Comparative Proteomic Analysis of Nuclear and Cytoplasmic Compartments in Human Cardiac Progenitor Cells. Functional Evaluation of IL1A and IMP3


 Clinical trials evaluating cardiac progenitor cells (CPC) demonstrated feasibility and safety, but no clear functional benefits. Therefore a deeper understanding of CPC biology is warranted to inform strategies capable to enhance their therapeutic potential. Here we have defined, using a label-free proteomic approach, the differential cytoplasmic and nuclear compartments of human CPC (hCPC). Global analysis of cytoplasmic repertoire in hCPC suggested an important hypoxia response capacity and active collagen metabolism. In addition, comparative analysis of the nuclear protein compartment identified a significant regulation of a small number of proteins in hCPC versus human mesenchymal stem cells (hMSC). Two proteins significantly upregulated in the hCPC nuclear compartment, IL1A and IMP3, showed also a parallel increase in mRNA expression in hCPC versus hMSC, and were studied further. IL1A, subjected to an important post-transcriptional regulation, was demonstrated to act as a dual-function cytokine with a plausible role in apoptosis regulation. The knockdown of the mRNA binding protein (IMP3) did not negatively impact hCPC viability, but reduced their proliferation and migration capacity. Analysis of a panel of putative candidate genes identified HMGA2 and PTPRF as IMP3 targets in hCPC. Therefore, they are potentially involved in hCPC proliferation/migration regulation.


Introduction
The adult mammalian heart has alow but intrinsic cardiomyocyte turnover [1][2][3]. The contribution of adult cardiomyocyte turnover to heart homeostasis and the origin of the new cells remain, however, unclear [4]. Determining the direct contribution of mature cardiomyocytes, by dedifferentiation/proliferation [4][5][6], as opposed to stem/progenitor cells (CSC/CPC) involvement, to heart homeostasis, remains a main focus of heart regeneration [4].
There is con ictive evidence for a reservoir of cardiac CSC/CPC residing in physiological niches that full lls many of the roles of any adult stem cell compartment [7]. The main difference between this atypical cardiac CSC/CPC compartment and most other adult stem/progenitor cell compartments would be the consolidated low turnover rate in the mammalian adult heart [1,2].
Adult murine CSC/CPC have been de ned primarily by the expression of cell surface markers (reviewed in [8]), but their diversity and approaches used for their detection has, however, hindered their unambiguous identi cation and molecular de nition [9]. Moreover, several recent lineage tracing studies using unique genetic reporters have yielded more con icting data [4,8].
Cardiosphere-derived cells (CDC) and c-kit pos CPC have been characterized and evaluated in pigs and humans. Results from transplantation studies in swine models of cardiac ischemic injury revealed a moderate but reproducible improvement in cardiac function [10][11][12][13]. Multiple lines of evidence from preclinical studies on the transplantation of human or swine CSC/CPC suggested that the mechanisms of action are mainly indirect (reviewed in [4]), resulting in durable bene ts despite low engraftment and cell survival of the transplanted cells [14,15]. Consistent with these data, a recent functional analysis in mice concluded that several cell-based therapies improve heart function after ischemia-reperfusion injury chie y through an acute sterile immune response of wound healing [16].
Based on promising preclinical results [10,12], a phase I/IIa clinical trial was carried out, using allogeneic hCPC, for the treatment of pacients with large cardiac infarcts (EudraCT 2013-001358-81; [23]). While the results demonstrated the feasibility and safety of the approach, no signi cant functional bene ts were demonstrated [24]. Given all of these data, a deeper understanding of hCPC biology and their behavior in response to acute or diffuse chronic damage might be critical for a better de nition of the mechanism of action of these therapies, which might lead to improvements in the current strategies based on hCPC.
With this in mind, here have compared, by a proteomics approach, the differential cytoplasmic and nuclear compartments of hCPC, hMSC and broblasts. From this analysis, we focused on two overexpressed nuclear proteins in hCPC, IL1A and IMP3 (IGF2BP3). IL1A was demonstrated to be a dualfunction cytokine with a plausible role in apoptosis regulation, and IMP3 regulated proliferation and migration of hCPC.
To validate the proteomic data, we compared P4HA1 and ASPH expression in hCPC, hMSC and broblasts by RT-qPCR analysis. The data con rmed P4HA1 overexpression in hCPC (Fig. 1e). This was also con rmed by western blotting, although the difference in hCPC P4HA1 expression was less pronounced when compared with human broblasts, and no differences were evident when compared with hMSC (Fig. 1f). When we analyzed the gene expression of ASPH, we found that it was not downregulated in hCPC ( Supplementary Fig. S1 online), as inferred from the proteomic analysis (Fig. 1e), suggesting a post-transcriptional regulation in hCPC. Additionally, we discarded the possibility that the expression differences found would be associated with the cardiac origin of the hCPC analyzed. We compared their diferential expression with total human heart samples con rming all results with the the sole exception of CDH5 ( Supplementary Fig. S1 online).
Comparative analysis of the human CPC nuclear compartment Analysis of the whole LF proteome in hCPC and hMSC yielded 446 and 514 nuclear proteins, respectively [20]; 95% of which could be included in a GO category. IPA analysis of the nuclear hCPC subproteome is shown (Fig. 2a). LF-proteomic analysis of puri ed nuclear fractions of hCPC and hMSC rendered 369 and 348 proteins, respectively (Supplementary Fig. S1 online). To con rm the proteins identi ed in the nuclear fraction, we compared with their representation in the cytoplasm (Fig. 2b,c). The comparative proteomic analysis of hCPC versus hMSC nuclear puri ed fractions revealed the potential differential expression of 27 proteins in hCPC (Fig. 2b,c). Of the more clearly overexpressed proteins in the hCPC nuclear compartment only IMP3 (also known as IGF2BP3), nestin and IL1A (Fig. 2b) also showed a parallel signi cant increase in mRNA expression in hCPC relative to hMSC. In addition, hCPC expressed lower nuclear levels of several proteins including Polymerase I and Transcript release factor (PTRF) (Fig. 2c); previous RNAseq studies [20] con rmed all denti ed proteins. PANTHER GO-Slim analysis of biological processes using the upregulated hCPC nuclear proteins indicated a strong involvement in cellular component morphogenesis and organization, as well as muscle contraction (Fig. 2d). PANTHER Pathway analysis also revealed an important representation of the ubiquitin proteasome pathway (Supplementary Aiming to validate the proteomics nuclear data we evaluated the expression of ASPHD1 and PTRF, which were up-and down-regulated, respectively, in hCPC versus hMSC, by proteomics. RT-qPCR analysis con rmed a signi cant differential expression of ASPHD1 in all hCPC isolates (hCPC1-3) in comparison with hMSC and broblasts (Fig. 2e). The preferential expression in hCPC was also con rmed by western blotting and by immuno uorescence (Fig. 2f,g). By contrast, PTRF downregulation in hCPC was not con rmed by RT-qPCR ( Supplementary Fig. S2 online), suggesting again a relevant post-transcriptional regulation.
Because a very signi cant fraction of regulatory nuclear proteins is expressed at low levels, below the detection limits of proteomics, we validated by RT-qPCR several transcriptional factors found up-and down-regulated, by RNAseq in hCPC compared with hMSC [20]. GATA4, SOX17, WT1 and GATA2 were robustly overexpressed in hCPC in comparison with hMSC ( Supplementary Fig. S2 online). The expression levels of TBX3 and MEF2C were also signi cantly higher in hCPC than in hMSC, but less pronounced. We also con rmed that HOXD8 and HOXA10 were barely expressed in hCPC in comparison with hMSC ( Supplementary Fig. S2 online).
IL1A is a dual-function cytokine in hCPC Two of the most differentially expressed genes in hCPC were IL1A (as mentioned above) and IL1B, whose overexpressions were also validated by RNAseq [20]. IL1A was found over-represented by comparative LF-proteomic analysis in the nuclear compartment of hCPC compared with hMSC (Fig. 2c). IL1A is produced as a precursor protein that yields a mature form and an N-terminal propeptide, containing a nuclear localization sequence that allows access to the nuclear compartment. In this way, IL1A is a well established "dual-function cytokine" that plays a role in the nucleus independently of its classical extracellular mediated effects [25][26][27].
To validate the proteomic analysis, we assessed the expression of IL1A and IL1B by RT-qPCR in hCPC and hMSC. The results clearly con rmed the overexpression of IL1A and IL1B in hCPC (6,788-and 1,409fold, respectively, Fig. 3a). We also tested the expression of other members of the IL1 signaling pathway. A signi cant increase (3.22-fold change) was found for the expression of the natural antagonist of IL1R1 (IL1RA) (Fig. 3a). Other main members of the IL1 pathway, IL1 receptor (IL1RI), IL38 and the secondary IL1 receptor (IL1R2) were, or not differentially expressed (IL1RI) or not detected. Immuno uorescence analysis of IL1A, with an antibody against the full size protein, revealed a highly preferential cytoplasmic location in basal conditions (homeostasis; Fig. 3b). However, while hCPC showed signi cantly higher levels of IL1A mRNA expression than hMSC (Fig. 3a), the latter showed higher levels of IL1A protein (Fig.  3b), suggesting an important lineage-speci c post-transcriptional regulation. We thus obtained nuclear and cytoplasmic fractions from hCPC and analyzed IL1RI, IL1A and IL1B expression by western blotting. The results com rmed an almost exclusive cytoplasmic location for all proteins analyzed ( Supplementary Fig. S3 online). The fraction of nuclear IL1A detected by proteomics (Fig. 2b) thus seems to be minor. Given the known immunoregulatory capacity of hCPC [17][18][19] and their de nition as an MSClike cell subpopulation [19], it is possible that IL1A could play a role in hCPC in homeostasis. We thus evaluated whether IL1A could contribute to the immunoregulatory capacity of hCPC. Thus, we cocultivated phytohemagglutinin-stimulated human CD3 T cells with control hCPC, hCPC silenced for the expression of IL1A (hCPC siIL1A), or negative-control tranfected cells (hCPC siNeg). We rst con rmed that the siIL1A (10 nM) could downregulate IL1A (~72%) in hCPC ( Supplementary Fig. S3 online). All cell populations (hCPC, hCPC siIL1A and hCPC siNeg) demonstrated similar immunoregulatory capacity at the higher cell doses analyzed (1:10-1:20), which was lost when lower doses were evaluated (1:40). Therefore no signi cant changes in immunoregulatory capacity were found (Fig. 3c), indicating that IL1A seems not to have a relevant role in the T cell immunoregulatory capacity of hCPC. We next evaluated whether IL1A could acts as a dual-function cytokine in hCPC, as has previously been described in other cell lineages [28]. We analyzed in hCPC the behavior of IL1A, IL1B and IL1RI expression in response to oxidative stress (see Methods). We found that IL1RI expression was equivalently and moderately reduced (~25%) by both treatments (preferential apoptosis or necrosis). IL1Aand IL1B expression were also decreased by both treatments, but to a much greater extent; IL1A expression was more pronouncedly reduced (90-95% reduction) when necrosis was induced (Fig. 3d). Of note, IL1A behaved differently to the two stmuli in hCPC and hMSC, as evaluated by immuno uorescence. Upon induction of apoptosis, IL1A protein was signi cantly upregulated in hCPC, whereas it was clearly reduced in hMSC (Fig. 3e); induction of necrosis provoked a major loss of IL1A in both cell types (Fig. 3e). Quanti cation of nuclear versus cytoplasmic localization by immuno uorescence of IL1A in hCPC, comparing homeostasis with the induction of apoptosis or necrosis, revealed a signi cant increase in the nuclear location of IL1A (co-localization with DAPI signal) after the induction of apoptosis (Fig. 3f). The opposite was found in hMSC where co-localization of IL1A with the nuclear compartment was poorer upon apoptosis induction, although it was augmented after necrosis induction (Fig. 3f). We also performed western blotting of subcellular compartments in hCPC subjected to apoptosis (Fig. 3g); in necrotic cells, expression of the three proteins was quite low and di cult to quantify because of the strong loss of cellular content. In agreement with the immuno uorescence study, a substantial fraction of IL1A and IL1B was found in the nuclear compartment ( Fig. 3g) whereas the subcellular localization of IL1RI was barely unchanged by the induction of apoptosis (compare with Supplementary Fig. S3 online).
Comparison of nuclear/cytoplasmic ratio in apoptosis versus homeostasis, in a representative western experiment, yielded an important increment for IL1A and IL1B (IL1A> IL1B), compared with a modest variation on IL1RI (Fig. 3g). Altogether, these data suggest that IL1A could be acting as dual-cytokine in hCPC with a potential role in the transcriptional regulation in apoptosis.
Functional evaluation of IMP3 in hCPC in homeostasis and in response to oxidative damage IGF2 is the predominant form of IGF in humans [29] and it binds to insulin-like growth factor 1 receptor (IGF1R), insulin-like growth factor 2 receptor (IGF2R; CD222) and the insulin receptor A isoform (IR-A). It seemed interesting that in addition to IGF2R, two additional members of the IGF2 pathway (insulin-like growth factor mRNA binding proteins 2 and 3; IMP2/IMP3) were identi ed as over-represented in the hCPC nuclear subproteome by LF-proteomics (Fig. 2b). RT-qPCR analysis validated the high levels of IMP3 expression in hCPC versus hMSC (>40-fold overexpression), but the opposite was observed for IMP2 (Fig. 4a).
IMP3 belongs to a family of mRNA-binding proteins that bind to multiple mRNAs in mammalian cells, including IGF2 [30]. Based on previous literature [29], we hypothesized that high levels of IMP3 would lead to a decrease in the autocrine bioavailability of IGF2, reducing the potential signaling through IGF2R and triggering senescence/apoptosis. We rst analyzed the impact of IMP3 knockdown in two independent hCPC isolates. Cells transfected with siIMP3 showed signi cantly reduced levels of IMP3 when compared with negative control or non-transfected control cells, analyzed both by RT-qPCR and western blotting (Fig. 4b,c). IMP3 silencing did not affect negativelly hCPC viability, 48 h posttransfection; in fact, IMP3-silenced cells showed a moderate increase in viability (Fig. 4d). We then analyzed the effects of IMP3 silencing on the response of hCPC to oxidative stress (500 mM H 2 O 2 , during 48 h) and evaluating apoptotic and necrotic cells with the Annexin V/propidium iodide ( Supplementary   Fig. S3 online). Neither of the two hCPC populations tested showed any remarkable difference in the percentages of homeostatic, apoptotic, late apoptotic or necrotic cells (Fig. 4e). Thus, IMP3 does not seem to play a critical role in the regulation of hCPC response to oxidative stress-mediated apoptosis.
Although IMP3 seems not to be essential for apoptotic responses we investigated IMP3 regulation in hCPC damage responses. We rst studied the impact of apoptosis or necrosis induction on the transcriptional activity of IMP3, IMP2 and IGF2R, and their subcellular localization. Neither apopotosis nor necrosis affected IGF2R expression; however, the induction of apoptosis (but not necrosis) promoted a signi cant decrease of IMP3 and IMP2 transcription in hCPC (Fig. 5a). Western blotting of hCPC showed that IGF2R and IMP3 were expressed at similar levels whereas IMP2 was expressed at apparently lower levels (Fig. 5b). Analysis of nuclear and cytoplasmic fractions by western blotting con rmed that a substantial fraction of IMP3 (22-40%), and to a lesser extent IMP2, was found in the nuclear fraction in homeostasis (Fig. 5b, left panel). After induction of apoptosis, both IMP2 and IMP3 showed an increased presence in the nuclear compartment with respect to the cytoplasmic compartment, whereas IGF2R was unchanged ( Fig. 5b; right panel); speci cally, we found an increase in nuclear IMP2 and IMP3 of 8.5-fold and 13-fold, respectively (Fig, 5c). These results were con rmed by immuno uorescence (Fig. 5d). We analyzed the nuclear versus cytoplasmic localization of the IMP3 uorescent signal and we con rmed that, upon induction of apoptosis, the nuclear pool of IMP3 singni cantly increases (co-localization coe cient referred to DAPI signal) (Fig. 5d).
Thus, apoptosis induction in hCPC triggers a signi cant decrease in IMP2 and IMP3 transcription, concomitant with an enrichement of both proteins in the nuclear compartment. These results suggest a non-essential role of IMP3 in gene expression regulation upon induction of oxidative stress-mediated apoptosis.

IMP3 regulates proliferation and migration of hCPC
Analysis of proliferation in IMP3-silenced cells and controls estimated by EdU (5-ethynyl-2'-deoxyuridine) incorporation during 12 h and 48 h post-transfection revealed that the knockdown of IMP3 signi cantly reduced proliferation in hCPC1 cells (about 2-fold) and a similar effect was found in hCPC3 cells (Fig.  6a,b). Thus, IMP3 seems not to be relevant for survival, but is likely involved in hCPC proliferation regulation.
We then evaluated the potential implication of IMP3 in cell motility, as previously proposed [31] using wound-healing assays. Monolayers of hCPC cells silenced or not for IMP3 were compared in their capacity to repair a wound during 24 h, as described [20]. As shown in Fig. 6c, both IMP3-silenced hCPC isolates demonstrated a statistically signi cant delay (at 9-12 h) in wound healing, albeit with different kinetics (Fig. 6c).
Finally, we analyzed a panel of candidate target genes previously reported to be regulated by IMP3 in heterologous models, or described as preferentially expressed in hCPC [20,22]. We used RT-qPCR to compare the levels of gene expression in hCPC silenced for IMP3 knockdown or in control cells. Figure 6d shows the results obtained with isolate hCPC1, which showed more robust IMP3 silencing.
Concerning genes involved in proliferation, we found that PTRF (also known as Cavin1 or Cavin-1) and HMGA2 (high-mobility group AT-hook 2) were signi cantly downregulated in siIMP3 cells (~50%). c-MYC, CDK6 and CD9 were moderately but not signi cantly downregulated (<20%). In relation to genes involved in apoptosis, we found that ICAM3 expression was signi cantly reduced (~60%) in IMP3-silenced cells, but unexpectedly not NEMO (inhibitor of nuclear factor kappa B kinase subunit gamma).
We also tested the consequence of IMP3 silencing for the expression of a small panel of transcriptional factors previously ( Supplementary Fig. S2 online) de ned in hCPC. All them, except SOX17 expression (~40% reduction), did not modify the expression after IMP3 knockdown. These results suggest that IMP3 might have a modest role in regulating hCPC fate-genes by regulation of SOX17. We additionally analyzed the potential impact of IMP3 knockdown on several genes associated in other cell types with self-renewal, such as Oct4, Dido3 and Mbd3, [20,22,32]. The expression of all three genes was unaffected by IMP3 silencing (Fig. 6d). These results suggest that IMP3 seems not to be mainly involved in the regulation of the undifferentiated state of hCPC. A similar analysis using hCPC3 yielded essentially identical but non-signi cant results.
Finally, because nuclear IL1A also regulates cell proliferation and migration [26,27], we sought to evaluate the potential involvement of IMP3 in the regulation of the IL1 pathway. IMP3-silenced hCPC and controls were evaluated by RT-qPCR 48 h post-transfection. Compared with control hCPC, IMP3-silencing failed to affect IL1A expression but enhanced IL1B expression ( Supplementary Fig. S3 online). IL1RA and IL1R1 were also unaffected by IMP3-silencing ( Supplementary Fig. S3 online), indicating that IMP3 is likely not involved in the regulation of IL1 pathway in hCPC.

Discussion
The adult mammalian heart has demonstrated a low but intrinsic turnover in mice [3] and in humans [1,2]. However, the source of this turnover remains controversial (reviewed by [4]). If cardiac stem/progenitor cell compartments exist, they must be atypical, and not comparable with other adult stem cell compartments with higher cell turnover. In this sense, SCA1 + PDGFRa + cells in the adventitial layer of arteries have recently been characterized as resident vascular stem cells that speci cally (monopotential) produce new smooth muscle cells only in response to severe injury [33]. In this sense, we have shown that a subpopulation of heart SCA1 + cells, characterized by the expression of high levels of Bmi1 (B-CPC), contains a resident non-cardiomyocyte progenitor population [34]. B-CPC, a heterogeneous population mostly PDGFRa-, demonstrated in vivo a dominant monopotential and bipotential lineage capacity, generating in the mid-term (6 months after labelling) predominantly endothelial cells (63%), but also smooth muscle cells (27%) and cardiomyocytes (10%) [35]. In addition, genetic depletion of B-CPC impairs heart functional recovery after acute myocardial infarction, without affecting functionality in homeostasis [36,37]. This is a similar scenario to that depicted for vascular arterial stem cells [33]. Thus, there are evidences from several models of the existence of atypical CSC/CPC populations in the heart that could be involved in heart cell turnover [14,32,33].
No equivalent marker to murine SCA1 has yet been de ned in humans. That being said, CDC and c-kit pos hCPC have been characterized and evaluated in preclinical studies, demonstrating modest therapeutic e cacy in acute ischemia models [10][11][12][13]. Based on previous studies, hCPC were de ned as an hMSClike population with con rmed immunoregulatory capacity [19,20,22]. Considering the promising, but not statistically signi cant, results of the clinical studies based on these cells [23,24] a more detailed description of hCPC populations might lead to a better understanding of their mechanism of action and, ultimately, more effective treatments.
Analysis of the most relevant cytoplasmic proteins over-represented in hCPC suggests that these cells might be well suited to mount a more effective response to hypoxia and active collagen metabolism. P4HA1 and P4HA2 are both are overexpressed (P4HA1 > P4HA2) in the cytoplasmic compartment of hCPC and are activated by hypoxia [38]. In addition, P4HA1 has a critical role in the breast cancer-derived metastatic lung niche, where it is regulated by α-ketoglutarate [39]. We also found PLOD1 to be signi cantly overexpressed in the cytoplasm, and PLOD2 in the nucleus. P4HA1 and P4HA2 are required for collagen deposition, whereas PLOD2 is required for extracellular matrix stiffening and collagen ber alignment. Furthermore, DHX9 is highly upregulated during activation of quiescent cells to collagenproducing cells [40]. In the context of different adult stem cell compartments, CKAP4 and DHX9, both overexpressed in the hCPC cytoplasm, have been related to differentiation regulation [41]. CKAP4 is a nucleoplasmic shuttle protein that acts as a high-a nity receptor for antiproliferative factor (APF) [42]. DHX9 is an ATP-dependent helicase of double-stranded RNA and DNA-RNA complexes [43] that has been also proposed as a RISC-loading factor [44]. Finally, a strong ALDH1A3 expression has been found in the cancer lung stem cell compartment [45]. Concerning the proteins that were found moderately downregulated in hCPC cytoplasm, ASPH and TXNDC5 are also implicated in proliferation and cell motility regulation [46,47] Given these data, hCPC might be more effective in response to hypoxia associated-damage showing an active remodeling of the extracellular matrix. In this sense, murine B-CPC demonstrated a high survival index both under severe oxidative damage and infarction [36,37].
Regarding the proteins preferentially expressed in the hCPC nuclear compartment, we con rmed high levels of expression of several cardiogenic transcriptional factors such as GATA4, SOX17, WT1, GATA2 and TBX3 ( Supplementary Fig. S2 online), with GATA4 and SOX17 more differentially expressed in comparison with hMSC. Many of the referred transcription factors show a comparable expression pro le by RNAseq in B-CPC murine cardiac progenitors, analyzed in non-expanded progenitors [32].
In addition, comparative proteomics analysis of enriched nuclear fraction yielded a panel of proteins more represented in hCPC than in hMSC. Among them, ASPDH1 that was also con rmed by RT-qPCR is poorly characterized. Finally, among the proteins under-represented in the hCPC nuclear compartment, it is noteworthy that levels of PTRF/cavin-1 have been directly associated with cell senescence. PTRF has been demonstrated to mediate in transcription pausing and termination, with the nal dissociation of the transcription complex [48].
Among the nuclear over-represented proteins in hCPC, IL1A and IMP3 were selected for further analysis.
IL1A is a pro-in ammatory cytokine with multiple immune-regulatory functions. It is mainly expressed as a cell-associated form and not actively secreted in healthy tissue, but its membrane-associated form is critically involved in cell senescence [49]. IL1A is one of the four (IL1A, IL33, HMGB1 and S100) "dualfunction cytokines" described in mesenchymal cells. These cytokines play a role in the nucleus independently of their extracellular-mediated effects, as a classical cytokines, and have been also called "damage-associated molecular pattern" molecules or alarmins [28]. Unlike IL1B, processed IL1A has a nuclear localization sequence and is tra cked to the nucleus, regulating cell proliferation and migration [26,27]. For example, in acute lymphocytic leukemia T cells, overexpression of the IL1A nuclear propeptide has been demonstrated to promote proliferation and reduce apoptosis, by NFkB and SP1 upregulation [50]. Analyses of hCPC in homeostasis demonstrated a strong post-transcriptional regulation of IL1A mRNA and a highly preferential cytoplasmic location of IL1A. We found that IL1A is not related with the immunoregulation capacity of hCPC but, upon induction of apoptosis, IL1A was clearly upregulated and a substantial nuclear fraction was found; this behavior was not paralleled by hMSC. We also found a similar intracellular pattern for IL1B, although less pronounced. Overall these results suggested that IL1A, and probably IL1B, as a more recently proposed member, could have dual-cytokine pro le in hCPC, playing a role in the regulation of response to apoptosis.
IMP3 is an mRNA binding protein that, among other functions, regulates IGF2 expression [30]. In the context of cancer, there are numerous examples of the critical role of IMP3 favoring chemoresistance, aggressiveness and metastasis [51,52]. In neural and pancreatic cancer cells, IMP2-and IMP3-bound transcripts are localized in cytoplasmic RNA granules that accumulate in dendrites or membrane protrusions, where they are preferentially translated [53]. In pancreatic ductal adenocarcinoma cells, IMP3 modulates miRNA-mRNA interactions [54]. However, IMP3 binding could result both in an enhanced expression of the target mRNA [55] or its destabilization [56]. Although different pathways and targets have been associated with the overexpression of IMP3 in cancer, few studies have addressed the role of IMP3 in healthy developmental processes; i.e. muscle growth is regulated by IMP3 levels, controlled by let-7b [57] and adult megakaryocyte development is also under the control of IMP3, by regulating P-TEFb [58].
hCPC in homeostasis show a clear overepresentation of IMP3, but not IMP2, in the nuclear compartment and induction of apoptosis provoked an enrichment in the nuclear compartment. IMP3 knockdown reduced hCPC proliferation and migration capacity, although it had no obvious impact on viability. IMP3 has been found to promote cell migration in glioma by increasing the levels of p65 protein (RELA; subunit of NF-κB heterodimer), but without modifying transcript levels [59]. In glioblastomas IMP3 also promotes cell proliferation, migration and invasion by inducing epithelial-mesenchymal transition [51].
Finally, we analyzed a panel of candidate target genes whose expression could be affected by the downregulation of IMP3 in hCPC. cMYC, CD44 and CDK6 were demonstrated previously to be targeted by IMP3 in mixed lineage leukemia, enhancing the half-life of the transcripts [55]. Silencing of IMP3 (siIMP3 cells), however, resulted in a moderate and non-signi cant downregulation in hCPC. By contrast, silencing of IMP3 led to a signi cant dowregulation of HMGA2 and PTPRF. HMGA2 is considered as an architectural transcription factor that is involved in growth regulation and tumorigenesis [60].
Interestingly, it has been demonstrated that IMP3 ribonucleoprotein complexes contain HMGA2 mRNA, preventing miRNA-directed mRNA decay during tumor progression [61]. In addition, it has been recently demonstrated that HMGA2 controls both, proliferation and migration / metastasis, in colon cancer [62]; analyses of other HMGA2 candidate genes associated with cell migration (ARF6, ARHGEF4) rendered negative results. It is also worth noting that HMGA2 mRNA is signi cantly overexpressed (17.8 fold) in hCPC versus hMSC [20]. By contrast, high PTRF expression levels correlate with an increased senescence in human broblasts [48]. Therefore, these data suggest that PTRF and HMGA2 are regulated by IMP3 and, consequently, could be involved in hCPC proliferation/migration regulation.
In conclusion, we have compared, using a label-free proteomic approach, the differential cytoplasmic and nuclear compartments of human CPC (hCPC) versus human mesenchymal stem cells (hMSC) and broblasts. Globally, hCPC, with a clear cardiogenic transcriptional factor pro le, are well suited to mount an effective response to hypoxia with active collagen metabolism. IL1A, characterized as a dual-function cytokine, seems to play a role in the regulation of the hCPC response to apoptosis caused by oxidative stress. Finally, IMP3 was demonstrated to be involved in hCPC proliferation and migration.

Ethical approval
Human CPC were obtained from human right atria appendage from adult donors, with no relevant cardiac pathology, and subjected to cardiac surgery with extracorporeal circulation; during the procedure this tissue is normally discarded during cannulation. Human CPC were isolated from human myocardial samples by c-kit immunoselection, as described [19].  [20,22].
Cells and culture conditions hCPC were maintained and expanded as previously indicated [20], essentially under equivalent conditions to those used in the CAREMI clinical trial (EudraCT 2013-001358-81 Label-free proteomics analysis hCPC3 protein levels were compared with those of hMSC19, essentially as previously described [20]. Cells were expanded to P7-P8, recovered and, after several washes in PBS, pellets (5-8 x 10 7 cells) were collected. Subcellular cytoplasmic and nuclear protein fractions (see Supplementary Methods section online) were obtained using the Qproteome Cell Compartment Kit (Qiagen, Barcelona, Spain). Samples (~500 mg) were digested using an in-gel digestion protocol, as described [20]. Tryptic peptides were dissolved in 0.1% formic acid (FA) and loaded on a liquid chromatography-mass spectrometry (LC-MS/MS) system for online desalting on C18 cartridges and further analysis by LC-MS/MS, using a reverse-phase nanocolumn (75 mm inner diameter × 50 mm, 3 mm-particle size, Acclaim PepMap 100 C18; Thermo Fisher Scienti c, San Jose, CA) in a continuous (0-30%) acetonitrile gradient consisting of B (90% acetonitrile, 0.5% formic acid), in 180 min, 30-43% in 5 min and 43-90% in 2 min. A ~200 nL/min ow rate was used to elute peptides from the nanocolumn to an emitter nanospray needle for real time ionization and peptide fragmentation onto an ion trap-orbitrap hybrid mass spectrometer (Orbitrap Elite, Thermo Fisher). Bioinformatic identi cation and analyses methods are described in Supplementary Methods section online. Relative representation of the different proteins identi cated was estimated by peptide-counting; three replicas were analyzed for each comparison. When indicated pathway analysis with PANTHER software [64] was carried out.

RT-qPCR analyses
Total mRNA was isolated as described [19]. cDNA rst strands were synthesized from total RNA (1 mg) with the SuperScript III First-Strand Synthesis System (Invitrogen). Genes of interest (see Supplementary Methods section online) were evaluated by quantitative RT-qPCR in a Mastercycler Ep-Realplex platform (Eppendorf, Hamburg, Germany), using Power SYBR Green reagents (Applied Biosystems, Foster City, CA). Cycle conditions were 95°C for 10 min, followed by 40 cycles of 95ºC for 15 s and 60°C for 1 min.
Quanti ed gene expression values were normalized against those of GUSB or GAPDH.

Western blotting
Cells were harvested in RIPA (radioimmunoprecipitation assay) lysis buffer, and equal amounts of lysate were separated by 10% SDS-PAGE. When indicated, cytoplasmic or nuclear fractions were obtained using the NE-PER™ Nuclear and Cytoplasmic Extraction kit (Thermo Fisher Scienti c). Proteins were transferred to PVDF membranes using the iBlot Dry Blotting System (Invitrogen). After incubation with primary and secondary antibodies, signals were developed using an ECL kit (GE Healthcare, Uppsala, Sweden). Supplementary Methods section online includes the list of all primary and secondary antibodies used.

Wound healing assay
For migration (scratch) assays, hCPC cells were cultured to con uence and starved in serum-free medium (24 h). The cell monolayer was then scraped with a pipette tip (t=0 h) and cultures were monitored (t=6-24 h) to evaluate their wound healing capacity. Images were acquired and migration rates were measured using ImageJ software (NIH, Bethesda, MD).

Statistics
Assays were performed three times and data expressed as mean ± SD; black lines summarize p-values (***<0.002, **<0.02, *<0.05) for hCPC versus broblasts or hMSC (one-way analysis of variance followed by the Bonferroni correction for multiple comparison).

Consent for publication
Not applicable