Photoreceptor nanotubes mediate the in vivo exchange of intracellular material

Abstract Emerging evidence suggests that intracellular molecules and organelles transfer between cells during embryonic development, tissue homeostasis and disease. We and others recently showed that transplanted and host photoreceptors engage in bidirectional transfer of intracellular material in the recipient retina, a process termed material transfer (MT). We used cell transplantation, advanced tissue imaging approaches, genetic and pharmacologic interventions and primary cell culture to characterize and elucidate the mechanism of MT. We show that MT correlates with donor cell persistence and the accumulation of donor‐derived proteins, mitochondria and transcripts in acceptor cells in vivo. MT requires cell contact in vitro and is associated with the formation of stable microtubule‐containing protrusions, termed photoreceptor nanotubes (PhNTs), that connect donor and host cells in vivo and in vitro. PhNTs mediate GFP transfer between connected cells in vitro. Furthermore, interfering with PhNT outgrowth by targeting Rho GTPase‐dependent actin remodelling inhibits MT in vivo. Collectively, our observations provide evidence for horizontal exchange of intracellular material via nanotube‐like connections between neurons in vivo.


Introduction
The mammalian retina is a central nervous system (CNS) structure that serves as a tractable system to model neural repair strategies for CNS disease (London et al, 2013). Like the rest of the CNS, the mammalian retina has little regenerative capacity; thus, degeneration of photoreceptors, first-order light-responsive neurons, results in irreversible vision impairment. Strikingly, photoreceptor transplantation into the adult retina can partially restore vision in preclinical mouse models of photoreceptor dysfunction (MacLaren et al, 2006;Lamba et al, 2009;Inoue et al, 2010;Pearson et al, 2012;Santos-Ferreira et al, 2015). However, we and others recently reported that donor neurons rarely integrate into the recipient retinal parenchyma and instead engage in bidirectional transfer of intracellular materials, including cytosolic GFP and Cre recombinase expressed from transgenes, as well as endogenous phototransduction proteins, which move into recipient (acceptor) photoreceptors in a process termed material transfer (MT) (Pearson et al, 2016;Santos-Ferreira et al, 2016;Singh et al, 2016;Decembrini et al, 2017;Ortin-Martinez et al, 2017). These observations suggest that cell transplantation-mediated rescue of photoreceptor function is indirect and at least partially mediated by the transfer of intracellular material to neurons in the recipient retina. New insights into the mechanics of cytosolic MT between neurons are essential to improve our capacity to design effective cell-based repair strategies in the CNS and to increase our foundational knowledge of interneuronal communication.
The direct exchange of cytoplasmic material is rapidly emerging as an important channel for intercellular communication in the CNS, facilitating the transfer of cytosolic nucleic acids, lipids, proteins and organelles between neurons and glia (Guescini et al, 2012;Davis et al, 2014;Thomas et al, 2017). Cells have evolved two main mechanisms to exchange labile intracellular material: the secretion of membrane-enclosed extracellular vesicles (EVs) (Colombo et al, 2014) and the formation of nanotubes, which are filopodia-like structures that act as cytoplasmic bridges between cells to allow cytosolic exchange (Yamashita et al, 2018). The role of nanotubes in the exchange of organelles (Wang & Gerdes, 2015) and pathogenic proteins (Victoria & Zurzolo, 2017) is well documented in vitro, but evidence that these structures mediate the exchange of intracellular material between neurons in vivo is lacking.
Here, we used photoreceptor transplantation to investigate how MT is mediated between neurons in vivo. We show that donor and acceptor photoreceptors are connected by nanotube-like protrusions, termed photoreceptor nanotubes ( Ph NTs), that mediate the transfer of intracellular proteins and organelles. Ph NTs contain microtubules, are stable after fixation, morphologically diverse and are able to mediate GFP transfer between pairs of connected photoreceptors. The outgrowth and consequently MT through Ph NTs are regulated by Rho GTPase-dependent actin remodelling. This study is the first to identify nanotube-like structures as primary conduits for the exchange of intracellular material between photoreceptors in vivo.

Efficacy of material transfer correlates with donor photoreceptor survival and is modulated by the host retina environment
Retinal photoreceptor transplantation involves transplanting GFP +rod photoreceptor precursor cells from Nrl::GFP mice , hereafter referred to as donor photoreceptors, into the subretinal space (SRS) of the adult mouse retina (Pearson et al, 2016;Santos-Ferreira et al, 2016;Singh et al, 2016;Decembrini et al, 2017;Ortin-Martinez et al, 2017;Fig 1A). Grafted GFP + -rod photoreceptors do not migrate into the recipient retina, and instead, their cell bodies remain sequestered in the SRS. Thus, the accumulation of fluorescent reporter proteins in adjacent photoreceptors in the recipients, hereafter referred to as acceptor photoreceptors, occurs through MT. To investigate the kinetics of MT and the relationship between donor and acceptor photoreceptors, we monitored the numbers of GFP + -donor and GFP + -acceptor photoreceptors over time in wild-type and Nrl À/À recipients, the latter being a mouse strain with enhanced levels of GFP transfer after transplantation (Santos-Ferreira et al, 2015;Decembrini et al, 2017;Ortin-Martinez et al, 2017). We found that there was a positive linear correlation between the numbers of GFP + -donor and GFP + -acceptor photoreceptors over time in both types of transplant recipients (Fig EV1A), indicating that there is a tight correlation between accumulation of GFP in acceptor photoreceptors and donor photoreceptor survival. We normalized the data as the ratio of GFP + -acceptor/GFP + -donor cells, hereafter referred to as MT index, which provided a rigorous metric to evaluate the phenomenon of MT and mitigated interrecipient transplant variability. Notably, the MT index in wild-type recipients increased from 14 days to reach a maximum by 21 days post-transplant (Fig 1B and C), coinciding with the time when the retinal detachment injury induced by transplant surgery was resolved ( Fig 1E). As expected, the MT index was $ 5-fold greater in Nrl À/À compared with wild-type recipients ( Fig 1C). Notably, because these data are normalized to donor cell number, the increase in MT Nrl À/À recipients is independent of donor photoreceptor survival. Taken together, our results show that MT is, in part, a function of the number of surviving donor photoreceptors, but also that recipient photoreceptors differ in their capacity to support MT from an identical pool of transplanted photoreceptors.
We noted a decline in the absolute numbers of GFP + -acceptor photoreceptors in both types of recipient mice by 90 days posttransplantation, even though the numbers of GFP + -acceptor and GFP + -donor photoreceptors were still tightly correlated ( Fig EV1A). A Photoreceptor transplantation schematic. B, C Analysis of GFP transfer after transplantation of Nrl::GFP donor photoreceptors into wild-type C57BL/6J (B) and Nrl À/À (C) recipient retinas after 14 (wt, n = 9; Nrl À/À , n = 7) and 21 (wt, n = 6; Nrl À/À , n = 4) and 90 (wt, n = 5; Nrl À/À , n = 4) days. Tissue is stained for GFP and nuclei (Hoechst). In this and all following figures, donor cells (D) are located above, and acceptor cells with GFP (A) are located below the outer limiting membrane (white dashed line), which delimits the apical border of the outer nuclear layer (ONL). In wild-type recipients, donor cells are typically located above the outer segment (OS) region (between the grey dashed lines). In Nrl À/À recipients, donor cells are located in the inner segment (IS) region. Plotted is the ratio of the numbers of GFP + photoreceptors in the ONL (acceptor photoreceptors) and GFP + donor photoreceptors in the subretinal space (SRS) over time. D Analysis of Cre recombinase transfer after transplantation of Crx::Cre donor photoreceptors into Nrl À/À recipients after 14 (n = 6), 21 (n = 5), 42 (n = 4), 90 (n = 6) and 150 (n = 6) days. The number of acceptor photoreceptors in the recipient retina that activated the Cre reporter gene is quantified. E Resolution of the transplant-induced retinal detachment in C57BL/6J recipient retinas (n = 11), as measured by the maximum detachment length (mm) over time. Fundoscopy (images on the left) was used to locate the region of cell delivery for optical coherence tomgraphy (OCT) imaging (green line) shown in the images on the right. Red lines correspond to the callipers used to determine the maximum length of the detachment at each time point. Before injection (B.I.) (n = 11), immediately post-injection (I.P.I.) (n = 10), 5 days post-injection (5 D.P.I.) (n = 11), 9 days post-injection (9 D.P.I.) (n = 11) and 21 days post-injection (21 D.P.I.) (n = 6). F Co-transplanted P3-4 Crx::Cre and P3-5 ROSA mT/mG donor photoreceptors into C57BL/6J recipient retinas show Cre reporter gene induction in donor cells located in the SRS 14 days post-transplantation.  One explanation for this parallel decline in both populations was that acceptor photoreceptors required constant replenishment of material from donor photoreceptors to maintain detectable levels of GFP. If this model were true, then the number of labelled acceptor photoreceptors should not decline in a situation where MT induces constitutive expression of a fluorescent reporter. To test this idea, we transplanted retinal cells from Crx::Cre mice  in which all photoreceptors express Cre recombinase, into adult Nrl À/À mice carrying a ROSA mT/mG Cre reporter, and monitored the Cre-dependent induction of membrane GFP (mG) expression in acceptor photoreceptors ( Fig 1D). The number of mG +acceptor photoreceptors reached a plateau by 21 days and did not decline even after 3 and 5 months post-transplant ( Fig 1D). From these data, we conclude that acceptor photoreceptors that receive transferred material are not selectively less viable and that MT requires constant replenishment of labile material from donor photoreceptors.

Material transfer is specific to photoreceptors and their precursors
While our findings show that accumulation of transferred material in acceptor photoreceptors depends on the presence of donor photoreceptors, it is not clear how selective this process is for photoreceptor donors, as it could simply be that acceptor photoreceptors have a high propensity to accumulate material from any type of cell. To address this issue, we compared the MT index after transplanting GFP + -human tumour cells (Fig EV1B-D). While engrafted cells were readily detected in the SRS, the incidence of MT was markedly low (Fig EV1B and C). We did similar experiments with embryonic cortical cells and observed only rare examples of GFP accumulation in acceptor photoreceptors ( Fig EV1D). Thus, efficient MT appears to be neuron lineage specific, as it is not recapitulated by tumour cells or cortical neurons. Finally, to investigate how photoreceptor maturation affects MT we also asked whether MT can occur between immature photoreceptors. We transplanted mixed retinal dissociates from P3-5 Crx:: Cre and ROSA mT/mG mice to the SRS of wild-type adult mice and tracked Cre transfer, marked by the induction of mG expression, within the bolus. We detected "recombined" mG + cells in the subretinal cell deposit, indicating that Cre is transferred between transplanted cells (Fig 1F). Since photoreceptor precursors do not fully mature after transplantation (Eberle et al, 2012), this result indicates that acceptor photoreceptors do not need to be mature for MT to occur and that this mode of cell-cell communication can occur outside of the neuroepithelial organization of the retina.

Material transfer is associated with the accumulation of donor-derived proteins and transcripts in acceptor cells
The accumulation of fluorescent reporter proteins in acceptor photoreceptors could be mediated by protein and/or transcript transfer. We used immunostaining and RNAscope to compare the levels of GFP protein and transcript, respectively, in donor and acceptor cells in transplanted wild-type and Nrl À/À recipients (Fig 2A and B). GFP fluorescence intensity and transcript levels were approximately 2-fold and 20-fold lower, respectively, in acceptor cells relative to donor cells in both types of recipients (Fig 2A-E). To investigate whether there was a similar relationship for the transfer of endogenously expressed proteins, we examined transfer of GNAT1, a rod photoreceptor specific gene, in Nrl À/À recipient photoreceptors, which do not express GNAT1. We found higher levels of GNAT1 protein compared with transcripts in Nrl À/À acceptor photoreceptors (Fig 2C-E). These findings indicate that mRNA can transfer between cells, but they do not distinguish whether the accumulation of donor-derived proteins in acceptor cells is due to direct protein transfer or translation from transferred mRNA. We reasoned that if transcripts were the primary cargo for transfer, then the accumulation of donor cell-derived nuclear and cytosolic proteins in acceptor cells should be similar. We used lentiviruses (Appendix Fig S1A) to express similar levels of cytosolic and nuclear localized GFP (nls-GFP) in transplanted donor cells and found that cytosolic GFP transfer was significantly increased relative to nls-GFP transfer (Fig 2F-K). Moreover, this difference in the transfer of cytosolic versus nuclear proteins was not associated with significant differences in the transcript levels in acceptor cells, despite similar levels of these transcripts in donor cells (Fig 2H and K). Taken together, these data support a model in which the accumulation of donor-derived material can be mediated by mRNA as well as direct protein transfer.

Mitochondria transfer in the context of photoreceptor transplantation
To ask whether MT extends to organelles, we investigated mitochondria transfer in vivo. We transplanted GFP + -donor photoreceptors from Nrl::GFP mice that were infected with lentiviruses to express Mito-DsRed, a mitochondrial targeting sequence fused to DsRed. At 21 days post-transplantation into adult retinas, we ▸ Figure 2. Transfer of donor-derived proteins and transcripts to host photoreceptors.
Data information: All data are presented as mean AE SEM; n.s. not statistically significant, ****P ≤ 0.0001, ***P < 0.001, **P < 0.01 and *P < 0.05; one-way ANOVA with Tukey's post hoc multiple comparison test. White dashed line delimits the apical border of the outer nuclear layer; grey dashed lines indicate the outer segment (OS) region. IS, inner segments. Scale bars: A-C, I, J: 10 lm. F, G: 50 lm. Source data are available online for this figure.

of 23
The EMBO Journal 40: e107264 | 2021 ª 2021 The Authors observed DsRed + -puncta in acceptor photoreceptors ( Fig 3A). While we cannot exclude the possibility that this result is a consequence of transcript and/or Mito-tagged protein transfer, we suggest that it reflects the transfer of donor mitochondria to acceptor cells for two reasons. First, all Mito-DsRed+ puncta in acceptor cells co-localized with a mitochondria marker, ATP5B (Fig 3B-D''). Second, the scenario where the mito-tagged protein is imported into acceptor cell mitochondria after transfer is unlikely as it would require engagement of the protein unfolding machinery (Pfanner & Truscott, 2002  Photoreceptor material transfer is bidirectional and requires cell contact To further characterize MT, we established an in vitro model where we tracked fluorescent protein exchange in co-cultures of GFP + (from Nrl::GFP mice) and membrane tethered tdTomato + (mT + ) (from ROSA mT/mG mice) photoreceptors (Appendix Fig S2A). After 3 days in vitro, we detected GFP + mT + cells in these co-cultures (Fig 4A), indicating that transfer of fluorescent proteins did occur in vitro. We then used flow cytometry to quantify the frequency of GFP + mT + photoreceptors under various conditions (Fig 4B-E and Appendix Fig  S2B). We found that MT in vitro is cumulative ( Fig 4C) and requires live cells, as it was not detected when viable cells were co-cultured with fluorescent debris from dead cells (Fig 4B, Appendix Fig S2D). Furthermore, MT did not require the presence of other neuronal cell types in the culture, as it occurred in co-cultures of purified photoreceptors ( Fig 4E). We also asked whether mitochondria could be exchanged by co-culturing retinal cell dissociates labelled with red and green MitoTracker dyes. Using imaging and flow cytometry, we found that cells exchanged MitoTracker-labelled puncta and that this exchange required live-labelled cells (Fig 4F and G,Appendix Fig S2C). Based on these data, we conclude that cytosolic and membrane-associated proteins, as well as mitochondria, can transfer between photoreceptor precursor cells in vitro. Extracellular vesicles can transfer proteins, nucleic acids and organelles between cells (Zappulli et al, 2016) and GFP-labelled donor photoreceptors from Nrl::GFP retinal dissociates secreted bilipidencased EVs that contained GFP protein (Fig EV2A and B). However, several lines of evidence indicate that EV exchange is not the primary mechanism of MT of proteins and mitochondria. We did not detect MT when photoreceptors expressing fluorescent proteins or labelled with MitoTracker dyes were separated by a transwell, indicating that MT in vitro required cell contact ( Fig 4G). EVs could still contribute to contact-dependent MT, particularly if they traffic cargo along protrusions, as has been shown in other contexts (Heusermann et al, 2016). However, pharmacologic inhibition of EV secretion had no effect on MT of fluorescent proteins in vitro and in vivo (Fig EV2D-F). Finally, MT was not increased following transplantation of GFP-labelled donor photoreceptors isolated from Nrl::GFP; Prphr2 À/À mice, which have genetically enhanced EV secretion (Salinas et al, 2017) (Fig EV2C). While these observations do not preclude a role for EVs in GFP transfer, they do suggest that uptake of randomly secreted EVs is not the primary mechanism of MT in vivo and in vitro.

Donor and acceptor photoreceptors engaging in material transfer are physically connected by cellular protrusions
We next examined transplanted retinas for evidence of physical contacts between donor and acceptor photoreceptors. Consistent with previous observations (Ortin-Martinez et al, 2017), we observed thin protrusions that appeared to connect GFP + donor and GFP + acceptor photoreceptors in transplanted retinas ( Fig 5A). Using a tissue clearing protocol for whole intact eyes (Gurdita et al, 2021), we next imaged single donor photoreceptors in the SRS by two-photon and light sheet fluorescence microscopy (Fig 5D and E,Movies EV3 and EV4). Consistent with our observations in whole mounts, we observed protrusions connecting GFP + -donor and GFP +acceptor photoreceptors in intact eyes (Fig 5D and E, Movies EV3 and EV4). Quantification across the three imaging modalities showed that in the majority of cases ( $ 80%), the apical regions of GFP + -acceptor photoreceptors were connected to donor photoreceptors via a protrusion (Fig 5F), revealing a previously unappreciated aspect of the donor-acceptor neuron interface. GFP +donor cell protrusions also contained transferable cargo, such as mitochondria and GFP mRNA ( Fig EV3). We therefore propose that these protrusions connecting donor and acceptor photoreceptors are relevant to MT, either serving as conduits for MT or serving a permissive adhesive function that allows for MT to occur over a short range.

Donor photoreceptor protrusions resemble immature neurites and can mediate GFP exchange in vitro
Transplanted photoreceptors have been reported to extend protrusions (Warre-Cornish et al, 2014), and these could be related to the apical or basal processes of mature rod photoreceptors, outer segments (OS) and axons, respectively (Sarin et al, 2018). While donor cells exhibited polarized staining with OS markers GNAT1 and Peripherin2 (Fig 6A and B), this was frequently oriented towards the retinal pigmented epithelium and not enriched in the protrusions connecting donor and acceptor cells, indicating that they are distinct from OS. Instead, these protrusions were similar to neurites, based on the presence of microtubules ( Fig 6A and C), tortuosity and spherule-like terminals at their distal tips ( Fig 6B). However, these protrusions appeared to be distinct from mature photoreceptor axons, as presynaptic markers were not enriched at their termini, rather these were scattered along the length of donor cell protrusions ( Fig 6D).
◀ Figure 4. Characterization of material transfer in primary photoreceptor cell culture.
White dashed line delimitates both cell somas and the protrusion connecting them (F). No exchange is observed when the two photoreceptor populations are separated by a transwell (n = 3), nor in co-cultures of live MTG-labelled and lysed MTR-labelled cells (n = 3) (G).
Data information: Abbreviations: DIV, days in vitro; GFP, green fluorescent protein; RFP, red fluorescent protein; and DIC, differential interference contrast microscopy. All data presented as mean AE SEM; n.s. not statistically significant, ****P ≤ 0.0001, ***P < 0.001, **P < 0.01 and *P < 0.05; one-way ANOVA with Tukey's post hoc multiple comparison test and unpaired t-test. Scale bars: 10 lm. Source data are available online for this figure. To investigate the capacity for these protrusions to mediate MT in vitro, we first confirmed that they formed ex vivo. In fixed ( Fig  EV4A and B) and live (Fig 7A and B) cultures, we detected microtubule + -protrusions that connected to photoreceptors and contained presynaptic markers ( Fig EV4B) and transferrable cargo, including mitochondria and GFP (Fig 7A and B, amd EV4A), along their length. These protrusions were similar in size and morphology to those we identified in vivo (Fig EV4C). To further characterize the types of protrusions that form in vitro, we analysed live Nrl::GFP cultures that were stained with vital actin and tubulin dyes. Protrusions could be grouped into two main categories: those that connected cells, which comprised the majority ( $ 60%), and those that were unconnected ( $ 40%). While nearly all protrusions stained with vital actin, the microtubule content was higher in connected (83%) compared with unconnected (38%) protrusions (Figs 7C and EV5). Analyses of fixed and live cultures revealed that the majority of the connected protrusions were suspended above the dish (Fig 7C, Movie EV5) and terminated with a distal swelling or spherule-like morphology (Figs 7C and EV5). In contrast, the majority of unconnected protrusions were attached to the substrate and terminated with a classical growth cone morphology (Movie EV6). Finally, live imaging analysis revealed that MitoTracker + puncta moved along connected protrusions (Fig 7B and E, and Movie EV7). To investigate whether connected protrusions could transfer material, we analysed the fluorescence recovery after photobleaching (FRAP) of Nrl::GFP cultures. Isolated photoreceptors (n = 14) recovered a small fraction of the original GFP fluorescence (4%) over the entire imaging interval of 250 s (Fig 7D), possibly due to redistribution of the remaining GFP in the cell. In contrast, photoreceptors that were connected to an adjacent cell via a protrusion (n = 21) recovered $ 15% of their 10 of 23 The EMBO Journal 40: e107264 | 2021 ª 2021 The Authors fluorescence over the entire imaging interval (Fig 7D). The recovery of GFP in the bleached cell was more rapid than would be expected for de novo protein synthesis (Kourtis & Tavernarakis, 2009) and was associated with a significant and reciprocal reduction in GFP intensity in the unbleached connected cell of the pair (n = 22) (Fig 7D). Taken together, cultured and transplanted photoreceptors form similar protrusions and in vitro analysis reveals that these connected protrusions are capable of trafficking mitochondria and transferring GFP between cells. Based on these results, we suggest that donor photoreceptors extend microtubule + -protrusions to form stable connections with acceptor photoreceptors in vivo. Furthermore, we provide evidence Data information: n.s. not statistically significant, ****P ≤ 0.0001 and ***P < 0.001; unpaired t-test and extra-sum-of-squares F-test (Y0 constrained to 0.0). Scale bars: 5 lm. Source data are available online for this figure.
ª 2021 The Authors The EMBO Journal 40: e107264 | 2021 that these protrusions do not resemble outer segments nor mature axons of photoreceptors. Rather, because they are suspended above the substratum of the dish and mediate the transfer of intracellular material between connected photoreceptors, we name them photoreceptor nanotubes ( Ph NT).

Rho GTPase actin remodelling regulates Ph NT extension and material transfer
The formation of membrane protrusions, including neurites, is regulated by the coordinated activity of Rho GTPases, where in general, Rac1 and Cdc42 promote, and Rho inhibits (Faix & Rottner, 2006;Arkwright et al, 2010;Ridley, 2011;Chen et al, 2012) neurite outgrowth. We have shown previously that inhibiting Rho kinase, a downstream RhoA effector, increases neurite outgrowth in photoreceptor precursors (Tsai et al, 2019). Given the similarities between Ph NTs and neurites, we assessed the role of Rho GTPase activity in MT in vitro and in vivo. Treatment with a Rho kinase inhibitor, Y27632, significantly increased GFP and mitochondria transfer in vitro (Fig 8A and B), suggesting a positive role for Rho signalling in MT. To investigate the role of actin remodelling on MT in vivo, we infected donor photoreceptors with lentiviruses expressing RHOA and dominant negative RAC1 (RAC1 DN) (Appendix Fig S3C  and D). After confirming that control, RHOA and RAC1 DN lentiviruses induced comparable levels of GFP expression in donor photoreceptors (Appendix Fig S3D) and that RAC1 DN expression induced growth cone collapse in vitro (Appendix Fig S3E), we transplanted these cells into the SRS of wild-type adult recipients and analysed GFP transfer after 21 days. We confirmed that the presence of GFP in the recipient retina was caused by MT and not integration of the transplanted cells using EdU labelling, which revealed the retention of transplanted donor photoreceptors in the SRS (Appendix Fig S3C). Compared with control lentivirus-infected cells, RHOA and RAC1 DN expression significantly reduced GFP transfer between donor and acceptor photoreceptors (Fig 8C). This effect was not due to reduced donor photoreceptor viability ( Fig 8C) but was associated with a significant reduction in the average and total Ph NT length in RHOA and RAC1 DN expressing donor photoreceptors (Fig 8C and D). We therefore conclude that Ph NT outgrowth in donor photoreceptors is promoted by Rac1 activity and that Ph NT formation is required for MT.

Discussion
In this report, we investigated the kinetics, cargo types and the cell biological processes that are involved in MT between donor and acceptor photoreceptors in vivo and in vitro. We find that MT correlates with donor neuron survival in vivo and transferred cargo includes protein, mRNA and mitochondria. The sites of GFP exchange between photoreceptors in vivo are characterized by the presence of nanotube-like protrusions, Ph NTs, that connect donor and acceptor cells, and genetic perturbation of the actin cytoskeleton reduces the outgrowth of these protrusions and reduces MT in vivo. Moreover, we demonstrate using FRAP that Ph NT protrusions mediate GFP transfer between photoreceptors in vitro. Based on these data, we conclude that Ph NTs are conduits for MT in vivo and in vitro. In a parallel study, Kalargyrou et al (2021) describe similar findings, where they show that nanotube-like protrusions mediate MT between photoreceptors in vitro and that Rho GTPasedependent cytoskeletal remodelling is required for MT from donor to host photoreceptors in vivo. Together these studies make a strong case for protrusion-based transfer of cytoplasmic contents between photoreceptors.
Nanotubes, referred to originally as tunnelling nanotubes (TNTs), were described as dynamic actin-containing protrusions that mediate electrical coupling and vesicle transfer between cells (Rustom et al, 2004). Subsequently, the definition of nanotubes has expanded to include protrusions with different cytoskeletal content, dimensions and context-specific functions (Wang et al, 2012). TNTs have been described in intact eye structures, including TNTs connecting immune cells in the cornea (Chinnery et al, 2008) and inter-pericyte TNTs in the retinal vasculature that mediate neural vascular coupling (Alarcon-Martinez et al, 2020). However, TNTlike connections between retinal neurons or glia in the intact retina have not been reported. The transplant-associated Ph NTs that we describe here share several characteristics with the microtubule + TNTs that connect PC12 cells (Wang & Gerdes, 2015) and cultured primary neurons (Wang et al, 2012), including morphology, size, stability after fixation and permissiveness for the exchange of small cytosolic molecules and larger organelles, including mitochondria. However, we cannot rule out an additional role for the classical actin-only dynamic TNTs in MT in our system. For instance, these types of TNTs could represent an intermediate structure in Ph NT formation or extend from the distal tips of Ph NTs to mediate MT, but as they are fragile, they would not be captured in fixed tissues. What also remains unclear is the mechanism of transfer at the contact sites of these protrusions on acceptor cells. We speculate transfer could involve transient membrane fusions that permit cytoplasmic continuity or the short-range release and uptake of vesicles. Distinguishing between these mechanisms will require deeper ultrastructural and functional analyses.
The striking parallels between nanotubes and neurites have been noted previously (Nussenzveig, 2019), and several lines of evidence support a neurite-like identity for the protrusions formed by photoreceptors that we describe here. The photoreceptor precursors used as donor neurons in our in vivo and in vitro assays are harvested at a stage in development where they extend neurites in vivo (Sarin et al, 2018) and they exhibit neurite-extension behaviour in heterotopic environments, including cell culture (Kljavin & Reh, 1991) and after transplantation to the adult retina (Tsai et al, 2019). Interestingly, photoreceptors in vitro can also form neurites with growth cones and whether these are the precursors to the Ph NT that connect cells remains to be determined. Consistent with this possibility is the demonstration that nanotubes that connect astrocytoma cells in vivo exhibit growth cone-like morphology at their tips (Osswald et al, 2015).
Nanotube-mediated transfer of neural degeneration diseasecausing proteins in vitro (Symons & Rusk, 2003) has contributed to the prevailing idea that a similar mechanism could be involved in the connectome-associated route of disease-causing protein spread in vivo (Ridley, 2006), including the inter-neuronal spread of misfolded alpha synuclein in Parkinson's disease and multiple system atrophy and misfolded tau protein in Alzheimer's disease and frontotemporal dementia (Uemura et al, 2020). Our work extends that concept and demonstrates that neuron MT is mediated However, in the case where the host photoreceptor layer is intact, there is no appreciable donor photoreceptor integration; thus, cell replacement is unlikely to mediate functional rescue in this context. MT of proteins that are missing in acceptor photoreceptors has been documented (MacLaren et al, 2006;Pearson et al, 2012Pearson et al, , 2016, which suggests that MT occurs at levels sufficient to complement single gene deficiencies in photoreceptors and restore function (MacLaren et al, 2006;Pearson et al, 2012). However, MT does not appear to be selective for functional deficiencies, as GNAT1 can transfer to acceptor photoreceptors in Nrl À/À mice (this study), which do not require this protein to function. In addition, we show that several types of cargo can be transferred including cytosolic and nuclear proteins, mRNA and organelles, which raises the question about the selectivity of this process and the function of transferred cargo in acceptor photoreceptors. As shown in this study, MT requires contact between donor and acceptor photoreceptors and thus, if optimized, could represent a therapeutic opportunity for gene complementation and possibly metabolic rescue by mitochondrial transfer. However, successful MT-based therapies will need to be long-lasting. Thus, optimizing donor photoreceptor survival, neurite formation and donor and acceptor photoreceptor contact, as well as regulation of cargo selectivity, are important avenues for future investigation. bottomed 96-well plates, as described above. For transwell experiments, 1 × 10 6 of one fluorescent population of retinal dissociates was cultured on 0.4 lm polycarbonate (PC)-membrane permeable inserts (10769-208, VWR International, LLC. Mississauga, ON, Canada) and 1 × 10 6 of a second fluorescent population on plasticbottomed 12-well plates.

Lentiviral infection
Retinal dissociates from P0-2 mice were resuspended in REM (Appendix Tables S3 and S4) and plated at a concentration of 3 × 10 7 cells in 10 ml of REM into 10 cm Primaria TM dishes (353801, BD Falcon, BD Biosciences, Oakville, ON, Canada). Lentiviruses were added to the plate at a titre to achieve > 80% infection efficiency. All cultures were maintained in 5% CO 2 -buffered incubators (50116048, Heracell TM 150i CO 2 , Thermo Fisher Scientific, Asheville, NC, USA) at 37°C. After 2 DIV, cells were de-adhered with Trypsin-EDTA (0.05% Trypsin with EDTA 4Na) (25300062, Thermo Fisher Scientific, Mississauga, ON, Canada) for 5 min and the enzymatic digestion was stopped with 10% BSA. Cells were washed in Ca 2+ / Mg 2+ -free PBS four times and filtered through a 40-lm cell strainer.
To confirm that GFP expression was not the result of lentivirus carryover in donor cell preparation, for every transplantation experiment we injected 2 ll of the final cell wash to the SRS in three different animals and analysed GFP expression after 21 days. Cells were resuspended at a concentration of 200,000 cells per ll in 0.005% DNAse in EBSS and kept on ice prior to transplantation or cell culture.

Lentivirus design, cloning and production
A list of all plasmids and expression vectors used in this study is provided in Appendix Table S5.

Subretinal injection
Dissociated cells were resuspended at a concentration of 50,000 (human tumour cell lines and cortical primary neurons) or 200,000 (primary retinal cells) cells per ll in 0.005% DNAse in EBSS and kept on ice prior to transplantation. Adult recipient C57BL/6J and Nrl À/À mice (6-8 weeks old) were anaesthetized using a mixture of ketamine (100 mg/ml, Ketalean) (8KET004D, Bimeda MTC Animal Health Inc. Cambridge, ON, Canada) at 50 mg/kg and medetomidine (1 mg/ml, Cepetor) (236 1506 0, Modern Veterinary Therapeutics LLC, Miami, FL, USA) at 1 mg/kg in sterile 0.9% NaCl (JB1300, Baxter Corp. Mississauga, ON, Canada) administered intraperitoneally. Pupils were dilated using 1% tropicamide (Mydriacyl) (0065-0355-03 Alcon, Mississauga, ON, Canada) drops, followed by application of a 0.2% hypromellose gel (Genteal Tears) (0078-0429-57, Alcon, Mississauga, ON, Canada) to maintain proper eye lubrication. For injections, the left eye was gently prolapsed and then immobilized using a customized latex dam to allow free blood circulation. The retina fundus was visualized under a dissection microscope, adjusting the reflexion index, for confirmation of cell deposit. A microscopy-guided microinjection was performed by a second surgeon located orthogonal to the viewing microscope. A scleral incision was made in the dorsal side, posterior to the limbus using a 30-gauge needle (305136, VWR International, LLC. Mississauga, ON, Canada). Next, a blunt 32-gauge needle (7762-05, Hamilton Company, Montreal, QC, Canada) was inserted tangentially into the SRS and advanced under the guidance of the person visualizing the fundus. Once the needle was located in the SRS, a small incision was made in the cornea to relieve intraocular pressure. reversed using an intraperitoneal injection of 1 mg/kg atipamezole (5 mg/ml Revertor) (236 1504 0, Modern Veterinary Therapeutics LLC, Miami, FL, USA). Animals were placed on a heating pad and monitored until fully recovered.

Drug administration
All mice transplanted with human cancer cells were treated with oral cyclosporin (C3662-5MG, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada) ad libitum (50 µg/g/day in drinking water) beginning 3 days before transplantation until the experimental endpoint. To inhibit extracellular vesicle secretion in vivo beginning one day before transplantation, animals were given intraperitoneal injections of the following compounds: GW4869 (D1692-5MG, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada), a neutral sphingomyelinase inhibitor, 2 mg/kg body weight, every 3 days; chloroquine (PHR1258-1G, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada), an autophagy and EV fusion inhibitor, 5 mg/kg body weight, daily; and amitriptyline (A8404-10G, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada), an inhibitor of ectosome secretion, 10 mg/kg body weight, daily. As a control, we injected a fourth group of animals with 0.9% NaCl (vehicle) daily. For in vitro experiments, culture medium was supplemented with GW4869 (20 lM diluted in dimethyl sulfoxide (DMSO) (276855-100ML, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada)), chloroquine (500 nM diluted in REM) and amitriptyline (500 nM diluted in REM). REM and REM-DMSO (same amount) were used as controls. Controls were run on the same days as their respective experimental treatments. To enhance neurite outgrowth in vitro, cultures were supplemented with Y27632 ROCK inhibitor (13624S, Cell Signaling Technologies, Danvers, MA, USA) (100 lM diluted in REM).

Optical coherence tomography imaging
Resolution of the retinal detachment induced by the subretinal cell delivery was monitored over time using a spectral-domain optical coherence tomography (SD-OCT) device (Bioptigen; Morrisville, NC, USA). Briefly, mice were anaesthetized and analysed using SD-OCT before, immediately after, and at 5, 9 and 21 days posttransplantation. Animals were excluded from the study if any complications affecting the opacity of the cornea or lens were observed. Retinal detachment was measured in SD-OCT images using the area of maximum length between the retina and the retinal pigmented epithelium, using callipers provided directly by the software (InVivoVue Clinic 1.2 QOP-60-05 revision C). The measurements correspond to the area of larger detachment in each eye that, consequently, is the area of maximum delivery of the transplanted cells during the injections.

Flow cytometric analysis
To prepare co-cultures for flow cytometric analysis, cells were detached from the wells with 100 ll of TrypLE (12604021, Thermo Fisher Scientific, Mississauga, ON, Canada) for 10 min at 37°C, followed by gentle trituration. Cell suspensions were centrifuged for 5 min at 1300 RPM and resuspended in 2% BSA and 0.005% DNase. To stain photoreceptors, cells were incubated with CD73-APC (0.05 lg/ll) or CD73-PE (127206, BioLegend, San Diego, CA, USA) (0.02lg/ll) for 20 min and then incubated with eFluor780 (501129035 Invitrogen, Thermo Fisher Scientific, Mississauga, ON, Canada) viability marker, 1:5 dilution for 10 min at room temperature (RT). Cells were then centrifuged and fixed in 4% paraformaldehyde aqueous solution (PFA) for 10 min followed by 1:1 addition of 10% BSA. Fixed cells were centrifuged and resuspended in Ca 2+ /Mg 2+ -free PBS and stored at 4°C until analysis. Analysis was performed on BD LSRFortessa X20 (5 laser) and LSRFortessa (4 laser) with high-throughput sampler option (HTS) using FACSDiva software.
Isolation and characterization of extracellular vesicles Extracellular vesicles were isolated from Nrl::GFP retinal dissociates harvested at P4; the retinas from six animals were pooled, dissociated with papain to obtain a single cell suspension and cultured for 24 h in REM at a density of 6 × 10 6 cells/ml in a total volume of 10 ml, in a controlled environment (37°C incubator, 5% CO 2 ). EVs were purified from the culture supernatant using a sequential centrifugation procedure described previously (Thery et al, 2006). In brief, the culture supernatant was centrifuged at 300 g for 10 min to remove the cells. The cell-free supernatant was then centrifuged at 2000 g for 10 min and then at 10,000 g for 30 min to remove dead cells and cell debris, respectively. The resulting supernatant was centrifuged at 100,000 g for 8 h at 4°C (Beckman Type 90 Ti).
The EV pellet was washed in PBS and subjected to another 100,000 g centrifugation for 1 h at 4°C. The supernatant was discarded, and the pellet obtained was resuspended in 50 ll PBS and stored at À80°C.

Tissue processing
Mice were euthanized by carbon dioxide inhalation and transcardially perfused with PBS for exsanguination and 4% PFA for fixation. Eyes were then marked with a silver nitrate stick ( (Nojima et al, 2017) where the incubation times were optimized for the size of the rodent eye. Eyes were incubated in 50% CUBIC-1/H 2 O for 2 h at 37°C before an overnight incubation in CUBIC-1 at 37°C followed by a wash in PBS for 2 h at RT. Eyes were then incubated in 50% CUBIC-2/PBS for 2 h at 37°C before an overnight incubation in CUBIC-2 at 37°C. Excess CUBIC-2 was removed using a Kimwipe, and samples were washed briefly in Type FF immersion oil (16212, Cargille Laboratories Inc. Cedar Grove, NJ, USA) for 10 min at RT before imaging. Whole-mounted retinas were incubated in CUBIC-1 for 5 min at RT before of a wash for 5 min at RT in PBS followed by a 10-min incubation in CUBIC-2 at RT, followed by removal of the excess CUBIC-2 solution and mounting in Type FF immersion oil.

Immunohistochemistry (IHC)
Cell cultures were fixed by adding 4% PFA to the wells for 10 min followed by three washes with PBS. Fixed cultures were blocked in 10% donkey serum (DS) (D9663-10ML, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada) in 0.5% Triton-X (Tx) (X100, T9284-500ML, Merck Millipore Sigma-Aldrich, Oakville, ON, Canada) in PBS for 1 h at RT followed by incubation with primary antibodies (Appendix Table S6) in 5% DS and 0.25% Tx in PBS overnight at 4°C. Cultures were washed three times with PBS and incubated with secondary antibodies (Appendix Ltd. Georgetown, ON, Canada) and 0.05% SDS. Membranes were blocked with 5% BSA in Tris-buffered saline (TBS), followed by an incubation for 1h at RT and overnight at 4°C with primary antibodies, and a subsequent incubation for 2 h at RT with secondary antibodies (Appendix Table S7). Membranes were subjected to three 10-min washes in TBS following primary and secondary antibody incubation.

Confocal imaging
Cell preparations, cryostat tissue sections and whole-mounted retinas were imaged using an LSM 780 (Carl Zeiss Inc. Thornwood, NY, USA) confocal microscope. Images were acquired at a minimum of 2,048 × 2,048 pixels resolution at a maximum of 5% laser intensity and a maximum 800 digital gain. 20×, 40× water immersion and 63× oil immersion objectives were used at a 1.0 Airy Unit pinhole size. Comparative images were taken using identical acquisition parameters. Live imaging was performed in a 37°C, 5% CO 2 equilibrated chamber. Images were acquired at intervals varying from 1 s to 5 min. Intact cleared eyes were imaged using the Ultramicroscope II light sheet microscope (LaVision BioTec To avoid including the distal tips of the protrusions in the measurement of fluorescence recovery, the ROI registered to monitor fluorescence recovery in the photobleached cell was smaller than the ROI used for photobleaching. Intensity values were normalized to the mean pre-photobleaching fluorescence intensity and the intensity value immediately post-photobleaching was used to normalize and represent all post-photobleaching values in terms of fluorescence recovery in each ROI. Statistical analyses were performed with an extrasum-of-squares F-test (Y0 constrained to 0.0) after fitting the curve in a non-linear regression with two-phase association.
Image processing Two-photon and light sheet fluorescence microscopy images were post-processed in Imaris 9.0.2 (Bitplane Inc. Zurich, Switzerland). When appropriate, 3D images were processed using a background subtraction, followed by a normalization algorithm before making adjustments to brightness and contrast in Surpass mode. Semiautomatic quantification of neurite length was performed using the Filaments module in Imaris 9.0.2 to trace GFP + protrusions. Neurites less than 5 µm in length were excluded from data analysis.

Quantification of Material Transfer index
The quantification of GFP + cells in transplanted retinas was performed, as previously described. Briefly, the number of GFP + cell bodies in the recipient outer nuclear layer (ONL) and the number of GFP + cells in the adjacent SRS were counted. The total number of GFP + cells per eye was determined by extrapolation, based on quantification of every 4 th section. Animals were excluded from the analysis if they showed lack of a cell deposit in the SRS or had evidence of a fulminant immune reaction, because macrophage infiltration is associated with reduced GFP exchange (West et al, 2010). We observed, by way of Pearson coefficient and linear regression analyses, a strong correlation between the number of GFP + cells in the recipient ONL (corresponding to recipient cells that received transferred GFP) and the number of GFP + cells in the SRS. Thus, we normalized the number of GFP + cells in the ONL to the number of GFP + cells in the SRS and refer to this parameter as material transfer index. In the Crx::Cre transplantation into ROSA mT/mG mice, mTdTomato switches to mGFP when Cre recombinase is transferred from the donor to the recipient. In these experiments, the number of mGFP + cells in the ONL was quantified by counting all GFP + cells where the entire cell was located in the recipient ONL.

Quantification using flow cytometry
Retinal cells were separated from debris through FSC-A vs SSC-A, and then singlets were discriminated by width vs height for both forward and side scatter. Live eFluor780 À cells were selected and further gated for photoreceptors by CD73 + . Biaxial plots were used to identify the GFP + vs tdTomato + populations or MTG + vs MTR + populations. Fluorescence minus one -controls were used to identify gates for each fluorophore. Single fluorescence controls of the same cell type analysed were used to calculate compensation for each individual experiment. At least 1,000 live cells and 600 CD73 + cells were analysed for each technical replicate. Data analysis indicates the percentage of double-positive cells (Q2) from total fluorescent eFluor À : CD73 + population (Q1+Q2+Q3). Each plotted value represents a triplicate mean for one biological replicate. At least three biological replicates were analysed for each condition. For pharmacological manipulations, the frequency of MT in treated co-cultures was normalized to untreated control conditions from the same experimental day.

Quantification of donor and acceptor photoreceptor connections in vivo
Donor-acceptor photoreceptor connectivity was evaluated from confocal, two-photon and light sheet microscopy images using Imaris 9.0.2 software. Whole transplanted eyes or whole-mounted retinas from transplanted eyes were imaged and GFP + acceptor 20 of 23 The EMBO Journal 40: e107264 | 2021 ª 2021 The Authors photoreceptors in the recipient ONL were identified. These GFP + photoreceptors were then classified into one of three groups based on the presence of a GFP + protrusion connecting donor and acceptor photoreceptors: (i) contiguous, defined as donor and acceptor connections with a gap of less than 3 µm, (ii) proximal, defined as donor and acceptor connected with a gap of between 3 and 5 µm, and (iii) not contiguous, defined as cells connected by a gap of greater than 5 µm. For graphical representation of the data, we compiled two categories, contiguous and proximal, as positive for contact.

Quantification of protrusions in live photoreceptor cultures
A systematic sampling of three images per dish was performed in 8 Nrl:GFP culture preparations after 3 DIV. A total of 24 3Dreconstructed images were analysed on Imaris 9.0.2 (Bitplane Inc., Zurich, Switzerland). Twelve of them were stained with SiR-Actin (100 lM) (CY-SC001, Cytoskeleton Inc. Denver, CO, USA) and 12 with ViaFluor â (50 lM) (70062, Biotium, Fremont, CA, USA). We identified a total of 590 and 278 GFP + protrusions extending from photoreceptors in actin and tubulin-stained cultures, respectively. Each protrusion was visually assessed for cytoskeletal marker positivity and labelled accordingly using the Spots module of the software. Then, by visual inspection, protrusions were manually classified as (i) connected, e.g., connecting two cells, or unconnected; (ii) suspended in the media or attached, e.g., in contact with the bottom of the dish at any point along their length; and (iii) distal tip morphology, e.g., straight or terminating with distal swelling (spherule-like). All data were collected and organized in Microsoft Excel (Office Excel, version 16.35, 2020; Microsoft Corp., Redmond, WA, USA) and analysed in GraphPad Prism version 7.0a sheets (GraphPad Software Inc. La Jolla, CA, USA).

Quantitation of protrusion outgrowth from transplanted donor photoreceptors in vivo
To quantify neurite length in transplanted donor photoreceptors, we used the Filaments module in Imaris 9.0.2 (Bitplane Inc., Zurich, Switzerland) to trace GFP + protrusions. Protrusions measuring less than 5 µm in length were excluded from the analysis. Average protrusion length was defined as the sum of the length of all the protrusions normalized to the total number of protrusions. Total protrusion length per cell was defined as the sum length of all the protrusions normalized to the number of cells with a protrusion.
Immunoblot densitometry Densitometric analysis of the protein was performed using ImageJ (National Institutes of Health, Bethesda, Maryland, USA). Briefly, the protein of interest was quantified by normalizing to GAPDH controls and the normalized protein amounts were compared between treatment groups.

Statistics and reproducibility
Experiments in this study were performed independently at least two times (in vivo) and at least three times (in vitro). No inconsistent results were excluded. Flow cytometric analyses of cultured retinal dissociates were performed on at least three independent biological replicates and with at least two technical replicates per condition, and samples were run on the same cytometers to reduce inter-experimental variability. All the data were collected and organized in Microsoft Excel spreadsheets (Office Excel, version 16.35, 2020; Microsoft Corp., Redmond, WA, USA). All the graphical representation and statistical analysis were performed in GraphPad Prism version 7.0a sheets (GraphPad Software Inc. La Jolla, CA, USA). Data are represented as mean AE SEM. Statistical analyses were performed using a one-way ANOVA with Tukey's multiple comparison test, extra-sum-of-squares F-test or unpaired t-test where appropriate. The P-values are represented as n.s. not statistically significant, ****P ≤ 0.0001, ***P < 0.001, **P < 0.01 and *P < 0.05.

Data availability
All data analyses for this study are provided in this manuscript and supplementary information files. This study includes no data deposited in external repositories.
Expanded View for this article is available online.