Exosomal Expression of CXCR4 Targets Cardioprotective Vesicles to Myocardial Infarction and Improves Outcome after Systemic Administration

Cell therapy has been evaluated to enhance heart function after injury. Delivered cells mostly act via paracrine mechanisms, including secreted growth factors, cytokines, and vesicles, such as exosomes (Exo). Intramyocardial injection of cardiac-resident progenitor cells (CPC)-derived Exo reduced scarring and improved cardiac function after myocardial infarction in rats. Here, we explore a clinically relevant approach to enhance the homing process to cardiomyocytes (CM), which is crucial for therapeutic efficacy upon systemic delivery of Exo. By overexpressing exosomal CXCR4, we increased the efficacy of plasmatic injection of cardioprotective Exo-CPC by increasing their bioavailability to ischemic hearts. Intravenous injection of ExoCXCR4 significantly reduced infarct size and improved left ventricle ejection fraction at 4 weeks compared to ExoCTRL (p < 0.01). Hemodynamic measurements showed that ExoCXCR4 improved dp/dt min, as compared to ExoCTRL and PBS group. In vitro, ExoCXCR4 was more bioactive than ExoCTRL in preventing CM death. This in vitro effect was independent from SDF-1α, as shown by using AMD3100 as specific CXCR4 antagonist. We showed, for the first time, that systemic administration of Exo derived from CXCR4-overexpressing CPC improves heart function in a rat model of ischemia reperfusion injury These data represent a substantial step toward clinical application of Exo-based therapeutics in cardiovascular disease.


Introduction
Over the past 20 years, cell-therapy approaches have been investigated as an exciting strategy to repair heart tissue after myocardial infarction (MI) [1][2][3]. Today the prevailing view is that stem and progenitor cells elicit their therapeutic effects via indirect paracrine mechanisms that result in the salvage of myocardium at risk, rather than cardiac regeneration by trans-differentiating in an attempt at enriching cells and Exo for hCXCR4 receptor. Transfection significantly increased expression of CXCR4 in cells (CPC CXCR4 ) in comparison to CPC transfected with empty backbone (CPC CTRL ), as verified by flow-cytometer (FC) and western blot (WB) analyses (Figure 1a,b). CXCR4 could be detected on Exo secreted by CPC CXCR4 (Exo CXCR4 ) and slightly onto surface of Exo derived from CPC CTRL (Exo CTRL ) (Figure 2a,b). The level of expression of CXCR4 in Exo CTRL was comparable to naïve expression of this receptor with no increment due to the transfection of backbone plasmid ( Figure A1e). 3 increased expression of CXCR4 in cells (CPC CXCR4 ) in comparison to CPC transfected with empty backbone (CPC CTRL ), as verified by flow-cytometer (FC) and western blot (WB) analyses (Figure 1a,b). CXCR4 could be detected on Exo secreted by CPC CXCR4 (Exo CXCR4 ) and slightly onto surface of Exo derived from CPC CTRL (Exo CTRL ) (Figure 2a,b). The level of expression of CXCR4 in Exo CTRL was comparable to naïve expression of this receptor with no increment due to the transfection of backbone plasmid ( Figure A1e). Quantitative data are relative expression levels (densitometry) of CXCR4 normalized for β-actin and showed as fold-changes to CPC CTRL (n = 6 patients/group) * p < 0.05 (two-tailed, paired Student t-test).
To verify the correct transmembrane orientation of the receptor in Exo CXCR4 , FC analysis was performed using two different antibodies directed against both the N-terminus (extraluminaldomain) and the C-terminus (intraluminal-domain) region in the absence of permeabilization. Only the anti-N-terminus CXCR4 antibody resulted in a positive staining (Figure 2c). Data was confirmed by Cryo Electron Microscopy (Cryo-EM) analysis and immunogold staining experiment, using anti-CXCR4 (N-terminus) as a primary antibody (Millipore) (Figure 2d).
In order to determine the presence of endogenous versus exogenous CXCR4 mRNA in Exo, we constructed 2 primer pairs that specifically amplified the hCXCR4 sequence only (pairs 1,2, Figure  A1) and the hCXCR4 sequence with the fused c-Myc tag (pairs 1-3, Figure 2e). The hCXCR4 mRNA was slightly detectable in Exo CTRL and dramatically increased in Exo CXCR4 . Only Exo CXCR4 contained the exogenous sequence as evidenced by the presence of the c-Myc tag amplicon (Figure 2e). By WB  CXCR4 . Quantitative data are relative expression levels (densitometry) of CXCR4 normalized for β-actin and showed as fold-changes to CPC CTRL (n = 6 patients/group) * p < 0.05 (two-tailed, paired Student t-test).
To verify the correct transmembrane orientation of the receptor in Exo CXCR4 , FC analysis was performed using two different antibodies directed against both the N-terminus (extraluminal-domain) and the C-terminus (intraluminal-domain) region in the absence of permeabilization. Only the anti-N-terminus CXCR4 antibody resulted in a positive staining (Figure 2c). Data was confirmed by Cryo Electron Microscopy (Cryo-EM) analysis and immunogold staining experiment, using anti-CXCR4 (N-terminus) as a primary antibody (Millipore) (Figure 2d).
In order to determine the presence of endogenous versus exogenous CXCR4 mRNA in Exo, we constructed 2 primer pairs that specifically amplified the hCXCR4 sequence only (pairs 1,2, Figure A1) and the hCXCR4 sequence with the fused c-Myc tag (pairs 1-3, Figure 2e). The hCXCR4 mRNA was slightly detectable in Exo CTRL and dramatically increased in Exo CXCR4 . Only Exo CXCR4 contained the exogenous sequence as evidenced by the presence of the c-Myc tag amplicon (Figure 2e). By WB analysis, Exo CTRL and Exo CXCR4 expressed similar levels of the Exo specific marker TSG-101 (Figure 2b), and tetraspanins (CD63, CD9, CD81) by FC analysis (Figure 2f). The expression of functional pregnancy-associated plasma protein-A (PAPP-A), previously showed to be functional into Exo-CPC [6], was similar in Exo CTRL and Exo CXCR4 ( Figure A1d). Moreover, the overexpression of CXCR4 does not impair size/concentration distributions of Exo, as assessed by Nanosight analysis (Figure 2g). analysis, Exo CTRL and Exo CXCR4 expressed similar levels of the Exo specific marker TSG-101 ( Figure  2b), and tetraspanins (CD63, CD9, CD81) by FC analysis (Figure 2f). The expression of functional pregnancy-associated plasma protein-A (PAPP-A), previously showed to be functional into Exo-CPC [6], was similar in Exo CTRL and Exo CXCR4 ( Figure A1d). Moreover, the overexpression of CXCR4 does not impair size/concentration distributions of Exo, as assessed by Nanosight analysis (Figure 2g).  Since our experimental hypothesis was based on the important role of the CXCR4-SDF-1α axis in cardiac repair, we empirically assessed the SDF-1α levels into myocardial tissue at different time-points following myocardial I/R injury ( Figure A2). The chemokine was picked at three h after reperfusion; this time-point was chosen for the in vivo tail vein injection of Exo (Figure 3a). Exo CXCR4 , but not Exo CTRL , improved left ventricular ejection fraction (LVEF) at 7 days (Figure 3b). Both types of Exo significantly improved LVEF at 4 weeks, as compared to PBS-injected control rats (75.0 ± 1.4% and 63.5 ± 2.5% for Exo CXCR4 and Exo CTRL , respectively, versus 53.7 ± 1.1% in the PBS group; p < 0.01 and p < 0.05 respectively). Exo CXCR4 was significantly more effective than Exo CTRL in this regard (p < 0.01). Parallel changes in end-systolic and end-diastolic LV volumes were observed (Figure 3c,d). Hemodynamic measurements showed that Exo CXCR4 significantly improved -dp/dt min. Although not statistically significant, a trend toward the dp/dt max improvement was also observed in Exo CXCR4 group. Exo CTRL improved -dp/dt min only slightly, as compared to the PBS group ( Figure 3e). From LV pressure/volume loops, the time constant of isovolumic LV pressure fall (Tau) properties was significantly reduced in hearts injected with Exo CXCR4 . The whole set of hemodynamic data indicated a global improved in diastolic relaxation ( Figure 3f) for hearts injected with Exo CXCR4 . Finally, hearts in the Exo CXCR4 group displayed smaller scars than those in the Exo CTRL or PBS groups ( Figure 4a). There also was a significant reduction of scar size in the Exo CTRL group vs. PBS. Both Exo CTRL and Exo CXCR4 increased blood vessel density ( Figure 4b) at a similar level.
6 Figure 3. In vivo cardio-protection after I/R. (a) Scheme depicting the study protocol. I/R model was induced by ligating the left descending anterior coronary that was then released after 30 min. Exo (2 × 10 11 total particles from a pool of n = 6 patients, in 300 μL PBS) were injected tail vein 3 h after reperfusion. (b) Exo CXCR4 but not Exo CTRL significantly improved LVEF at 7days after I/R, compared to PBS control group. At 28 days after I/R, Exo CXCR4 , and to a lesser extent Exo CTRL , significantly improved LVEF compared to control group (n = 6 rats/group), # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons; test was performed at each time-point). (c,d) Exo CTRL reduced both left ventricle (LV) end-systolic volume (LVESV) and left ventricle end-diastolic volume (LVEDV) was significantly less than modified Exo CXCR4 (n = 6 rats/group), # p < 0.05 (one-way analyses of variance- In vivo cardio-protection after I/R. (a) Scheme depicting the study protocol. I/R model was induced by ligating the left descending anterior coronary that was then released after 30 min. Exo (2 × 10 11 total particles from a pool of n = 6 patients, in 300 µL PBS) were injected tail vein 3 h after reperfusion. (b) Exo CXCR4 but not Exo CTRL significantly improved LVEF at 7days after I/R, compared to PBS control group. At 28 days after I/R, Exo CXCR4 , and to a lesser extent Exo CTRL , significantly improved LVEF compared to control group (n = 6 rats/group), # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons; test was performed at each time-point). (c,d) Exo CTRL reduced both left ventricle (LV) end-systolic volume (LVESV) and left ventricle end-diastolic volume (LVEDV) was significantly less than modified Exo CXCR4 (n = 6 rats/group), # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons; test was performed at each time-point). (e) Hemodynamic measurements showing a decreased ventricular compliance in the Exo CXCR4 by quantitative analysis of -dp/dt min and dp/dt max (n = 6 rats/group for Exo CTRL and Exo CXCR4 ; n = 5 rats/group PBS) # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons; test was performed at each time-point). (f) From LV pressure/volume loops, the time constant of isovolumic LV pressure fall (Tau) value, which reflects diastolic relaxation properties, was calculated. LV relaxation was improved and time constant reduced in the Exo CXCR4 (n = 6 rats/group) group compared with the Exo CTRL (n = 6 rats /group) and PBS (n = 5 rats/group) # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tuke-Kramer multiple comparisons test for paired parametric comparisons). 7 ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons; test was performed at each time-point). (e) Hemodynamic measurements showing a decreased ventricular compliance in the Exo CXCR4 by quantitative analysis of -dp/dt min and dp/dt max (n = 6 rats/group for Exo CTRL and Exo CXCR4 ; n = 5 rats/group PBS) # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons; test was performed at each time-point). (f) From LV pressure/volume loops, the time constant of isovolumic LV pressure fall (Tau) value, which reflects diastolic relaxation properties, was calculated. LV relaxation was improved and time constant reduced in the Exo CXCR4 (n = 6 rats/group) group compared with the Exo CTRL (n = 6 rats /group) and PBS (n = 5 rats/group) # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tuke-Kramer multiple comparisons test for paired parametric comparisons;). Exo reduce scar size at 28 days after I/R. The reduction in scar size is more pronounced in Exo CXCR4 compared to Exo CTRL treated group (n = 5 rats/group PBS, n = 6 rats/group Exo CTRL and Exo CXCR4 ). (b) Exo CTRL and Exo CXCR4 . increased blood vessel density. SMA, smooth muscle actin (green); TnI, Troponin I (red); nuclear counterstaining (blue); n = 4 rats/group PBS, n = 6 rats/group Exo CTRL and Exo CXCR4 ). # p < 0.05 (oneway analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons).

Exo CXCR4 Is Delivered More Efficiently than Exo CTRL to CM in Ischemic/Reperfused Rat Hearts Ex Vivo
To assess whether the superior in vivo cardioprotective effect of Exo CXCR4 is mediated by increased cardiac homing as a result of SDF-1α upregulation in the injured heart, rat hearts were subjected to in vivo I/R injury. Three hours later, hearts were harvested and perfused retrogradely in a Langendorff-mode ( Figure 5a). Equal amounts of Exo CXCR4celmiR39 or Exo CTRLcelmiR39 (see section 4) were added to the perfusates in the two groups, and hearts were continuously perfused for 2 h. CM were then isolated from hearts perfused ex vivo, and cel-miR39 levels were determined in the isolated CM as a measure of Exo delivery. Exo CXCR4celmiR39 perfused hearts showed increased cel-miR39 levels Exo reduce scar size at 28 days after I/R. The reduction in scar size is more pronounced in Exo CXCR4 compared to Exo CTRL treated group (n = 5 rats/group PBS, n = 6 rats/group Exo CTRL and Exo CXCR4 ). (b) Exo CTRL and Exo CXCR4 . increased blood vessel density. SMA, smooth muscle actin (green); TnI, Troponin I (red); nuclear counterstaining (blue); n = 4 rats/group PBS, n = 6 rats/group Exo CTRL and Exo CXCR4 ). # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons).

Exo CXCR4 Is Delivered More Efficiently than Exo CTRL to CM in Ischemic/Reperfused Rat Hearts Ex Vivo
To assess whether the superior in vivo cardioprotective effect of Exo CXCR4 is mediated by increased cardiac homing as a result of SDF-1α upregulation in the injured heart, rat hearts were subjected to in vivo I/R injury. Three hours later, hearts were harvested and perfused retrogradely in a Langendorff-mode ( Figure 5a). Equal amounts of Exo CXCR4celmiR39 or Exo CTRLcelmiR39 (see Section 4) were added to the perfusates in the two groups, and hearts were continuously perfused for 2 h. CM were then isolated from hearts perfused ex vivo, and cel-miR39 levels were determined in the isolated CM as a measure of Exo delivery. Exo CXCR4celmiR39 perfused hearts showed increased cel-miR39 levels compared with both Exo CTRLcelmiR39 and PBS-perfused control hearts (fold-increase vs. PBS: 162.6 ± 42.6 and 7.7 ± 1.7 for Exo CXCR4celmiR39 and Exo CTRLcelmiR39 respectively; Figure 5b). In addition, hCXCR4/c-Myc mRNA was detectable in adult CM from hearts perfused with Exo CXCR4celmiR39 , but not in those receiving Exo CTRLcelmiR39 (Figure 5c). These results indicate that following in vivo I/R injury Exo CXCR4 is delivered to CM in injured hearts more efficiently than Exo CTRL .  Figure 5b). In addition, hCXCR4/c-Myc mRNA was detectable in adult CM from hearts perfused with Exo CXCR4celmiR39 , but not in those receiving Exo CTRLcelmiR39 (Figure 5c). These results indicate that following in vivo I/R injury Exo CXCR4 is delivered to CM in injured hearts more efficiently than Exo CTRL .

Exo CXCR4 Shows Superior Ex Vivo Cardioprotective Effects Compared with Exo CTRL
Having previously showed that Exo act mostly through MAP kinases pathway [6], we evaluated whether ERK1/2 phosphorylation was improved in adult CM isolated from ex vivo experiments. CM isolated from infarcted and perfused hearts were lysed for WB analysis. Exo CXCR4celmiR39 was able to significantly induce ERK1/2 phosphorylation after 2 h of perfusion as compared to PBS perfused hearts (Figure 5b), whereas Exo CTRLcelmiR39 was less effective in this regard.
To confirm cytoprotective role of Exo a subset of isolated adult CM were plated on glass coverslip and soon after fixed for TUNEL staining. Both Exo CTRLcelmiR39 and Exo CXCR4celmiR39 increased the number of surviving CM after 2 h of perfusion as compared to PBS group. The number of adult CM undergoing apoptosis was lower in Exo CXCR4 than Exo CTRL , although not statistically significant ( Figure 3c).

Exo CXCR4 Shows Superior In Vitro Cardioprotective Effects Compared with Exo CTRL
Exo CTRL is cytoprotective in neonatal HL-1 [6,7]; here we compared its cytoprotective activity with that of Exo CXCR4 using the in vitro model of staurosporine-induced death in HL-1 cardiomyocytes. Exo CXCR4 was more effective than Exo CTRL in preventing staurosporine-induced death in HL-1 cells (Exo CXCR4 94.7 ± 2.6%; Exo CTRL 80.9 ± 1.8%) (Figure 6a,b). Similar results were obtained in human iPSC-derived CM, as Exo CXCR4 was more effective than Exo CTRL at preventing staurosporine-induced death in these cells (Exo CXCR4 73.0 ± 2.7%; Exo CTRL 58.6 ± 1.2%; Figure A3). Furthermore, the effect of Exo CXCR4 in HL-1 cells was not antagonized by ADM3100, indicating that is was independent from SDF-1α ( Figure 6c). Similar intracellular levels of celmiR39 was observed in normal or stress conditions when Exo CXCR4celmiR39 and Exo CTRLcelmiR39 were added to the media of HL-1 cells; indicating that the presence of CXCR4 on Exo surface did not increase in vitro uptake (Figure 6d). This finding is probably related to the absence of an SDF-1α gradient in this in vitro model, as opposed to the in vivo situation, as shown by WB in vitro analysis ( Figure A2b). Both Exo CTRL and Exo CXCR4 induced ERK1/2 phosphorylation after staurosporine-induced damage. Compared with Exo CTRL , Exo CXCR4 treated HL-1 showed higher levels of phosphorylated ERK1/2 in both staurosporine-induced ( Figure 7a) and normal ( Figure A4c) conditions. The effect was prevented by adding the kinases chemical inhibitor U0126 (Figure 7b). In parallel, U0126 was able to abolish Exo CXCR4 protective effect on HL-1 CM (Figure 7d). This finding is in line with our previous data [6], showing that the ERK pathway is involved in the anti-apoptotic effect mediated by Exo.
10 Figure 6. In vitro bioactivities of Exo CXCR4 . (a,b) Exo CTRL , and Exo CXCR4 inhibit staurosporine-induced death in HL-1 CM (green) (n = 6 patients/group for Exo CTRL and Exo CXCR4 ; n = 5 repeated measurements for PBS and FBS) # p < 0.01 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparison tests for paired parametric comparisons). Exo CXCR4 exhibit superior pro-survival effect. (c) Exo CXCR4 superior pro-survival effect in HL-1 CM is not affected by the addition of AMD3100 as specific CXCR4 antagonist. (d) In vitro Exo uptake. Realtime PCR analyses of cel-miR39 (mature sequence) levels in HL1 CM incubated with Exo CTRLcel-miR39 and Exo CXCR4cel-miR39 during normal or stress conditions. Data are 2 −ΔΔCt values vs. PBS (n = 5 patients/group).  and Exo CXCR4 were added to the medium in stress-conditions (n = 4 patients/group) # p < 0.01 (Kruskal-Wallis test nonparametric multiple comparisons with Dunn's multiple comparison test); * p < 0.05 (two-tailed, paired Student t-test). (b) ERK1/2 phosphorylation is prevented by adding the kinases chemical inhibitor U0126. (n = 5 patients/group for Exo CTRL and Exo CXCR4 ; n = 5 repeated measurements for PBS) # p < 0.01 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons). (c) Human recombinant SDF-1α is able to stimulate CXCR4 receptor in mouse HL-1 CM only after these cells have uptaken the human receptor from Exo (n = 5 patients/group); * p < 0.05 (two-tailed, paired Student t-test). (d) Exo CXCR4 are significantly less effective in preventing staurosporine cell death when HL1 CM are treated with the kinases chemical inhibitor U0126, (n = 5 patients/group for Exo CXCR4 and and Exo CXCR4 were added to the medium in stress-conditions (n = 4 patients/group) # p < 0.01 (Kruskal-Wallis test nonparametric multiple comparisons with Dunn's multiple comparison test); * p < 0.05 (two-tailed, paired Student t-test). (b) ERK1/2 phosphorylation is prevented by adding the kinases chemical inhibitor U0126. (n = 5 patients/group for Exo CTRL and Exo CXCR4 ; n = 5 repeated measurements for PBS) # p < 0.01 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons).
(c) Human recombinant SDF-1α is able to stimulate CXCR4 receptor in mouse HL-1 CM only after these cells have uptaken the human receptor from Exo (n = 5 patients/group); * p < 0.05 (two-tailed, paired Student t-test). (d) Exo CXCR4 are significantly less effective in preventing staurosporine cell death when HL1 CM are treated with the kinases chemical inhibitor U0126, (n = 5 patients/group for Exo CXCR4 and Exo CXCR4 +U0126; n = 4 repeated measurements for PBS ) # p < 0.05 (one-way analyses of variance-ANOVA with post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test for paired parametric comparisons). Cell vitality was not affected by U0126 alone (n = 4 repeated measurements).

Exo CXCR4 Transfers CXCR4 mRNA and Protein into CM In Vitro
We hypothesized that the superior cytoprotective in vitro effect of Exo CXCR4 was due to the transfer of bioactive CXCR4 to target cells that may enhance the previously described cytoprotective effect of Exo (which also includes PAPP-A mediated protease activity) [6]. We tested the presence of exogenous hCXCR4 in murine HL-1 CM at mRNA and protein levels. The hCXCR4-mRNA was detected in HL-1 CM exposed to Exo CXCR4 but not in those exposed to Exo CTRL . Moreover, the c-Myc nucleic acid sequence within the hCXCR4-c-Myc construct was detected by RT-PCR in cells exposed to Exo CXCR4 but not in those exposed to Exo CTRL ( Figure A4a). This observation was made both under basal conditions and after staurosporine-induced cell stress ( Figure A4a). The c-Myc tag was detectable at the protein level in CM treated with Exo CXCR4 , demonstrating the translation of the hCXCR4-c-Myc fusion construct, but not in those receiving Exo CTRL ( Figure A4b). These results demonstrate the ability of Exo CXCR4 to deliver the mRNA encoding for the exogenous hCXCR4, which may result in the translation of hCXCR4 protein in mouse CM. Because hCXCR4 bioactivity requires the presence of SDF-1α [18], we used a species-specific human recombinant SDF-1α protein (hDF-1α) that does not activate the downstream ERK signal in mouse HL-1 cells in order to assess hCXCR4 bioactivity in these cells (Figure 7c). HL-1 cells pre-treated with Exo CXCR4 , but not pre-treated with Exo CTRL , exhibited increased ERK1/2 phosphorylation after the subsequent exposure to hDF-1α (Figure 7c). These results indicate that Exo CXCR4 delivers bioactive hCXCR4 to target cells.

Discussion
Exo have emerged as a therapeutic tool with the potential to replace cellular approaches in the treatment of cardiovascular disease [6,7,27,28]. Compared with cell therapy, Exo offer several potential advantages, as they can be stored at −80 • C without a substantial loss of functional activity [6,7,29], they are effective after systemic delivery [30], with a dosage that is not limited by microvascular plugging [31], and they are poorly immunogenic [32]. With respect to therapeutic applications, safety and efficacy of Exo depend on their bio-distribution that influence selective and effective uptake, respectively, by target cells after systemic delivery [33]. In fact, a given Exo is taken up with different efficiencies by different recipient cell types [34]. Systemically delivered Exo is largely up taken by the liver, and to lesser extents by spleen and lungs [35][36][37], although in vivo Exo bio-distribution also depends on the producing cell type [12]. In animal models of ischemic heart disease, IM Exo administration was beneficial in several studies [6][7][8], whereas IC Exo administration was tested in a single study [15]. The latter found that IM delivered Exo from CPCs, but not IC delivered Exo from the same producing cells were beneficial in a pig MI model, mainly due to comparatively lower Exo retention in the heart using the IC route of administration. From a translational perspective, IM Exo injection would hardly be feasible for treating acute MI patients, thus, improving Exo retention by the heart after IC or systemic administration is a crucial goal. Recently studies have attempted to engineer transmembrane proteins of Exo using "cardiac-homing peptide" to increase targeting capability to cardiac tissues [38][39][40]. These study showed that homing peptides displayed on the external part of Exo, although not directly implicated into mechanism of action of Exo, promote the specificity and efficiency of systemic delivery of Exo.
Our experimental approach leverages on the important role of the CXCR4-SDF-1α axis in cardiac repair. SDF-1α is markedly upregulated within 1 h of myocardial ischemia, resulting in a chemoattractant gradient for CXCR4 expressing leukocytes and stem cells [41]. Within the heart, SDF-1α expression was demonstrated for both CM and fibroblasts a few hours after acute MI [41]. In light of the above, we hypothesized that beside the physiological ischemia-induced cardiac SDF-1α expression, overexpressing CXCR4 in Exo would result in an enhanced homing of Exo to the injured myocardium, and thus, increased cardioprotective effects. To test this hypothesis, we genetically modified CPCs to overexpress CXCR4 and then isolated the secreted Exo. CXCR4 was detectable on Exo CXCR4 but not on Exo CTRL released from naive CPCs. Exo homing to, and uptake by, CM was studied ex vivo in isolated-perfused hearts after in vivo I/R injury. Exo CXCR4 were markedly more effective than Exo CTRL in this regard. It should be emphasized that Exo uptake was specifically studied in CM isolated from ischemic rat hearts, as opposed to the total heart cell population. We did not evaluate and discriminate Exo uptake by other cell types within the ischemic heart, such as fibroblast and endothelial cells. This represents a limitation of the study, however, we have evidence showing that the benefit of Exo CXCR4 and its superior effect as compared to Exo CTRL results in an increased number of surviving cardiomyocytes (Figures 4a and 5c) rather than improved angiogenesis (Figure 4b), for which we did not find any significant difference between the groups. Exo uptake measured here includes both homing to I/R heart tissue and CM binding and uptake. These different components of the global "uptake" process cannot be discriminated. For instance, enhanced Exo homing to injured myocardium would increase Exo bioavailability in the extracellular compartment, and therefore on the CM surfaces, and result in an increase in total Exo uptake by CM, even in the case when the intrinsic Exo uptake efficiency (i.e., the intrinsic efficiency of the process at a single particle level) was unaffected. In vivo, we similarly found superior functional benefits in terms of scar size as well as LV systolic and diastolic function for Exo CXCR4 compared with Exo CTRL . Efficient Exo CXCR4 homing to I/R myocardium, as demonstrated in ex vivo isolated-perfused hearts, most likely played an important role in vivo, too.
Although the increased concentration of SDF-1α cytokine within the heart tissue may enhance bio-availability of Exo in close proximity of CM surface, it does not have a direct effect in mediating the uptake in vitro, where the experimental conditions failed to recapitulate in vivo cytokines gradient (Figure 6a). Indeed, the use of a specific CXCR4 antagonist (AMD3100) has no effect on increasing Exo uptake, neither inhibits protection, in vitro. On the other hand, the transfer of active and SDF-1α-responsive receptor from Exo to target cells, as shown in the Figure 7c, plays a crucial role leading to activation of protective signaling pathways, and in addition is of the already described mechanisms via PAPP-A-mediated IGF-1 release [6]. As expected, the degree of Exo-mediated improvement of cardiac function, previously shown by our group upon IM injection into infarcted hearts [6], could not be matched by the tail-vein injection, likely due to larger in vivo biodistribution. Indeed, at seven days after tail-vein injection of Exo CTRL we did not achieve statistically significant improvement compared to saline group (Figure 7), as was the case by IM injection in previous studies [6,7]. By enhancing homing of Exo to CM via CXCR4 expression on the vesicle surface, we achieved comparable functional improvements using IM and systemic routes of administrations ( Figure 3). Previous studies have shown that mesenchymal stem cells (MSCs) overexpressing surface CXCR4 improve migration and engraftment in response to SDF-1α gradients in various disease models [42][43][44]. Taking a methodological approach similar to the one used here, Kang et al. [25] recently showed that Exo secreted from CXCR4 overexpressing MSCs mediated antiapoptotic and proangiogenic effects in vitro, while also indirectly enhancing MSC-mediated CM survival in vivo.
Here, we investigated for the first time the direct effects of systemic Exo CXCR4 injection and established Exo homing/uptake as a crucial mechanism by which CXCR4 increases Exo-CPC delivery to CM. Exo CXCR4 not only transfer their exogenous protein to the target cells, they are also able to transfer the mRNA, and this confirmed data obtained from different groups [45]. The relative contribution of bioactive CXCR4 versus the transduction of its mRNA by Exo receiving cells has not been detailed in the present study. From the functional point of view, Exo CXCR4 have superior in vitro and in vivo anti-apoptotic effects in CM compared to Exo CTRL . These results have major translational implications for treatment of acute MI patients, owing to the major advantages associated with systemic drug administration [25], which can be performed upon recanalization of the culprit artery.

Cell Culture
Right cardiac atrial appendage specimens were obtained from patients (n = 13) who underwent surgical repair of heart valve disease and gave written informed consent. Protocols used in this study were approved by local Ethical Committee for Clinical Research (Comitato Etico Cantonale, Bellinzona, Switzerland; Rif. CE 2923), and study was performed according to the Declaration of Helsinki. CPCs were derived from atrial specimens, as described [6]. Briefly, atrial specimens were cultured as primary ex vivo tissue explants in basic Iscove's modified Dulbecco's medium (IMDM; Lonza, Basel, Switzerland) supplemented with 20% fetal bovine serum (FBS) and 1% v/v penicillin-streptomycin (Life Technologies, Europe BV, Zug, Switzerland). The cellular outgrowth of atrial explants was collected within 14 days and seeded into fibronectin-coated 25 mL flasks, forming a monolayer of CPCs that were passaged twice by trypsinization before Exo purification from conditioned media. Mouse HL-1 cardiomyocyte cell line (LSU Health Sciences Center, New Orleans, LA, USA) was grown in Claycomb Medium (Sigma, Merck KGaA, Darmstadt, Germany) supplemented with 100 µM norepinephrin, 10% FBS, 1% v/v penicillin/streptomycin, and 4 mM L-glutamine [46]. AMD3100 (Millipore, Merck KGaA) at the concentration of 10 µM was used as specific CXCR4 antagonist. For experiments performed to establish the bioactivity of human CXCR4 into target cells (Figure 7c), confluent HL-1 were treated with human recombinant SDF-1α chemokine (hDF-1α, Sigma) at the concentration of 100 ng/mL for 40 min. The kinases chemical inhibitor U0126 monoethanolate (Sigma) at the concentration of 20 µM was used to inhibit ERK1/2 phosphorylation; iPS-derived CM were obtained as previously described [45]. Briefly CPC were reprogrammed through Sendai virus infection using OCT3/4, SOX2, KLF4, and MYC (CytoTune™-iPS 2.0 Sendai Reprogramming Kit; Thermo Fisher Scientific, Europe BV, Zug, Switzerland), according to manufacturer's instructions. Cardiac differentiation of iPSCs was induced using the GibcoTM PSC Cardiomyocyte Differentiation Kit (Thermo Fisher Scientific). Differentiated cells were maintained in RPMI (Gibco, Thermo Fisher Scientific) supplemented with B27 (Gibco) at 37 • C and 5% of CO2 and changed every two days.

CPC Engineering to Overexpress CXCR4
The human sequence of CXCR4 (NM_001008540) was cloned into PCMV3-C-Myc expression plasmid (Sino Biological Inc., Wayne, PA, USA) to encode the hCXCR4 cDNA fused with c-Myc tag under the transcriptional control of a CMV promoter (PCMV3-Cx-Myc). This vector was used to transfect CPCs (CPC CXCR4 ). Control CPC (CPC CTRL ) were transfected with an empty backbone plasmid. Briefly, CPCs were gently collected by trypsinization and re-suspended in normal growth medium at the density of 1 × 10 5 cells/mL. The transfection agent Lipofectamine 3000 was diluted in Opti-MEM I medium (both from Thermo Fisher Scientific) and left for 10 min at RT. PCMV3-Cx-Myc plasmid or empty backbone were diluted in Opti-MEM I medium and P3000 reagent (2 µL/µg DNA). After incubation, the mix with Lipofectamine 3000 Reagent was added to the mix, containing the diluted plasmid and incubated for an additional 10 min at RT. The mixture was then gently rocked and incubated at 37 • C, 5% CO2. Two days after transfection, cells were washed several times with PBS and the growth medium was exchanged with Exo-producing medium (DMEM high-glucose, Thermo Fisher Scientific).

Particle Purification
Media conditioned by CPC CTRL and CPC CXCR4 were obtained by culturing cells in Exo-producing medium in absence of serum for 7 days [6]. For particle purification, conditioned medium (10 mL) was centrifuged at 3000 g for 15 min, filtered through a 0.2 µm membrane (BD Biosciences, Allschwil, Switzerland), centrifuged at 10000 g for 15 min, and then ultracentrifuged at 100000 g for 4 h. Exo pellets were re-suspended in 100 µL PBS, pH 7.4, and stored at −80 • C. Particle number and size were measured using NanoSight technology (Malvern Instruments, Malvern, UK).

In Vitro and Ex Vivo Exo Uptake by CM
To assess in vitro and ex vivo cellular uptake of Exo by CM, both CPC CXCR4 and CPC CTRL were transfected with a pre-miRNA precursor for cel-miR-39 (Thermo Fisher Scientific), a C. elegans miRNA not present in mammalian cells. Intracellular cel-miR39 levels can be used as a measure of cellular uptake of cel-miR39 containing exosomes. Exo released from these cells are referred hereby as Exo CXCR4celmiR39 (from CPC CXCR4 ) and Exo CTRLcelmiR39 (from CPC CTRL ). In vitro HL-1 were incubated with Exo CXCR4celmiR39 or Exo CTRLcelmiR39 . CTRLcelmiR39 in normal conditions or after treatment with staurosporine. Ex vivo isolated rat hearts were perfused in a Langendorff-mode using perfusates supplemented with Exo CXCR4celmiR39 or Exo CTRLcelmiR39 , and cel-miR-39 levels were measured in HL-1 and primary adult rat CM using real-time RT-PCR. The level of expression cel-miR39 in the CPC or Exo was consistent with dose-response efficiencies of transfection [6].

Levels of Exogenous mRNA into CM
PCR was performed using 2 pairs primers that specifically amplified the hCXCR4 sequence only or the hCXCR4 sequence fused with c-Myc tag (1,2 for human CXCR4-mRNA; 1-3 c-Myc-mRNA; see scheme in Figure A1). The murine sequence of CXCR4 receptor was not amplified by primers ( Figure A4), thus allowing the identification of mRNA carried by human derived Exo (hCXCR4-mRNA), or overexpressed CXCR4 (c-Myc-mRNA).

In Vitro CM Viability Assay
To assess HL-1 CM viability, cells were seeded at a 6.25 × 10 4 cells/cm 2 density in 96-well plates. After 24 h, cell death was induced by 1µM staurosporine (Sigma). Exo CTRL or Exo CXCR4 (isolated from conditioned media of cells from n = 6 patients) were added on top of staurosporine at the 3 × 10 6 particles/cm 2 concentration, as assessed by Nanosight [6]. In some of the tested conditions, the role of SDF-1α was tested by adding AMD3100 (10 µM) as CXCR4 specific antagonist. Twelve hours later, cells were stained with Cellstain™ Double Stain Kit (Dojindo EU, München, Germany) for 30 min at 37 • C and counted under a fluorescence microscope.
Functional data were validated in human iPS-derived CM. Cell death was induced in iPS-CM by 1 µM staurosporine (Sigma). Exo CTRL or Exo CXCR4 (isolated from conditioned media from n = 5 patients) were added on top of staurosporine at the 3 × 10 6 particles/cm 2 concentration, as assessed by Nanosight. Twelve hours later, cells were stained with Cellstain™ Double Stain Kit (Dojindo) for 30 min at 37 • C and counted under a fluorescence microscope.

RNA Extraction, Reverse Transcription and Real-Time PCR
Cells or exosomes were lysed with TRI Reagent (Sigma), and chloroform was added following company instructions. The pellet was air-dried, re-suspended in DEPC water, and quantified with NanoDrop™ 2000c (Thermo Fisher Scientific). For real-time amplification of cel-miR-39 sequence, 5ng of specific RNA was reverse-transcribed using the Taqman miRNA reverse stem-loop primers (Applied Biosystems, München, Germany) with a T100 Termal Cycler (Bio-Rad). Real-time PCR for specific cell-miR-39 sequence was performed by adding to 1.33 µL cDNA: 1 µL 20× specific primer (Applied Biosystems), 10 µL TaqMan Universal PCR Master Mix, and 7.67 µL DEPC water. Amplification and detection of specific products were performed in triplicate using a CFX Connect™ Real-Time PCR Detection System (Bio-Rad). Negative controls were included in each RT-PCR assay. The threshold cycle (Ct) of each gene was automatically defined and normalized to the control miR16 (∆Ct value) for CPCs, HL-1 CM and exosomes, and to the control U87 or snoRNU202 for rats primary CM. Data are shown as 2 −∆∆Ct values.

Immunofluorescence and Histology
Cells were fixed with paraformaldehyde 4%, sucrose 2%, permeabilized with 0.5% Triton for 30 min and blocked with 5% serum for 30 min at 25 • C. Samples were then incubated with primary antibodies in 5% serum direct against CXCR4 (Abcam 1/100), c-Myc tag (Abcam 1/200) overnight at 4 • C. After 3× wash with PBS, each 10 min, Alexa Fluor secondary antibodies (Life Technologies, 1/1000) were used for detection. Images were acquired by a confocal microscopy system C2 Plus (Nikon). Harvested hearts were cryo-preserved in optimal cutting temperature (OCT) medium at −80 • C. For morphometric analysis, hearts were cut in 8-µm sections, and stained with Massontrichrome (HT15 Trichrome Stain Kit; Sigma) for measurement of scar size. On each heart section, scar area was determined by tracing the infarct borders manually using ImageJ software (NIH Image, Bethesda, MD, USA). Six sections distributed at fixed steps from the cardiac basis to the apex were analyzed for each heart and values were averaged. Anti-SMA (1:100) (Santa Cruz, CA, USA), smooth muscle actin, was used to stain vessels and TnI (1:100) (Santa Cruz, Santa Cruz, USA), Troponin I, as marker for CM, on each heart section. The number of vessels and surface area were quantified using ImageJ software. Six sections for each rat were used for the analysis and values were averaged.

SDF-1α ELISA
To assess the amount of SDF-1α protein on heart tissues of infarcted rats compared to control, ELISA Kit for rat SDF1/CXCL12 was used (LS Bio, Seattle, WA, USA). Briefly, samples were directly immobilized on to the wells of the microtiter plate. The detection (primary) antibody was added to the wells for binding to its specific antigen for 1 h at 37 • C. After 3 washes, the Detection (secondary) Ab was added and plate incubated for 30 min at 37 • C. A colorimetric substrate (Extra-sensitive TMB) was used for the assay read-out. The accumulation of the coloured product is proportional to the specific antigen present in each well. The results were quantitated with a microtiter plate reader at 450 nm absorbance.

TUNEL Assay
Adult CM isolated from ex vivo experiments were plated on glass coverslips and fixed with paraformaldehyde 4%, permeabilized with 0.2% Triton for 5 min, and equilibrate at room temperature for 5-10 min with Equilibration Buffer. Samples were then incubated with a solution of Nucleotide Mix and rTdT Enzyme in a humidified chamber for 1 h at 37 • C and immersed in 2× SSC for 15 min at room temperature to terminate the reactions. After three washes to remove unincorporated fluorescein-12-dUTP, cells were incubated with a primary antibody against alpha sarcomeric actinin (1:100) and subsequently with an anti-mouse secondary antibody 568 (1:1000), both from Abcam and with DAPI, and immediately analyzed under a fluorescence microscope. A positive control slide using DNase I was included in each experiment.

Animal Experiments
Experimental protocols were approved by the Animal Care Committee of Canton Ticino, Switzerland (TI-08-18). All procedures conformed to the Directive 2010/63/EU of the European Parliament. A model of myocardial ischemia with reperfusion was used. Ischemia Reperfusion (I/R) was induced in healthy male Wistar rats (250-300 g body weight) anesthetized with a cocktail of Ketamine (Ketasol 100, 100 mg/kg) and Xylazine (Rompun 2%, 80 mg/kg), intubated and ventilated. The left anterior descending artery was ligated near its origin with a 6-0 prolene suture. The coronary ligature was released after 30 min, and exosomes were injected in the tail-vein 3 h after reperfusion with a total of 300 µL of PBS containing 2 × 10 11 particles of the following vesicle populations: pool of Exo-CPC CTRL (n = 6) and pool of Exo-CPC CXCR4 (n = 6) or PBS. The chest was then closed, pneumothorax was reduced, and the rats were treated with Meloxicam during postsurgical recovery. Sedated rats underwent transthoracic echocardiography at 7 and 28 days post-MI using the Vevo2100 echocardiography system (VisualSonic System 2100, FUJIFILM VisualSonics, Toronto, Canada) equipped with a 15-MHz linear transducer. Under ECG monitoring of heart rate, 2D images of the hearts were acquired in long-axis views at the level of the greatest LV diameter. LV ejection fraction (LVEF) was measured using Simpson's analysis.
Hemodynamic analysis was performed at day 28 using a Millar pressure-volume conductance catheter. Rats were anesthetized with an intraperitoneal injection of ketamine (Ketasol 100, 100 mg/kg) and xylazine (Rompun 2%, 80 mg/kg). The trachea was cannulated, and the animal connected to a positive-pressure volume-controlled rodent ventilator. The catheter was introduced through the right carotid artery into the ascending aorta and then into the LV cavity. After hemodynamics measurements, the animal was sacrificed.

Statistical Analyses
Results are presented as means ± SE of n independent experiments or patients from which Exo producing cells were derived. Each single "n" is specified in the figure legends. Statistical analyses were performed using InStat, version 3.0 (GraphPad Software, Inc., San Diego, CA, USA). The hypothesis that data came from a normally distributed population was assessed using the Kolmogorov Smirnov normality test. The assumption that data were sampled from populations with identical SDs was tested using the method of Bartlett. The Student t-test was used to compare two normally distributed populations. One-way analyses of variance (ANOVA) with subsequent post-hoc multiple comparisons using the Tukey-Kramer multiple comparisons test were used for parametric comparisons. The Kruskal-Wallis test was used for nonparametric multiple comparisons. To perform post-hoc pairwise comparisons following Kruskal-Wallis test, Dunn's non-parametric multiple comparisons procedure was used. Differences with probability values p < 0.05 were considered statistically significant.