TNF, acting through inducibly expressed TNFR2, drives activation and cell cycle entry of c-Kit+ cardiac stem cells in ischemic heart disease


  • Author contributions: R.S.A.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; W.L., J.W., R.W., and C.S.: collection and assembly of data; J.Y., T.J.S., and P.W.: data analysis and interpretation; M.G., Q.H., A.H.L., G.T., Y.H., and W.M.: provision of study material or patients; J.S.P.: conception and design, data analysis and interpretation, editing and final approval of manuscript; J.R.B: conception and design, data analysis and interpretation, editing and final approval of manuscript, and financial support.


TNF, signaling through TNFR2, has been implicated in tissue repair, a process that in the heart may be mediated by activated resident cardiac stem cells (CSCs). The objective of our study is to determine whether ligation of TNFR2 can induce activation of resident CSCs in the setting of ischemic cardiac injury. We show that in human cardiac tissue affected by ischemia heart disease (IHD), TNFR2 is expressed on intrinsic CSCs, identified as c-kit+/CD45/VEGFR2 interstitial round cells, which are activated as determined by entry to cell cycle and expression of Lin-28. Wild-type mouse heart organ cultures subjected to hypoxic conditions both increase cardiac TNF expression and show induced TNFR2 and Lin-28 expression in c-kit+ CSCs that have entered cell cycle. These CSC responses are enhanced by exogenous TNF. TNFR2−/− mouse heart organ cultures subjected to hypoxia increase cardiac TNF but fail to induce CSC activation. Similarly, c-kit+ CSCs isolated from mouse hearts exposed to hypoxia or TNF show induction of Lin-28, TNFR2, cell cycle entry, and cardiogenic marker, α-sarcomeric actin (α-SA), responses more pronounced by hypoxia in combination with TNF. Knockdown of Lin-28 by siRNA results in reduced levels of TNFR2 expression, cell cycle entry, and diminished expression of α-SA. We conclude that hypoxia-induced c-kit+ CSC activation is mediated by TNF/TNFR2/Lin-28 signaling. These observations suggest that TNFR2 signaling in resident c-kit+ CSCs induces cardiac repair, findings which provide further understanding of the unanticipated harmful effects of TNF blockade in human IHD. Stem Cells 2013;31:1881-1892


Tumor necrosis factor (TNF, also known as TNF-α) is an inflammatory mediator produced during ischemia and implicated in ischemic myocardial damage and the development of heart failure [1, 2]. Circulating levels of TNF are higher in acute myocardial infarction patients who developed heart failure compared to patients who do not [3], with a direct correlation between functional capacity, survival, and circulating TNF levels [3, 4]. Experiments in mice had suggested that TNF neutralization could be beneficial [4], but clinical trials using TNF antagonists have been unsuccessful or even harmful [5]. This discrepancy is unresolved.

TNF exerts its effects by binding to two distinct cell surface receptors, TNFR1 (p55 or CD120a) and TNFR2 (p75 or CD120b) [6]. TNFR1 is the main receptor subtype in most cell types in resting tissues, including the heart, and its downstream signaling pathways have been studied extensively. Engagement of TNFR1 exacerbates, whereas engagement of TNFR2 ameliorates myocardial damage [7]. We have previously reported that expression of TNFR1 is decreased while that of TNFR2 is increased on both human kidney and heart cells in the setting of transplant rejection [8–12]. The precise signals that cause these changes are not known, but TNF and other proinflammatory cytokines can increase the expression of TNFR2 through new transcription, whereas TNFR1 is more commonly downregulated by these same stimuli [13, 14]. These observations have led us to hypothesize that an altered expression and activation of TNFRs in human myocardium could contribute to the failure of anti-TNF therapeutics in ischemic heart disease (IHD). However, the cell types and mechanisms activated by TNF signaling through TNFR2 that improve cardiac function are unknown.

Recent compelling evidence has accumulated suggesting that the adult heart has regenerative potential [15, 16], mediated by resident cardiac stem cells (CSCs) with the ability to self-renew and differentiate [17]. CSCs expressing c-kit have been isolated from human and murine heart tissue and are significantly increased in number in myocardial infarction and in acute ischemia-induced cardiac injury [18–21]. Results from recent phase I trials of autologous c-kit+ CSCs in patients with ischemic cardiomyopathy are encouraging [22].

Lin-28 is a RNA-binding protein essential for maintaining the pluripotency, growth, and survival of embryonic stem cells [23, 24]. Lin-28 has been cited as a marker of “stemness” [25]. Importantly, Lin-28 has been shown to mediate mouse skeletal myogenesis [26]. While barely detectable in mouse resting muscle [27], Lin-28 expression was strongly upregulated during regeneration of skeletal muscle fibers, followed by downregulation upon completion of regeneration. This biphasic expression pattern of Lin-28 was recapitulated in cultured adult mouse primary myoblasts where Lin-28 involved in post-transcriptional regulation of insulin-like growth factor 2 was extremely low during proliferation and was dramatically induced as early as 24 hours following induction of differentiation. The high level of Lin-28 expression lasted for several days until completion of differentiation when it returned to the basal level suggesting a role in differentiation of muscle tissue [26]. Likewise, in adult mouse small intestine, Lin-28 protein was neither detected in the basal stem cells, nor terminally differentiated villous cells, but it was detected in a patch of cells between the basal stem cells and the mature villous cells [27]. Thus, during regeneration of mouse skeletal muscle and small intestinal villi, induction of Lin-28 expression marks activation and differentiation of resident stem cells. Indeed, recent studies have shown significant function of baseline levels of Lin-28 in lentiviral-infected rat cardiomyoblasts in regulation of pre-miR-1 [28]. However, while it is not known definitively whether induction of Lin-28 also occurs in human myocardial regeneration, increased expression of Lin-28 in c-kit expressing CSCs is likely to indicate early activation and possible commitment of these cells to the myocyte lineage.

Here we report that in human heart tissue affected by IHD there is both increased TNF expression by various cell populations and induction of TNFR2 expression on intrinsic c-kit+CD45VEGFR2 CSCs. By multicolor immunofluorescence microscopy increased expression of TNFR2 on these cells correlates both with increased expression of Lin-28, marker of activation and potentially of commitment to the cardiac lineage, and with evidence of cell cycle entry. Similar changes in c-kit+ CSCs can be recapitulated in wild-type (WT) (but not TNFR2 null) mouse cardiac organ cultures and in c-kit+ CSCs isolated from mouse hearts subjected to hypoxia. These observations are consistent with the idea that TNFR2 signaling in resident c-kit+ CSCs induces cardiac repair and may help to explain the unanticipated harmful effects of TNF blockade in human IHD.

Materials and Methods

Human Heart Tissue

Twelve human heart samples were obtained under protocols approved by the Institutional Review Boards of Yale University and of the New England Organ Bank. These samples comprised two groups of six specimens each: (a) normal myocardium (NM) from patients who died for reasons other than cardiovascular diseases with no history of coronary artery disease and (b) myocardium affected by IHD from patients with clinical and pathological evidence of prior myocardial infarctions with advanced heart failure sufficient to be listed for cardiac transplantation. These samples did not differ significantly by age, that is, they were age-matched with mean ± SEM for NM of 47.67 ± 6.43 years old and IHD of 51.33 ± 3.35 years old; two-tailed t test with p value = .06244, that is, [mt] .05. The characteristic of patients' samples is presented in supporting information Table 1. All samples were processed for histology and analyzed by immunohistochemistry, immunoblotting, in situ hybridization (ISH), and quantitative real-time polymerase chain reaction (qRT-PCR).

Murine Heart Organ Cultures

Sacrifice of mice was performed under a protocol approved by the Yale Institutional Animal Care and Use Committee. Pieces of heart tissue from WT C57BL/6 and TNFR2−/− (B6.129-tnfrsflb) mice, purchased from Jackson Laboratory (Bar Harbor, ME, http://www.jax.or) were obtained immediately from surgically excised specimens. Duplicate <1 mm3 fragments of tissue were placed in flat-bottomed 96-well tissue culture plate (Appleton Woods Limited, Birmingham, U.K.) in complete culture medium M199 and incubated at room air plus 5% CO2 or in hypoxic condition in 1% O2 and 5% CO2 in a controlled environment chamber (MACS-MG-1000 Anaerobic workstation, Don Whitley Scientific, U.K.) maintained at a humidified temperature of 36°C ± 1°C with or without recombinant murine TNF (rmTNF) (Abingdon, Oxfordshire, UK, for 0, 3, 6 or 18 hours. A dose-response curve showed that both TNFRs were activated in the same concentration range. An optimal concentration of 10 ng/mL was used in all reported experiments. Multiple randomized samples from each patient were used to obtain parallel group comparisons and to assess the reliability and reproducibility of these assays. Some cultures were incubated in media alone (untreated) or pretreated with 10 ng/mL rmTNF with or without various concentrations (150, 300, and 600 µM) of pimonidazole hydrochloride (hypoxyprobe-1) (HPI, Burlington) to monitor low oxygen condition. Cultures were then harvested and either snap-frozen in isopentane-cooled liquid nitrogen or fixed in 4% formaldehyde for paraffin-wax embedding. Five µm-thick paraffin sections of all the samples were stained with hematoxylin and eosin (H&E) for morphological analysis, and the diagnosis in all cases was verified independently by two experienced pathologists and was based entirely on examination of routinely stained slides.


Paraffin-wax sections of NM, IHDM, and murine heart organ cultures were immunostained for TNF, TNFR1, or TNFR2, and α-sarcomeric actin (α-SA, marker for cardiomyocytes [CMs]) as previously described [8, 9, 12]. To assess the presence of cardiac precursor cells in human and mouse heart we have used anti-c-kit (CD117) [18] and anti-α-SA or -CD45 (pan-leukocyte marker) or -VEGFR2 (also known as flk-1 in mice or Kinase Insert Domain Receptor in humans) [17]. Parallel sections were coimmunostained for c-kit and Lin-28 or TNFR2 or phospho-Histone H3S10 (pH3S10) (nuclear protein involved in the cell cycle), followed by fluorochrome-conjugated secondary antibodies and Hoechst 33342 for nuclei detection before viewing on a Leica TCS-SPE confocal microscopy. Mouse neural stem cells were used as positive controls for c-kit and Lin-28 [29, 30] and negative controls included replacement of the primary antibodies with isotype-matched antisera. See supporting information data for detailed method and antibodies/reagents used.

Detection of Hypoxyprobe-1 in Murine Heart Organ Cultures

Exposure of murine heart organ cultures to low oxygen conditions was assessed using anti-hypoxyprobe-1 antibody as previously described [31]. See supporting information data for detailed method.

ISH and qRT-PCR and Immunoblotting

Paraffin-wax sections of NM and IHDM were hybridized with digoxigenin-labeled anti-sense probes specific for human c-kit and Lin-28 and murine organ cultures with probes specific to mouse TNF, TNFR1, and TNFR2 (MWG-Biotech, U.K.) as previously described [10, 11]. Gene expression was visualized using alkaline-phosphatase/5-bromo-4-chloro-3-indolylphosphate/nitro blue tetrazolium substrate (Sigma-Aldrich, U.K. Corresponding sense probes were used as negative controls. Probe sequences are provided in supporting information text. For qRT-PCR, total RNA was isolated using Taqman expression systems. TaqMan_Gene Expression Assay ID: TNFR1 (Hs00533560), TNFR2 (Hs00153550), TNFR (Hs99999043), c-kit (Hs00174029), and Lin-28 (Hs00702808) (Applied Biosystems, Foster City, CA, RPLPO (Hs99999902_m1) and TBP (Hs99999910_m1) were used for normalization. Heart tissue homogenates from same study groups were analyzed by immunoblotting as previously described [11] using antibodies to TNFR1, TNFR2, TNF, pH3S10, Lin-28, c-kit and normalized to β-actin. Immunoblots were quantified using Image J software version v1.47k. See supporting information data for detailed methods.

Processing, Isolation, and Culture of C-Kit+ Cells from the Adult Murine Heart

c-kit+ CSCs were enzymatically dissociated from 30 adult mice hearts (10–13 week old) with NOD or C57BL6 background into single cell suspension using a previously described method [32]. Hearts were dissected into two halves and rinsed thoroughly with Ca2+-Mg2+-free phosphate-buffered solution (PBS) (Invitrogen, Paisley, U.K., Enzymatic dissociation of the hearts was carried out using collagenase II (∼600 U/mL; Lorna Laboratories, U.K.), DNase I (∼12 U/mL, Invitrogen, U.K.), and HBSS on a GentleMACs Disassociator (Miltenyi Biotec, Bergisch Gladbach, Germany, according to the manufacturer's instructions. Tissue was then incubated on a MACsMix rotator (Miltenyi) for 30 minutes at 37°C, centrifuged at 300g for 5 minute cell suspension, passed through 40 µm nylon mesh to remove cell clumps, rinsed with HBSS and centrifuged at 300g for 10 minutes. Single cell suspension was suspended in cardiosphere growth medium (35% complete Iscove's Modified Dulbecco's Medium/65% Dulbecco's modified Eagle's medium [DMEM]-Ham F-12 mix containing 2% B27, 0.1 mmol/L 2-mercaptoethanol, 10 ng/mL epidermal growth factor [EGF], 20 ng/mL basic fibroblast growth factor [bFGF], 40 nmol/L cardiotrophin-1, 40 nmol/L thrombin, antibiotics, and l-Glu) [32]. The cells were cultured on sterile Petri dishes and incubated for 3–5 days and collected using 0.1% trypin/0.2% EDTA, and enrichment of the c-kit+ cells was achieved by magnetic cell sorting system (MACS), Miltenyi Biotech). A total of 6 × 108 cells were magnetically labeled with CD117 Microbeads in PBE buffer [PBS, 2 mM EDTA, and 0.5% BSA] for 15 minutes at 4°C according to the manufacturer's instructions and passed through the MACS column placed in the magnetic field of a MiniMACS Separator. The magnetically labeled CD117+ cells were retained within the column and eluted as the positively selected cell fraction. Enriched CD-117+ cells approximately 6 × 105 cells per milliliter were seeded in poly-l-lysine-coated eight-well slide chambers (Thermo-scientific, Essex, U.K.) in a differentiation medium (IMDM supplemented with 10% fetal calf serum, 100 U/mL penicillin G, l-glutamine DMEM-Ham F-12 mix containing 2% B27, 0.1 mmol/L 2-mercaptoethanol, 10 ng/mL epidermal growth factor [EGF], 20 ng/mL basic fibroblast growth factor [bFGF], 40 nmol/L cardiotrophin-1, 40 nmol/L thrombin, l-glutamine,10 µM 5′-azacytidine [33], 100 nM oxytocin [33], and 40 nM thrombin [32]). All reagents were from Invitrogen, except 5′-azacytidine and 2-mercaptoethanol (Sigma-Aldrich, Gillingham, U.K.,, EGF and bFGF were from R&D Systems, Oxford, U.K. ( After 2 days the medium was replaced with fresh growth medium. Once confluent, the CD117+ cells were incubated in media alone in normoxic conditions (untreated controls) or treated with 10 ng/mL recombinant mouse TNF (R&D systems) and/or exposed to hypoxia for 6 hours.

RNA Interference

c-kit+ CSCs from adult mouse hearts were transfected with control siRNA (siRNAControl; cat∼sc-37007; Santa Cruz Biotechnology, Santa Cruz, CA, or On-TARGET plus Non-targeting siRNA cat∼D-001810-01-05; Dharmacon) or with RNA targeting mouse Lin-28 (siRNALin-28-1; cat∼sc-106990; Santa Cruz Biotechnology) or TARGET plus mouse Lin28siRNA-SMART-pool (siRNALin-28-2; cat∼L-051530-01; Dharmacon) at 40 nM in the presence of TurboFect Transfection Reagent (Fermentas, Cambridge, U.K.) according to the manufacturer's instructions for 48 hours prior to treatment with rmTNF (10 ng/mL) and/or hypoxia for 6 hours. Cells were immunostained using R-Phycoerythrin-conjugated anti-CD-117 or rabbit anti-c-kit antibody and biotin-conjugated anti-Lineage marker (comprising a panel of monoclonal antibodies that recognizes all mature hematopoietic lineages) and a mouse monoclonal antibody to Lin-28, pH3S10, anti-TNFR2, or anti-α-SA, followed by secondary antibody with anti-rabbit or anti-mouse-Northern Light-498 (NL−498) or NL−557 (R&D Systems) or anti-Streptavidin-Fluorescein isothiocyanate or Texas Red (Vector Laboratories, Peterborough, U.K.,, and examined on a Leica SPE confocal microscope (Leica Microsystems, Knowlhill, Milton Keynes, U.K.). Image for each fluorophore was acquired sequentially using the same constant acquisition time and settings rather than simultaneously to avoid crosstalk between channels. We have examined the specificity of the Lin-28siRNAs using the mouse teratocarcoma cell line P19 [29].

Data Analysis

All results are expressed as mean ± SEM unless otherwise stated. Positive cells in NM and IHDM human samples were counted in 10 randomly chosen fields of view at ×40 Mag (with each field of view containing at least 280 cells, most of which are CMs approximately 84% and the remaining are interstitial cells (ICs) and vascular structures). Sections were viewed using a Nikon OPTIPHOT-2 (Nikon, Surrey, U.K.) in a blind-folded manner. Statistical differences analyses were performed using Student's t test or ANOVA followed by Bonferroni's correction in GraphPad Prism version 5.02. Isolated c-kit+CSCs from murine hearts were scored in a similar manner (with each field of view containing at least 30 cells). The total number of positive cells from 10 random fields was divided by the total cell numbers to generate the % of positive cells. Each mouse heart experiment was repeated three times and the same statistically significant differences between experimental groups were observed in all three independent experiments although the absolute values varied.


TNFR2 and pH3S10 Are Increased in IHDM

Specimens of NM showed no abnormalities on H&E stained sections (supporting information Fig. 1A), whereas sections of IHDM displayed CM damage and/or loss, areas of fibrosis containing cellular debris, and disrupted vasculature with numerous small, round ICs present within the fibrotic regions (supporting information Fig. 1B). The expression of TNF and TNFRs in NM and IHDM were examined by immunofluorescence (IF), ISH, and qRT-PCR (Fig. 1A, 1B, quantified in 1C; supporting information Fig. 2). Consistent with our previous findings [8], TNFR1 protein and mRNA were constitutively expressed in CMs, identified by expression of α-SA, in NM; TNFR1 was also present in some microvessels (possibly vascular endothelial cells [VECs] and/or perivascular cells), occasional ICs, and in rare fibroblasts (sections not shown). In contrast, TNFR2 and TNF protein and transcripts were mainly confined to ICs and microvessels. Compared to NM, sections of IHDM demonstrated reduced expression of TNFR1 in CMs, but a strong signal in VECs and ICs, and increased TNFR2 expression in CMs, VECs, and ICs. Similar to NM, TNF expression in IHDM specimens was mainly confined to VECs and ICs (supporting information Table 2). By qRT-PCR (Fig. 1C and supporting information Fig. 2), tissue extracts demonstrated a statistically significant increase in TNFR1 transcripts in NM compared to IHDM (***p < .0001). In contrast, TNF (*p < .05) and TNFR2 (***p < .0001) transcripts were both markedly increased in IHDM compared to NM. Collectively, these data indicate that TNF and TNFRs are differentially expressed in NM and IHDM and, more specifically, that TNFR2 expression increases with ischemic injury.

Figure 1.

Combined immunofluorescence and in situ hybridization for TNF, TNFRs, and pH3S10 in NM and IHDM. (A): NM show TNFR1 protein and mRNA expression in CMs, also positive for α-SA (marker for CMs), in VECs and in ICs while TNFR2 and TNF protein and mRNA are confined to VECs and ICs but absent in CMs. (B): In contrast in IHDM, TNFR1 protein and mRNA is downregulated in CMs and mainly present in VECs and ICs. In comparison, TNFR2 and TNF protein and mRNA expression is upregulated in CMs, VECs, and ICs. (C): Quantitative real-time polymerase chain reaction of tissue extracts from same study groups show strong signal for TNFR1 mRNA in NM versus IHDM, with increased TNF and TNFR2 mRNA in IHDM versus NM. *, p < .05; ***, p < .0001. Expression (D, E) and quantification (F) of pH3S10 show a rare signal in NM, more pronounced in IHDM with a higher level of expression in VECs and ICs versus CMs.*, p < .05; , p < .05. (G): Colocalization of TNFR2 and pH3S10 is seen in CMs, VECs, and ICs in IHDM. (H): Immunoblot of the same study groups show upregulation of TNF, TNFR2, and pH3S10 and downregulation of TNFR1 in IHDM with β-actin as loading protein. (I): Densitometric analysis, normalized to β-actin levels, with one sample for each protein set to 1. Results expressed as mean ± SD; n = 4 per study group. ×40 magnification, scale bars = 50 µm; (E)-zoomed ×1.22, scale bar = 50 µm; (D)-scale bar = 75 µm; (G)-zoomed ×2.39, scale bar = 75 µm; confocal images; n = 6 per study group with similar findings in two other independent experiments; immunoblot; n = 4 per study group. Abbreviations: CMs, cardiomyocytes; ICs, interstitial cells; IHDM, ischemic heart disease myocardium; NM, normal myocardium; pH3S10, phospho-histone H3S10; TNF, tumor necrosis factor; VECs, vascular endothelial cells; α-SA, α-sarcomeric actin.

We recently demonstrated an association between increased TNFR2 expression and cell cycle entry of CMs and VECs in injured myocardium of cardiac allografts [8]. We therefore compared the expression of pH3S10, an indicator of cell cycle entry, in NM and IHDM. A negligible level of staining was detected in CMs in NM (Fig. 1D). In contrast, IHDM showed a marked increase in pH3S10 in α-SA-positive CMs and in VECs and ICs with the highest level of pH3S10 detected in VECs and in ICs (p < .001) (Fig. 1E, quantified in 1F; supporting information Table 2). Colocalization of pH3S10 and TNFR2 is detected in CMs, VECs, and in ICs (Fig. 1G). A similar pattern and frequency of colocalization for TNFR2 and Ki67 or Proliferating Cell Nuclear Antigen was seen in IHDM (supporting information Fig. 3). Immunoblotting confirmed increased levels of expression of TNF, TNFR2, and pH3S10 in extracts of myocardium from IHD specimens compared to NM (Fig. 1H, quantified in Fig. 1I).

CSCs Positive for C-Kit Are Increased and Coexpress Lin-28 in IHDM

Experimental and clinical studies in humans and animals indicate the presence of precursor cells within the heart that have the capacity to reconstitute and restore cardiac function in damaged myocardium [15, 16, 34]. To determine whether the small round ICs observed in our heart samples are potential cardiogenic precursors, sections of NM or IHDM were immunostained for expression of c-kit and α-SA (Fig. 2A). In NM, only occasional ICs showed strong signal for c-kit while negative for α-SA (≤3 cells in high power field at ×40 magnification). In contrast, a striking increase in c-kit+ cells (≥12 cells per hpf) was observed in sections of IHDM mainly confined to ischemic zones. No signal for c-kit was detected in negative controls when anti-c-kit antibody was replaced with an isotype-matched nonspecific antibody (supporting information Fig. 4A) and a strong signal present in mouse neural stem cells (positive controls) (supporting information Fig. 4B). The majority of these c-kit+ cells are small round ICs that lack CD45 or VEGFR2 [17] (Fig. 2B, quantified in Fig. 2C) and are putative CSCs. In skeletal muscle and small intestinal epithelium, activated stem cells express Lin-28 [26, 27]. We therefore determined whether Lin-28 is expressed in IHDM and whether it colocalizes with c-kit by IF. Only a few Lin-28+ ICs cells were present in NM compared to IHDM, which showed an increased number of ICs positive for Lin-28, negative for α-SA or CD45 or VEGFR2 (supporting information Fig. 5A–5E, quantified in supporting information Table 2). No signal for Lin-28 was detected in IHDM when anti-Lin-28 antibody was replaced with an isotype-matched nonspecific antibody but a strong signal was seen in mouse neural stem cells (positive control) (supporting information Fig. 5D, 5E). Of particular interest, Lin-28 is detected in c-kit+ CSCs (≤ 8 cells per hpf; Fig. 3A, quantified in Fig. 3B). We interpret c-kit+CD45VEGFR2 small round ICs that express Lin-28 as being activated resident CSCs. Immunoblotting confirmed increased expression of c-kit and Lin-28 in IHDM compared to NM (Fig. 3C, quantified in Fig. 3D). Gene expression was concordant with protein expression; transcripts of both markers were detected in NM, with a striking increase expression in IHDM (***p < .0001) (Fig. 3E–3G; supporting information Fig. 5F). Most c-kit+ CSCs in IHDM were also TNFR2+, with only a rare CSC TNFR2, with some c-kit+ CSCs also positive for pH3S10 (Fig. 4A, quantified in Fig. 4B), providing evidence of cell cycle entry. These findings by IF are supported by changes in protein expression detected by immunoblotting, by ISH analysis of tissues and by qRT-PCR measurements of transcripts in tissue extracts. Collectively, these observations are consistent with an association between TNFR2 expression, cell cycle entry, and activation of resident CSCs found in cardiac tissue from patients with IHD and suggest the hypothesis that TNF signaling through induced TNFR2 on c-kit+ CSCs results in stem cell activation and cell proliferation.

Figure 2.

Colocalization of c-kit and α-sarcomeric actin or CD45 or VEGFR2 in NM and IHDM. (A): NM show a few c-kit+ CSCs, negative for α-SA, noticeably increased in IHDM (arrows). (B): c-kit+ cardiac stem cells (CSCs) are CD45 and VEGFR2 (arrows), with an occasional c-kit+ CSC, CD45+ and VEGFR2+, with the latter also detected in some blood vessels (Bv, open arrows). (C): Quantification of c-kit+/CD45 and c-kit+/VEGFR2 CSCs. **, p < .001; Bars = mean ± SEM; scale bars = (A), 50 µm; (B), upper panel 10 µm; lower panel 50 µm; n = 6 per study group with similar findings in two other independent experiments. Abbreviations: Bv, blood vessel; CM, cardiomyocyte; IHDM, ischemic heart disease myocardium; NM, normal myocardium; α-SA, α-sarcomeric actin.

Figure 3.

Protein and gene expression for c-kit+ and Lin-28 in NM and IHDM. (A): Sections of IHDM show some CSCs strongly c-kit+/Lin-28+ (arrows), with a few cells c-kit+/Lin-28 (open arrows). (B): Quantification of immunostaining show increased number of c-kit+/Lin-28+ CSCs in IHDM versus NM (***, p < .0001). (C): Immunoblot show increased c-kit and Lin-28 in IHDM with β-actin used as loading protein. (D): Densitometric analysis, normalized to β-actin levels, with one of the samples for each protein set to 1, n = 4 per study group. (E, F): Quantification of in situ hybridization and qRT-PCR show an increased level of c-kit and Lin-28 transcripts in IHDM versus NM (***, p < .0001). (G): ISH of NM and IHDM show CSCs positive for c-kit and Lin-28 mRNA (arrowheads); scale bars = (A), 25 µm; (G), 75 µm; n = 6 per study group with similar findings in two other independent experiments. Abbreviations: CSCs, cardiac stem cells; IHDM, ischemic heart disease myocardium; NM, normal myocardium; qRT-PCR, quantitative real-time polymerase chain reaction.

Figure 4.

Colocalization for c-kit and TNFR2 or pH3S10 in NM and IHDM. (A): TNFR2 or pH3S10 is seen in some c-kit+ CSCs in IHDM (open arrows), with a few c-kit+/TNFR2 or c-kit/pH3S10+CSCs (arrows). (B): Quantification of immunostaining show increased c-kit+/TNFR2+and c-kit+/pH3S10+ cardiac stem cells (CSCs) in IHDM versus NM (***, p < .0001), scale bars; 25 µm; n = 6 per study group with similar findings in two other independent experiments. Abbreviations: IHDM, ischemic heart disease myocardium; NM, normal myocardium.

Hypoxia and/or TNF Induce Upregulation of TNF and TNFR2 in CMs in Murine Heart Organ Cultures

We next analyzed the effects of stimuli associated with ischemia on expression of TNF and TNFRs in murine heart organ cultures. Cultures were incubated either in normoxic or hypoxic conditions at various time points (0, 3, 6, or 18 hours). Low oxygen conditions were confirmed by immunostaining for hypoxyprobe-1 (supporting information Fig. 6). Normoxic cultures showed a constitutive expression of TNFR1 in CMs, VECs, and ICs, with TNF or TNFR2 expression mainly confined to VECs and ICs (sections not shown). In contrast, exposure to hypoxia for up to 18 hours resulted in a diminished expression of TNFR1 in CMs but induced an increase in TNF expression in VECs and ICs and increased TNFR2 expression in VECs, ICs, and CMs (supporting information Table 3). To assess whether the effects of hypoxia can be induced by TNF, we added exogenous TNF to normoxic or hypoxic cultures. Addition of exogenous TNF in normoxic cultures induced TNF expression in VECs and ICs and markedly increased TNFR2 expression in CMs, VECs, and ICs at all the times points, with an even greater increase in the intensity of staining observed in cultures incubated in hypoxia in combination with TNF (supporting information Table 3; supporting information Fig. 7A). Gene expression analysis by ISH (supporting information Fig. 7B) correlated with protein expression showing increased levels of TNF mRNA in VECs and ICs following TNF treatment, indicating that exogenous TNF was inducing endogenous TNF production in these cells. TNF and/or hypoxia resulted in a similar induction of TNF expression in WT and TNFR2−/− mice (supporting information Fig. 7C). Hypoxic conditions caused a time-dependent expression of TNFR2 mRNA in VECs, ICs, and CMs and TNF in combination with hypoxia induced a more marked increase in TNFR2 mRNA expression in all three cell types. As expected, no TNFR2 mRNA was detected in TNFR2−/−cultures. These data suggest that hypoxia-induced TNF could be responsible for the up-regulation of TNFR2 and the downregulation of TNFR1 observed in heart cells in the setting of ischemia.

Hypoxia and/or TNF Induce Expression and Activation of c-kit+ CSCs and Expression of TNFR2 and Cell Cycle Entry in c-kit+ CSCs in Organ Cultures of Cardiac Tissue from WT but not TNFR2−/− Mice

We next assessed the presence of resident CSCs in mouse hearts and whether hypoxia and TNF effect their activation. Analysis of murine heart tissue confirmed the presence of a population of interstitial c-kit+ CSCs that lacked CD45 and VEGFR2, which accounted for approximately 45% of the c-kit+ population (Fig. 5A). Hypoxia induced an increase in the percentage of c-kit+ cells expressing Lin-28 in a time-dependent manner with the most marked increase in expression occurred after 18 hours of treatment with hypoxia and TNF in combination (Fig. 5B; Table 1). Following 18 hours of culture under control conditions 17% of c-kit+ cells expressed Lin-28 and after 18 hours of treatment under hypoxic conditions with TNF 69% of c-kit+ cells expressed Lin-28. The number of CSCs expressing Lin-28 was negligible in hearts from TNFR2−/− mice compared WT cultures, and TNF and hypoxia did not alter expression of this activation marker (Table 1). Hypoxia and/or TNF also increased the percentage of c-kit+ cells expressing TNFR2 and entering cell cycle. Following 18 hours of culture under control conditions 1% of c-kit+ cells expressed TNFR2, and 3% stained positive for pH3S10. After 18 hours of treatment under hypoxic conditions with TNF, 26% of c-kit+ cells expressed TNFR2 and 28% stained positive for pH3S10 (Table 1; Fig. 5C). In contrast to WT, pH3S10expression was negligible in TNFR2−/− hearts (Table 1). These data indicate that both TNF and hypoxia induce activation of c-kit+ CSCs and that the presence of TNFR2 is important for TNF-mediated activation of CSCs and subsequent cell cycle entry.

Figure 5.

Hypoxia and/or TNF induce activation of c-kit+CSCs in murine heart organ cultures. (A): Representative confocal images show c-kit+/CD45/VEGFR2 CSCs, (open arrows), with a few c-kit+/CD45+ (arrows), and VEGFR2+Bv. (B): TNF plus hypoxia result in a sequential induction of Lin-28 in c-kit+ CSCs, with the highest level of expression in cultures incubated for 18 hours versus 0, 3, or 6 hours. (C): Cultures incubated in TNF plus hypoxia for 18 hours show induction of TNFR2 and pH3S10 in c-kit+ CSCs (open arrows), with a few c-kit+/TNFR2 CSCs (arrows). Scale bars = 50 µm; n = 3 per study group with similar findings in two other independent experiments. Abbreviations: Bv, blood vessel; CSCs, cardiac stem cell; ICs, interstitial cells; TNF, tumor necrosis factor

Table 1. Quantitative analysis of two-color immunofluorescence for c-kit and Lin-28 or pH3S10 or TNFR2 in murine heart organ cultures from WT or TNFR2−/− mice incubated either in normoxic or hypoxic conditions, with or without TNF for 18 hours
 % of c-kit cells expressing cellsNormoxiaHypoxia
  1. Percentages represent the % of c-kit+ cells that express Lin-28, pH3S10, or TNFR2. In comparison to expression in WT, TNFR2−/− mice show a diminished level of all the three markers. Data analyzed by ANOVA and Bonferroni's correction. Values are mean ± SEM; n = 3 per genotype.

  2. Bold, p<.05 vs. normoxia (−TNF).

  3. a

    p<.001 vs. TNF or hypoxia alone.

  4. b

    p<.01 vs. normoxia (+TNF).

  5. Abbreviations: pH3S10, phospho-Histone H3S10, TNF, tumor necrosis factor; WT, wild type.

WTLin-28+17 ± 2%48 ± 2.7%62 ± 1.6%b69 ± 1.7%a
 pH3S10+3 ± 0.2%18 ± 1.5%22 ± 2.1%b28 ± 1.5%a
 TNFR2+1 ± 0.1%13 ± 1.3%19 ± 1.2%b26 ± 2.0%a

Hypoxia and/or TNF-Induced Expression of Lin-28 and TNFR2 Result in Activation and Differentiation of Isolated Mouse c-kit+ CSCs

To investigate the roles of Lin-28 and TNFR2 in activation and differentiation of CSCs, c-kit+ cells were isolated from adult mouse heart and cultured under control conditions. These cells were negative for lineage marker (Lin), pan leukocyte marker CD45, CM marker α-SA, and hematopoietic progenitor marker CD133. Very few c-kit+ cells incubated in media alone (untreated) expressed Lin-28 (1%), TNFR2 (2.2%), or pH3S10 (1.5%) (supporting information Fig. 8A). Treatment with TNF or hypoxia alone for 6 hours resulted in a noticeable induction of Lin-28 (4.2%), TNFR2 (8.0%), and pH3S10 (6%), which were more pronounced with hypoxia in combination with TNF (c-kit+/Lin-28+ ∼10%, c-kit+/TNFR2+ ∼13%, and c-kit+/pH3S10+ ∼14%) (supporting information Fig. 8B, 8C). Hypoxia and TNF also induced expression of α-SA in some cells (3.5%), which then showed weak or no signal for c-kit. Addition of differentiation agents (5-azacytidine [32] and oxytoxin [33]) to culture medium prior to exposure to hypoxia or to TNF resulted in a more pronounced expression of Lin-28, TNFR2, or pH3S10 (Lin-28 ∼14%, TNFR2 ∼17%, and pH3S10 ∼20%), with a slight increase in α-SA expression approximately 7% in some cells, negative for c-kit. Exposure to hypoxia in combination with TNF resulted in an even higher expression for all the four proteins (Lin-28 ∼40%, TNFR2 ∼36% and, pH3S10 ∼46%, α-SA ∼16%), with identifiable sarcomeres detected in the α-SA+ cells. Expression of α-SA strongly suggests that c-kit+ cells represent myogenic precursors, and their differentiation is associated with loss of c-kit and expression of Lin-28 and TNFR2. A functional role for Lin-28 was demonstrated by Small interfering RNA knockdown using two siRNALin-28s. A reduction in Lin-28 in siRNALin-28-1 transfected cells versus siRNAControl was accompanied by a reduction in TNFR2 (13%), pH3S10 (6.6%), and α-SA expression (1.5%) versus TNFR2 (32%); pH3S10 (41%); α-SA expression (15%) in cells transfected with siRNAControl (Fig. 6). Similar findings were observed using siRNALin-28-2 (supporting information Fig. 9) and knockdown efficiency of about 80% was observed with both the siRNALin-28s in P19 cells (supporting information Fig. 10). Collectively these data support the role of TNF, Lin-28, and TNFR2 in activation and differentiation of c-kit+ CSCs to a cardiogenic lineage.

Figure 6.

c-kit+CSCs isolated from murine hearts transfected with either siRNAControl or siRNALin28-1 and left in media alone in normoxia (UT) or treated with TNF plus hypoxia for 6 hours. (A): UT siRNAControl-transfected cells, Lin, and α-SA, show negligible levels of Lin-28, TNFR2, and pH3S10. (B): In contrast, cells exposed to TNF plus hypoxia shows an increased expression of c-kit, Lin-28, TNFR2, pH3S10, and α-SA. (C): UT siRNALin28-1 transfected cells show comparable levels of c-kit, TNFR2, and pH3S10expression to siRNAControl-transfected cells. (D): In contrast, exposure to TNF plus hypoxia result in a significant diminished level of Lin-28, accompanied by a reduction in TNFR2, pH3S10, and α-SA compared to siRNAControl-transfected cells. Blue nuclei with Hoechst 33342; Scale bars = 50 µm; n = 3 with similar findings in two other independent experiments. Abbreviations: pH3S10, phospho-Histone H3S10; siRNA, Small interfering RNA; TNF, tumor necrosis factor; UT, untreated; α-SA, α-sarcomeric actin.


We have previously reported upregulation of TNFR2 in association with CM cell cycle entry in human cardiac allograft rejection [8]. In this study, we investigated the expression of TNF and TNFR2 in IHDM and gained insights into their role in ischemic injury using c-kit+ cells isolated from mouse hearts and an organ culture model [8, 12, 35] of WT and TNFR2−/− mice hearts incubated in hypoxic conditions as a model of ischemic injury. We report several new findings. In NM TNFR1 protein and transcript, but not TNFR2 or TNF, are strongly expressed in CMs, with TNFR2 and TNF expression seen only in VECs and ICs. In IHDM, TNFR1 expression is reduced, and TNFR2 is upregulated consistent with the response to injury in other tissues [9, 12, 36]. TNFR2 upregulation in CMs, VECs, and ICs in IHDM is associated with increased expression of markers of cell cycle entry, consistent with a role for TNFR2 in tissue repair. Our data reaffirm findings of Higuchi et al. [7] using transgenic mice with TNF-induced cardiomyopathy who have shown that ablation of the TNFR2 gene exacerbates heart failure and reduces survival, whereas ablation of TNFR1 blunts heart failure and improve survival.

We observed increased numbers of ICs in ischemic areas in IHDM, the majority of which express the CSC marker c-kit and lack leukocyte (CD45) and VEC (VEGFR2) markers. We propose that small round ICs which express c-kit but lack CD45 and VEGFR2 are resident CSCs. Many c-kit+ CSCs in ischemic areas coexpress TNFR2, Lin-28, and nuclear pH3S10, findings that potentially implicate TNFR2 in activation of c-kit+ CSCs and CSCs cell cycle entry. These observations in cardiac tissue from patients with IHD can be replicated in murine cardiac tissue and in c-kit+ cells isolated from mouse hearts cultured under hypoxic conditions with or without TNF. Most importantly, TNF and hypoxia increase expression and cell cycle entry of c-kit+/Lin28+ CSCs in WT but not TNFR2−/− mice, and these changes are more pronounced in cultures exposed to hypoxia in combination with TNF. We were unable to perform fluorescence-activated cell sorting and Western blot analysis on isolated c-kit+ cells due to low number of cell yield. To address whether inhibition of Lin-28 plays a functional role in the activation of c-kit+ CSCs, we reduced Lin-28 expression using Lin-28-mouse specific siRNA prior to exposure to hypoxia and TNF. We observed suppression of Lin-28 led to a decrease in the population of c-kit+ cells expressing pH3S10. This was also accompanied by a noticeable reduction in TNFR2 and α-SA expression. A previous genome-wide study in human embryonic stem cells have suggested that Lin28 may regulate the expression of at least three TNFR superfamily members (TNFRSF12A, RNFRSF10C, and TNFRSF1B) at post-transcriptional level, as revealed by the significant enrichment of the respective mRNAs in Lin28-containing ribonucleoprotein complexes (supporting information Table S1 in Peng et al. [24]). Collectively, these results imply that Lin-28, induced by TNF signaling through TNFR2, then contributes to enhanced TNFR2 expression and c-kit+ CSC proliferation and/or differentiation to the cardiogenic lineage [37].

Our identification of CSCs in the adult human heart and their increase in ischemic injury is consistent with previous studies [15, 16, 19-21, 32]. CSCs can lead to CM regeneration and restoration of cardiac function [15, 21, 22, 34, 38]. In particular, human c-kit+ CSCs have the ability to self-renew, are clonogenic and multipotent, and able to differentiate into CMs and, to a lesser extent, into smooth muscle cells and VECs [17]. Initiation of the reparative response is thought to be associated with cardiac release of proinflammatory cytokines [39] and chemokines [40]. Our new findings indicate that in ischemic injury local production of TNF coupled with hypoxia-mediated induction of TNFR2 on resident c-kit+ CSCs trigger factors that activate c-kit+ CSCs in their niche, causing them to undergo transient proliferation and eventually differentiation into mature CMs. Following our previous observations of the presence of a population of CMs that enter cell cycle [8], it is not unreasonable to speculate, from our data here, that these myocytes may originate from activation of c-kit+ CSCs.


The primary results in this study are from human tissue but a murine heart organ culture model and c-kit+ cells isolated from mouse hearts has provided a useful alternative tool in gaining insights in the effects of TNF and hypoxia [35]. In conclusion, we have provided evidence for the first time that c-kit+ CSCs can express TNFR2 and that binding of TNF to this receptor can result in CSC activation and cell cycle entry in association with Lin-28. These findings provide a potential mechanism for cardiac repair that could have clinical application. Although TNF inhibition has been reported to be beneficial in certain models of myocardial ischemia [4], substantial clinical evidence contradicts this notion [5, 41]. Our new data suggests that a reduction in TNFR2/Lin-28-mediated c-kit+ CSCs activation could explain the harmful effects on TNF blockade in IHD and that TNFR1 blockade in a manner that spares TNFR2 signaling may be a better solution for the treatment of heart failure.


We are grateful to Drs. Sarah Howlett, Maja Wallberg (Cambridge Institute of Medical Research, Cambridge), Jane Goodall, Tim Fitzmaurice, and Alexi Crosby (Department of Medicine, Cambridge University) for kindly providing murine hearts. We also thank Mrs. Suzanne Diston for her help with the references and the Tissue Bank (Department of Histopathology, Cambridge University Hospital NHS Trust) for their assistance in tissue processing. This work was supported by the British Heart Foundation, the National Institute for Health Research Cambridge Biomedical Research Centre and the U.S. National Institutes of Health (R01-HL036003).

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflict of interest