTL (Table 2), 7 significant eQTLs map-ping within the Chr2 collagen QTL (Table 3), and 12 mapping within the 1.5 LOD-interval of the Chr18 collagen covariate locus (Table 4). The transcripts with a suggestive eQTL map-ping to the Chr2, Chr8, and Chr18 loci are listed in Table I in the Data Supplement.

As causative transcripts are expected to correlate with the amount of collagen, we performed a Spearman correlation test on the identified transcripts. Of the 31 transcripts with signifi-cant eQTLs mapping to the Chr8 and Chr2 loci, 5 (Pdlim3, Tmem66, Rad23A, Lsm6m, and Tnks) and 1 (Gpr158) tran-scripts, respectively, correlated with the relative amount of collagen (nominal significance, P<0.05; Tables 2 and 3 and Figure 3B and 3C). Of the 12 transcripts with eQTLs under-lying the Chr18 collagen covariate locus, the levels of Fgf1, assessed by 2 probes, showed a suggestive correlation with collagen (P=0.05 and P=0.09; Table 4 and Figure 3D).

To identify possible covariates for the identified eQTLs, we performed MQM on all expressed transcripts, similar to the procedure followed for the mapping of collagen loci. This analysis revealed that the final model for the transcript of one of the Chr8 eQTLs, namely Pdlim3¸ had a significant (FDR<0.05) covariate at the exact same position (rs6361186) as the Chr18 covariate identified for collagen deposition (see Figure 3D for genotype effects of rs6361186 on the expression level of Pdlim3). Furthermore, a suggestive eQTL (see Table I in the Data Supplement) at the Chr2 locus, for the Itga6 tran-script, and the Fgf1 eQTL also had a significant (FDR<0.05) covariate at rs6361186 (Figure 3D). In summary, we identi-fied a genetic covariate at rs6361186 on Chr18 using an addi-tive genetic model of cardiac collagen deposition, which was shared by the models for Pdlim3, Itga6, and Fgf1.

In total 12 012 covariates for eQTL models were identi-fied for all transcripts tested. Of these models, 248 (out of the 26 838 transcripts tested, 0.92%) had a genetic covariate at rs6361186 on Chr18. The presence of this Chr18 covari-ate in 13% of the QTL models (4 out of 32; 1 collagen QTL and 31 eQTLs mapping at the collagen QTLs) identified in relation with LV collagen, thus, represents a 13-fold overrep-resentation compared with random expectation (P<10−9). An overview of the loci involved, the correlating transcripts with underlying eQTLs, and the covariates identified is given in Figure 4 for clarity.

Figure 1. Collagen levels as function of genotype of the founder strains and F2 mice 129P2-Scn5atm1Care/+, 129P2-WT, FVBN/J-Scn5atm1Care/+, and FVBN/J-WT n=4; F2_Scn5a-tgCare1 n=66. Boxplot of the geometric mean of the percentage of collagen in left ventricular protein isolates ±25 to 75 percentiles.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

Lodder et al Genetics of Cardiac Collagen Deposition 793

Effect of FGF1 on the Chr2 and Chr8 Candidates in Neonatal Cardiac FibroblastsAs the collagen QTL model improved by including the covari-ate on Chr18, we next explored this association in vitro. Fibro-blast growth factor 1, Fgf1 (also known as acidic fibroblast growth factor or aFgf) was the only Chr18 covariate locus gene for which the transcript both had a cis-eQTL at this locus and showed a (suggestive) correlation to cardiac collagen deposition. Fgf1 is known to be involved in an anti-fibrotic effect in the lungs,17,18 and increased expression of FGF1 has been reported in patients with idiopathic cardiomyopathy.19 We therefore investigated the possible effect of Fgf1 on the expression levels of the genes with eQTLs overlapping the Chr2 and Chr8 collagen QTLs, in cultured neonatal cardiac fibroblasts from wild-type FVBN/J mice. Isolated cardiac fibroblasts were cultured for 48 hours with and without human

recombinant FGF1 in the culture medium after which RNA was isolated and transcript levels of the eQTL cognate genes were determined by qRT-PCR (Figure 5).

Of the 6 genes with an eQTL at the Chr8 locus, and for which the transcript was correlated to collagen, only Pdlim3 was reduced (P=0.028) in the presence of FGF1. This is in accordance with the finding that Pdlim3 was the only tran-script at Chr8 that had a significant covariate at the Chr18 Fgf1 locus. Of the genes with eQTLs at Chr2, the transcripts of Itga6 and Gpr158 were increased (P=0.0495 and P=0.025, respectively) in the presence of FGF1. No effects on the tran-script levels of Fgf1, Col1a1, or Col3a1 were observed.

An overview of the genes involved and the direction of their involvement is given in Figure 6. Note the opposite effects of human recombinant FGF1 on the final collagen levels through the different intermediate proteins.

Table 1. Overview of the Identified Collagen QTLs

Locus Marker Chr Pos, cM Size, cM Pos, bp (GRCm38) Size, Mb No of Genes LOD Score

Chr8 dominant QTL Upper border rs13479716 8 42.2 21.6 39709256 55.0 672 2.6

Peak rs3672639 8 50.2 61221786 4.1

Lower border rs4137596 8 63.8 94689598 2.0

Chr2 additive QTL (with Chr18 covariate)

Upper border rs3713997 2 0.0 22.6 3233903 25.1 574 3.0

Peak rs3713997 2 0.0 3233903 3.0

Lower border rs13476399 2 22.6 28289138 1.1

Chr18 covariate region Upper border rs13483183 18 0.0 52.1 3516541 82.3 781 1.5

Peak rs3720827 18 32.4 63290216 2.5

Lower border mCV24836796 18 52.1 85858806 0.6

Borders in cM and bp, number of genes in the locus, and LOD scores of the 1.5 LOD-drop intervals of the 3 identified loci.LOD indicates logarithm of the odds; and QTL, quantitative trait loci.

Figure 2. (Multiple) quantitative trait loci (QTL) mapping (MQM) of collagen logarithm of the odds (LOD) score plots for dominant QTL mapping (black) and additive multiple QTL mapping with correction for Chr18 rs6361186 (grey; A), genome-wide and detailed plots of Chr2 (B), and Chr8 (C) n=66.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

794 Circ Cardiovasc Genet December 2014

DiscussionWe here applied an integrative genomics approach to iden-tify novel genetic factors underlying deposition of collagen in heart. Using a genetically and phenotypically diverse mouse population (129.FVB.F2_Scn5atgCare1/+), we identified 2 novel QTLs (Chr2 and Chr8) and 1 covariate region (Chr18) for cardiac collagen. Overlaying these results with eQTL data obtained in the same F2-cross and with data on correlation between cardiac transcript abundance and deposited collagen, we identified a possible novel network of genes with opposing effects on cardiac collagen levels.

Collagen is a normal component of the cardiac extracellu-lar matrix. However, upon injury or stress (such as a myocar-dial infarction, hypertrophy, and age), the heart remodels and fibrotic scars are formed, a process in which collagen depo-sition is central. Fibrotic tissue hampers the normal function of the heart by reducing contractility and electric conductiv-ity.20 The identification of genes involved in collagen depo-sition will provide insight into the mechanisms determining the development of fibrosis in the setting of cardiac injury or stress.

MQM was developed as a technique to identify additional genetic loci after compensating for variation explained by other genetic loci (covariates).12,13 This allowed us to identify a genetic covariate locus on Chr18 for collagen deposition.

Similarly, MQM of cardiac gene expression data identified genetic covariates of eQTLs.

Integration of gene transcript MQM mapping and correla-tion data between transcript abundance and collagen depo-sition pointed to candidate genes at the collagen QTL loci. Although extensive validation is required, using this integra-tive genomics approach, we could uncover a potential regula-tory network, where Fgf1 appears to both suppress and induce collagen deposition through Pdlim3 (Chr8) and Gpr158 (Chr2), respectively. This results in a net positive correla-tion between Fgf1 and cardiac collagen in the F2-cross, but in a balanced effect on collagen deposition in both parental strains and in the cultured neonatal cardiac fibroblasts. Thus, FVBN/J mice that have relatively higher levels of Fgf1 in the presence of concurrent higher levels of Gpr158 and lower lev-els of Pdlim3 have the same net amount of LV collagen as 129P2 mice that have lower levels of Fgf1 with concurrent lower levels of Gpr158 and higher levels of Pdlim3. In the F2 population, however, as the different alleles are now seg-regating, a much larger spread in the levels of collagen can be observed (Figure 1), allowing the genetic dissection of these loci. These results provide an example of the power of such integrative genomic strategies to identify opposing effects using a genetically segregating population, rather than only typing the founding inbred mouse lines.

Figure 3. Collagen levels and expression levels as function of genotype at the candidate loci in the F2 mice percentage of collagen left ventricular (LV) protein isolates (mean±SE; A) and normalized log2 expression levels (mean±SE) of the candidate genes (B–D) per geno-type at the respective loci: green homozygous for the 129P2 allele, turquoise heterozygous, and blue homozygous for the FVBN/J allele (group sizes for the respective genotypes AA (129P2), AB (het), BB (FVBN/J): rs3713997 (chr2) AA=25; AB=52; BB=32; rs3672639 (chr8) AA=31; AB=57; BB=20; rs6361186 (chr18 cofactor) AA=23; AB=65; BB=21).

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

Lodder et al Genetics of Cardiac Collagen Deposition 795

Although the Scn5a mutation carried by the mutant mice we studied leads to enhanced fibrosis at old age,9 collagen deposition at the age of killing (12–14 weeks) did not differ between mutant and wild-type mice of both strains (Figure 1). The variability in the levels of collagen in the LV tissue that

were detected by the Sircol assay in the F2 mice, therefore, likely reflect normal variation in collagen deposition. More experimental work will be required to test whether the putative gene network identified in this study also plays a role in modu-lation of collagen deposition in the setting of cardiac stress. Furthermore, it is interesting in future studies to conduct simi-lar studies on collagen deposition with aging or in the setting of cardiac disease because these may uncover additional loci.

The human fibroblast growth factor family consists of 22 members. Fibroblast growth factors are multifunctional pro-teins with a wide variety of effects, including cell division, embryological patterning, and remodeling.21 Although Fgf2 (basic fibroblast growth factor) is a known factor in the devel-opment of cardiac fibrosis,22 the role of FGF1 (acidic fibroblast growth factor) in the heart has remained obscure. FGF1 was shown, however, to be overexpressed in cardiac tissue from patients with diverse idiopathic cardiomyopathies.19 Pdlim3 is a PDZ and LIM domain-containing protein that is associated with the Z-disc of cardiac myocytes. Recently, a potentially pathogenic mutation (M60Tfs1X) in PDLIM3 was reported in a patient with dilated cardiac myopathy.23 Mice deficient in Pdlim3 (Alp) develop right ventricular cardiomyopathy dur-ing development.24 Pdlim3 enhances the capacity of α-actinin to cross-link actin filaments, suggesting that destabilization of actin-anchorage sites in cardiac muscle lies at the basis of the observed phenotype.25 The data in our F2-cross show a posi-tive correlation between Pdlim3 transcript levels and collagen deposition in the heart, indicating that tight regulation of this protein is essential for a proper balance between maintenance of the extracellular matrix and the formation of fibrotic scars. Gpr158, a relatively unstudied orphan member of class C of the large G-protein coupled receptor family, was recently implicated in the formation of tight junctions in ocular tra-becular meshwork cells.26 To our knowledge, no connection between Gpr158 and either fibrosis or cardiac function has been described to date.

Table 3. Significant eQTLs Mapping Within the 1.5 LOD-Interval of the Chr2 Collagen QTL

Max LOD Score

Spearman Corr P Value Chr. Position Cis/Trans Gene Symbol

36.8 0.16 0.34 2 26832106 Cis 5930434b04rik

26.2 0.33 0.04* 2 21751924 Cis Gpr158

18.9 0.11 0.52 2 32455108 Cis St6galnac4

14.7 −0.10 0.56 2 26170000 Cis Gpsm1

13.3 0.00 0.98 2 26219159 Cis Snapc4

8.3 0.09 0.61 2 29598384 Cis Trub2

6.746 −0.15 0.38 2 6110159 Cis Echdc3

eQTL indicates expression quantitative trait loci.*The single transcript, Gpr158, which correlated with the relative amount of

collagen.

Table 4. Significant eQTLs mapping at the Chr18 covariate locus

Max LOD Score

Spearman Corr P Value Chr. Position Cis/Trans Symbol

86.0 0.29 0.09* 18 38999147 Cis Fgf1†

32.6 0.32 0.05* 18 39006739 Cis Fgf1†

32.4 0.09 0.60 18 32091161 Cis Lims2

26.2 −0.01 0.95 18 16747386 Trans Cdh2

25.7 −0.03 0.87 18 12313932 Trans Riok3

25.6 0.09 0.60 18 32101561 Cis Lims2

18.0 0.04 0.83 18 12314818 Trans Riok3

16.4 −0.12 0.48 18 39531826 Cis Arhgap26

10.2 0.03 0.88 18 32101561 Cis Lims2

9.0 −0.22 0.19 18 57549118 Trans Prrc1

8.5 −0.01 0.96 18 39531826 Cis Arhgap12

eQTL indicates expression quantitative trait loci; and LOD, logarithm of the odds.

*The single transcript, Fgf1, with suggestive correlation with collagen deposition.

†The corresponding 2 probes also had a genetic covariate at Chr18 rs6361186.

Table 2. Significant eQTLs at the Chr8 Collagen QTL

Max LOD Score

Spearman Corr P Value Chr Position Cis/Trans Gene Symbol

124.3 −0.23 0.17 8 85287669 Cis Zfp330

120.1 −0.25 0.14 8 84855860 Cis Il15

106.4 −0.25 0.13 8 84855771 Cis Il15

46.9 0.19 0.27 8 47004324 Cis Pdlim3*

37.5 0.40 0.01† 8 47418224 Cis Pdlim3

36.2 −0.11 0.52 8 85958484 Cis Scoc

34.2 −0.23 0.17 8 87606025 Cis BC056474

25.9 −0.15 0.37 8 83505084 Cis Usp38

19.5 0.39 0.02† 8 35217617 Trans Tmem66

19.5 0.14 0.41 8 81862661 Cis Smad1

18.8 0.10 0.55 8 87621335 Cis Man2b1

16.3 0.10 0.57 8 87410594 Cis Gcdh

15.8 −0.38 0.02† 8 87725910 Cis Rad23a

15.7 0.21 0.21 8 86188640 Cis Gipc1

14.0 −0.24 0.16 8 87864469 Cis Prdx2

13.6 0.33 0.05† 8 81328764 Trans Lsm6

13.4 0.20 0.23 8 90695790 Cis Heatr3

10.4 −0.47 3E-03† 8 87725910 Cis Rad23a

9.1 −0.07 0.66 8 86247615 Cis Cd97

8.9 −0.18 0.27 8 23328998 Cis Vps36

8.8 −0.07 0.68 4 133047239 Trans 2300002D11Rik

7.6 −0.24 0.15 8 87598296 Cis Dhps

7.4 −0.15 0.37 8 86247913 Cis Cd97

6.9 −0.35 0.03† 8 35894643 Cis Tnks

The 24 significant eQTLs mapping within the 1.5 LOD-interval of the Chr8 collagen QTL.

eQTL indicates expression quantitative trait loci; LOD, logarithm of the odds; and QTL, quantitative trait loci.

*The single eQTL with a cofactor at Chr18 rs6361186, Pdlim3.†The 6 transcripts displaying a correlation with collagen deposition.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

796 Circ Cardiovasc Genet December 2014

Strength and LimitationsBoth the sample size and the genetic variability of the 129.FVB.F2-Scn5atm1Care/+ cross are limited; undoubtedly, addi-tional loci and genes underlying cardiac collagen deposi-tion remain to be unravelled. Additional genetic differences

between 129P2 and FVBN/J affecting cardiac collagen depo-sition may be present at the loci identified in this study. Fur-ther dissection of these loci in additional crosses or selected recombinant inbred lines with the founding haplotypes of these 2 strains could help to identify these additional genetic

Figure 4. Scheme of the mapped collagen and expression quantitative trait loci (QTLs) with the multiple QTL mapping (MQM) cofactors. Solid lines indicate QTL effects, dotted lines the identified cofactors, the effects on collagen are indicated in green, the eQTL effects in red, and the loci in blue. Spearman correlation between the transcripts and collagen are indicated (ρ and P values) in the respective red boxes.

Figure 5. Effect of human fibroblast growth factor (hFGF1) on the Chr2 and Chr8 expression quantitative trait loci (eQTL) candidates in neonatal cardiac fibroblasts normalized expression levels (mean±SE) of the candidate genes in neonatal cardiac fibroblasts in the pres-ence (n=6; red) and absence (n=6; blue) of recombinant hFGF1 in the culture medium. *Significant differences (P<0.05) by nonparametric Kruskal–Wallis tests.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

Lodder et al Genetics of Cardiac Collagen Deposition 797

factors by reducing the haplotype size through additional recombinations.

The potential network described in this study will need to be validated in detail in model systems in vivo, with targeted deletion or overexpression of the different partners in the pro-posed network. It will be interesting to see how these targeted interferences will affect the phenotypic readout (ie, collagen deposition and possibly cardiac stress response) in different genetic backgrounds (eg, different combinations of alleles at the Chr2 and Chr8 loci).

Integrative genomics approaches, such as the study described here, are potentially powerful to identify genetic loci and genes underlying cardiovascular (disease) phenotypes. The challenge for such studies lies in the step from locus to gene and in the functional follow-up studies to validate the findings. Apart from studying expression level differences in the tissue and popu-lation under investigation (ie, eQTLs; see Ref. 27 for a recent review on harvesting the power of this approach), whole genome resequencing in many different inbred mouse (and rat) strains is now underway, potentially allowing the fast identification of causal genetic variants, including protein function chang-ing variants. Furthermore, deep phenotyping of (novel) genetic reference panels, such as the collaborative cross, the diversity outbred, and longer existing panels, such as the BXD and the BXH/HXB in rat, will allow the generation of phenotypic net-works and identification of the genetic controls of these.28

ConclusionsIn conclusion, using MQM, we here mapped a putative novel genetic network that modulates collagen deposition in mouse LV myocardium. These results demonstrate the potential of integrative genomics on whole-genome genetic and tran-scriptional data in combination with careful phenotyping in a genetically diverse murine population, to unravel not only the genes underlying QTL, but their possible associations in biological networks as well.

Sources of FundingThis study was funded by a Netherlands Heart Foundation Established Investigator Fellowship to C.R. Bezzina (NHS 2005T024). The Centre for BioSystems Genomics (CBSG) and the Netherlands Consortium of Systems Biology (NCSB), both of which are part of the Netherlands Genomics Initiative/Netherlands Organisation for Scientific Research (D. Arends); the European Union 7th Framework Programme under the Research Project PANACEA (Quantitative Pathway Analysis of Natural Variation in Complex Disease Signaling in C. elegans) [222936 R.J.].

DisclosuresNone.

References 1. Bezzina CR, Rook MB, Groenewegen WA, Herfst LJ, van der Wal AC,

Lam J, et al. Compound heterozygosity for mutations (W156X and R225W) in SCN5A associated with severe cardiac conduction distur-bances and degenerative changes in the conduction system. Circ Res. 2003;92:159–168.

2. Coronel R, Casini S, Koopmann TT, Wilms-Schopman FJ, Verkerk AO, de Groot JR, et al. Right ventricular fibrosis and conduction delay in a patient with clinical signs of Brugada syndrome: a combined electrophysi-ological, genetic, histopathologic, and computational study. Circulation. 2005;112:2769–2777.

3. Royer A, van Veen TA, Le Bouter S, Marionneau C, Griol-Charhbili V, Léoni AL, et al. Mouse model of SCN5A-linked hereditary Lenègre’s dis-ease: age-related conduction slowing and myocardial fibrosis. Circulation. 2005;111:1738–1746.

4. Remme CA, Baartscheer A, Verkerk AO, Engelen M a., Wilde AAM, van Rijen H V., et al. Genetic background determines magnitude of late sodi-um current, extent of intracellular Na+ and Ca2+ dysregulation, and sever-ity of cardiomyopathy in murine sodium channelopathy. Heart Rhythm. 2009;6:1686.

5. Lodder EM, Scicluna BP, Milano A, Sun AY, Tang H, Remme CA, et al. Dissection of a quantitative trait locus for PR interval duration identi-fies Tnni3k as a novel modulator of cardiac conduction. PLoS Genet. 2012;8:e1003113.

6. Petretto E, Sarwar R, Grieve I, Lu H, Kumaran MK, Muckett PJ, et al. Integrated genomic approaches implicate osteoglycin (Ogn) in the regula-tion of left ventricular mass. Nat Genet. 2008;40:546–552.

7. Keane TM, Goodstadt L, Danecek P, White MA, Wong K, Yalcin B, et al. Mouse genomic variation and its effect on phenotypes and gene regula-tion. Nature. 2011;477:289–294.

8. Scicluna BP, Tanck MW, Remme CA, Beekman L, Coronel R, Wilde AA, et al. Quantitative trait loci for electrocardiographic parameters and ar-rhythmia in the mouse. J Mol Cell Cardiol. 2011;50:380–389.

9. Remme CA, Scicluna BP, Verkerk AO, Amin AS, van Brunschot S, Beek-man L, et al. Genetically determined differences in sodium current charac-teristics modulate conduction disease severity in mice with cardiac sodium channelopathy. Circ Res. 2009;104:1283–1292.

10. Arends D, Prins P, Jansen RC, Broman KW. R/qtl: high-throughput mul-tiple QTL mapping. Bioinformatics. 2010;26:2990–2992.

11. Remme CA, Verkerk AO, Nuyens D, van Ginneken AC, van Brunschot S, Belterman CN, et al. Overlap syndrome of cardiac sodium channel disease in mice carrying the equivalent mutation of human SCN5A-1795insD. Circulation. 2006;114:2584–2594.

12. Arends D, Prins P, Broman K, Jansen R. Tutorial-Multiple-QTL Mapping (MQM) Analysis. 2010;1–39.

13. Jansen R. Quantitative trait loci in inbred lines. In: Balding DJ, Bishop M, Cannings C, eds. Handbook of Statistical Genetics. 3rd ed. Chichester, UK: John Wiley & Sons; 2007.

14. Teunissen BE, van Amersfoorth SC, Opthof T, Jongsma HJ, Bierhuizen MF. Sp1 and Sp3 activate the rat connexin40 proximal promoter. Biochem Biophys Res Commun. 2002;292:71–78.

15. Ramakers C, Ruijter JM, Deprez RH, Moorman AF. Assumption-free analysis of quantitative real-time polymerase chain reaction (PCR) data. Neurosci Lett. 2003;339:62–66.

16. Ruijter JM, Pfaffl MW, Zhao S, Spiess AN, Boggy G, Blom J, et al. Evalu-ation of qPCR curve analysis methods for reliable biomarker discovery: bias, resolution, precision, and implications. Methods. 2013;59:32–46.

17. Ramos C, Becerril C, Montaño M, García-De-Alba C, Ramírez R, Checa M, et al. FGF-1 reverts epithelial-mesenchymal transition induced by TGF-{beta}1 through MAPK/ERK kinase pathway. Am J Physiol Lung Cell Mol Physiol. 2010;299:L222–L231.

18. Becerril C, Pardo A, Montaño M, Ramos C, Ramírez R, Selman M. Acidic fibroblast growth factor induces an antifibrogenic phenotype in human lung fibroblasts. Am J Respir Cell Mol Biol. 1999;20:1020–1027.

19. Tomita Y, Kusama Y, Seino Y, Munakata K, Kishida H, Hayakawa H. In-creased accumulation of acidic fibroblast growth factor in left ventricu-lar myocytes of patients with idiopathic cardiomyopathy. Am Heart J. 1997;134:779–786.

20. Creemers EE, Pinto YM. Molecular mechanisms that control interstitial fi-brosis in the pressure-overloaded heart. Cardiovasc Res. 2011;89:265–272.

Figure 6. Scheme of direction of the effects of identified genes on each other. Directions of the effects of Pdlim3, Itga6, and Gpr158 are based on the direction of the correlation coefficient between the expression levels and collagen in the LV samples; directions of the effects of Fgf1 on Pdlim3, Itga6 and Gpr158 are based on the F2 data in combination with the in vitro validation.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

798 Circ Cardiovasc Genet December 2014

21. Goetz R, Mohammadi M. Exploring mechanisms of FGF signal-ling through the lens of structural biology. Nat Rev Mol Cell Biol. 2013;14:166–180.

22. Detillieux KA, Sheikh F, Kardami E, Cattini PA. Biological activities of fibroblast growth factor-2 in the adult myocardium. Cardiovasc Res. 2003;57:8–19.

23. Arola AM, Sanchez X, Murphy RT, Hasle E, Li H, Elliott PM, et al. Muta-tions in PDLIM3 and MYOZ1 encoding myocyte Z line proteins are in-frequently found in idiopathic dilated cardiomyopathy. Mol Genet Metab. 2007;90:435–440.

24. Georges-Labouesse E, Messaddeq N, Yehia G, Cadalbert L, Dierich A, Le Meur M. Absence of integrin alpha 6 leads to epidermolysis bullosa and neonatal death in mice. Nat Genet. 1996;13:370–373.

25. Pashmforoush M, Pomiès P, Peterson KL, Kubalak S, Ross J Jr, Hefti A, et al. Adult mice deficient in actinin-associated LIM-domain protein reveal a developmental pathway for right ventricular cardiomyopathy. Nat Med. 2001;7:591–597.

26. Patel N, Itakura T, Gonzalez JM Jr, Schwartz SG, Fini ME. GPR158, an orphan member of G protein-coupled receptor Family C: glucocorticoid-stimulated expression and novel nuclear role. PLoS One. 2013;8:e57843.

27. Rotival M, Petretto E. Leveraging gene co-expression networks to pin-point the regulation of complex traits and disease, with a focus on cardio-vascular traits. Brief Funct Genomics. 2014;13:66–78.

28. Andreux PA, Williams EG, Koutnikova H, Houtkooper RH, Champy MF, Henry H, et al. Systems genetics of metabolism: the use of the BXD murine reference panel for multiscalar integration of traits. Cell. 2012;150:1287–1299.

CLINICAL PERSPECTIVEWith aging and in cardiac disease, fibrosis caused by collagen deposition is increased, impairing contractility and providing a substrate for arrhythmia. Identifying the genetic factors underlying this process is an essential step to develop novel interven-tion strategies aimed at preventing pathogenic remodeling. This study provides a first step on this road. Harvesting the strong potential of the genetic diversity between different inbred mouse, we here mapped a possible novel genetic network modulat-ing collagen deposition in mouse left ventricular myocardium, thus providing new candidates for future functional studies.

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

Moerland, Michael W.T. Tanck, Michiel E. Adriaens and Connie R. BezzinaElisabeth M. Lodder, Brendon P. Scicluna, Leander Beekman, Danny Arends, Perry D.

DepositionIntegrative Genomic Approach Identifies Multiple Genes Involved in Cardiac Collagen

Print ISSN: 1942-325X. Online ISSN: 1942-3268 Copyright © 2014 American Heart Association, Inc. All rights reserved.

Dallas, TX 75231is published by the American Heart Association, 7272 Greenville Avenue,Circulation: Cardiovascular Genetics

doi: 10.1161/CIRCGENETICS.114.0005372014;7:790-798; originally published online September 12, 2014;Circ Cardiovasc Genet. 

http://circgenetics.ahajournals.org/content/7/6/790World Wide Web at:

The online version of this article, along with updated information and services, is located on the

http://circgenetics.ahajournals.org/content/suppl/2014/09/12/CIRCGENETICS.114.000537.DC1.htmlData Supplement (unedited) at:

  http://circgenetics.ahajournals.org//subscriptions/

is online at: Circulation: Cardiovascular Genetics Information about subscribing to Subscriptions: 

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

  document. Permissions and Rights Question and Answer information about this process is available in the

requested is located, click Request Permissions in the middle column of the Web page under Services. FurtherCenter, not the Editorial Office. Once the online version of the published article for which permission is being

can be obtained via RightsLink, a service of the Copyright ClearanceCirculation: Cardiovascular Geneticsin Requests for permissions to reproduce figures, tables, or portions of articles originally publishedPermissions:

by guest on May 14, 2016http://circgenetics.ahajournals.org/Downloaded from

SUPPLEMENTAL MATERIAL

Supplementary Table 1 Suggestive eQTLs underlying the three loci *, significiant correleations with collagen; # significant genetic covariate at Chr18

locus max LOD score

spearman corr p value Chr. position cis/ trans Symbol

2 6.473 -0.15 0.39 2 71694494 trans ITGA6* 2 6.469 -0.08 0.62 2 49630170 trans KIF5C 2 5.118 0.16 0.36 2 41033074 trans AQP7 2 5.025 0.04 0.79 2 52254728 trans ARL5A 2 4.668 -0.32 0.05 2 32258002 cis PTGES2 2 4.649 -0.03 0.85 17 12648874 trans SLC22A1 2 4.582 0.06 0.73 2 55188134 trans MOBKL3 2 4.566 -0.05 0.76 2 29052506 cis NTNG2 2 4.549 0.19 0.27 2 50597040 trans 4632404H22RIK 2 4.492 -0.44 0.01* 2 89926932 trans V1RB8 2 4.326 0.05 0.77 2 83287541 trans LCA5 2 4.304 -0.01 0.95 2 50580644 trans D630004A14RIK 2 4.231 0.10 0.54 2 89507810 trans OLFR1251 2 4.154 0.01 0.93 2 104626064 trans TMTC2 2 4.031 0.21 0.20 2 30416039 trans PPP2R4 2 4.025 0.20 0.22 2 53868544 trans PDLIM4 8 6.56 -0.12 0.47 8 86520071 cis Prkaca 8 6.50 0.05 0.77 8 86520168 cis Prkaca 8 6.18 -0.14 0.42 8 86247913 cis Cd97 8 6.13 -0.21 0.21 8 31089471 trans Dusp26 8 6.06 -0.09 0.61 8 96492548 cis Gnao1 8 6.04 0.01 0.96 8 86201446 cis Pkn1 8 5.51 -0.10 0.57 0 35375741 trans Ppp1r3b 8 5.03 -0.26 0.13 8 84926305 cis Il15 8 4.96 -0.11 0.50 3 19888033 trans Cp 8 4.95 0.21 0.21 8 86735036 cis Zswim4 8 4.49 -0.18 0.28 8 97346731 cis Ciapin1 8 4.47 -0.36 0.03 8 86770441 cis D8Ertd738e 8 4.27 0.03 0.85 8 47039034 cis 1700029J07Rik 8 4.20 0.25 0.14 11 83257127 trans 1700020L24Rik 8 4.20 0.38 0.02* 11 119464219 trans 4932417H02Rik 8 4.14 -0.08 0.64 0 27938695 trans Pld1 8 4.10 0.09 0.61 6 141930618 trans EG625716 8 4.03 0.21 0.21 6 113458484 trans Il17re 8 4.01 -0.02 0.89 8 114211100 cis Ctrb1 8 4.00 -0.28 0.09 8 96492013 cis Gnao1 8 4.00 0.06 0.72 7 100518144 trans Plekhb1

18 6.33 0.16 0.34 18 38477304 cis RNF14 18 6.18 0.07 0.68 15 85645836 trans PKDREJ 18 5.69 -0.13 0.45 18 34567301 cis PKD2L2 18 5.23 0.08 0.64 8 47039034 trans 1700029J07RIK 18 5.10 0.32 0.05 18 53841504 cis CCDC100 18 4.95 -0.08 0.65 18 67808490 cis PSMG2 18 4.48 0.11 0.53 18 60526159 cis DCTN4 18 4.44 0.16 0.35 18 12168689 trans 2400010D15RIK 18 4.35 -0.33 0.05 15 75428526 trans GPIHBP1 18 4.20 -0.10 0.55 X 132806214 trans NGFRAP1 18 4.11 0.23 0.18 10 90580838 trans SLC25A3