Author + information
- Received April 22, 2015
- Revision received May 27, 2015
- Accepted May 30, 2015
- Published online August 11, 2015.
- Eleni Tseliou, MD∗,
- Joseph Fouad†,
- Heidi Reich, MD∗,
- Leandro Slipczuk, MD∗,
- Geoffrey de Couto, PhD∗,
- Mark Aminzadeh, MD∗,
- Ryan Middleton, MS∗,
- Jackelyn Valle, BS∗,
- Liu Weixin, MS∗ and
- Eduardo Marbán, MD, PhD∗∗ ()
- ↵∗Reprint requests and correspondence:
Dr. Marbán, Cedars-Sinai Heart Institute, 8700 Beverly Boulevard, Los Angeles, California 90048.
Background Cardiosphere-derived cells mediate therapeutic regeneration in patients after myocardial infarction and are undergoing further clinical testing for cardiomyopathy. The beneficial effects of cardiosphere-derived cells are mediated by the secretion of exosomes and possibly other extracellular membrane vesicles (EMVs).
Objectives This study sought to investigate the effect of cardiosphere-derived EMVs (CSp-EMVs) on fibroblasts in vitro and tested whether priming with CSp-EMVs could confer salutary properties on fibroblasts in vivo.
Methods CSp-EMVs were isolated from serum-free media conditioned for 3 days by cardiospheres. Dermal fibroblasts were primed with CSp-EMVs for 24 h followed by exosomal micro-ribonucleic acid profiling. In vivo, we injected CSp-EMV−primed or −unprimed dermal fibroblasts (or CSp-EMVs) in a chronic rat model of myocardial infarction and defined the functional and structural consequences.
Results CSp-EMVs amplified their own biological signals: exposure of “inert” fibroblasts to CSp-EMVs rendered the fibroblasts therapeutic. Intramyocardially injected CSp-EMV−primed (but not unprimed) fibroblasts increased global pump function and vessel density while reducing scar mass. CSp-EMV priming caused fibroblasts to secrete much higher levels of stromal-cell-derived factor 1 and vascular endothelial growth factor and dramatically changed the micro-ribonucleic acid profile of fibroblast-secreted EMVs in vitro. The priming was followed by significant angiogenic and cardioprotective effects.
Conclusions CSp-EMVs alter fibroblast phenotype and secretome in a salutary positive-feedback loop. The phenotypic conversion of inert cells to therapeutically active cells reveals a novel mechanism for amplification of exosome bioactivity.
Paracrine signaling mediates the therapeutic benefits of cardiospheres (1) and their progeny (2) (cardiosphere-derived cells [CDCs]). Extracellular membrane vesicles (EMVs), ranging from nano-sized exosomes to micro-scale ectosomes, are constitutively secreted by many cell types and play a central role in intercellular communication (3–5). EMVs from CDCs contain a distinctive payload (6) and serve as agents of regeneration and cardioprotection (5,7–10); the micro-ribonucleic acids (miRNA) within these EMVs have been proposed to mediate the beneficial effects of CDCs after myocardial infarction (MI) (6).
Although EMV secretion is well characterized, the precise target cell type remains undefined. In reductionist systems, membrane vesicles are known to exert direct anti-apoptotic effects on cardiomyocytes and to stimulate tube formation by endothelial cells (6,8); in contrast, little is known about EMVs’ effects on fibroblasts, which feature prominently in scar formation and wound healing post-MI. Herein we describe the beneficial effects of cardiosphere-derived EMVs (CSp-EMVs) and report a novel mechanism whereby EMVs amplify their own bioactivity: priming dermal fibroblasts (DF) with CSp-EMVs suffices to convert inert fibroblasts to therapeutically active cells in vitro and in vivo.
Female Wistar Kyoto rats (n = 54) 5 to 6 weeks of age were used for in vivo experiments.
Human dermal fibroblasts (hDFs), human cardiospheres (hCSps), and human cardiosphere-derived extracellular vesicles (hCSp-EMVs) were used for in vitro assays. To avoid any possible confounding effects of xenotransplantation, we used rat DFs (rDFs) and rat CSp-EMV (rCSp-EMVs) for in vivo experiments.
To create MIs, animals underwent permanent left anterior descending artery ligation. Four weeks later they underwent a second survival thoracotomy with animals randomly assigned to intramyocardial border zone injection using 1 of 4 treatments: 1) rCSp-EMV− derived from 2 million cells (n = 16); 2) 2 million rDFs (n = 12); 3) 2 million rDFs incubated overnight with rCSp-EMVs, then washed (rCSp-EMV DF, n = 16); or 4) vehicle (phosphate-buffered saline [PBS]; n = 10). The animals were monitored for an additional 4 weeks followed by endpoint functional and histological studies.
Baseline transthoracic echocardiography was performed 28 days post-MI (2 days before the second thoracotomy). Briefly, long-axis images were used to measure left ventricular end-systolic and end-diastolic volumes and ejection fraction. Short-axis M-mode images at the level of the papillary muscle were used to measure end-systolic diameter. Follow-up echocardiographic analysis was performed 4 weeks post-injections followed by euthanasia.
CDCs were isolated from male Sprague Dawley and Brown Norway rats and cultured in Iscove's Modified Dulbecco's Medium (IMDM, Life Technologies, Carlsbad, California) supplemented with 20% fetal bovine serum and antibiotics. To form cardiospheres, 15 million CDCs were incubated with IMDM supplemented with penicillin/streptomycin and 0% fetal bovine serum in ultra-low attachment dishes. Three days later, the conditioned medium was collected and processed for EMV isolation (Figure 1A).
An expanded Methods section is available in the Online Appendix.
Pooled data are expressed as mean ± SE. Statistical analysis was performed using factorial analysis of variance followed by a Tukey post-hoc analysis of mean differences or with paired Student t test, indicated in figures by lines connecting compared values. A value of p ≤ 0.05 was accepted as significant.
EMV characterization and internalization
EMVs were isolated from serum-free medium conditioned by hCSps over a period of 3 days. The final pellet contained ∼12 × 109/ml of 175 ± 12-nm diameter vesicles by nanoparticle tracking analysis (3) (NTA, NanoSight Ltd., Amesbury, Wiltshire, United Kingdom) (Figure 1B). Flow cytometry revealed that these vesicles expressed tetraspanins characteristic of exosomes such as CD63, CD9, and CD81 (schematic in Online Figure 1B and pooled data in Figure 1C). Adding the final pellet to hDFs resulted in vesicle internalization as observed by confocal microscopy (Figures 1D and 1E). For quantification of dose-dependent vesicle internalization, images were obtained 6 (Figures 1F to 1H), 12 (Figures 1I and 1J), and 24 h (Figures 1K and 1L) after the addition of hCSp-EMVs. Higher concentrations of added particles (20 to 40 × 109) resulted in significantly higher numbers of vesicle-laden cells, with >90% of cells positive as early as 6 h (Figures 1G, 1I, and 1K). Individual cells accumulate particles more rapidly at higher concentrations (Figures 1H, 1J, and 1L). The fluorescence per cell reached a plateau with all groups equal in intensity after 24 h of incubation (Figure 1L) despite persistent differences in percent uptake at steady state (Figure 1K). Minimal background due to free-dye internalization was observed in the cells incubated with serum containing the lipophilic dye only (Online Figure 2). Thus, at low vesicle concentrations, cells either take up vesicles or they do not, with comparable capacities among transduced cells. This finding provides indirect evidence against a stochastic process such as membrane fusion, but is consistent with more active mechanisms of EMV uptake (endocytosis or receptor-mediated uptake).
Validation of in vitro biological activity of EMVs
Experiments performed in vitro to assay the bioactivity of hCSp-EMVs on hDFs revealed dose-dependent suppression of phosphorylated small mothers against decapentaplegic homolog (smad)2/3, smad4, and snai1, a zinc finger transcription factor and master regulator of epithelial-mesenchymal transition (Figures 2A to 2C). These antifibrotic signaling changes mirror those described for CSp-conditioned media (11). To look for potential conversion of fibroblast phenotype, we evaluated the expression of fibroblast-specific protein 1 (FSP1), discoidin domain receptor 2 (DDR2), CD105, and CD90, after 24 h of hCSp-EMV incubation. Representative flow cytometry plots (Figure 2D) and pooled data (Figure 2E) reveal significant attenuation of both FSP1 and DDR2, but no effects on CD105 or CD90 expression after single exposures to hCSp-EMVs. Immunohistochemistry confirmed the reduced expression of FSP1, but also showed enhanced expression of smooth muscle actin (SMA) (Figures 2F and 2G). The secretome of hDFs also changed after exposure to hCSp-EMVs: primed hDFs secreted much higher levels of stromal-cell-derived factor 1 (SDF-1) and vascular endothelial growth factor (VEGF) than unprimed hDFs (Figures 2H and 2I). Both changes are salutary in the post-MI setting, insofar as SDF-1 promotes homing and activation of endogenous stem cells (13), whereas VEGF stimulates angiogenesis (1). Similar changes were seen in hDFs treated with CSp-exosomes isolated by ultracentrifugation (Online Figure 3A), a complementary isolation method to the default precipitation approach. Ultracentrifugation enriches EMVs and particularly exosomes while excluding protein complexes and other debris (12). The congruence of the findings with the 2 isolation methods supports the conjecture that EMVs, and particularly exosomes, are primarily responsible for the biological effects investigated here.
Cardioprotective and angiogenic effects of hCsp-EMV−primed DFs
Conditioned media from hCSp-EMV−primed hDFs and hCSp-EMVs per se reduced cardiomyocyte apoptosis after oxidative stress, unlike hDF-EMVs (representative fluorescence activated cell sorting [FACS] plots in Figure 3A to 3C and pooled data in Figure 3D). Additionally, in an in vitro matrigel angiogenesis assay, conditioned media from hCSp-EMV−primed DFs exerted an angiogenic effect, which was as strong as that of hCSp-EMV alone; both treatments induced significantly more tube formation than hDF-EMVs (representative microscope images in Figures 3E to 3G and pooled data in Figure 3H). Thus, hCSp-EMV priming confers on hDFs the ability to stimulate angiogenesis and to protect cardiomyocytes against stress-induced apoptosis. Enhanced angiogenesis was also observed using hCSp-exosomes isolated by ultracentrifugation (Online Figures 3B to 3D), once again indicating that a preparation enriched in exosomes can recapitulate the beneficial effects seen with EMVs isolated by precipitation.
Distinctive miRNA profiles of hCSp-EMV−primed DFs
We previously reported that hCDC-derived EMVs, identified as exosomes, express a unique miRNA payload that at least partially accounts for the in vivo regenerative capacity of CDCs (6). Indeed, this and other reports have led to the conjecture that vesicles affect gene expression of recipient cells by miRNA transfer (14–16). To see if this mechanism might be operating here, we first investigated the global miRNA content of hCSps and compared it to that of hDFs. A number of miRNAs were enriched in hCSps relative to hDFs: Figure 4A highlights those that are most abundantly overexpressed in hCSps. We then compared the miRNA profiles of CSp-EMV−primed and unprimed hDFs. Figure 4B reveals that hCSp-EMV−primed hDFs express very different miRNAs than unprimed hDFs (Figure 4B). The pattern only partially resembles that of the cells of origin (hCSps; compare to Figure 4A) or of the vesicles secreted by hCSps (Figure 4C), hinting that simple passive transfer of vesicular miRNAs cannot fully account for the distinctive miRNA profile of hCSp-EMV−primed hDFs. Finally, we compared the miRNA profiles of vesicles secreted by primed and unprimed hDFs by collecting media produced 24 h after priming by hCSp-EMVs or vehicle. The miRNA profiles of primed and unprimed hDFs differed enormously (Figure 4D). The miRNAs secreted by hCSp-EMV−primed hDFs include several that are enriched in hCSp-EMVs themselves (notably miRNA-146a, which was highlighted by Ibrahim et al. (6) and is elevated in all therapeutically active groups here), but the patterns are otherwise quite individual. Thus, priming with hCSp-EMVs leads to fundamental changes in hDF miRNA expression profiles and hDF-secreted vesicles. The distinctive miRNA profiles in hCSp-EMV−primed hDFs and their membrane vesicles argue against the possibility that the changes merely reflect accumulation and subsequent “regurgitation” of miRNAs transferred in hCSp-EMVs.
rCSp-EMV−primed fibroblasts reverse remodeling
We have presented evidence that hCSp-EMV−primed hDFs secrete SDF-1 and VEGF, exert anti-apoptotic and angiogenic effects in vitro, and express distinctive miRNAs. Therefore, we questioned whether rCSp-EMVs themselves, as well as rCSp-EMV−primed DFs, might confer therapeutic benefits in vivo in a rat model of chronic MI. One month after permanent left anterior descending ligation, animals underwent intramyocardial injection. To assess particle biodistribution qualitatively, animals (n = 3) injected with dye-labeled rCSp-EMVs were euthanized 1 h post-injection and selected organs were imaged. Approximately 20% of the injected rCSp-EMVs were found in the heart; the lungs also exhibited obvious uptake, with less in other organs (Online Figure 4). This percentage of retention in the heart at 1 h compares favorably with that seen with intramyocardially injected cells (17). Minimal intensity was detected after the free dye control injections only (data not shown).
For long-term physiological experiments, animals were randomly allocated to 1 of the following 4 groups: vehicle (PBS), unprimed rDFs, rCSp-EMVs, or rCSp-EMV−primed rDFs with treatment at 1 month post-MI (Figure 5A). One month later, echocardiography revealed improved ejection fraction in the rCSp-EMVs and the rCSp-EMV−primed rDFs groups compared to either vehicle or unprimed rDFs (Figure 5B). This finding primarily reflected differences in left ventricular end-systolic diameter (Figure 5C). Histological analysis (of samples exemplified in Figure 5D) showed significant reductions in scar mass (Figure 5E) and enhanced infarct wall thickness (Figure 5F) in the rCSp-EMVs and rCSp-EMV−primed rDF groups. The structural and functional improvements seen with rCSp-EMVs and rCSp-EMV−primed rDFs were comparable to those reported with rCSp injection in this model (11).
Finally, to test the in vivo angiogenic capacity of CSp-EMVs and CSp-EMV−primed cells, we quantified capillaries (bounded by von Willebrand factor−positive cells) and microvessels (bounded by SMA-positive cells) (Figure 6A). Analysis of serial images from the apex to the base revealed greater capillary density in the rCSp-EMVs and rCSp-EMV−primed groups compared to both controls in all 3 zones evaluated (infarct, border, and remote) (Figures 6B to 6D, left panel). Microvessel density was likewise increased, but only in the infarct zone (Figures 6B to 6D, right panel). Another mechanism underlying CSp regenerative efficacy is cardiomyocyte proliferation (13,18). Bromodeoxyuridine incorporation revealed enhanced deoxyribonucleic acid synthesis after exposure to rCSp-EMVs and rCSp-EMV-rDFs compared to PBS and unprimed rDFs, validating previous reports (6) (Online Figures 5A and 5B). Interestingly, this effect tended to be more prominent after rCSp-EMV−only injections compared to rCSp-EMV-rDFs. Finally, no changes in cardiomyocyte diameter were observed (Figures 7A to 7C).
We have observed remarkable effects of EMVs from hCSps on fibroblasts. DFs are venerable controls for cardiac cell therapy; their injection neither improves nor aggravates adverse remodeling after MI (1,19). DF-produced exosomes are likewise inert (6). In contrast, cardiospheres and their progeny trigger functional recovery and structural improvements in various ischemic and nonischemic models of heart failure (20–22). This beneficial effect was recently attributed to secreted exosomes (6). Although interaction of EMVs with endothelial cells and cardiomyocytes has been reported (5,6,23,24), our data support a strong, previously unappreciated bioactivity of CSp-EMVs on fibroblasts and other cardiac cell types (Central Illustration). More specifically, we report that, in the restricted environment of in vitro priming with hCSp-EMVs, hDFs exert a dose-dependent downregulation of the transforming growth factor−β cascade and increased secretion of SDF-1 and VEGF. Remarkably, these primed fibroblasts promote angiogenesis and inhibit cardiomyocyte apoptosis in vitro, whereas in vivo they can attenuate remodeling and improve function to levels equivalent to those reported with rCSps (11).
Vesicular transfer of miRNAs mediates cell-cell communication in different biological systems (25–27). However, the miRNA cargo of EMVs does not necessarily reflect passive loading with RNAs in the parent cell; selective enrichment mechanisms appear to be at play (15). This selective miRNA payload may be a crucial determinant of bioactivity on the recipient population. Indeed, we found a distinct miRNA signature in primed versus unprimed hDFs that does not reflect passive release of internalized hCSp-EMVs. Therefore, internalization of hCSp-EMVs leads to downstream, biologically significant changes in miRNA vesicular cargo released by the recipient hDFs. Additionally, because fibroblast-derived EMVs enriched in miRNAs do not improve recovery in vivo (28), the cargo transition described here may provide promising clues to pathways involved in reverse remodeling.
Many studies report an innate plasticity of fibroblasts that allows them to acquire a cardiomyocyte or endothelial phenotype after exposure to either transcription factors or small molecules (29–32). These direct reprogramming approaches may constitute a promising therapeutic strategy (33). Our data indicate that single-dose priming with CSp-EMVs converted DFs to a less fibrotic phenotype with functional properties equal to those of CSp-EMVs. Interestingly, the transforming growth factor−β pathway, which provides key signals in cellular conversion (34,35), was significantly downregulated (Figures 2A to 2C). We do not yet know if EMVs or an EMV subgroup (e.g., exosomes) suffice to durably reprogram DFs to a fully-distinct cell type, but our data do indicate that inert fibroblasts can be functionally converted both in vitro and in vivo for a sufficient duration to shape therapeutic activity.
Crosstalk between endothelial cell-derived vesicles containing miRNAs and the surrounding myocardium has been reported (24). Here we showed that, beyond strictly paired cell-cell communication, divergent transportation of biologically significant signals takes place between hCSp-EMV−primed DFs and recipient human umbilical vein endothelial cells or cardiomyocytes. These findings in culture may help to rationalize the efficacy of a relatively low number of cells injected in vivo: transplanted cells secrete EMVs, which interact with the surrounding tissue and convert it into a more salutary milieu, in a positive feedback loop (2,36).
Xenogeneic transplantation of CSps post-MI elicits detrimental immunological sequelae, although allogeneic CSps are effective and immunologically innocuous (2). Here we used allogeneic rCSp-EMVs to evaluate potency in vivo and allogeneic hCSp-EMVs for in vitro experiments. Evaluation of the immunological sequelae of allogeneic and xenogeneic EMVs is beyond the scope of the present study, but it seems likely that CSp-EMVs may be even less immunogenic than CSps, as they express far fewer surface antigens and, unlike cells, cannot react dynamically to immunological cues.
The cardiovascular field has previously tended to focus on exosomes, rather than EMVs as a whole (37,38). Exosomes, nano-sized particles produced by the endolysosomal pathway, are but one class of EMVs (39). The isolation method used here relies on precipitation by polyethylene glycol, which pulls down all EMVs, not just exosomes, and also macromolecules including RNA-protein complexes (39,40). Thus, it is possible that some of our results might be attributable to larger, biologically distinct microvesicles or apoptosomes or even to non-EMV entities (e.g., free RNA or protein). Nevertheless, the size distribution (Figure 1B), antigenic composition (Figure 1C), and contents (Figure 4) of our default preparations are most consistent with a preponderance of exosomes (41). Key phenotypic changes are recapitulated by EMVs isolated by ultracentrifugation (Online Figure 3), the gold standard method for enriching exosomes (12), further implicating exosomes as mediators of the beneficial effects.
CSp-EMVs alter fibroblast phenotype and secretome in a salutary positive-feedback loop: exposure of “inert” fibroblasts to CSp-EMVs rendered the fibroblasts therapeutic. Intramyocardially injected CSp-EMV−primed (but not unprimed) fibroblasts increased global pump function and vessel density while reducing scar mass. The phenotypic conversion of inert cells to therapeutically active cells reveals a novel mechanism for amplification of exosome bioactivity.
COMPETENCY IN MEDICAL KNOWLEDGE: CSp-EMVs and particularly their exosomes convert inert dermal fibroblasts into therapeutically active cells capable of decreasing scar and improving function in a chronic myocardial infarction model. This represents a mechanism of amplification whereby a small number of transplanted cells can exert widespread, durable benefits in injured myocardium.
TRANSLATIONAL OUTLOOK: Further animal and clinical experiments are warranted to explore the potential therapeutic roles of CSp-EMVs and CSp-EMV−primed fibroblasts for diseases in which scar reduction, angiogenesis, and cardiomyocyte proliferation could be beneficial.
This work was supported by the NIH (1R01HL124074-01) and the Cedars-Sinai Board of Governors Heart Stem Cell Center. Dr. Marbán owns equity in and serves as an unpaid advisor to Capricor. Based on this work, Dr. Tseliou received the Herman K. Gold Young Investigator's Award in Molecular and Cellular Cardiology at ACC.2015. All other authors have no relationships relevant to the content of this paper to disclose.
- Abbreviations and Acronyms
- cardiosphere-derived cell
- cardiosphere-derived extracellular membrane vesicle
- dermal fibroblasts
- dermal fibroblast-derived extracellular membrane vesicle
- extracellular membrane vesicle
- myocardial infarction
- stromal-cell-derived factor
- vascular endothelial growth factor
- Received April 22, 2015.
- Revision received May 27, 2015.
- Accepted May 30, 2015.
- American College of Cardiology Foundation
- Chimenti I.,
- Smith R.R.,
- Li T.S.,
- et al.
- Tseliou E.,
- Pollan S.,
- Malliaras K.,
- et al.
- Sahoo S.,
- Klychko E.,
- Thorne T.,
- et al.
- Heinrich E.M.,
- Dimmeler S.
- Barile L.,
- Lionetti V.,
- Cervio E.,
- et al.
- Gray W.D.,
- French K.M.,
- Ghosh-Choudhary S.,
- et al.
- ↵Smith JA, Ng KS, Mead BE, et al. Extracellular Vesicles. Commercial potential as byproducts of cell manufacturing for research and therapeutic use. Cell Therapies Trends and Technologies 2015;13(4)s. Available at: http://www.bioprocessintl.com/manufacturing/cell-therapies/extracellular-vesicles-commercial-potential-as-byproducts-of-cell-manufacturing-for-research-and-therapeutic-use/. Accessed June 2, 2015.
- Malliaras K.,
- Ibrahim A.,
- Tseliou E.,
- et al.
- Terrovitis J.V.,
- Smith R.R.,
- Marban E.
- Smith R.R.,
- Barile L.,
- Cho H.C.,
- et al.
- Johnston P.V.,
- Sasano T.,
- Mills K.,
- et al.
- Messina E.,
- De Angelis L.,
- Frati G.,
- et al.
- Aminzadeh M.A.,
- Tseliou E.,
- Sun B.,
- et al.
- Nam Y.J.,
- Song K.,
- Luo X.,
- et al.
- Sahoo S.,
- Losordo D.W.
- Hirsch E.,
- Hilfiker-Kleiner D.,
- Balligand J.L.,
- et al.
- Thery C.,
- Amigorena S.,
- Raposo G.,
- et al.
- Van Deun J.,
- Mestdagh P.,
- Sormunen R.,
- et al.