MicroRNA-21/PDCD4 Proapoptotic Signaling From Circulating CD34⁺ Cells to Vascular Endothelial Cells: A Potential Contributor to Adverse Cardiovascular Outcomes in Patients With Critical Limb Ischemia Free

The chemokine stromal-derived factor 1 (SDF-1) participates in cardiovascular repair through the mobilization of bone marrow (BM)-derived CD34⁺ progenitor cells that express the CXCR4 receptor. CD34⁺CXCR4⁺ cells positively interact with the vascular endothelium by releasing trophic soluble factors and extracellular vesicles (EVs). Risk factors, ageing, and age-related diseases compromise this homeostatic mechanism by perturbing the BM microenvironment (1,2). Interestingly, both biased myelopoiesis and deficit/dysfunction of CD34⁺ cells are associated with an increased risk of cardiovascular morbidity and mortality (3–10).

We showed that CD34⁺ cell migration predicted cardiovascular mortality in patients with type 2 diabetes (T2D) undergoing revascularization of critical limb ischemia (CLI) (10). Phenotypic changes in CD34⁺ cells may cause systemic vascular damage in these high-risk patients through antiangiogenic and proapoptotic miRNAs (miRs) (10–13).

The current study investigated 1) if CD34⁺ cells predict cardiovascular mortality long term and 2) how CD34⁺ cells cause vascular damage instead of repair.

See Supplementary Material online for supplementary tables and figures.

We performed a phone survey of T2D-CLI patients recruited in the NCT01269580 study, which demonstrated that migration of peripheral blood (PB) CD45^dimCD34⁺CXCR4⁺KDR⁺ cells predicted cardiovascular death after angioplasty (10). At 6 years, 15 patients were lost to follow-up and were therefore excluded. Baseline characteristics of the remaining 104 patients are summarized in Supplementary Table 1.

Clinical characteristics of a new cohort comprising 47 control subjects and 41 patients with T2D, of which 30 affected by CLI, are reported in Supplementary Table 2. T2D and CLI were defined according to the American Diabetes Association and TASC 2007, respectively. Exclusion criteria: acute disease/infection, immune diseases, current/past hematological disorders or malignancy, unstable angina, recent (within 6 months) myocardial infarction or stroke, liver failure, renal failure, and pregnancy. All participants signed an informed consent to donate the BM leftovers from the femoral head otherwise discarded during hip replacement surgery (control subjects and subjects with T2D without complications) or BM aspirates of the iliac crest performed ad hoc for the study (patients with T2D-CLI).

The studies received ethical authorization from the IRCCS MultiMedica (PROT 20/2010), Bristol University (REC14/SW/1083 and REC14/WA/1005), and Santa Maria Ca’ Foncello Hospital (DDG 2333/2017) (13,14). BM-CD34⁺ cells were assessed in situ or following immunomagnetic bead isolation as previously described (13). Some assays were conducted using the CD34⁺ cell–derived conditioned medium (CCM) and respective EVs collected using Exospin kit (Cell Guidance System) following the manufacturer’s instructions.

miR profiling was conducted at Exiqon Services, Vedbæk, Denmark. Total RNA (30 ng) was reverse-transcribed using the miRCURY LNA Universal reverse transcriptase (RT) miR PCR, Polyadenylation and cDNA synthesis kit (Exiqon). The amplification was performed in a LightCycler 480 Real-Time PCR System (Roche) in 384-well plates. A total number of 372 miRs were tested. A cut-off (C_t >37) was used to consider a miR as expressed and included in the subsequent analyses.

For biological validation and expressional studies, RNA was extracted from cultured cells, cell-derived CCM/EVs, plasma, and plasma-derived EVs using miRNeasy Mini Kit (Qiagen) following the manufacturer’s instructions. Quantitative RT-PCR was performed with the QuantStudio 6 Flex Real-Time PCR using miR TaqMan probes (Thermo Fisher) and long noncoding RNA (ncRNA)/gene primers listed in Supplementary Table 3. miR-21 was silenced in control BM-CD34⁺ cells by transfecting them with 50 nmol/L anti-miR-21–5p (AM10206; Ambion), while controls were transfected with a nontargeting sequence or scramble (SCR) (AM4611, Silencer Negative Control; Ambion), using transfection reagents (Genlantis) according to the manufacturer’s protocol.

PB–mononuclear cell (MNC) in vitro migration assays using SDF-1 as a chemoattractant were performed using transwell chambers as described (10). Migrated cells were stained for surface antigens CD45 (FITC; BD Biosciences), CD34 (PeCy7; BD Biosciences), and CXCR4 (APC; BD Biosciences), then fixed in Met-OH, permeabilized in PBS-Tween 0.1%-BSA 0.5%, and stained for intracellular antigen PDCD4 (Abcam) with secondary antibody anti-rabbit-phycoerythrin (BD Biosciences). Cells were analyzed using a FACSCanto flow cytometer with the FACSDiva software (both from BD Biosciences).

BM-CD34⁺ cells or human umbilical vascular endothelial cells (HUVECs) were stained with 7-aminoactinomycin D (7-AAD)/Annexin-V and analyzed on FACSCanto II flow cytometer (BD Biosciences). The 7-AAD⁻/Annexin V⁺ combination identified apoptotic cells. Caspase-3/7 Assay (Promega) was performed on HUVECs exposed to CD34⁺ cells CCMs.

HUVECs were transfected with 50 nmol/L siPDCD4 (Dharmacon) or nontargeting sequence (AM4611, Silencer Negative Control; Ambion).

Paraffin-embedded BM sections were incubated with the following: rabbit anti-PDCD4 (1:200; abcam), mouse anti-CD34 (1:5; Dako), and rabbit anti-4-hydroxy-2-nonenal (4-HNE) (1:500; Bioss). Nuclei were counterstained with DAPI (1 μg/mL) (Sigma-Aldrich). Images were acquired with a fluorescent microscope (DM6 B; Leica Microsystems) at 63× magnification and analyzed using Fiji software. Cultured HUVECs were fixed with 4% paraformaldehyde (PFA) (Electron Microscopy Science), permeabilized with 0.3% triton (Sigma-Aldrich), and stained with rabbit anti-PDCD4 (1:200). Nuclei were counterstained with DAPI (1 μg/mL) (Sigma-Aldrich). Microphotographs were captured using a Zeiss microscope equipped with digital image processing software (AxioVision Imaging System).

BM sections were stained with rabbit anti-PDCD4 (1:200) by BenchMark ULTRA system (Roche). Ten images for each sample were acquired with an optical microscope (Eclipse E800; Nikon) at 40× magnification and quantified by using ImageJ software.

BM-CD34⁺ cells, their CCM, or isolated EVs were cocultured with HUVECs (1:1 ratio) on Matrigel (Corning Incorporated Life Sciences) (15).

Continuous variables were expressed as mean ± SD or SE as indicated, tested for normality by the Kolmogorov-Smirnov test, and compared using parametric (t test or ANOVA) or nonparametric tests (Wilcoxon or Kruskal-Wallis), as appropriate. Categorical variables were expressed as frequency and percentage and were compared by χ² test or Fisher exact test. A P value <0.05 was considered statistically significant. SAS (version 9.4), R (version 3.4.4), and GraphPad Prism (version 7) were used for analyses and graphics.

In study 1, cumulative incidences of events were drawn overall and for data stratified by cells (above versus below the median) that significantly differed between participants with or without events. This analysis considered the competitive causes of the event (16); specifically, in the case of cardiovascular death, other causes of death were considered as a competitive event, and vice versa. Comparisons between incidence curves were assessed fitting the proportional subdistribution hazards regression model (17). Time-to-event was defined as the time from revascularization to death (cardiovascular or for other causes). Patients lost to follow-up were excluded from the analyses. The 15th day of a given month and the month of June were imputed if the day or month of follow-up was missing, respectively. Incidence rate and 95% CI at 3 years and 6 years of follow-up were calculated for cardiovascular death and for other causes of death.

To evaluate the association between basal cell counts and migratory activity and risk of death, the event-specific hazard ratio (HR) and 95% CI was calculated. HRs associated with cell migration were evaluated for a 1-year increase, for the presence of a history of coronary artery disease, and for a 0.01-unit increase in the percentage of CD45^dimCD34⁺CXCR4⁺KDR⁺ migrated cells toward SDF-1 over total MNCs. All models were performed for the presence of investigated variable, if dichotomous, and for a 1-unit increase of continuous variables, if not otherwise specified. A multivariable regression model was subsequently implemented, adjusting for prognostic features that were found significantly associated with the event in the univariate analysis.

Supplementary Table 1 illustrates clinical/laboratory data of the 104 T2D-CLI patients who completed the 6-year follow-up.

Three outcomes were considered: no event (N = 54), cardiovascular death (N = 32), and other causes of death (N = 18). Age at recruitment was the only clinical data that differed among the three outcomes (P = 0.0067) (Supplementary Table 4). Regarding CD45^dimCD34⁺CXCR4⁺KDR⁺ cells, migration toward SDF-1 (experimental setting illustrated in Fig. 1Ai) was higher in the cardiovascular death group compared with no event group or the other causes of death group (P = 0.0312), whereas there was no difference in PB levels of CD45^dimCD34⁺CXCR4⁺KDR⁺ cells or in the migration of total MNCs and CD45^dimCD34⁺CXCR4⁺KDR⁺ cells exposed to the SDF-1 vehicle (Supplementary Table 5 and Supplementary Fig. 1).

As shown in Fig. 1Aii and Supplementary Table 6, patients with values of SDF-1–migrated cells greater than or equal to the median had higher cumulative incidence of cardiovascular death compared with those with values less than the median (P = 0.0012). Cell migration was associated with an increased cardiovascular risk (HR 1.10, 95% CI 1.04–1.17, P = 0.0005, data not shown), which was further confirmed by a multivariable Cox analysis simultaneously assessing the effect of age and prevalence of coronary artery disease (Supplementary Table 7).

To obtain mechanistic insights into the observed association, we studied the functional characteristics of BM-CD34⁺ cells from a new series of patients with T2D or T2D-CLI and control subjects without diabetes.

In line with previous studies (13), flow cytometry analyses demonstrated that CD34⁺ cells immune-magnetically sorted from the BM of T2D-CLI patients have a 2.5-fold higher abundance of 7-AAD⁻/Annexin V⁺ events compared with controls (Fig. 1B).

Moreover, using an in vitro Matrigel assay, we demonstrated that the coculture of HUVECs with T2D-CLI BM-CD34⁺cells resulted in a lower number of branches compared with the coculture of HUVECs and control BM-CD34⁺ cells (Fig. 1C). A marked decrease in endothelial network formation was also observed when incubating HUVECs with BM-CD34⁺ cell–derived CCM or EVs from patients with T2D-CLI (Fig. 1C). These data demonstrate that BM-CD34⁺ cells from T2D-CLI patients have reduced viability and can transfer destabilizing signals to endothelial cells through factors secreted as soluble molecules or packaged in EVs.

Diabetes influences the expression of several miRs in hematopoietic cells (11–13). An unbiased miR profiling of CD34⁺ cells isolated from the BM of control subjects (N = 6) and T2D patients with (N = 6) or without CLI (N = 7) identified a suppressive effect of T2D on the quantity of expressed miRs (Supplementary Fig. 2A and B), which might be attributable to the downregulation of Dicer (Supplementary Fig. 2C and D), an endoribonuclease involved in miRNA maturation, as also described by others (18).

Supplementary Table 8 shows that 56 miRs were commonly expressed in the three groups. Moreover, 2 sets of 11 different miRs were shared by control subjects and patients with T2D-CLI or control subjects and patients with T2D, respectively. Of those not shared, 49 were unique to control subjects and 1, namely miR-146a, to T2D-CLI, whereas none was exclusive to T2D. As shown in Supplementary Table 9, 18 miRs were differentially expressed in cells from T2D subjects, with or without CLI, compared with control subjects, with 2 of them, miR-21 and miR-30e, being shared by the two T2D groups. The heatmap diagram in Fig. 2A indicates a marked separation of miR expression in patients with T2D-CLI compared with control subjects, while values from patients with T2D were spread between control subjects and patients with T2D-CLI. Among the miRs that showed a differential expression between patients with T2D-CLI and control subjects, we studied a set of six miRs (miR-125a, miR-222, let-7e, miR-93, miR-21, and miR-30e) known to control cell survival, differentiation, hematopoiesis, and angiogenesis (Supplementary Table 10). To validate the profiling results, we performed single PCR analyses for the six miRs of interest on BM-CD34⁺ cells isolated from a new set of donors (control subjects, N = 5; patients with T2D, N = 4; patients with T2D-CLI, N = 4) using an Applied Biosystem platform and normalizing miR expression to U6snoRNA, which showed a stable expression among the three groups. Three miRs from the set of choice, miR-125a, miR-21, and miR-30e, were significantly modulated in the new cohort (Fig. 2B).

miR-21, one of the most highly expressed miRs in mammalian cells, is modulated in cardiovascular disease (19–23). However, little is known regarding the expression of miR-21 in hematopoietic progenitor cells. Data of RT-PCR confirmed the downregulation of mature miR-21 in sorted T2D-CLI BM-CD34⁺ cells (Fig. 2Ci), whereas the levels of the corresponding pri-miR were similar to controls (data not shown). miR-21 was also found in CD34⁺ cell–derived CCMs and EVs, but no difference was seen when comparing patients with T2D-CLI and control subjects (Fig. 2Ci). Interestingly, miR-21 levels were decreased in plasma and EVs isolated from the PB of T2D-CLI patients (Fig. 2Cii).

Long ncRNAs can act as miR sponges, thereby interfering with regulation of miR targets. We investigated the expression of several long ncRNAs reportedly implicated miR-21 modulation (24–28). As shown in Fig. 2Ci, TALNEC2 was upregulated in T2D-CLI BM-CD34⁺ cells and respective CCM; TUG1 was decreased in cells and increased in the CCM; MEG3 was not altered; and TCONS was downregulated in CCM. These data suggest that TALNEC2 could inhibit miR-21 at intracellular level, and, together with TUG1, at extracellular level. Moreover, all the studied sponges were upregulated either in plasma, EVs, or both, suggesting they may synergize in inhibiting miR-21 in the circulation (Fig. 2Cii).

The proapoptotic factor PDCD4 is a validated target of miR-21 (19). In line with the miR-21 downregulation, we found higher PDCD4 mRNA levels in T2D-CLI BM-CD34⁺ cells (Fig. 2D). Likewise, in situ immunohistochemistry confirmed the higher expression of PDCD4 in the BM (Fig. 2E). Altogether, these data point at a novel molecular mechanism involving the downregulation of miR-21 and induction of PDCD4 in T2D-CLI CD34⁺ cells.

CD34⁺CXCR4⁺ cells represented a small fraction (1.5 ± 0.2%) of the total CD34⁺ cell population in PB of T2D-CLI patients. PDCD4 was more abundant in the CD34⁺CXCR4⁺ (77.9 ± 0.4%) than in the CD34⁺CXCR4⁻ cell fraction (5.8 ± 0.2%). Moreover, SDF-1–stimulated migration of PB-MNCs resulted in an enrichment of cells expressing both CXCR4 and PDCD4 in the migrated fraction (Supplementary Fig. 3).

Next, we sought confirmation of a direct link between T2D-CLI–associated miR-21 downregulation and BM-CD34⁺ cell dysfunction. To this aim, we silenced miR-21 in control BM-CD34⁺ cells using an anti-miR strategy. The effective miR-21 knockdown (Fig. 3A) was associated with PDCD4 upregulation (Fig. 3B) and increased apoptotic events compared with SCR-treated cells (Fig. 3C). Moreover, miR-21 silencing conferred antiangiogenic properties to CD34⁺ cells as well as to the CD34⁺ cell–derived CCM and EVs (Fig. 3D).

We next assessed if the negative crosstalk between CD34⁺ cells and endothelial cells involves the direct transfer of miR-21 or PDCD4 or is mediated by associated factors. To this purpose, using a protocol illustrated in Fig. 3E, we measured the relative expression of miR-21 and PDCD4 in HUVECs exposed to the CCM from control CD34⁺ cells (either naive or transfected with anti-miR-21 or SCR) or to the CCM from T2D-CLI CD34⁺ cells. Interestingly, plotting the expressional values of miR-21 and PDCD4 from the four groups demonstrated an inverse relationship between the miR and its target (Fig. 3E). This indicates that modulation of the miR-21/PDCD4 duo in BM-CD34⁺ cells can induce similar expressional changes in the exposed endothelial cells. Next, employing a protocol illustrated in Fig. 3F, we demonstrated that PDCD4 silencing in HUVECs remarkably reduced the apoptosis caused by the exposure to the CCM from T2D-CLI CD34⁺ cells (Fig. 3F). A dedicated ELISA could not detect PDCD4 in the CD34⁺ cell–derived CCM. These data suggest that CD34⁺ cells induce apoptosis in HUVECs through different paracrine mechanisms.

The presence of miR-21 sponges in the CCM from T2D-CLI CD34⁺ cells suggested that they could act paracrinally to inhibit miR-21 in endothelial cells. Therefore, we assessed TALNEC2, TUG1, MEG3, and TCONS in HUVECs either naive or exposed to the CCM from CD34⁺ cells from control or T2D-CLI subjects. Results confirmed the downregulation of miR-21 and the induction of PDCD4 by the T2D-CLI CCM and demonstrated that TUG1 was the only upregulated long ncRNA in conditioned HUVECs (Fig. 4A).

Reciprocal interactions exist between long ncRNAs and reactive oxygen species (ROS) production and scavenging (29). Relevant to our study, H₂O₂ and hypoxia reportedly induced TUG1 in cardiomyocytes, thereby increasing ROS production and apoptosis (30). Hence, we hypothesized that oxidative stress could be involved in the transfer of proapoptotic signaling to endothelial cells. This was confirmed by the following evidence: 1) T2D-CLI BM-CD34⁺ cells showed elevated in situ 4-HNE staining, a marker of oxidative damage and lipid peroxidation (Fig. 4B); 2) treatment of HUVECs with increasing H₂O₂ concentrations induced miR-21 inhibition and PDCD4 upregulation (Fig. 4C); and 3) ROS scavenging with N-acetylcysteine (NAC) blocked the miR-21 downregulation in HUVECs exposed to the CCM from T2D-CLI CD34⁺ cells (Fig. 4Di) or to the CCM from anti-miR-21-transfected control CD34⁺ cells (Fig. 4Dii). In parallel, NAC abrogated the induction of PDCD4 by the CCMs (Fig. 4Diii and iv). It also counteracted the apoptosis of HUVECs exposed to the CCM from T2D-CLI CD34⁺ cells (Fig. 4Ei) but was unable to prevent the apoptosis of HUVECs exposed to the CCM from miR-21-silenced CD34⁺ cells (Fig. 4Eii). These findings indicate that ROS in association with TUG1 is the likely mediator for transmission of an altered miR-21/PDCD4 balance from T2D-CLI CD34⁺ cells to endothelial cells (Fig. 4F). They also point to the possibility that the total disruption of miR-21 impacts on additional proapoptotic inducers that are independent of ROS.

In an extended follow-up of T2D-CLI patients, CD34⁺ cell migration maintained a predictive value at 6 years after angioplasty. In these critical patients, CD34 cells responsive to SDF-1 chemoattraction exert negative effects on the vascular endothelium through a mechanism involving miR-21 inhibition and PDCD4 upregulation.

In the high-migratory group, cardiovascular mortality accrued during the first years of follow-up, with the difference versus the low-migratory group remaining unchanged later. Most patients with the highest cardiovascular risk had already died at that stage, resulting in a reduction of the target for prediction. Moreover, mortality for other causes can act as a strong opposer in an elderly population. Nonetheless, the biomarker maintained its validity in a multivariable analysis accounting for the age of participants.

Reduction and dysfunction of stem/progenitor cells is associated with and predicts adverse outcomes of diabetic complications (8). Thus, at first glance, our data appear counterintuitive. We hypothesized that subfractions of BM-CD34⁺ cells could become antiangiogenic and proapoptotic due to the adverse metabolic milieu they are exposed to in the BM and in circulation. Once they enter the circulation, these cells may convey pathogenic signals to the vascular endothelium, thereby accelerating ischemic complications.

Accumulating evidence indicates that CLI aggravates the remodeling effect of T2D on the BM niche and induces a senescent phenotype in CD34⁺ cells, which may be, at least in part, attributable to alteration in miR biogenesis, expression, and degradation (13,18). The observed reduction of expressed miRs in BM-CD34⁺ cells from T2D patients may be attributable to a block in the miR processing, as suggested by Dicer downregulation. Among differentially expressed miRs, we focused on miR-21 because of its involvement in cardiovascular disease (31–34). miR-21 downregulation in T2D-CLI CD34⁺ cells was associated with upregulation of the programmed cell death protein PDCD4, a validated inhibitory target of miR-21, and with alterations in viability and proangiogenic activity of CD34⁺ cells. Silencing miR-21 reproduced the same phenotype in CD34⁺ cells of subjects without diabetes. Moreover, data of an SDF-1 migration assay on T2D-CLI PB-MNCs demonstrated the high coexpression of PDCD4 and CXCR4 within the migrated CD34⁺ cell fraction. This finding provides a key of interpretation for the link between CD34⁺CXCR4⁺ cell migration and cardiovascular death in the follow-up study on T2D-CLI patients.

miR-21 is the most abundant miRNA in macrophages and its downregulation has been associated with induction of atherosclerosis, plaque necrosis, and vascular inflammation (35). Silencing of miR-21 in macrophages increases the expression of mitogen-activated protein kinase kinase 3, thereby leading to the activation of the p38-CHOP and c-Jun N-terminal kinase signaling pathways and triggering macrophage apoptosis (35). Additionally, miR-21–silenced macrophages are unable to remove apoptotic cells, which contributes in delaying the resolution of inflammation (35). The miR-21 inhibitory target PDCD4 acts as a tumor suppressor protein involved in programmed cell death. Recent cardiovascular research has shown that PDCD4 is upregulated in coronary arteries of atherosclerotic rats, where it participates in the formation of coronary plaques, through destabilization of vascular smooth muscle cells and promotion of inflammatory chemokines (36). Nonetheless, the regulation of miR-21/PDCD4 interaction is only partially known. Prostaglandins reportedly act as inducers of miR-21 expression and suppressors of PDCD4 protein, whereas cyclooxygenase 2 inhibitors produce opposite effects (37). Moreover, several long ncRNAs can act as sponges for miR-21 (24–28). For instance, miR-21 is a direct target of MEG3, and in hypoxic vascular cells, MEG3 interferes with miR-21 modulation of PTEN resulting in cell proliferation and migration (27). Likewise, studies on H9c2 cells showed that TALNEC2 modulates miR-21/PDCD4 expression under hypoxia, aggravating its consequences (25). TCONS is an endothelium-associated long ncRNA involved in plaque progression. Binding of miR-21 to TCONS reduces its expression and, for this reason, was proposed as potential treatment to improve endothelial dysfunction and plaque stabilization (28). Finally, TUG1 was proposed to interact miR-21 and to modulate endothelial cell apoptosis (38). Both oxidative stress and ischemia induce the sponging activity of TUG1, thereby stimulating intracellular ROS accumulation and aggravating the ischemic injury (30). Interestingly, we found that the above long ncRNAs were modulated in CD34⁺ cells and PB of T2D-CLI patients. Moreover, exposure of HUVECs to the CCM of T2D-CLI CD34⁺ cells induced the expression of TUG1 and PDCD4, while suppressing miR-21. This mechanism may be ROS dependent when considering that the inhibition exerted by NAC resulted in improved cell survival. We propose that CD34⁺ cells recruited from PB may exert a sponge-dependent inhibition of the interaction between miR-21 and PDCD4, thereby sustaining vascular damage. Elevated PB levels of miR-21 sponges could strengthen this cellular crosstalk.

These findings have clinical and therapeutic implications, but they also raise new questions. Currently, the long-term prognosis of CLI is based on clinical parameters (39). Our study suggests that the assessment of the CD34⁺ cells profile may help identify high-risk patients for whom more aggressive treatments are necessary. Although more sophisticated than traditional biomarkers, cellular biomarkers can be very useful in helping us understand the complex interplay among cellular systems in inducing diabetic vascular disease. In this respect, circulating cells offer a more feasible means for molecular profiling than BM cells, also considering the variability in cell composition of different BM sites. This last aspect represents a limitation of our study, as BM samples were obtained from the femoral head or iliac crest aspirates with the intention of not interfering with clinical practice and patient care.

PDCD4 might represent a valuable biomarker and therapeutic target in ischemic disease. The latter assumption is indirectly supported by the established benefit of prostaglandin E1, an inhibitor of PDCD4, in the treatment of limb ischemia. Novel treatments targeting upstream modulators of PDCD4, including miR-21 and related sponges, might be also considered for treatment of CLI. Finally, autologous BM-CD34⁺ cells are currently used in clinical trials of CLI patients. Our study calls for caution in using CD34⁺ cells that carry a proapoptotic and antiangiogenic molecular signature, e.g., low miR-21/high PDCD4. New investigation is needed to determine if this signature can be exploited to increase the safety and efficacy of the cell therapy approach.

Acknowledgments. The authors thank Dr. Simona Rodighiero, Imaging Unit of the Department of Experimental Oncology, European Institute of Oncology, Milan, Italy, for the support in the generation of the microscopy data and immunofluorescence data presented herein.

Funding. Funding and financial support was obtained from the Italian Ministry of Health, Ricerca Corrente to the IRCCS MultiMedica, Foundation Cariplo (2016-0922), and a British Heart Foundation program grant. Study 1 was also supported by Diabetic ONLUS Association, section of Treviso.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. G.S. and P.M. acquired funding, analyzed data, and wrote the manuscript. E.S. researched data relative to both study 1 and study 2. E.T. performed the statistical analyses for study 1. D.M. isolated cells and performed the in vitro molecular analyses of study 2. O.C. conducted tissue and cells immune characterization. D.F.-M. isolated cells and provided biological samples. F.C. and P.O. provided human BM samples and contributed to the discussion. A.P. acquired funding and contributed to the discussion. A.F., P.M.S., and L.S. were involved in collection of the human BM samples and recording the clinical data. M.S. acquired funding, provided human BM samples, and corrected the manuscript. G.S. and P.M. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Keystone Symposia: Novel Aspects of Bone Biology (E3), Snowbird, UT, 13–16 June 2018 and at the European Society of Cardiology Congress, Paris, France, 31 August to 4 September 2019.