Developmental Heterogeneity of Cardiac Fibroblasts Does Not Predict Pathological Proliferation and ActivationNovelty and Significance
Rationale: Fibrosis is mediated partly by extracellular matrix–depositing fibroblasts in the heart. Although these mesenchymal cells are reported to have multiple embryonic origins, the functional consequence of this heterogeneity is unknown.
Objective: We sought to validate a panel of surface markers to prospectively identify cardiac fibroblasts. We elucidated the developmental origins of cardiac fibroblasts and characterized their corresponding phenotypes. We also determined proliferation rates of each developmental subset of fibroblasts after pressure overload injury.
Methods and Results: We showed that Thy1+CD45−CD31−CD11b−Ter119− cells constitute the majority of cardiac fibroblasts. We characterized these cells using flow cytometry, epifluorescence and confocal microscopy, and transcriptional profiling (using reverse transcription polymerase chain reaction and RNA-seq). We used lineage tracing, transplantation studies, and parabiosis to show that most adult cardiac fibroblasts derive from the epicardium, a minority arises from endothelial cells, and a small fraction from Pax3-expressing cells. We did not detect generation of cardiac fibroblasts by bone marrow or circulating cells. Interestingly, proliferation rates of fibroblast subsets on injury were identical, and the relative abundance of each lineage remained the same after injury. The anatomic distribution of fibroblast lineages also remained unchanged after pressure overload. Furthermore, RNA-seq analysis demonstrated that Tie2-derived and Tbx18-derived fibroblasts within each operation group exhibit similar gene expression profiles.
Conclusions: The cellular expansion of cardiac fibroblasts after transaortic constriction surgery was not restricted to any single developmental subset. The parallel proliferation and activation of a heterogeneous population of fibroblasts on pressure overload could suggest that common signaling mechanisms stimulate their pathological response.
Cardiac fibrosis underlies the pathological response of the mammalian heart to certain insults, such as a pressure overload injury, and this process is largely mediated through cardiac fibroblasts. The functions and characteristics of cardiac fibroblasts in scar formation and tissue remodeling are being actively explored.1 However, the study and analysis of cardiac fibroblasts are limited by imprecise definitions for this cell type and, concomitantly, by lack of specific markers to aid in their identification. Several other interstitial cell types, such as mesenchymal and inflammatory cells, express certain markers commonly used for identifying cardiac fibroblasts. One such example is fibroblast-specific protein 1 that is expressed in both fibroblasts and macrophages in the liver.2 Therefore, it is imperative to delineate and validate a robust set of markers to isolate cardiac fibroblasts.
Editorial, see p 602
In This Issue, see p 601
It is thought that injury induces the expression of α-smooth muscle actin (α-SMA) in a fraction of cells expressing fibroblast markers leading to their activation. These activated cells, termed myofibroblasts, secrete extracellular matrix to increase collagen content and hence cause histological fibrosis. Accumulating evidence supports the existence of heterogeneity among cardiac fibroblasts.1 For example, several developmental origins have been suggested for cardiac fibroblasts, such as the bone marrow (BM),3 embryonic endothelium,4 circulation,5 neural crest,6,7 and epicardium.8 Nevertheless, it is not known whether certain fibroblast subtypes, on the basis of their developmental origin, have a propensity toward proliferation and activation after profibrotic cardiac injuries. Indeed, it remains unclear whether fibroblasts from different lineages exhibit phenotypic or transcriptional disparities. Understanding the developmental origin of cardiac fibroblasts may enable identification of fibroblast subsets that are directly involved in fibrosis and may thus help to develop novel treatment strategies to target fibrosis.
Herein, we define a set of surface markers that distinguish cardiac fibroblasts from other cell types in the heart. We use fate-mapping models, BM transplantation, and parabiosis studies in the setting of physiological aging as well as pressure overload injury to determine the developmental sources of cardiac fibroblasts and to elucidate whether pathological activation of fibroblasts is developmentally determined. We demonstrate that cardiac fibroblasts derive primarily from the epicardium, whereas a smaller fraction originates from the endothelial lineage, and a minority from the neural crest. Our work further shows that circulating, hematopoietic, and stromal cells do not contribute to the cardiac fibroblast pool. Interestingly, we did not observe any differences in proliferation rate among cardiac fibroblasts of different developmental origins after pressure overload injury. Rather, our findings show that fibroblasts from distinct lineages have a similar phenotype and gene expression profile in a given context, such as aging and injury.
An expanded Materials and Methods is available in the Online Data Supplement.
Pax3Cre/+, Tie2Cre/+, Wt1CreERT2/+, Myh11Cre/+-GFP, Vav1Cre/+, CAG-dsRed, and C57Bl/6 mice strains were obtained from the Jackson Laboratory and have been described previously. Tbx18Cre transgenic mice, Myh6-GFP, and R26RmT/mG reporter mice were gifts from Sylvia Evans (San Diego, CA), Deepak Srivastava (San Francisco, CA), and Liqun Luo (Stanford, CA), respectively. All procedures were performed with the approval of the University of California, Los Angeles Animal Research Committee or the Institutional Animal Care and Use Committee at Stanford University. Two operators blinded to the genotype and experimental design performed all animal surgeries.
Isolation of Cardiac Fibroblasts
Mice were injected with heparin and euthanized, then the heart was dissected out and perfused with Hanks’ balanced salt solution. The hearts were cut into small pieces and digested with Liberase Blendzyme TH and TM (Roche) in Medium 199 plus DNase I and polaxamer at 37°C for 1 hour. Cells were passed through a 70-μm cell strainer (BD Falcon) and centrifuged. The cell pellet was resuspended in staining buffer (3% fetal bovine serum in Hanks’ balanced salt solution) containing the relevant Thy1+HE− (HE refers to hematopoietic and endothelial lineages) surface marker antibodies (Online Table I) and incubated in the dark for 30 minutes at room temperature.
Mice of the same weight, size, and sex were paired ≥2 to 3 weeks before the surgery, after being observed for at least a week to make sure that they were compatible.9 Mice were anesthetized by isoflurane in O2. Toe pinch was used as an indicator of the pain response. After shaving the corresponding lateral aspects of each mouse, the site was disinfected using an alcohol swab and povidone/iodine. Matching skin incisions are made from the olecranon to the knee joint of each mouse, and the subcutaneous fascia was bluntly dissected to create ≈0.5 cm of free skin. One partner had the procedure done on the right side and the other on the left. The right olecranon of 1 animal is attached to the left olecranon of the other by a single 4-0 silk suture and tie. The partners’ knee joints were similarly connected. The dorsal and ventral skins were then approximated by staples or by continuous suture, and the animals were warmed with heating pads and monitored until recovery. Parabiotic pairs were housed 1 pair per cage and given acidified water (pH 2.5). After 4 weeks of anastomosis, blood samples from each animal in a parabiont pair were analyzed using flow cytometry. Animal pairs with <90% blood chimerism were excluded from our studies. Biological replicates were used, and the experiment was repeated 3 times in the laboratory (n=4–6 per sample; minimum n=3 to observe statistical significance, with more to account for possible animal death).
C57BL/6 mice (H2b, Thy1.1+CD45.2+dsRed+) served as donors for hematopoietic stem cells (HSCs), green fluorescent protein (GFP)-expressing or dsRed-expressing C57BL/6 mice (H2b; Thy1.2+CD45.2+GFP+) were used as donors for nonhematopoietic BM cells for transplantation into C57BL/6 (H2b, Thy1.1+CD45.2+) recipients. Donors were 6 to 12 weeks old, recipients were 8 weeks old at transplant. BM was flushed from tibiae and femora into Hanks’ balanced salt solution/2% fetal bovine serum, enriched for c-Kit (3C11) cells by magnetic column separation (CD117 MicroBeads, MACS Separation Columns LS; Miltenyi Biotec, Auburn, CA), and KTLS (c-Kit+, Thy1.1lo, lineage marker−, Sca-1+)-HSC were purified by FACS (fluorescence-activated cell sorting)-sorting, selecting for c-Kit+, Thy1.2lo-neg, Sca-1+, Lin− (CD3, CD4, CD5, CD8α, B220, Gr1, Mac1, and Ter119). A total of 1000 FACS-purified HSCs were infused per recipient mouse. For cotransfer of nonhematopoietic cells, BM was flushed from tibiae, femora, and pelvis. Nonhematopoietic cells were extracted by magnetic column depletion of CD45+ cells (CD45 MicroBeads; Miltenyi Biotec). C57BL/6.CD45.2 recipients received a lethal 1050 cGy dose total body γ-irradiation, ≈5 hours before tail-vein injection of a radioprotective dose of 1000 KTLS-HSC. In cotransfer experiments, 5×106 CD45− stromal/nonhematopoietic cells were injected simultaneously with the HSCs. Biological replicates were used, and the experiment was performed 3 times.
Statistical testing was performed with Microsoft Excel version 12.2.8 and GraphPad Prism. Results are presented as mean±SEM and were compared using a 2-tailed Student t test or 2-way ANOVA (significance was assigned for P<0.05).
Thy1+HE− Cells Exhibit a Fibroblast Phenotype In Vitro and In Vivo and Encompass the Majority of Cardiac Fibroblasts
We applied a panel of surface markers to identify cardiac fibroblasts: Thy1 as an inclusive surface protein (Thy1+) given its association with cardiac fibroblasts,10,11 and exclusion of hematopoietic cells (CD45−Ter119−), macrophages (CD11b−),2 and endothelial cells (CD31−). In accordance with the reported proportion of cardiac fibroblasts in mice,12 this combination (hereafter called Thy1+HE−) marked ≈30% of the cells in adult mouse hearts (Figure 1A). Using cardiomyocyte-specific and smooth muscle reporter transgenic mice, we show that cardiomyocytes and vascular smooth muscle cells are excluded from the Thy1+HE− population (Online Figure IA and IB). In comparison with age-matched whole heart controls, Thy1+HE− cells exhibit markedly elevated expression of fibroblast- and extracellular matrix (ECM)-associated genes (Figure 1B), whereas cardiomyocyte and vascular cell–related genes are expressed at low levels (Figure 1B, inset). Because individual fibroblast markers may be found on multiple cell types, we hypothesized that individual fibroblasts may not express every fibroblast-associated marker. We quantified the fraction of Thy1+HE− cells from whole hearts coexpressing known fibroblast markers using flow cytometry, and we demonstrate an enriched expression of collagen I (Col1), discoidin domain-containing receptor 2 (DDR2), and platelet-derived growth factor receptor α (PDGFRα) in the Thy1+HE− population compared with the whole heart control (Figure 1C). In addition, immunostaining of Thy1+HE− cells isolated from freshly digested hearts showed robust expression of DDR2 and PDGFRα (Online Figure IC). Notably, expression of α-SMA or CD146 (markers associated with smooth muscle cells and pericytes, respectively) was rarely detected in the Thy1+HE− cells (Online Figure IC). Sorted Thy1+HE− cells were cultured in vitro for several passages (Online Figure ID), and they retained expression of fibroblast-associated ECM and structural proteins. Cultured cells also expressed the fibroblast markers Thy1, DDR2, and PDGFRα (Figure 1C), further demonstrating the fibroblast phenotype of Thy1+HE− cells (fibroblast-specific protein 1 marked a heterogeneous population of cells, including some hematopoietic cells. We therefore restricted our analysis to the aforementioned fibroblast-associated markers).
In culture, treatment of Thy1+HE− cells with transforming growth factor-β1 resulted in upregulation of α-SMA in a dose-dependent manner (Figure 1D).13 Furthermore, RNA-seq of Thy1+HE− cells isolated 7 days after transaortic constriction (TAC) or sham operation revealed an increased expression of ECM and ECM regulatory genes, activated fibroblast genes, and cell cycle genes in the TAC relative to sham specimens (Online Figure IE, IF, and IG), as would be predicted following fibrosis.14 This provides additional validation for the use of the Thy1+HE− panel of markers to isolate cardiac fibroblasts.
To determine whether the Thy1+HE− population encompasses all cardiac fibroblasts, we isolated Thy1−HE− cells from adult wild-type (WT) mouse hearts and examined the expression of the fibroblast-associated markers DDR2 and PDGFRα. We observed rare cells that costained for both markers, suggestive of a fibroblast phenotype (Online Figure IH). Further fractionation revealed the Thy1−HE− cells to include CD146+ cells (pericyte-associated marker) and CD105+ cells (mesenchymal stem cell marker), which stained for DDR2 and PDGFRα (Online Figure II). However, the absolute number of such cells was orders of magnitude lower than the Thy1+HE− population. These data demonstrate that few cardiac fibroblasts are excluded from the Thy1+HE− population.
Cell Proliferation and Programmed Cell Death After TAC Injury
Next, we sought to determine whether increased fibrosis on pressure overload TAC injury occurs through fibroblast proliferation as opposed to an increase in ECM production from a fixed number of fibroblasts (Online Figure IIA and IIB). Adult mice (8 weeks old) underwent TAC or sham operation and were analyzed. Masson’s trichrome staining of the respective tissue revealed sparse global fibrosis at 4 days after injury that increased by day 7 and persisted at 2 weeks (Online Figure IIC) in the TAC hearts. Using flow cytometry, we observed an increase in the number of Thy1+HE− cells as well as an increase in the fraction of Thy1+HE− cells on TAC injury relative to sham (Figure 2A; Online Figure IID). In accordance, ≈10% of the Thy1+HE− cells derived from the operated mice were bromodeoxyuridine (BrdU+) at 7 days after TAC and remained at this level at 4 weeks, compared with 2% in the sham mice (P=0.008; n=3 mice/group/time point; Figure 2A). Immunohistochemistry of TAC specimens showed BrdU+ cells that costained with the fibroblast markers DDR2 and PDGFRα. BrdU+ cells also costained with α-SMA and Col1 (Figure 2B). Together, these data suggest that fibroblast proliferation in response to TAC occurs primarily during the first week after injury.
To determine whether apoptosis of fibroblasts occurs after TAC injury, we analyzed for Annexin V and propidium iodide (PI) expression on Thy1+HE− cells from WT mice at 1, 4, 7, and 14 days after injury. We observed a significant increase in fibroblasts undergoing late apoptosis (defined by Annexin V+PI+ staining) 4 days post TAC (Online Figure IIE) compared with sham. At all other time points, the extent of apoptosis was equivocal between the 2 surgical groups (Online Figure IIE). It is possible, however, that nonapoptotic pathways (necrosis, elimination by the reticulo-endothelial system, etc) also lead to clearance of fibroblasts. We therefore examined for the presence of necrotic cells among the Thy1+HE− fibroblasts by analyzing the uptake of PI. We observed rare necrosis of fibroblasts at the time points studied (7 and 14 days after sham and TAC), with no significant increase in necrosis in the TAC relative to the sham fibroblasts (Online Figure IIF). These data further support that the first week after pressure overload injury is the most dynamic period with respect to cell turnover. Fibroblast apoptosis peaks during this time but is counteracted by a more robust generation of new fibroblasts. The time course of these cellular changes suggests a discrete temporal period for potential therapeutic interventions.
BM and Circulating Cells Do Not Generate Cardiac Fibroblasts During Development, Aging, or After Pressure Overload Injury
Hematopoietic cells have been reported to contribute to the formation of scar tissue after cardiac injury.5 To rigorously test this hypothesis, we transplanted GFP-labeled whole BM cells or red fluorescent protein (RFP)-labeled HSCs along with CD45-depleted GFP-labeled BM stromal cells into adult mice after myeloablation.15 Eight weeks after engraftment, TAC and sham operations were performed (Figure 3A). We observed labeled cells grossly in both surgical groups16 (Online Figure IIIA). However, there were no GFP+ nor RFP+ cells in the Thy1+HE− population 30 days after TAC and sham surgery in either transplant model (Figure 3B). Moreover, HSC-derived RFP+ cells expressed the hematopoietic marker CD45 rather than fibroblast markers (Figure 3C).
To further confirm whether circulating cells can be a cell of origin for cardiac fibroblasts, we surgically paired WT and GFP-labeled female mice to achieve parabiosis (Figure 3D). Blood chimerism was obtained by 4 weeks after anastomosis (Online Figure IIIB and IIIC) and GFP+ cells were observed in the hearts of both mice. To determine whether pressure overload injury facilitates fibroblast or fibroblast-like cell generation by circulating cells, we performed TAC injury on the WT mouse of WT-GFP parabiont pairs 4 weeks after surgical anastomosis. Heart/body weight ratio demonstrated cardiac hypertrophy after TAC injury (Online Figure IIID). Flow cytometric analysis (Figure 3E) and immunohistochemistry (Figure 3F; Online Figure IIIE) at 7 and 30 days after injury did not reveal any GFP+ circulating cells expressing the Thy1+HE− set of markers; instead, the hematopoietic marker CD45 could be detected in GFP+ cells (Figure 3F).
Analysis of the BM transplantation and parabiosis models excludes a hematopoietic or BM stromal contribution to cardiac fibrosis during pressure overload injury. However, a potential developmental contribution of hematopoietic progenitors or lineages to cardiac fibroblasts cannot be excluded; such a phenomenon could reconcile the transplantation data with previous studies.17 To examine whether cardiac fibroblasts have an embryonic hematopoietic origin, we generated VavCre/+;R26RmT/mG mice, in which GFP labels hematopoietic progenitors and their progeny (Online Figure IIIF).18 We did not observe any GFP expression in the Thy1+HE− cell compartment (Online Figure IIIG). Collectively, these distinct models exclude a hematopoietic, BM stromal, or circulating cell origin for cardiac fibroblasts during development, physiological aging, or on pressure overload.19
Endothelial-Derived Cells Generate Cardiac Fibroblasts
It has been reported that endothelial-to-mesenchymal transition (EndMT) contributes to cardiac fibrosis on TAC injury.4 Whether EndMT, which is known to occur during development to generate heart valves,20 also contributes to cardiac fibroblasts during aging and injury remains unclear. We generated Tie2Cre/+;R26RmT/mG mice, in which Tie2-derived progeny is indelibly labeled with GFP (Figure 4A). At 7 and 30 days after sham and TAC operations, ≈20% of Thy1+HE− cells were GFP+ in both groups (P=0.35; n=5 mice/TAC/time point and n=3 mice/sham/time point; Figure 4B). Therefore, the relative proportion of endothelial-derived fibroblasts remains the same despite proliferation after pressure overload, suggesting that fibroblast division on injury is not restricted to the EndMT-derived fibroblast subset.
Immunohistochemistry demonstrated that GFP+ cells expressed DDR2 and PDGFRα, as well as Col1 and Vim, in both sham and TAC hearts (Figure 4C; Online Figure IVA and IVB). Tie2-derived fibroblasts were found in association with most blood vessels, whereas no single vascular territory exhibited a disproportionate number of GFP+ fibroblasts. Some GFP+ cells, predominantly in the fibrotic areas of TAC hearts, colocalized with α-SMA expression, demonstrating the capacity of this lineage to express myofibroblast markers (Figure 4C).
Pax3-Derived Cells Contribute to Cardiac Fibroblasts
The neural crest is a migratory population of cells that originates from the dorsal aspect of the neural tube. Neural crest–derived lineages have been fate-mapped using the Pax3Cre transgenic model21 and the cardiac neural crest has been postulated to contribute to the cardiac mesenchyme.22 Thy1+HE− fibroblasts from WT hearts after sham and TAC operations did not express Pax3, as evidenced by RNA-seq analysis (data not shown). We therefore generated Pax3Cre/+;R26RmT/mG mice to ascertain the cardiac contribution of neural crest cells (Figure 5A). We observed many GFP+ cells in the outflow tract, consistent with the previously described contribution of neural crest cells (Figure 5B).22
The fraction of Thy1+HE− cells that were GFP+ remained ≈5% in both TAC and sham groups at 7 and 30 days post injury (Figure 5C; P=0.4; n=5/TAC/time point and n=3/sham/time point). Most GFP+ cells expressed the neural markers, Smi32 and NF160, on immunohistochemistry (Online Figure VA). However, we also observed rare, scattered GFP+ cells in the atria and ventricles that stained with Col1 and Vim (Figure 5D). We did not detect expression of Pax3 in the GFP+ cells at both the mRNA (data not shown) and protein levels in adult hearts (Online Figure VB), which argues against ectopic expression of Pax3 in cardiac fibroblasts. However, we cannot exclude transient Pax3 expression in cardiac fibroblasts during development. These data support a limited contribution of Pax3-expressing neural crest cells to cardiac fibroblasts, which warrants further inquiry.
Epicardial Cells Form Cardiac Fibroblasts During Development But Not After Pressure Overload Injury
It has been reported widely that cardiac fibroblasts may be derived from the epicardium,23 which forms from the migrating cells of the transient proepicardium after E9.5 in mice.24 We generated Tbx18Cre/+;R26RmT/mG mice (Figure 6A) to determine whether epicardium-derived fibroblasts preferentially proliferate in response to pressure overload injury.25 The GFP+ cells in the Thy1+HE− population remained at ≈75% in both sham and TAC groups at 7 and 30 days post injury (P=0.35; n=5 mice/TAC/time point and n=3 mice/sham/time point; Figure 6B). We observed dispersed GFP+ cells that costained with Vim, DDR2, PDGFRα, and Col1 in sham and TAC heart sections (Figure 6C; Online Figures VIA, VIIA, and VIIB). We also observed GFP-labeled cells whose phenotype (CD146+CD31−) and anatomic location (intimate proximity to CD31+ endothelial cells of small capillaries) is characteristic of pericytes (Figure 6C). In the TAC hearts we also observed a significant amount of GFP+ cells remote from the vasculature (ie, not associated with blood vessels) that costained with α-SMA. Moreover, the majority of epithelial-to-mesenchymal transition (EMT)–derived fibroblasts were observed throughout the interstitium of left and right ventricular free walls.
To determine whether there is reactivation of the epicardium and subsequent EMT on pressure overload injury, we administered tamoxifen to adult Wt1CreERT2/+;R26RmT/mG mice26 before a sham or TAC operation, followed by tamoxifen delivery during the week after surgery (Online Figure VIB and VIC). On analysis 1 week after the sham operation, we observed rare epicardial and myocardial GFP expression, which did not colocalize with expression of fibroblast markers (Figure 6D; Online Figure VIE). Notably, 1 week after the operation, we observed a 5-fold increase in the abundance of GFP+ cells in both the epicardium and the myocardium of TAC when compared with that in sham hearts (Online Figure VID). As in the sham model, these cells did not express fibroblast markers or α-SMA: many were CD146+CD31+ (endothelial markers) and a smaller fraction were CD146+CD31− (pericyte markers; Figure 6D), suggesting that in contrast to the effects of ischemic injury,27 Wt1-expressing cells may not generate fibroblasts in adulthood or on pressure overload injury. These data therefore argue against EMT as the signaling mechanism that underlies proliferation in the epicardial-derived fraction of fibroblasts after injury. On the contrary, the observed pattern of costaining suggests the following explanations: (1) endogenous endothelial expression of WT1 is induced after injury, (2) expression of WT1 in pericytes, and (3) potential generation of pericytes from the epicardium in adulthood and after injury (the expression of Wt1 in endothelial cells has been reported previously).27
Characterization of Pax3-, Tie2-, and Tbx18-Derived Fibroblasts In Vitro and In Vivo
To compare the phenotypic characteristics and gene expression profiles of fibroblasts from different developmental sources, we isolated GFP+ Thy1+HE− cells from each of the 3 Cre models and expanded them in culture. All fibroblast subsets exhibited a spindle-like morphology characteristic of mesenchymal cells and expressed Col1 (Figure 7A). On addition of transforming growth factor-β1 to the culture medium, fibroblasts from all subsets expressed the activated fibroblast marker, α-SMA (Figure 7B). In vitro BrdU uptake assays showed that the relative proliferation rate (based on BrdU optical density) is similar among all 3 developmental subsets at individual time points after BrdU exposure, and this rate increases in parallel in all subsets over time (Online Figure VIIIA).
We next studied the anatomic distribution of each fibroblast subset in sham and TAC hearts 7 days after injury (Figure 7C; Online Figure VIIIB): Tbx18-derived fibroblasts seemed evenly distributed throughout all cardiac chambers, whereas Tie2-derived fibroblasts exhibited a trend toward preferential localization in the ventricles and septum. The majority of Pax3-derived fibroblasts were localized in the right atrium. Interestingly, the relative abundance of each subset with respect to individual chambers did not change on TAC injury.
To directly measure the in vivo proliferative capacity of the fibroblast subsets on TAC injury, we administered BrdU to Tbx18Cre/+;R26RmT/mG and Tie2Cre/+;R26RmT/mG mice starting at the time of surgery until analysis 7 days later (Figure 7D). The fraction of BrdU+ fibroblasts in the sham group was similar between the 2 models (≈5% for Tbx18 versus ≈6% for Tie2; P=0.13; n=3 mice/group) and significantly increased after TAC injury (≈13% for Tbx18, ≈15% for Tie2; P=0.07; n=3 mice/group). Similar percentages of BrdU+ fibroblasts were observed in WT hearts (Figure 2A). These data directly confirm that on pressure overload injury, there is a balanced proliferation in each developmental subset of fibroblasts. Collectively, these findings suggest that the developmental origin affects the anatomic distribution of cardiac fibroblasts but does not result in phenotypic differences with respect to the expression of fibroblast-associated genes, the propensity toward activation in response to appropriate stimuli, or their proliferation rate.
Tbx18- and Tie2-Derived Fibroblasts Have Similar Transcriptional Profiles Upon Pressure Overload Injury
In an effort to determine phenotypic similarities among the different fibroblast developmental subsets, we performed RNA-seq of Tie2-derived and Tbx18-derived fibroblasts isolated 7 days post TAC or sham operation. To quantify and compare the gene expression level within and between the sham and TAC groups (ie, Tie2-derived and Tbx18-derived fibroblasts), we calculated and normalized the expression level of the genes through reads per kilobase of exon model per million mapped reads as described previously.28 The total number of expressed genes (reads per kilobase of exon model per million mapped reads, ≥1) in Tie2 sham and Tbx18 sham was 9198 and 8845, respectively (Figure 7E). Among them, 1950 genes in Tie2 sham and 2277 genes in Tbx18 sham showed a high level of expression (reads per kilobase of exon model per million mapped reads, ≥11), with 1851 common genes between these 2 groups (Figure 7F). Of 100 79 genes expressed in Tie2 TAC, 1828 genes showed a high level of expression, whereas 2711 genes of 8968 genes expressed in Tbx18 TAC were highly expressed (Figure 7F). Among the common genes expressed in sham and TAC fibroblasts were fibroblast- and fibrosis-related markers, whose expression increased after TAC operation (Figure 7G; Online Table III). For example, fibroblast-related genes, such as Col1a1, Col1a2, Postn, Thy1, and DDR2, were expressed at approximately the same level between Tie2 sham and Tbx18 sham fibroblasts; expression of these genes increased in the respective TAC fibroblasts. ECM homeostasis-related genes showed a dynamic variation between sham and TAC fibroblasts: Sparc, Timp1, Mmp14, and Mmp2 showed upregulation in TAC compared with sham fibroblasts, whereas Timp2 and Mmp2 expression levels did not show a significant change (Online Table III). Upregulation of transforming growth factor, tumor necrosis factor, LIF (leukemia inhibitory factor) signaling components on TAC injury suggests a role for these factors during and after pressure overload injury and cardiac hypertrophy. These data further attest that developmentally distinct fibroblasts have parallel gene expression patterns that change similarly in response to pressure overload injury, providing evidence for comparable phenotypes.
The lack of cardiac fibroblast–specific markers has made it challenging to investigate the lineage origin of these cells and their response to injury. The previously reported markers are either not specific for fibroblasts because they are expressed by other interstitial cells or they only mark a subpopulation of fibroblasts. Here, we identified a panel of surface markers (Thy1+HE−) and validated its use for prospective identification of cardiac fibroblasts using several approaches: (1) immunohistochemical and flow cytometric analysis using a combination of fibroblast markers in the context of their anatomic location, (2) global- and fibroblast-specific gene expression analysis, (3) in vivo proliferation in response to injury, and (4) in vitro characterization. This set of core surface markers may form the basis for further refinement and fractionation of cardiac fibroblasts to delineate phenotypically or pathologically distinct fibroblast subsets in future studies.
We used several experimental tools to demonstrate that the progeny of BM and circulating cells can migrate to the heart, especially in the context of injury,16 but they do not adopt a fibroblast phenotype. Use of the Thy1+HE− markers may exclude fibrocytes, which are putative circulating hematopoietic-derived cells that participate in the process of fibrosis and scar formation in the heart after injury.29 However, given previous reports of inflammatory cells that express fibroblast markers but do not deposit ECM (and thus phenotypically are not fibroblasts), better characterization of this cell population is needed, especially given the results from our BM transplantation, parabiosis, and the lineage-tracing hematopoietic models described above.2 Moreover, a recent study demonstrated the near-complete absence of hematopoietic-associated genes (eg, CD45, CD34) in cardiac fibroblasts.30 However, it is possible that pressure overload injury may not be the ideal injury model to study the possible recruitment of circulating progenitors to the heart (to generate fibroblasts), perhaps, because of a less severe inflammatory response as compared with other injury models.
Although it has been shown previously that epicardial cells generate cardiac fibroblasts, our work demonstrates that although EMT is an important developmental process, its induction after injury does not generate fibroblasts (unlike during development). We found rare cells staining for epicardial markers after TAC injury dispersed throughout the myocardium rather than localized to the fibrotic regions, but this likely represents ectopic expression by interstitial and endothelial cells as these cells did not coexpress fibroblast markers. It is possible, however, that epicardial cells undergo EMT on injury and generate mesenchymal cells that are confined to the subepicardial location without migration to the area of fibrosis. It is technically possible that inefficient recombination in the used inducible system or the short time course before analysis may have led to underrepresentation of fibroblasts being generated from EMT in the Wt1-CreERT2 model (although we show that proliferation peaks by 1 week post injury).
Our data corroborate some of the findings of Zeisberg et al4 in identifying endothelial-derived fibroblasts as a subset of cardiac fibroblasts but differs in the absence of evidence for active EndMT on TAC injury. We show that the relative proportion of endothelial-derived fibroblasts remains the same after injury in the setting of a global increase in fibroblasts. By itself, this finding could certainly be explained by EndMT occurring after pressure overload, as this could lead to increased labeling using the Tie2Cre model. However, based on the data obtained with the Tbx18Cre and Pax3Cre models, as well as the fact that the percentage of endothelial-derived fibroblasts remains the same post injury, it can be inferred that proliferation of existing fibroblasts, rather than EndMT, is the primary response to pressure overload. The differences between the 2 studies could be explained by the use of different transgenic fate-mapping mouse models. In our study, we used the Tie2Cre rather than the Tie1Cre model. Developmental studies have demonstrated that the Tie1 promoter drives gene expression in endothelial cells from embryonic day E10 until birth as well as in a small fraction of hematopoietic cells (≈12%–20% of the adult erythroid, myeloid, and lymphoid cells).31 Although the expression pattern is similar to Tie1Cre mice, Tie2 expression starts as early as embryonic day E8.5, hence making it potentially a more suitable mouse model to lineage trace a greater proportion of endothelial cells.32 Moreover, experimental differences such as the length of time from injury to analysis as well as the use of fibroblast-specific protein 1 and α-SMA as markers for the identification of cardiac fibroblasts4 could explain the discrepancies observed between the 2 studies. Fibroblast-specific protein 1 marks the myeloid lineage in addition to fibroblasts,33 which may confound interpretation of fibroblast lineage studies given the abundance of BM-derived cells after TAC injury.
Recently, a comprehensive study by Moore-Morris et al34 was published that corroborates our findings herein using a collagen1a1-GFP reporter mouse, which could identify cardiac fibroblasts based on the expression of GFP. Although they used an alternate method for labeling cardiac fibroblasts, they used similar genetic models (Tbx18Cre, Wt1Cre, Wt1CreER, Tie2Cre, and others) and sought to clarify the embryonic lineages that generate fibroblasts in the heart. In line with our data, they showed that the epicardial and endothelial lineages are the primary contributors (ie, no apparent contribution from hematopoietic cells). On aortic banding injury of adult mice, moreover, they did not observe reactivation of EMT or EndMT; furthermore, fibroblasts of epicardial and endothelial origins had similar gene expression patterns in the setting of aortic banding. Therefore, the independent report by Moore-Morris et al that takes advantage of a unique transgenic reporter model for cardiac fibroblasts validates the findings in this study.
In spite of their developmental heterogeneity, fibroblasts from disparate subsets proliferate at a parallel rate on TAC. The developmental programs that generate fibroblasts in utero do not seem to mediate proliferation after injury, as confirmed by Moore et al. These data suggest that ontogeny does not determine the pathological proliferation of cardiac fibroblasts elicited by pressure overload. It is more likely that a shared mechanism stimulates division of embryonically distinct fibroblast subsets. Moreover, the data derived from the RNA-seq studies could lead to a precise understanding of the signaling pathways that regulate each developmental subset and the entire fibroblast population in response to injury.
Prospective isolation of cardiac fibroblasts using our panel of surface markers may facilitate future studies to characterize signaling pathways that regulate the response of this pathologically important cell type to injury and especially to determine the molecular mechanisms that drive proliferation on injury. Identification of such mechanisms would be therapeutically relevant, as inhibiting fibroblast proliferation may prevent deposition of scar tissue and, thereby, minimize fibrosis.35,36
We acknowledge Konstantina-Ioanna Sereti for critical reading of the article and Dr Matt Schibler for assistance with confocal images. We would also like to acknowledge Dr Xinmin Li and Jian Zhou of the University of California, Los Angeles (UCLA) Clinical Microarray Core with their assistance in RNA sequencing. Flow Cytometry experiments were performed in UCLA Broad Stem Cell Research Center Flow Cytometry Core Resource. S.R. Ali, S. Ranjbarvaziri, and R. Ardehali conceived the project and designed the experiments. S.R. Ali, S. Ranjbarvaziri performed the majority of the experiments and analyzed data from all experiments with R. Ardehali. Figures were prepared by S.R. Ali, S. Ranjbarvaziri, and M. Talkhabi. M. Talkhabi, P. Kamran, and Z. Tang assisted with immunohistochemistry, A. Subat assisted with experimental breeding, and A. Hojjat assisted with RNA-seq experiments. K.S. Volz performed the BrdU experiments. P. Zhao performed the surgeries. A.M.S. Müller performed bone marrow transplantation. K. Red-Horse provided Tbx18Cre/+;R26RmT/mG mice. S.R. Ali, S. Ranjbarvaziri, and R. Ardehali wrote the article.
Sources of Funding
This work was supported in part by a grant from the California Institute of Regenerative Medicine (RC1-00354-1; to R. Ardehali) and from the American Heart Association (AHA-BGA 12BGIA8960008; to R. Ardehali). S.R. Ali was supported by the Howard Hughes Medical Institute Medical Research Fellowship, Stanford Medical Scholars Program, Paul and Daisy Soros Fellowship, and AHA Student Scholarship in Cardiovascular Disease. Confocal laser scanning microscopy was performed at the California NanoSystems Institute Advanced light microscopy/Spectroscopy Shared Resource Facility at UCLA, supported with funding from National Institutes of Health-National Center for Research Resources shared resources grant (CJX1-443835-WS-29646) and National Science Foundation Major Research Instrumentation grant (CHE-0722519).
In June 2014, the average time from submission to first decision for all original research papers submitted to Circulation Research was 15 days.
The online-only Data Supplement is available with this article at http://circres.ahajournals.org/lookup/suppl/doi:10.1161/CIRCRESAHA.115.303794/-/DC1.
- Nonstandard Abbreviations and Acronyms
- α-smooth muscle actin
- bone marrow
- epithelial-to-mesenchymal transition
- endothelial-to-mesenchymal transition
- hematopoietic stem cell
- propidium iodide
- transaortic constriction
- Received February 21, 2014.
- Revision received July 15, 2014.
- Accepted July 17, 2014.
- © 2014 American Heart Association, Inc.
- Österreicher CH,
- Penz-Österreicher M,
- Grivennikov SI,
- Guma M,
- Koltsova EK,
- Datz C,
- Sasik R,
- Hardiman G,
- Karin M,
- Brenner DA
- Ogawa M,
- LaRue AC,
- Drake CJ
- Bergwerff M,
- Verberne ME,
- DeRuiter MC,
- Poelmann RE,
- Gittenberger-de Groot AC
- Gittenberger-de Groot AC,
- Vrancken Peeters MP,
- Mentink MM,
- Gourdie RG,
- Poelmann RE
- Kamran P,
- Sereti K-I,
- Zhao P,
- Ali SR,
- Weissman IL,
- Ardehali R
- Sirish P,
- Li N,
- Liu JY,
- Lee KS,
- Hwang SH,
- Qiu H,
- Zhao C,
- Ma SM,
- López JE,
- Hammock BD,
- Chiamvimonvat N
- Manabe I,
- Shindo T,
- Nagai R
- Stadtfeld M,
- Graf T
- Ross R,
- Everett NB,
- Tyler R
- Rinkevich Y,
- Mori T,
- Sahoo D,
- Xu PX,
- Bermingham JR Jr.,
- Weissman IL
- Furtado MB,
- Costa MW,
- Pranoto EA,
- Salimova E,
- Pinto AR,
- Lam NT,
- Park A,
- Snider P,
- Chandran A,
- Harvey RP,
- Boyd R,
- Conway SJ,
- Pearson J,
- Kaye DM,
- Rosenthal NA
- Gustafsson E,
- Brakebusch C,
- Hietanen K,
- Fässler R
- Braren R,
- Hu H,
- Kim YH,
- Beggs HE,
- Reichardt LF,
- Wang R
- Kong P,
- Christia P,
- Saxena A,
- Su Y,
- Frangogiannis NG
- Friedman SL,
- Sheppard D,
- Duffield JS,
- Violette S
Novelty and Significance
What Is Known?
Cardiac fibroblasts are a developmentally heterogeneous population, reported to have embryonic origins from the hematopoietic system, endothelium, epicardium, and neural crest.
Cardiac fibroblasts play a key role in regulating normal myocardial integrity, as well as reverse remodeling that occurs after injury.
Reactivation of certain developmental gene programs might prime a subset of fibroblast to be preferentially activated after myocardial injury.
What New Information Does this Article Contribute?
We characterize a combination of surface markers that can be used to prospectively identify and isolate the majority of cardiac fibroblasts using FACS (fluorescence-activated cell sorting).
The cardiac fibroblast pool is primarily derived from the epicardial and endothelial lineages, with no ostensible contribution from hematopoietic or circulating cells.
On injury, cardiac fibroblasts from different lineages exhibit similar proliferation rates and gene expression patterns, suggesting that cardiac fibroblasts are functionally identical in spite of distinct developmental origins.
Studies of cardiac fibroblasts have been limited by a lack of universally accepted markers that enable accurate identification and characterization of this developmentally heterogeneous population. Moreover, previous reports implied that a subset of developmentally distinct fibroblasts is preferentially activated on injury. We used a panel of surface markers to isolate cardiac fibroblasts and studied their developmental origins using lineage-tracing experiments. We found that most cardiac fibroblasts are derived from the epicardium and the endothelium during development. Using multiple models, we show that hematopoietic cells do not generate cardiac fibroblasts. Importantly, we discovered that epicardial- and endothelial-derived fibroblasts proliferate at a similar rate and have a similar pattern of gene expression in response to pressure overload–induced stress/injury. These findings suggest that developmental origin of cardiac fibroblasts does not affect their pathological propensity; rather, cardiac fibroblasts may be functionally similar. Our findings suggest that therapeutic targets for pathological cardiac fibrosis should aim to identify the common signaling pathways that are activated on injury in all fibroblasts.