Photoreceptor protection by mesenchymal stem cell transplantation identifies exosomal MiR-21 as a therapeutic for retinal degeneration

Photoreceptor apoptosis is recognized as one key pathogenesis of retinal degeneration, the counteraction of which represents a promising approach to safeguard visual function. Recently, mesenchymal stem cell transplantation (MSCT) has demonstrated immense potential to treat ocular disorders, in which extracellular vesicles (EVs), particularly exosomes, have emerged as effective ophthalmological therapeutics. However, whether and how MSCT protects photoreceptors against apoptotic injuries remains largely unknown. Here, we discovered that intravitreal MSCT counteracted photoreceptor apoptosis and alleviated retinal morphological and functional degeneration in a mouse model of photoreceptor loss induced by N-methyl-N-nitrosourea (MNU). Interestingly, effects of MSCT were inhibited after blockade of exosomal generation by GW4869 preconditioning. Furthermore, MSC-derived exosomal transplantation (EXOT) effectively suppressed MNU-provoked photoreceptor injury. Notably, therapeutic efficacy of MSCT and EXOT on MNU-induced retinal degeneration was long-lasting as photoreceptor preservance and retinal maintenance were detected even after 1–2 months post to injection for only once. More importantly, using a natural occurring retinal degeneration model caused by a nonsense mutation of Phosphodiesterase 6b gene (Pde6bmut), we confirmed that MSCT and EXOT prevented photoreceptor loss and protected long-term retinal function. In deciphering therapeutic mechanisms regarding potential exosome-mediated communications, we identified that miR-21 critically maintained photoreceptor viability against MNU injury by targeting programmed cell death 4 (Pdcd4) and was transferred from MSC-derived exosomes in vivo for functional regulation. Moreover, miR-21 deficiency aggravated MNU-driven retinal injury and was restrained by EXOT. Further experiments revealed that miR-21 mediated therapeutic effects of EXOT on MNU-induced photoreceptor apoptosis and retinal dysfunction. These findings uncovered the efficacy and mechanism of MSCT-based photoreceptor protection, indicating exosomal miR-21 as a therapeutic for retinal degeneration.


Introduction
Photoreceptor apoptosis is recognized as one of the major form of photoreceptor death and one important contributor to visual loss in retinal degenerative disorders [1,2], such as inherited retinal dystrophies encompassing the retinitis pigmentosa (RP) [3]. Recently, mesenchymal stem cell transplantation (MSCT) has demonstrated immense potential to ameliorate a variety of ocular dysfunctions [4,5]. Particularly, MSCT retards retinal degeneration with the prevention of loss of retinal ganglion cells (RGCs) and retinal pigment epithelium (RPE) [6][7][8][9]. However, despite evidence indicating that MSCT rescues photoreceptor deficiency in mice with genetic defects [9][10][11], it remains unknown whether MSCT is efficient in directly protecting photoreceptors against specific cell death stimuli in vivo, e.g., light exposure and certain phototoxins [12,13]. Elucidating potential effects and underlying mechanisms of MSCT counteracting photoreceptor apoptosis would therefore provide promise for establishing novel therapeutics of retinal degeneration.
Emerging evidence suggests that paracrine effects of MSCT to create beneficial microenvironments for functional recovery of recipient cells represent the crucial mechanisms underlying MSC-based disease therapeutics [14,15]. Correspondingly, subretinal or intravitreal transplanted MSCs exert profound trophic effects on the retina, which constitute soluble neuroprotective cytokines [8,16] and extracellular vesicles (EVs) carrying retina-regulatory molecules [17,18]. In particular, intravitreal delivery of MSC-secreted exosomes, EVs of endocytic origin with less than 100 nm diameter in size [19], has been reported to promote survival of RGCs after traumatic optic neuropathy [17]. Notably, exosomes are increasingly documented as physiological regulators of retinal homeostasis [20] and important indicators of retinopathies [21], including being involved in photoreceptor degeneration upon insufficient release [22]. Nevertheless, whether exosomal transplantation (EXOT) shows efficacy to ameliorate photoreceptor loss and improve degenerative retinopathies has not yet been evaluated.
It is widely accepted that EVs mediate intercellular communications through exchange of proteins, mRNAs, and mostly the microRNAs, which are~22-nucleotide-long noncoding RNAs that negatively regulate expression of target genes at the posttranscriptional level [23]. Especially, exosomal transfer of microRNAs has been extensively reported to mediate effects of MSCT in treating a range of disorders [24][25][26]. In this regard, studies have identified individual microRNAs that play important roles in regulating ocular and retinal cell function, including those essential for photoreceptor maturation, maintenance, and visual function [27][28][29]. Furthermore, microRNA dysregulation has been revealed to be associated with photoreceptor death in the degenerating retina [30,31]. Importantly, a pioneer study has reported diminished effects of MSCderived exosomes on preventing traumatic loss of RGCs after knockdown of Argonaute-2, a key microRNA effector molecule, suggesting microRNA-dependent mechanisms for retinal protection [17]. However, there is still a lack of data unraveling specific microRNA(s) that regulates photoreceptor apoptosis and supports retinal therapeutics.
In this study, we aimed to investigate whether MSCT protects photoreceptors against apoptosis in counteracting retinal degeneration, and whether exosomal microRNA transfer serves as the therapeutic mechanism.

Intravitreal MSCT counteracts MNU-induced photoreceptor apoptosis and alleviates retinal degeneration
Allogeneic MSCs applied in this study was isolated from mouse bone marrow and was confirmed for their identity and suitability for therapeutic usage according to current criteria ( Fig. S1A-C) [32,33]. To investigate whether MSCT serves as an effective therapeutic for photoreceptor loss in retinal degeneration, the phototoxin N-methyl-Nnitrosourea (MNU) was used to establish a photoreceptorspecific injury model in mice [13], and MSCs were delivered intravitreally (Fig. 1A). Biodistribution analyses confirmed that PKH26-labeled MSCs engrafted in the eye with specific location in the photoreceptor-nuclei-residing outer nuclear layer (ONL) (Fig. 1B, C).
As expected, intraperitoneal injection of MNU-triggered remarkable degeneration of the retinal tissue after 3 d, in which the ONL underwent substantial loss (Fig. 1D, E). Importantly, intravitreal MSCT significantly counteracted detrimental effects of MNU, leading to preserved integrity of retinal tissues with improved ONL thickness at 3 d (Fig. 1D, E). To further examine function of photoreceptors, we performed electroretinography (ERG) analyses, in which scotopic (dark-adapted, mixed rod-and cone-mediated) and photopic (light-adapted, cone-mediated) responses are both recorded for depicting a-wave and/or b-wave, respectively reflecting hyperpolarization of photoreceptors and the subsequent depolarization of bipolar cells postsynaptic to photoreceptors [34,35]. As detected, scotopic ERG analyses demonstrated protection of MSCT on the retinal electrophysiological function against MNU damages, quantified by higher amplitudes of both a-wave and b-wave ( Fig. 1F-H). Also, photopic ERG responses were significantly protected by MSCT (Fig. 1I, J), which was confirmed by S-opsin staining exhibiting ameliorated loss of cone photoreceptors by MSCT against MNU (Fig. 1K). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) analysis further showed that therapeutic effects of MSCT on MNU-provoked retinal damages were based on inhibition of the apoptosis of photoreceptors in the ONL (Fig. 1L, M). As verified by in vitro experiments, MNU-induced apoptosis of the 661W cone photoreceptor cell line and that indirect co-culture of MSCs with 661W cells in the Transwell system counteracted detrimental effects of MNU, which also indicated paracrine mechanisms (Fig. 1N, O). Protective effects of MSCT on MNU-treated retina were continuously prominent at 7 d, when both morphological and electrophysiological alterations were dramatically restrained ( Fig. S2A-G). These findings suggested that intravitreal MSCT effectively counteracts MNU-induced photoreceptor apoptosis and alleviates retinal degeneration.
Exosomes mediates therapeutic effects of MSCT on MNU-provoked retinal photoreceptor damages Next, we investigated whether therapeutic effects of MSCT on MNU-provoked retinal damages were indeed mediated by paracrine mechanisms, particularly exosomes. GW4869, which blocks generation of exosomes [36,37] as verified here (Fig. S1G), was applied for preconditioning of MSCs before transplantation. Notably, GW4869 pretreatment did not affect the biodistribution of injected MSCs in the retina (Fig. S1H), nor influence the viability or the differentiation of MSCs (Fig. S1I-K). Hematoxylin and eosin (H&E) staining images demonstrated that the application of GW4869 efficiently suppressed therapeutic effects of MSCT on MNU-induced retinal damages, resulted in failure of transplanted MSCs to preserve ONL thickness at 3 d ( Fig. 2A, B). GW4869-dampened therapeutic efficacy of MSCT was also significant in both scotopic and photopic ERG analysis, in which MNU-triggered loss of retinal function was not rescued at 3 d ( Fig. 2C-G). Similar phenomena were observed at 7 d after MNU injection, when transplantation of GW4869-pretreated MSCs demonstrated weaker protective effects compared to normal MSCs (Fig. S3).
To direct evaluate the effects of exosomes, we collected exosomes from cultured MSCs by serial centrifuges and ultracentrifuges, and verified their identity as EVs according to morphological characteristics and specific markers ( Fig. S1D-F). Importantly, intravitreal transplantation of exosomes remarkably counteracted detrimental effects of MNU on the retina, leading to maintained ONL thickness and ERG function at 3 d despite MNU injection ( Fig. 2H-N). The protective effects of EXOT against MNU injury were also prominent at 7 d ( Fig. S2H-L) with remarkable preservance of cone photoreceptors and their function in the retina (Fig. S2M-O). Furthermore, to examine whether exosomes indeed regulate survival of photoreceptors, we analyzed TUNEL staining in retinal tissues in vivo and confirmed that GW4869 preconditioning prevented MSCT to alleviate MNU-induced photoreceptor apoptosis, while EXOT mimicked the protective effects of MSCT on photoreceptor viability ( Fig. 2O, P). In vitro cellular apoptosis assay additionally proved that exosomal release determined beneficial effects of co-cultured MSCs on MNU-treated 661W cells (Fig. 2Q, R). Collectively, these findings indicated that exosomes mediates therapeutic effects of MSCT on MNU-provoked retinal photoreceptor damages.
MSCT and EXOT exert long-term effects to protect photoreceptor and retina against degeneration We continued to investigate whether a single intravitreal injection of MSCs and exosomes could have a long-lasting protection of photoreceptors. We found that at 4 weeks in the experiment, MSCT and EXOT were still effective in preserving the ONL thickness and retinal scotopic ERG responses ( Fig. 3A-E), with the protection of cone photoreceptor function in photopic ERG analysis (Fig. 3F, G) and S-opsin + cone photoreceptor survival (Fig. 3H). Even at 8 weeks post to MNU challenge, MSCT and EXOT showed efficacy to protect the retina from degeneration ( Fig. 3I-M), as also confirmed by preserved cone photoreceptors and their function ( Fig. 3N-P). These results suggested longterm therapeutic effects of MSCT and EXOT despite only one-time injection.
We further applied a genetic mouse retinal degeneration model caused by a nonsense mutation of Phosphodiesterase 6b gene (Pde6b mut ), as previously established to mimick RP [38], and treated the mice with MSCs and exosomes at 2week old (Fig. 4A). We discovered that the Pde6b mut mice at 4 weeks of age exhibited fast retinal degeneration, with substantial loss of ONL thickness (Fig. 4B, C), electrophysiological function (Fig. 4D-H), and cone photoreceptors (Fig. 4I). Expectedly, MSCT and EXOT prevented the above degeneration ( Fig. 4B-I). Importantly, MSCT and EXOT significantly alleviated alterations of retinal morphology, loss of electrophysiological function, and diminished cone photoreceptors of Pde6b mut mice at 8 weeks of age ( Fig. 4J-Q). These effects can be attributed to MSCT and EXOT rescue of photoreceptor apoptosis at the early stage of Pde6b mut mice at 3-week old (Fig. 4R). Collectively, these findings indicated long-term therapeutic effects of intravitreal MSCs and exosomes in a natural occurring retinal degeneration model.

miR-21 deficiency aggravates MNU-driven retinal injury and can be restrained by EXOT
In the next experiments, we aimed to decipher how MSCderived exosomes maintained retinal photoreceptor integrity. Using CD63-EGFP plasmid-transfected MSCs, we firstly confirmed that injected MSCs were able to release exosomes in vivo to the ONL (Fig. 5A). Direct tracing of PKH26-labeled exosomes further verified that after intravitreal transplantation, the exosomes were concentrated and linear along the posterior edge of the ONL, suggesting specific targeting of photoreceptors gathering around photoreceptor nuclei (Fig. 5B). In vitro tracing confirmed that 661W photoreceptors were capable of uptaking MSCderived exosomes, suggesting exosome-mediated cellular communications (Fig. 5C).
The long-term effects of MSCT and EXOT on the retina indicated epigenetic therapeutic mechanisms, among which microRNA regulation may exist [26]. It has also been well documented that exosomes transfer microRNAs for the regulation of recipient cell function in therapeutics [24]. To figure out the potential microRNAs mediating therapeutic effects of exosomes in this study, we took a look on the expression changes of several retina-regulating microRNAs upon the phototoxin MNU treatment [29,[39][40][41]. Quantitative real-time polymerase chain reaction (qRT-PCR) analysis showed that of the selected microRNAs, only miR-21 expression was downregulated by MNU in the retina (Fig. 5D). To investigate whether decline of miR-21 expression contributed to the pathogenesis of retinal degeneration, we applied mice deficient for miR-21 (miR-21 −/− ) [42,43] and challenged them with MNU. As expected, MNU induced more prominent retinal damages in miR-21 −/− mice compared to wild-type (WT) mice at 3 d, leading to more loss of ONL thickness and worse electrophysiological function of the retina (Fig. 5E-K). Interestingly, miR-21 −/− mice per se did not show retinal deficiency, indicating that effects of miR-21 were in responses to cell death stimuli ( Fig. 5E-K). TUNEL staining further demonstrated that miR-21 −/− mice after MNU injection developed higher apoptotic rate of the photoreceptors compared to WT mice with MNU treatment (Fig. 5L, M). Importantly, MNU-induced retinal damages in miR-21 −/− mice were also rescued by EXOT, exhibiting improved ONL thickness and retinal function not only at 3 d ( Fig. 5N-T) but also at 7 d ( Fig. S4A-G). Notably, EXOT also helped photoreceptors to survive after MNU challenge in miR-21 −/− mice (Fig. S4H, I). These results suggested that miR-21 deficiency aggravates MNU-driven retinal injury and can be restrained by EXOT.

miR-21 protects photoreceptors against MNUtriggered apoptosis and can be transferred from MSCs for functional regulation
Next, we intended to verify miR-21 as a crucial regulator of photoreceptor survival before evaluating its function to mediate therapeutic effects of EXOT. As depicted, direct transfection of miR-21 mimics and its negative control (NC) into cultured 661W cells was applied (Fig. 6A), and the efficacy of transfection was confirmed by qRT-PCR (Fig. 6B). Flow cytometric analysis demonstrated that To examine that whether miR-21 can be transferred from MSCs to photoreceptors, we transfected MSCs with Cy3labeled miR-21 mimics before co-culture, and detected the Cy3 fluorescent signal surrounding the nuclei of co-cultured 661W cells (Fig. 6K). Further in vivo tracing analysis after injection of Cy5-labeled miR-21-transfected MSCs confirmed that miR-21 can be transferred to photoreceptors in the retina (Fig. 6L). These data collectively suggested that miR-21 protects photoreceptors against MNU-triggered apoptosis and can be transferred from MSCs for functional regulation.

Exosomal counteraction of MNU damages on photoreceptors involves inhibition of the miR-21 target Pdcd4
We then investigated the functional target of miR-21 in regulating photoreceptor apoptosis. qRT-PCR screening of reported miR-21 targets [44][45][46][47] in both the retinal tissues and 661W cone photoreceptors demonstrated that the mRNA expression levels of Pdcd4, which is a valid target of miR-21 and has been reported for neuroprotection [48], were consistently upregulated by MNU treatments, correlated with downregulated miR-21 expression (Fig. 7A, B). Further evaluation showed that not only in the mRNA levels but also in the protein levels, Pdcd4 expression was induced by MNU but suppressed by EXOT in the retina, suggesting pathological and therapeutic responses (Figs. 7C-E and S5A). Examination of Pdcd4 expression in 661W cells confirmed that EXOT restored MNU-triggered upregulation of Pdcd4 (Figs. 7F-H and S5B). To verify that Pdcd4 expression was indeed regulated by miR-21, we analyzed 661W cells transfected with miR-21 mimics, and found that miR-21 mimics repressed Pdcd4 expression at the protein level (Figs. 7I, J and S5C). To prove that Pdcd4 was functionally important for photoreceptor survival, we transfected 661W cells with a small interfering RNA (siRNA) for Pdcd4 (siPdcd4) and challenged them with MNU (Figs. 7K, L and S5D). Flow cytometric analysis demonstrated that transfection of siPdcd4, but not its NC, effectively protected 661W cells against MNU-induced apoptosis (Fig. 7M, N). These data suggested that exosomal counteraction of MNU damages on photoreceptors involves inhibition of the miR-21 target Pdcd4.
Exosomal miR-21 serves as an effective therapeutic for MNU-induced photoreceptor apoptosis and retinal degeneration The above results prompted us to decipher whether miR-21 mediated therapeutic effects of EXOT on retinal degeneration. By using MSCs derived from WT and miR-21 −/− mice, we identified that MSCs from WT mice secreted miR-21-carrying EXO, while EXO released by MSCs from miR-21 −/− mice failed to show miR-21 expression (Fig. S1L). We discovered that donor miR-21 deficiency remarkably diminished effects of EXOT to preserve morphological and functional integrity of the retina, leaving the ONL thickness and both scotopic and photopic ERG amplitudes unrecovered at 3 d (Fig. 8A-G) and 7 d (Fig. S6A-G). TUNEL analysis confirmed that miR-21 −/− EXOT was less efficient to protect photoreceptors against MNU-induced apoptosis, whereas after transfection of miR-21 mimics into miR-21 −/− MSCs, EXOT regained the efficacy to improve photoreceptor survival after MNU treatment (Fig. 8H-K). Furthermore, miR-21 mimics significantly rescued miR-21 −/− EXOT to overcome the failure in counteracting MNUinduced morphological and electrophysiological damages of the retina (Figs. 8L-R and S6H-N). A diagram showing the synopsis of our findings is provided (Fig. 8S). Taken together, these data indicated that exosomal miR-21 serves as an effective therapeutic for MNU-induced photoreceptor apoptosis and retinal degeneration.

Discussion
Photoreceptor apoptosis represents one key pathogenesis of retinal degeneration and one of the ultimate reasons of visual loss with only few available treatments [1,2]. Whether and how MSCT protects photoreceptors against apoptotic injuries remains unknown. Here, we discovered that intravitreal MSCT long-lastingly counteracted photoreceptor apoptosis and alleviates retinal degeneration, which was exerted through exosomes for intercellular communication between injected MSCs and photoreceptors. We further identified that miR-21, which critically maintained photoreceptor viability by targeting Pdcd4, was transferred from MSCs and mediate therapeutic effects of EXOT. Our findings uncover the efficacy and mechanism of MSCT-based photoreceptor protection, which also indicates exosomal miR-21 as a therapeutic for retinal degeneration.
Degenerative retinal diseases are significant causes of incurable sight loss but the therapeutic options are currently limited [49]. Despite their very different and multifactorial etiologies, the various retinal diseases lead to the same endresult for visual impairment, namely the loss of photoreceptors [1,2]. Nevertheless, photoreceptors fail to regenerate upon injuries, which therefore demands cell replacement or rescue technologies [50]. In this regard, transplantation of embryonic stem cell-differentiated RPE to support photoreceptors has obtained encouraging results to alleviate retinal degeneration from phase I/II trials [51]. For photoreceptors per se, however, although that the apoptotic pathogenesis have been studied, targeting photoreceptor apoptosis or replenishing photoreceptors as potential approaches to retinal degeneration are still in developing [1,2]. Extensively efficient to ameliorate retinal traumatic, ischemic, and oxidative injuries [7,18] as well as retarding inflammatory and diabetic retinopathies [8,52], MSCT represents promising cell-based therapeutics which also rescues photoreceptor deficiency in mice with genetic defects [9][10][11].
Notably, there are studies reporting retinal damages and ocular complications induced by intravitreal or subretinal MSCT, indicating that stem cell delivery should still be applied with caution [53,54]. In this study, we for the first time revealed that MSCT protected photoreceptors against specific apoptotic stimulus in vivo, and further identified EXOT as cell-free alternative strategy to overcome potential drawbacks of cell transplantation with the beneficial effects preserved. Moreover, we discovered that a single intravitreal injection of MSCs or exosomes has a long-lasting protection of photoreceptors in both the MNU pharmacological model and the Pde6b mut genetic model with ameliorated loss of either scotopic (mixed rod and cone responses) or photopic (cone-mediated) ERG responses, which therefore provides promise for establishing novel therapeutics of retinal degeneration.
It is recognized that MSCT maintains tissue homeostasis through either paracrine effects to establish beneficial microenvironments or through inhabitation in recipient tissues to replenish deficient cells [14,55]. After intravitreal transplantation, exogenous MSCs have been traced to remain in the vitreous body, in which no retinal incorporation has been reported [56][57][58]. Nevertheless, it has been claimed in MSCT treating a mouse model of RP that transplanted MSCs morphologically integrated into the RPE, while not being detected in other retinal layers [10]. Additionally, in vitro experiments indicated regulation of MSCs by the retinal microenvironment, in that RPE cellconditioned medium and photoreceptor outer segments stimulate differentiation of MSCs toward the RPE cell phenotype, but the in vivo functional evidence of MSC replacing retinal cells is lacking [59]. Here, we showed that the biodistribution of injected MSCs was primarily at the ONL, indicating close functional correlationships with recipient photoreceptors and potential needs of damaging photoreceptors for recovery. However, MSCs were able to release exosomes for therapeutic effects at the ONL, and effects but not biodistribution of MSCs were diminished upon blockade of exosomal generation. Therefore, it can be expected that MSCT improves retinal integrity and function via paracrine effects, as also confirmed by the identification of various neuroprotective and anti-inflammatory factors secreted, such as nerve growth factor, basic fibroblast growth factor, and tumor necrosis factor alpha-stimulated gene-6 [8,16]. It has additionally been reported that MSC-derived EVs enhance functional recovery of retina after ischemia-reperfusion damages and optic nerve crush injuries, and that transplanted EVs can be uptaken by RGCs and microglia [17,18]. In this study, we further revealed that EXOT efficiently and continuously counteracted phototoxin-induced and Pde6b mut retinal degeneration by targeting photoreceptors, adding another dimensional to the current knowledge of paracrine-based MSC therapy of retinal disorders. Whether the injected MSCs can also proliferate and differentiate toward photoreceptors in vivo remain to be investigated in future studies. Shuttle of microRNAs by EVs have been widely proved as significant contributors to tissue homeostasis and feasible therapeutics for diseases [23][24][25]. In the ocular system, it has been reported that EV-mediated microRNA communication may influence posttranscriptional regulation of retinal development, and that microRNAs carried by circulating EVs can be used as prognostic biomarkers for retinopathy [20,21]. microRNA-dependent mechanisms have also been implicated in MSC-derived EVs to treat corneal fibrosis and optic nerve injury [17,60]. In this study, we first identified an individual microRNA, miR-21, as a contributor to photoreceptor viability in vivo and a mediator of therapeutic effects of EXOT on retinal dysfunction. Notably, miR-21 has been previously documented mainly as a negative regulator of retinal health, such as inhibition of neovascularization in the ischemic retina by targeting tissue inhibitor of metalloproteinase 3 [61], induction of Müller cell gliosis after optic nerve crush by regulating glial fibrillary acidic protein [39], promotion of autoimmune uveoretinitis by targeting interleukin-10 [62], and facilitation of the progression of retinoblastoma by targeting phosphatase and tensin homolog [63]. Besides, previous studies have shown that miR-21 can induce cell apoptosis by targeting S-phase kinase-associated protein 2 and B-cell lymphoma-2 [64,65]. Our results showing miR-21 protection against retinal disorders only join with another study demonstrating that the miR-21/Pdcd4 axis regulates MSC-induced neuroprotection in glaucoma, indicating target-specific regulatory effects based on Pdcd4 inhibition [48]. Actually, miR-21 inhibition of Pdcd4 expression for antiapoptotic effects has additionally been reported in other systems [66,67]. The detailed function and mechanisms of Pdcd4 in regulating photoreceptors and retinal degeneration in vivo remains to be elucidated in future studies.

Animals
Animal experiments were performed following the Guidelines of Intramural Animal Use and Care Committees of Fourth Military Medical University, Xi'an Jiaotong University and the ARRIVE guidelines. Twelve-week WT C57BL/6 mice and miR-21 −/− mice (C57BL/6 background) (weight, 20-22 g; three male or female mice per cage) were purchased from the Jackson Laboratory, as we have used in previous studies [42,43]. Two-week-old male or female Pde6b mut mice with a nonsense mutation of Pde6b gene on a C57BL/6J background that result in the failure of protein production [38] were obtained from TC at Department of Clinical Medicine, Fourth Military Medical University. Mice were maintained with good ventilation and a 12-h light/dark cycle, and were kept feeding and drinking ad libitum.

Cell culture
Isolation and culture of MSCs from mouse bone marrow were as previously described [68]. Briefly, whole bone marrow cells were seeded, incubated overnight, and rinsed with phosphate buffer saline (PBS) to remove the nonadherent cells. The adherent cells were cultured with alpha-Minimum Essential Medium supplemented with 20% fetal bovine serum (FBS), 2-mM L-glutamine, 100-U/ml penicillin, and 100-g/ml streptomycin (all from Invitrogen, USA) at 37°C in a humidified atmosphere of 5% CO 2 . Primary MSCs were digested with 0.25% trypsin (MP Biomedicals, USA) and passaged for the following experiments after seeding at appropriate densities.
Culture of the 661W cell line was according to our previous study and verified for identity and noncontamination [70]. The 661W cell line was derived from mouse retinal tumors and has been characterized previously to be of cone photoreceptor cell lineage [71]. The 661W cell line was cultured in Dulbecco's Modified Eagle Medium supplemented with 10% FBS, 2-mM L-glutamine, 100-U/ml penicillin, and 100-g/ml streptomycin (all from Invitrogen, USA) at 37°C in a humidified atmosphere of 5% CO 2 . For co-culture of MSCs with 661W cells, 661W cells were seeded in 12-well plates as the bottom of a Transwell system (0.4-μm pore size; Corning, USA), while MSCs at the first passage were added into the upper chamber of the Transwell system for 48 h.

Chemical treatments
MNU-provoked retinal degeneration model was established accordingly [70,72] by intraperitoneal injection of MNU (Sigma-Aldrich, USA) at 40 mg/kg. MNU was freshly dissolved in sterile saline. Mice were sacrificed at indicated time points according to study design. In vitro treatment of 661W cells with MNU was performed at a concentration of 200 μg/ml for 6 h. For blockade of MSC generation of exosomes [36,37], 10-μM GW4869 (Sigma-Aldrich, USA) was applied in MSC culture for 24 h, which was initially dissolved in DMSO into a stock solution of 5 mM and was diluted in culture media. The effects of GW4869 on MSC viability and osteogenesis were determined after wash-out procedures.

Collection and identification of exosomes
Collection of MSC-derived exosomes was performed as stated before [73]. Briefly, MSCs at nearly confluence of the first passage were cultured in exosome-depleted media (complete media supernatant after overnight centrifugation at 100,000 g) for 48 h. Exosomes from supernatants were then isolated by the differential centrifugation protocol at 4°C: 300 g for 10 min, 3000 g for 10 min, 10,000 g for 20 min, 100,000 g for 70 min, followed by washing with PBS for another centrifugation at 100,000 g for 70 min. Quantification of exosomes were performed by determining the concentration of total proteins using the Pierce BCA Protein Assay (Thermo Fisher Scientific, USA).
Exosomes were identified according to the criteria of EVs [74]. The number and size distribution of exosomes was quantitated using dynamic light scattering with a Nanosizer TM instrument (Malvern Instruments, UK) [75]. Transmission electron microscopy (TEM) [76] was performed on whole mounts of isolated exosomes using 2.5% glutaraldehyde, postfixed with 1% osmium tetroxide, embedded and stained with uranyl acetate and lead citrate, and observed using a TEM microscope (JEM-1230; JEOL, Japan). Purified exosomes were further characterized by western blot using anti-CD63 and anti-CD81 antibodies [73], as stated below.

Intravitreal injection
MSCs at the first passage were collected and suspended in PBS at 1 × 10 7 cells/ml and were injected at 1 μl into each vitreous chamber with a 33-G Hamilton syringe (Hamilton Company, USA) at 6 h after MNU treatment or at 2-week old of Pde6b mut mice. The collected exosomes were adjusted to a protein concentration of 1 μg/μl and were also injected at 1 μl into each vitreous chamber at 6 h after MNU treatment or at 2-week old of Pde6b mut mice. For the intravitreal delivery, the injection site was selected just posterior to the limbus and the needle was injected under anesthesia (100 mg/kg ketamine plus 20 mg/kg xylazine intraperitoneally), which was retracted after 2 min to minimize backflow [17].

Retinal histology and morphology
At sacrifice, retinal tissues were rapidly isolated, fixed overnight with 4% paraformaldehyde, and embedded in paraffin. 5-μm serial sections were prepared (RM2125; Leica, Germany) and underwent H&E staining for tissue histology and morphology, according to previous reports [70]. Quantification of ONL thickness was determined using the ImageJ 1.47 software along the vertical meridian of the eyeball from nasal to temporal side and through the optic nerve head (ONH) [77], and statistically analyzed as area under the curve (AUC).

TUNEL and IF staining
At sacrifice, retinal tissues were rapidly isolated, fixed in 4% paraformaldehyde, cryoprotected with 30% sucrose, and embedded in the optimal cutting temperature (OCT) compound. The specimens were snap-frozen and sectioned into 15-μm sagittal sections (CM1950; Leica, Germany). Sections were then underwent TUNEL assay using DeadEnd TM Fluorometric TUNEL System according to the manufacturers' instructions (Promega, USA), counterstained by Hoechst 33342 (Sigma-Aldrich, USA) [32]. The images were further analyzed using the ImageJ 1.47 software from at least five consecutive microscopic fields for TUNEL + cells over total ONL cells. For IF analyses on cone photoreceptors, sections were blocked with 5% bovine serum albumin (BSA) (Sigma-Aldrich, USA) dissolved in PBS for 1 h at room temperature, and stained with a rabbit antimouse S-opsin primary antibody (ABN1660; Millipore, USA) overnight at 4°C at a concentration of 1:1000. After washing with PBS, sections were stained with an AF647conjugated goat-anti-rabbit secondary antibody for 1 h at room temperature, and were counterstained with Hoechst 33342 (Sigma-Aldrich, USA). Quantification of the percentages of positively stained area over total retinal fields was performed using the ImageJ 1.47 software.

ERG analysis
The UTAS Visual Diagnostic System with a Big Shot Ganzfeld (LKC Technologies, USA) was employed. Mice were dark-adapted overnight before ERG analysis. Under dim red light conditions, anesthesia was induced by an intraperitoneal injection of 80 mg/kg ketamine and 10 mg/kg chlorpromazine, and mice were then lightly secured to a stage to ensure a stable position for recording. Cornea was anesthetized with a drop of 0.5% proxymetacaine and was kept moist with physiological saline. Platinum circellus record electrodes were placed on each cornea, while reference and ground electrodes were respectively located in the mouth and inserted in the tail. White flashes with the intensity of 3.0 cd.s/m 2 were applied for stimulating the responses. The band-pass (1-300 Hz) was used to amplify the recorded signals. The line noise was wiped off by a 50-Hz notch filter. Totally 10 scotopic (dark-adapted, mixed rod-and cone-mediated) and 60 photopic (light-adapted, cone-mediated) responses were recorded for analyzing the amplitudes of a-wave and/or b-wave by ERG View v4.380 R software [78].
Transfection of microRNA mimics, siRNA, and plasmids The mimics and NC of miR-21, and the siRNA and NC of Pdcd4 were designed by RiboBio (Guangzhou, China) and were transfected into cells at final concentrations of 50 nM, according to previous studies [42,43]. The Lipofectamine 2000 reagent (Invitrogen, USA) was used for transfection according to the manufacturer's recommendation. CD63-EGFP plasmids were purchased from Hanheng (Shanghai, China) and were also transfected into MSCs using the Lipofectamine 2000 reagent (Invitrogen, USA). The transfected MSCs or 661W cells were further used for downstream experiments after 48 h incubation.
In vitro apoptosis assay 661W cells were harvested and evaluated by FITCconjugated Annexin V and propidium iodide (PI) double staining according to the manufacturer's instruction of Annexin V Apoptosis Detection Kit I (BD Biosciences, USA). Cell apoptosis was detected with a flow cytometer (CytoFLEX; Beckman Coulter, USA) equipped with the CXP 2.1 software. The percentages of early apoptotic (FITC + PI − ) plus late apoptotic (FITC + PI + ) cells were expressed as apoptotic percentages [55,79].

Tracing of MSCs, exosomes, and miR-21
For tracing of MSCs and exosomes, PKH26 (Sigma-Aldrich, USA) was used according to manufacturer's instructions [73]. For biodistribution of MSCs in the eye, freshly dissected eyes were scanned using the Xenogen IVIS imaging system (Xenogen IVIS Lumina II; PerkinElmer, USA) at 6 h after injection of PKH26-labeled MSCs. For in vivo tracing in the retina, retinal tissues were harvested at 24 h after injection of PKH26-labeled MSCs or exosomes. CD63-EGFP plasmidtransfected MSCs were also injected for in vivo examination of exosome release. For tracing of MSC-derived miR-21 in vivo, Cy5-labeled miR-21 mimics (RiboBio, China) was transfected into MSCs using the Lipofectamine 2000 reagent (Invitrogen, USA) before injection. The retina were then fixed in 4% paraformaldehyde, cryoprotected with 30% sucrose, and embedded in the OCT compound. The specimens were snap-frozen and sectioned into 15-μm sagittal sections (CM1950; Leica, Germany) and counterstained by Hoechst 33342 (Sigma-Aldrich, USA). For in vitro tracing, 661W cells were treated by PKH26-labeled exosomes for 48 h, fixed in 4% paraformaldehyde, blocked by 5% BSA, and immunostained with a rabbit anti-mouse primary antibody at a concentration of 1:100 for β-tubulin (ab18207; Abcam, UK) overnight at 4°C. After washing with PBS, the sections were stained with a FITC-conjugated goat-anti-rabbit secondary antibody and counterstained by Hoechst 33342 (Sigma-Aldrich, USA). For tracing of MSC-derived miR-21 into 661W cells in vitro, Cy3-labeled miR-21 mimics (RiboBio, China) was transfected into MSCs using the Lipofectamine 2000 reagent (Invitrogen, USA) before MSCs being added in the upper chamber of the Transwell system. MSCs and 661W cells were then co-cultured for 48 h, and 661W cells were fixed in 4% paraformaldehyde and counterstained by Hoechst 33342 (Sigma-Aldrich, USA).

qRT-PCR
Total RNA was extracted from retinal tissues or cells by direct adding of Trizol Reagent (Takara, Japan) and purified by phenol-chloroform extraction. For mRNA, cDNA synthesis was performed using the PrimeScript TM RT Reagent Kit (Takara, Japan), and the primer sequences of the genes detected in this study were listed in the Supplementary Information (Table S1). For microRNA, reverse transcription primers and qRT-PCR primers were designed by RiboBio (Guangzhou, China) as the Bulge-loop TM miRNA qRT-PCR Primer Sets. qRT-PCR detection was performed using the SYBR Premix Ex Taq II Kit (Takara, Japan) by a Real-Time System (CFX96; Bio-Rad, USA). The relative expression level of each gene was obtained by the cycle number after normalizing against Glyceraldehyde-3-phosphate dehydrogenase (Gapdh, for mRNA) and RNU6 (for microRNA) abundances using the 2 −ΔΔCT method [42,43].

Western blot
Traditional western blot was performed as previously described [79]. Whole lysates of exosomes were prepared using the Lysis Buffer (Beyotime, China). Proteins were extracted, loaded on sodium dodecyl sulfate-polyacrylamide gels, transferred to polyvinylidene fluoride membranes (Millipore, USA), and blocked with 5% BSA (Sigma-Aldrich, USA) in PBS with 0.1% Tween for 2 h in room temperature. The membranes were incubated overnight at 4°C with the following primary antibodies: a rabbit antimouse primary antibody at a concentration of 1:1000 for CD63 (sc-5275; Santa Cruz Biotechnology, USA); and a rabbit anti-mouse primary antibody at a concentration of 1:1000 for CD81 (sc-70803; Santa Cruz Biotechnology, USA). The membranes were then incubated with peroxidase-conjugated goat anti-rabbit secondary antibodies at a concentration of 1:40000 for 1 h in room temperature. The blotted bands were visualized using an enhanced chemiluminescence Kit (Amersham Biosciences, USA) and a gel imaging system (5500; Tanon, China).
Capillary-based immunoassay was performed for retinal and 661W cell lysates using the Wes-Simple Western method with the anti-rabbit detection module (Pro-teinSimple, USA) [80]. Protein expression was measured by chemiluminescence and quantified as AUC using the Compass for Simple Western program (ProteinSimple, USA). Proteins were detected with the following primary antibodies: a rabbit anti-mouse primary antibody for Pdcd4 (9535S; Cell Signaling Technology, USA) and a rabbit antimouse primary antibody for Gapdh (5174S; Cell Signaling Technology, USA) at concentrations of 1:1000 (both from Cell Signaling Technology, USA).

Sample processing
Each experiment was performed in at least three independent times and data represented biological replicates. Sample size of animals was determined based on preliminary tests for variation to ensure adequate power for analysis. Samples were blindly allocated to each of the experimental group based on randomization and were also blind for data analysis. No samples were excluded from the results.

Statistical analysis
The results are represented as median ± range or as box (25th, 50th, and 75th percentiles) and whisker (range) plots for indicated experiments. The data were analyzed using non-parametric methods, i.e. the Mann-Whitney U tests for two group comparisons or the Kruskal-Wallis tests for multiple group comparisons, in the GraphPad Prism 5.01 software for not meeting normal distribution. Values of P < 0.05 were considered to be statistically significant.
Author contributions CLD, CBH, and STL contributed equal to the experimental performing, data acquisition and analysis, and manuscript drafting. NZ, LHB, and FZ contributed to data analysis and interpretation. YCX and TC contributed to data interpretation. BDS, XRY, and CHH contributed to the study conception and design, data interpretation and manuscript revision.

Conflict of interest The authors declare that they have no conflict of interest
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.