Author + information
- Received May 26, 2016
- Revision received July 18, 2016
- Accepted July 19, 2016
- Published online December 26, 2016.
- Malin K.B. Jonsson, PhDa,
- Robin J.G. Hartman, MSca,b,
- Matthew Ackers-Johnson, PhDa,c,
- Wilson L.W. Tan, BSca,c,
- Bing Lim, MD, PhDd,e,
- Toon A.B. van Veen, PhDf and
- Roger S. Foo, MDa,c,∗ ()
- aHuman Genetics, Genome Institute of Singapore, Singapore
- bUniversity of Utrecht, Utrecht, the Netherlands
- cCardiovascular Research Institute, National University Health Systems, National University of Singapore, Singapore
- dMerck Research Laboratories, Translational Medicine Research Centre, Singapore
- eCancer Stem Cell Biology, Genome Institute of Singapore, Singapore
- fDepartment of Medical Physiology, Division of Heart & Lungs, University Medical Center Utrecht, Utrecht, the Netherlands
- ↵∗Reprint requests and correspondence:
Dr. Roger S. Foo, Genome Institute of Singapore, 60 Biopolis Street, Genome #02-01, Singapore 138672, Singapore.
• The interplay between cardiomyocytes and cardiac fibroblasts is increasingly being recognized as important in cardiac disease.
• Fetal and adult cardiac fibroblasts influence their neighboring cardiomyocytes in different ways. A genome-wide comparison of the 2 reveals that they share >80% of gene transcripts.
• Motif analysis of empirical regulatory elements located next to differentially expressed genes led to identification of key differential regulators of fibroblast identity.
• STAT1 and PLAGL1 were identified and validated as key transcription factors to maintain the adult cardiac fibroblast phenotype. Loss of either factor led to a significant change in phenotype, including smaller cell size, apoptosis, reduced turnover, and down-regulated collagen gene expression.
Cardiovascular disease remains the number one global cause of death and presents as multiple phenotypes in which the interplay between cardiomyocytes and cardiac fibroblasts (CFs) has become increasingly highlighted. Fetal and adult CFs influence neighboring cardiomyocytes in different ways. Thus far, a detailed comparison between the two is lacking. Using a genome-wide approach, we identified and validated 2 crucial players for maintaining the adult primary human CF phenotype. Knockdown of these factors induced significant phenotypical changes, including senescence and reduced collagen gene expression. These may now represent novel therapeutic targets against deleterious functions of CFs in adult cardiovascular disease.
Although cardiomyocytes (CMs) occupy most of the tissue volume and provide the mechanical force delivered by the heart, they are largely outnumbered by nonmyocyte cells (30% vs. 70%), part of which are cardiac fibroblasts (CFs) (1,2). Cross-sectional confocal microscopy of ventricular tissue reveals that each CM is in the direct vicinity of at least 1 CF (3), reflecting a significant role for CFs in the heart; that is, to create and hold a supportive environment for CMs, such as by regulation of the extracellular matrix (ECM). More than simply a “scaffold cell,” CFs are understood to communicate with CMs in 3 different ways. The first method is through direct cell-to-cell contact, in which the formation of adherens junctions (cadherins) and gap junctions (connexins) play a crucial role (4). The second method is by paracrine or autocrine secretion of growth factors such as fibroblast growth factor (FGF)-2/basic FGF and transforming growth factor–β or important cytokines such as interleukin (IL)-1β and the IL-6 family, including leukemia inhibitory factor and cardiotrophin-1 (5). In the third method, cells indirectly relay signals via the ECM by modulating its composition and quantity by secretion or degradation of the ECM building blocks (4,6).
Common features of fibroblasts are the lack of a basement membrane, profound granular material in the cytoplasm scattered along a large Golgi apparatus, and a substantial rough endoplasmic reticulum (7). A unique marker for fibroblasts or CFs is still lacking, although both fetal and adult CF express periostin, discoidin domain receptor 2, and vimentin (8). Another feature of CFs is their ability to transform into an active state; the myofibroblast. Myofibroblasts express smooth muscle cell markers (e.g., smooth muscle actin [SMA]) and may contract. They have also been implicated in wound contraction, fibrosis, and scar healing and are a source of cytokines and growth factors, such as IL-6 and transforming growth factor–β (9).
Fibroblasts are abundant throughout all tissues in the body, and the population is heterogeneous, with diverse appearances and functions depending on where the cells reside (10). Apart from the expression of common core fibroblast genes, a specific gene expression profile involving the cardiogenic transcriptional network has been described uniquely for CFs (11). Furthermore, regional differences exist in which CFs from the atrium and the ventricle express different cardiogenic transcription factors (TFs) (11,12). Important differences have also been found between rat CFs from the embryonic heart compared with the adult heart, with a differential response in insulin-like growth factor–induced collagen production (13). The influence of CFs on co-cultured CMs also varies depending on age. Embryonic CFs increase proliferation of CMs, whereas adult CFs induce hypertrophy (6). Notably, this finding is consistent with the growth of a fetal or neonatal heart, which depends on the proliferation of CMs, whereas the adult heart responds to stress by hypertrophy. The diversity of the CF population is also underlined by the finding that CF may be derived from at least 3 sources: from the proepicardium, from endocardial- or epicardial-to-mesenchymal transformation (EMT), and from bone marrow (14–16).
Despite increasing attention to the potential role that CF may play in novel disease therapeutics (17), a detailed genome-wide characterization of the genetic and epigenetic profiles of CF has yet to be performed. In the present study, we conducted next-generation sequencing experiments with primary ventricular fetal human cardiac fibroblasts (fHCFs) and adult human cardiac fibroblasts (aHCFs) to carefully map their respective transcriptomic and epigenomic profiles, with the aim of unraveling key players responsible for maintaining the CF phenotype.
Detailed Methods are available in the Supplemental Appendix. aHCFs were purchased from PromoCell (Heidelberg, Germany; catalog no. C-12375, lot no. 3042901.1, lot no. 397Z030.3 and lot no. 1051601.5), and fHCFs were purchased from Cell Applications, Inc. (San Diego, California; catalog no. 306-05g, lot no. 1916, and lot no. 2112). The fibroblasts were of ventricular origin, and all donors were free of cardiac disease. Profiling experiments were performed with human cardiac fibroblasts (HCFs) between passages 3 and 4. Cellular, molecular, and next-generation sequencing experiments were performed as described in the Supplemental Appendix. Primer sequences for reverse transcription–quantitative polymerase chain reaction (RT-qPCR) are listed in Supplemental Table 1.
Ribonucleic acid–sequencing (RNA-seq) data were aligned to the Hg19 reference genome by using TopHat version 2.0.11 (Center for Computational Biology at Johns Hopkins University, Baltimore, Maryland). The transcriptome was assembled by using Cufflinks version 2.2.1 (Laboratory for Mathematical and Computational Biology at UC Berkeley; Computational Genomics at the Institute of Genetic Medicine at Johns Hopkins University, Baltimore, Maryland; Caltech, Pasadena), and differential expression of genes was analyzed in GENCODE gene annotation version 16 (Wellcome Trust Sanger Institute, Cambridge). Differentially expressed genes were selected on the basis of q value (false discovery rate–corrected p value) and log2fold-change produced by Cuffdiff. Chromatin immunoprecipitation–sequencing (ChIP-seq) and assay for transposase accessible chromatin–sequencing (ATAC-seq) peaks were called by using DFilter (18). Filter types for H3K4me3 and ATAC were zero mean, whereas H3K27me3 was interrogated by using a nonzero mean. Filter width settings for H3K4me3, H3K27me3, and ATAC were as follows: 8 to 10 kb, -ks = 100, -bs = 100; 1.5 to 3 kb, -ks = 25, -bs = 100; and 4 to 5 kb, -ks = 50, -bs = 100, respectively. Expected peak cutoff p values for H3K4me3, H3K27me3, and ATAC were set at 1E-6, 1E-3, and 1E-2 according to Dfilter recommendations. Peak annotation was performed by using ChIPpeakAnno (19). Known motifs and de novo motifs from ChIP-seq data were obtained by HOMER Motif Analysis (20).
Two-tailed unpaired Student t tests were used to assess significance in the RT-qPCR and immunostaining analysis. Results are presented as mean ± SD, and statistical significance is denoted in the graphics as p ≤0.05.
Cellular characterization of CFs
fHCFs and aHCFs were characterized with the use of immunocytochemistry and flow cytometry. Clinical characteristics of the cell isolates are listed in Table 1. Immunocytochemical labeling was performed for vimentin and Ki67 (proliferation marker). Representative images are shown in Figure 1A and are quantified in Figure 1B. Image analysis of more than 2,000 cells in independent cultures showed that fHCFs were consistently smaller than aHCFs (802 ± 46 μm2 vs. 2,127 ± 131 μm2) and proliferated faster. Ki67-positive cells for fHCFs and aHCFs were 63% and 52%, respectively, representing higher proliferation in the former. A higher turnover of fHCFs compared with aHCFs was also substantiated by a shorter doubling time of 34.6 h and 68.9 h, respectively (data provided by the supplier). More than 98% of cells were α-SMA negative (Supplemental Figure 1), ensuring that the majority of the study CFs were not activated to myofibroblasts by stress conditions of cell culture. Flow cytometry results proved that both fHCFs and aHCFs expressed THY-1 and PDGFRA. Moreover, FACS also allowed us to exclude the possibility that the study cells were contaminated with other cells of endothelial (PECAM1 negative) or hematopoetic (CD45 negative) origin (Figure 1C).
RNA-seq was performed for both fHCFs and aHCFs (n = 3) (Supplemental Figure 2A). We used the stringent cut-off of fragments per kilobase of exon per million fragments mapped >1 to call genes expressed in both aHCFs and fHCFs. A total of 13,400 genes were expressed in both, with 1,639 and 1,354 being exclusively expressed in fHCFs and aHCFs, respectively (Figure 2A). Thus, aHCF and fHCF transcriptomes exhibited marked similarities, raising the important question of whether specific transcriptomic differences might define their unique phenotypic characteristics. Figure 2B shows a heat map of exclusively expressed genes, which expectedly segregate the 2 sets of fHCF and aHCF samples. Supplemental Appendix 2 lists all genes expressed in the 2 sets of HCFs and their corresponding expression values. Correlation comparisons between sample replicates in Figure 2C corroborate the distinction between aHCFs and fHCFs (correlation of fHCF replicates in blue is ∼0.98, and the same for aHCF replicates in red, contrasting against lower correlations compared across fHCFs and aHCFs in purple).
We next sought to interpret gene expression signatures that distinguished between fHCFs and aHCFs and started by performing Ingenuity Pathway Analysis (IPA, Qiagen GmbH, Hilden, Germany). Supplemental Table 2 displays the highest significant genes with a log fold-change aHCF > fHCF of >4 or <–4. Using the list of differentially expressed genes, a selection was made according to manual curation for relevance on the basis of results of a thorough literature search. Selected genes were validated by using RT-qPCR (Supplemental Figure 2B) and biologically replicated in 3 additional independent HCF cell isolates (2 for adult and 1 for fetal) (Supplemental Figure 2C). Mean gene expression in aHCFs versus fHCFs by using RT-qPCR supported the RNA-seq data for all 24 genes tested.
The top 1,000 differentially expressed genes on the basis of their log-fold difference were fed into IPA to perform a comparison analysis. Canonical pathway analysis of the lists revealed that genes from ephrin receptor signaling, IL-8 signaling, and Notch signaling were statistically enriched in the fetal transcriptome, whereas IL-6 signaling, among others, was enriched in the adult transcriptome. Figure 2D provides the IPA summary for the top canonical pathways and the physiological system development and functions with their respective p values. Both transcriptomes contained genes involved in tissue development and cardiovascular system development, potentially representing a common phenotype for CFs regardless of adult or fetal origins. In contrast, immune cell trafficking was an aHCF hallmark, and embryonic development was an fHCF hallmark. Gene ontology analysis for molecular functions, cellular processes, and biological processes was also consistent with the IPA pathways, and results are listed in Supplemental Figure 3. All together, we curated a list of genes to display those that were: 1) highly expressed in both aHCFs and fHCFs; 2) highly expressed in fHCFs but not in aHCFs; or 3) highly expressed in aHCFs but not in fHCFs (Figure 2E). Examples of genes differentially overexpressed in aHCFs include FGF7, ITGA8, CRYAB, ELN, and TNFRSF11B, whereas genes overexpressed in fHCFs include IL1B, POSTN, HBEGF, and VCAM.
Epigenetic characteristics: Chromatin accessibility and histone marks
ATAC-seq was performed to assess chromatin accessibility. Quality control of ATAC-seq libraries verified the expected nucleosomal pattern both before and after sequencing (Supplemental Figures 4A and 4B). A total of 63,829 peaks were identified for fHCFs and 77,614 for aHCFs by using criteria as detailed in the Methods section. The HCF epigenome was further characterized by mapping active and repressed gene regulatory domains using chromatin immunoprecipitation followed by next-generation sequencing (ChIP-seq) for histone marks of H3K4me3 (representing active promoters) and H3K27me3 (representing repressed genomic domains). A total of 22,662 and 25,449 peaks were detected for H3K4me3 in fHCFs and aHCFs, respectively, whereas 8,446 and 6,330 were detected for H3K27me3. By integrating the ATAC-seq loci with the 2 sets of histone ChIP-seq loci, we confirmed that ATAC-seq mainly overlaid active promoters (H3K4me3) but not repressed domains (H3K27me3) (Figure 3A). Analysis for the loci of ATAC, H3K4me3, and H3K27me3 revealed that all 3 signals were enriched around transcription start sites of genes (Figure 3B). We further integrated transcriptome (RNA-seq), ATAC-seq, and histone ChIP-seq datasets and confirmed that loci of H3K4me3 and ATAC-seq corresponded consistently with each other, representing loci of open chromatin and active promoters. In contrast, these loci differed from H3K27me3 peaks, representing repressed domains. Similarly, highly expressed genes corresponded with ATAC-seq and H3K4me3 loci but not with H3K27me3 (Figure 3C). As examples, the epigenomic and transcriptomic profiles of 2 genes, ELN (highly expressed in aHCFs) and POSTN (highly expressed in fHCFs), are shown in Supplemental Figures 5A and 5B. Table 2 provides the expression profiles of some expressed genes along with their epigenetic profile.
Identification of TFs from motif analysis
Having confirmed the consistency and validity of our genome-wide transcriptomic and epigenomic profiles, we investigated whether TF regulators could be identified with an upstream role unique to aHCFs or fHCFs. A de novo motif analysis was performed with Homer on H3K4me3 loci (taken to represent active gene promoters) associated with differentially expressed genes in aHCFs and fHCFs. Two motifs and their corresponding TFs were enriched in differential H3K4me3 loci of each cell type (Figure 4A). Notably, the 2 TFs identified in fHCFs (SOX17 and SOX4) and the 2 for aHCFs (PLAGL1 and STAT1) were also strongly differentially up-regulated in their corresponding cells on the basis of their corresponding RNA-seq data. We accurately replicated these findings further in 3 (2 adult and 1 fetal) additional independent HCF cell isolates (Supplemental Figure 6A). These TFs now represent novel regulators that may play an important role in their respective cells.
TF validation by GapmeR knockdown
The therapeutic potential of targeting aHCFs in adult cardiac disease led us to choose PLAGL1 and STAT1 for validation by in vitro knockdown using GapmeR oligonucleotides. We performed this procedure in all 3 aHCF lines (Table 1). PLAGL1 and STAT1 knockdown led to >80% reduced expression levels when targeted with GapmeRs (Figure 4B, Supplemental Figure 6B). Interestingly, PLAGL1 knockdown also led to reduced STAT1 levels, suggesting an upstream effect of PLAGL1 on STAT1 that has not been previously reported. Negative A GapmeR construct, vehicle transfection, and nontransfected control specimens were used in all assessments. Cells were stained with antibodies against vimentin and Ki67 and analyzed for cell size, roundness, cell cycle, and cell number (n = 3). Cell size, morphology, or cell number was unchanged in all control specimens (Figures 4C and 4D). Instead, knockdown of STAT1 significantly increased cell roundness, reduced cell number without increasing cell proliferation, and knockdown of either TF changed aHCF cell size (Figure 4D, replicated in independent cell isolates in Supplemental Figure 6C). Changes at the transcriptional level revealed that whereas VIM and SMA were unchanged, both STAT1 and PLAGL1 knockdown caused significant down-regulation of other genes previously noted as being characteristic of the aHCF phenotype (ELN, TNFRSF11B, COL1A1, and COL1A2). Conversely, HBEGF associated with the fetal gene profile was up-regulated (Figure 5A, also replicated in independent cell isolates in Supplemental Figure 6D).
To test whether knockdown fibroblasts underwent mesenchymal epithelial transition (21), key markers for epithelial cells (CDH1, KRT9, and ATP2C2) were quantified, but these remained poorly expressed or undetected, rejecting the possibility that this transition had taken place (Supplemental Figure 6D). We next assessed for apoptosis by using the ApopTag-Red assay. Baseline apoptosis of aHCFs in culture was 3% (untreated cells, n = 5,237), 5% (cells treated with control vehicle lipofectamine alone, n = 4,992), and 6% (cells treated with control negative A GapmeR, n = 3,710). In contrast, knockdown of PLAGL1 or STAT1 led to significant up-regulation of apoptosis to 29% in PLAGL1 knockdown cells (n = 2,500) and 156% in STAT1 knockdown cells (n = 2,739; >100% apparent apoptosis is due to automated detection of multiple ApopTag-Red–stained granules in many cells) (Figure 5B).
Interestingly, apart from the classical ApopTag-Red–positive staining of nuclei to mark apoptotic cells, an overwhelming number of STAT1 knockdown cells were also noted that showed cytoplasmic fragmentation labeling. This finding was replicated more than 3 times and was absent both in PLAGL1 knockdown cells and in DNase1-treated positive control cells. At this time, we considered the possibility that cytoplasmic labeling represents advanced apoptosis in STAT1 knockdown cells or evidence of widespread mitochondrial damage.
Furthermore, a significant up-regulation of the classical marker for fibroblast senescence MMP3 (1,400-fold compared with control), as well as a subtle up-regulation of RB1 and TP53 (also markers of cell senescence), were noted in STAT1 knockdown aHCFs (Figure 5A). Significant up-regulation of MMP3 was replicated in 1 of 2 of the additional aHCF cell isolates. The third cell line instead showed a significant up-regulation of RB1 (Supplemental Figure 6D). High-power magnification of 4′,6-diamidino-2-phenylindole–stained nuclei showed formation of heterochromatin foci in 9% of STAT1 knockdown aHCFs (n = 5,595), whereas this outcome was not seen in untreated or positive control cells and in <1% after PLAGL1 knockdown (n = 4,899). Cells with heterochromatin foci were frequently positive for ApopTag-Red (Figure 5C), although not all ApopTag-Red–positive cells showed heterochromatin foci. Instead, we observed a striking exclusivity in which nuclei displaying the heterochromatin foci were never positive for Ki67 (Figure 5D). Together, these findings suggest that whereas 1 subset of knockdown aHCFs undergoes apoptosis, another appears to become senescent. Furthermore, STAT1 and PLAGL1 knockdown led to a significant reduction in collagen COL1A1 and COL1A2 gene expression (Figure 5A). These results were again supported by RT-qPCR in all additional sets of aHCF cells.
Results of previous studies indicated that fHCFs and aHCFs are different in that they influence neighboring CMs in distinct ways (6,13). In the present study, we have further defined their distinct differences by performing a thorough characterization of primary fetal and adult human ventricular CFs by using cellular, molecular, and genome-wide sequencing approaches. Our results show that fHCFs are smaller and proliferate more quickly than their adult counterpart. We hypothesized that these characteristics are explained by differences in their transcriptomes and epigenomes. Indeed, genome-wide analyses of RNA-seq, ATAC-seq, and histone ChIP-seq led us to 2 TFs, highly expressed in aHCFs but not in fHCFs. We performed functional validation to conclude that they regulate aHCF cell size, cell cycle re-entry, and cell survival. Knockdown of either of the 2 TFs led to a destabilization of the aHCF phenotype and reduced gene expression of collagen genes, a major constituent of pathological fibrosis in the adult heart.
RNA-seq confirms fibroblast identity and identifies differential signaling pathways
We first confirmed that “core” genes that define a fibroblast and that are typically used as fibroblast markers, such as VIM, THY1, DDR2, and COL1A2, do not differ in expression between fHCFs and aHCFs. According to IPA analysis, 3 top functions found in the overlapping transcriptomes of the fHCFs and aHCFs are tissue development, cardiovascular system development and function, and organismal development, validating our cardiac cell source and supporting the conclusion that fHCF and aHCF transcriptomes do not differ substantially. Conversely, this finding implied that important differences would be decidedly more subtle. Nonoverlapping gene functions are embryonic development and organismal survival for fHCFs, reflecting their fetal origin (22,23), and immune cell trafficking and skeletal and muscular system development and function for aHCFs, suggesting how aHCFs may have a different role in the adult human heart compared with their fetal counterpart. We used 2 additional HCF lines to validate the RNA-seq to show that differential gene expression was confirmed across independent biological replicates. Among these are likely to be bona fide markers that can be used to distinguish between aHCFs and fHCFs.
Novel and interesting genes identified
Differences in transcriptomes between cell types may help us to understand and explain their differential functions. For example, we expected that the fHCF transcriptome would contain genes that reveal their role for the developing heart and for inducing appropriate heart growth, whereas the aHCF transcriptome should comprise genes involved in the maintenance of a heart that is already matured. We therefore dissected the gene pathways and functions specific to each differentially expressed transcriptome for this analysis. Genes of the ephrin-receptor and ephrin b–signaling pathways were enriched in fHCFs. Indeed, ephrin B2 is important in the developing heart for ventricular chamber morphogenesis (24). Coherent to this process is the importance of Notch signaling; we also found that the Notch pathway is present in fHCFs but not aHCFs. Ephrin signaling has further been implicated for the development of vasculature (25). The finding of IL-8 signaling and ephrin signaling in fHCFs therefore suggests a role for fHCFs in vascular development as well. VCAM-1 is among the highest differentially expressed genes in fHCFs, and VCAM-1 has known importance for cardiac development because Vcam1 null mice embryos do not survive past E13.5 due to retarded heart growth (23). Interestingly, ITGA4 is also among the highest differentially expressed genes in fHCFs, and ITGA4 is an endogenous ligand for VCAM-1, creating the opportunity for inter-fHCF communication that might guide cardiac development. We also noted that analysis of gene ontology cellular processes revealed more profound plasma membrane components in fHCFs, such as integrins and other receptors, whereas aHCF-specific genes seem to focus on extracellular components. HBEGF is yet another gene that is highly expressed in fHCFs and lacking in aHCFs. HBEGF has been reported to promote cardiomyocyte proliferation in a paracrine fashion in a mouse model (6), and our data support its importance exclusively in fHCFs.
Among the top differentially expressed genes in aHCFs is FGF7. A gene network of FGF signaling in IPA also showed that 10 of the 15 genes involved in FGF signaling were highly differentially expressed in aHCFs, with other members of the network, including FGF2 and FGF14. Although it is well established that the FGF superfamily of growth factors is important in pathophysiological processes within the cardiac environment (26), a specific role for FGF7 has yet to be defined. Another gene more highly expressed in aHCFs was TNFRSF11B encoding for osteoprotegerin, a mesenchymal marker for EMT, from which a large population of adult CFs is proposed to derive (27). Overexpression of TNFRSF11B in rat CFs increased the expression of fibrosis-related proteins after treatment with angiotensin II, whereas its knockdown produced the reverse effect (28). ELN (elastin) is a third gene highly expressed in aHCFs. Elastin is a major contributor to the ECM, and it functions together with fibrillin to provide elasticity to a range of tissues (29). Interestingly, overexpression of ELN in cells that were transplanted to infarcted cardiac tissue in rats ameliorated the formation of scar tissue, compared with infarcted cardiac tissue that received vector-transduced cells (30). A necessity for crosslinking of elastin fibers is the presence of lysyl oxidases (29), and LOXL4 (lysyl oxidase–like 4) is also one of the highest differentially expressed genes in aHCFs.
Putative regulatory TFs for fetal HCFs determined by using motif analysis
TFs determined by using de novo motif analysis add another layer of information to the regulation of the differential transcriptomes between aHCFs and fHCFs. We performed motif analysis by using active promoters, demarcated according to H3K4me3, of differentially expressed genes for either aHCFs or fHCFs, thus specifying TFs for the respective transcriptomes. Top motifs were SOX17 and SOX4 in fHCFs. Importantly, these TFs were also all highly expressed genes in fHCFs compared with aHCFs.
SOX17 is also expressed in endothelial cells, where it regulates angiogenesis under the influence of Notch signaling (31). Our results show that SOX17 is expressed in fHCFs as well. SOX4 is a master regulator of EMT and is needed for proper formation of the cardiac outflow tract (32,33). Together, the abundant expression of these 2 TFs in fHCFs suggests that they may maintain prefate determination functions, and/or that fHCFs have a role in the development of the vasculature of the heart, similar to the ephrin and IL-8 signaling discussed earlier.
Regulatory TFs in aHCFs found by motif analysis and validated in vitro
The top 2 TFs identified for aHCFs were PLAGL1 and STAT1, and their corresponding messenger ribonucleic acid was indeed significantly highly expressed in aHCFs compared with fHCFs. PLAGL1 is an important TF in the developing heart, but its role has only been linked to CMs rather than CFs (34). In our experiments, knockdown of PLAGL1 in aHCFs led to reduced cell size and increased apoptosis. STAT1 activation reportedly increases fibroblast activation and collagen synthesis induced by high glucose levels (35). In our experiments, STAT1 knockdown in aHCFs also led to reduced cell size and apoptosis. Gene expression analysis after knockdown showed that VIM and SMA levels remained unchanged, indicating that STAT1 knockdown cells were not transforming to myofibroblasts. Instead, markers associated with the adult phenotype were down-regulated, and fetal markers were up-regulated. Together with their smaller size and increased apoptosis, we interpret these changes to represent a de-stabilization of the aHCF phenotype and viability.
We also found evidence for cell senescence based on the very high up-regulation of the classical fibroblast senescence marker MMP3 and concomitant up-regulation of other senescence markers (RB1 and TP53) (36,37). Consistent with this finding was the formation of heterochromatin foci in knockdown cells. Cells with heterochromatin foci were mutually exclusive from Ki67-positive cells, consistent with the description of senescence-associated heterochromatin foci that have already been extensively described (38). Concurrently, the expression of collagen genes COL1A1 and COL1A2 as well as TNFRSF11B were strongly down-regulated, suggesting that STAT1 knockdown cells have a diminished ability to stimulate profibrotic processes. Importantly, these findings were confirmed in independent sets of aHCF cells from different donors.
Overall, this report describes for the first time important genome-wide differences between fHCFs and aHCFs. Signals from loci of active promoters led to the identification of important TFs responsible for the differential roles of HCFs. Although the exclusive use of human material limited this research to in vitro studies, these findings nevertheless reveal interesting new explanations for the different ways in which fibroblasts act in the human adult and fetal heart. Our data further suggest that this source of HCFs may be useful for translational research, such as for disease modeling of cardiac fibrosis, or for other uses, such as improving CM maturation in differentiation protocols from human embryonic stem cells. Our knockdown studies for PLAGL1 and STAT1 suggest their importance for aHCF viability and phenotype maintenance. Further in vivo studies will be needed to determine whether they do indeed represent potential drug targets for steering away the malevolent behavior of CFs during adult heart disease progression.
COMPETENCY IN MEDICAL KNOWLEDGE: We report important genome-wide differences between fHCFs and aHCFs and identified key players responsible for maintaining the adult HCF phenotype. Our data suggest that this source of HCFs is useful in translational research, such as for disease modeling of cardiac fibrosis.
TRANSLATIONAL OUTLOOK: The transcription factors identified (PLAGL1 and STAT1) may be potential drug targets for steering away the malevolent behavior of HCFs during progression of adult heart disease.
The authors are grateful to Zenia Tiang for assisting with library preparation for RNA-seq; to Shivaji Rikka for assistance with the Operetta high contents analysis system; and to Dr. Justus Stenzig for assistance with flow cytometry. They also highly appreciate the generosity of Professor Mark Richards and Dr. Yei-Tsung Chen, who provided the third adult HCF line aHCFs (3).
For supplemental material including tables and figures, please see the online version of this article.
The authors acknowledge the support from the Netherlands CardioVascular Research Initiative: the Dutch Heart Foundation, the Dutch Federation of University Medical Centres, the Netherlands Organisation for Health Research and Development, and the Royal Netherlands Academy of Sciences (CVON-PREDICT). Dr. Jonsson has received an EMBO long-term fellowship grant (EMBO ALTF 304-2012). Funding was received by Dr. Foo from the Singapore National Medical Research Council and by Drs. Foo and Lim from the Biomedical Research Council, Agency for Science, Research and Technology. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- adult human cardiac fibroblast
- assay for transposase accessible chromatin
- assay for transposase accessible chromatin–sequencing
- cardiac fibroblast
- chromatin immunoprecipitation–sequencing
- extracellular matrix
- epithelial-to-mesenchymal transformation
- fibroblast growth factor
- fetal human cardiac fibroblast
- human cardiac fibroblast
- Ingenuity Pathway Analysis
- ribonucleic acid–sequencing
- reverse transcription–quantitative polymerase chain reaction
- transcription factor
- Received May 26, 2016.
- Revision received July 18, 2016.
- Accepted July 19, 2016.
- The Authors
- Camelliti P.,
- Borg T.K.,
- Kohl P.
- Pinto A.R.,
- Ilinykh A.,
- Ivey M.J.,
- et al.
- Kakkar R.,
- Lee R.T.
- Snider P.,
- Standley K.N.,
- Wang J.,
- Azhar M.,
- Doetschman T.,
- Conway S.J.
- Souders C.A.,
- Bowers S.L.,
- Baudino T.A.
- Chang H.Y.,
- Chi J.T.,
- Dudoit S.,
- et al.
- Furtado M.B.,
- Costa M.W.,
- Pranoto E.A.,
- et al.
- Burstein B.,
- Libby E.,
- Calderone A.,
- Nattel S.
- Ali S.R.,
- Ranjbarvaziri S.,
- Talkhabi M.,
- et al.
- Haudek S.B.,
- Xia Y.,
- Huebener P.,
- et al.
- Acharya A.,
- Baek S.T.,
- Huang G.,
- et al.
- Kwee L.,
- Baldwin H.S.,
- Shen H.M.,
- et al.
- Smith C.L.,
- Baek S.T.,
- Sung C.Y.,
- Tallquist M.D.
- Shen A.,
- Hou X.,
- Yang D.,
- et al.
- Kielty C.M.,
- Sherratt M.J.,
- Shuttleworth C.A.
- Mizuno T.,
- Mickle D.A.,
- Kiani C.G.,
- Li R.K.
- Lee S.H.,
- Lee S.,
- Yang H.,
- et al.
- Yuasa S.,
- Onizuka T.,
- Shimoji K.,
- et al.
- Rodier F.,
- Campisi J.
- Young A.R.,
- Narita M.