Context-dependent compensation among phosphatidylserine-recognition receptors

Phagocytes express multiple phosphatidylserine (PtdSer) receptors that recognize apoptotic cells. It is unknown whether these receptors are interchangeable or if they play unique roles during cell clearance. Loss of the PtdSer receptor Mertk is associated with apoptotic corpse accumulation in the testes and degeneration of photoreceptors in the eye. Both phenotypes are linked to impaired phagocytosis by specialized phagocytes: Sertoli cells and the retinal pigmented epithelium (RPE). Here, we overexpressed the PtdSer receptor BAI1 in mice lacking MerTK (Mertk −/− Bai1 Tg) to evaluate PtdSer receptor compensation in vivo. While Bai1 overexpression rescues clearance of apoptotic germ cells in the testes of Mertk −/− mice it fails to enhance RPE phagocytosis or prevent photoreceptor degeneration. To determine why MerTK is critical to RPE function, we examined visual cycle intermediates and performed unbiased RNAseq analysis of RPE from Mertk +/+ and Mertk −/− mice. Prior to the onset of photoreceptor degeneration, Mertk −/− mice had less accumulation of retinyl esters and dysregulation of a striking array of genes, including genes related to phagocytosis, metabolism, and retinal disease in humans. Collectively, these experiments establish that not all phagocytic receptors are functionally equal, and that compensation among specific engulfment receptors is context and tissue dependent.


BAI1-transgene reduces corpse accumulation in the testes of Mertk −/− mice.
To determine whether distinct PtdSer receptors play unique roles in the process of engulfment, we designed a genetic approach to determine whether the overexpression of one receptor could rescue for the loss of another. We crossed MerTK null (Mertk −/− ) and BAI1-overexpressing (Bai1 Tg ) mice to generate Mertk −/− Bai1 Tg mice. The rationale for choosing Mertk −/− mice is that they have two in vivo phenotypes associated with impaired phagocytosis. First, Mertk −/− mice exhibit accumulation of apoptotic germ cells in the testes due to impaired Sertoli cell phagocytosis 10,31 . Second, these mice exhibit profound retinal degeneration due to failed clearance of photoreceptor outer segments (POS) by the RPE 2,14,32 . We elected to overexpress Bai1 in an attempt to rescue the phenotypes in Mertk −/− mice, as BAI1 overexpression can enhance PtdSer-dependent apoptotic cell clearance by multiple phagocytes, including intestinal epithelial cells and Sertoli cells. Furthermore, Bai1 Tg mice had been previously generated and characterized 1,5,18,33 .
Sertoli cells are the specialized phagocytes of the testes and promote routine phagocytosis of apoptotic germ cells. Sertoli cells utilize both BAI1 and MerTK during the phagocytosis of apoptotic germ cells 1,10 . Mertk −/− mice exhibit accumulation of apoptotic corpses at baseline 10 and Bai1 −/− mice exhibit apoptotic corpse accumulation following testicular torsion 1 . Furthermore, Bai1 Tg mice exhibit a decrease in corpse accumulation following torsion 1 . Given the endogenous role for MerTK and BAI1 in the testes, we initially tested PtdSer receptor compensation in Mertk −/− Bai1 Tg mice in the context of Sertoli cells.
Prior to evaluating apoptotic cell accumulation in the testes, we confirmed that Sertoli cells endogenously express Mertk, Bai1 and components of the BAI1 signaling pathway: Elmo1, Dock180 and Rac1 (Fig. 1A) 4 . Importantly, expression of Bai1, Elmo1, Dock180 and Rac1 did not differ between Mertk +/+ and Mertk −/− mice (Fig. 1A). In addition, we confirmed that the Bai1 Tg was expressed by Sertoli cells and that the BAI1-Tg properly localized to the surface of Mertk −/− Sertoli cells (Fig. 1B). To determine whether Bai1 overexpression could rescue phagocytic deficits in Mertk −/− Sertoli cells, we evaluated apoptotic corpse accumulation in testes that had undergone surgical torsion and those that underwent sham surgery 6,8 . Corpse accumulation was quantified by counting cleaved-caspase 3 positive cells in testicular cross-sections (Figs 1C, S1). As was previously reported, Mertk −/− mice exhibited a slight but significant increase in apoptotic corpse accumulation at baseline (Fig. 1C,D) 10 . While the Mertk −/− Bai1 Tg mice trended towards fewer apoptotic corpses at baseline, this was not statistically significant (Fig. 1D). Analysis of injured testes revealed that Mertk −/− mice had substantially more apoptotic cell accumulation than control littermates (Fig. 1C,D). Interestingly, corpse accumulation in Mertk −/− Bai1 Tg mice was significantly decreased compared to Mertk −/− mice. In fact, the corpse numbers in Mertk −/− Bai1 Tg mice were reduced to the number in Mertk +/+ mice (Fig. 1C,D). These data suggest that in the context of testicular torsion, overexpression of Bai1 can reduce the number of apoptotic corpses that seem to accumulate in Mertk −/− Sertoli cells.

RPE express the Bai1
Tg and BAI1 signaling pathway in Mertk −/− Bai1 Tg mice. Previous studies have established that POS expose PtdSer that is subsequently recognized by the RPE 13 and Mertk −/− mice exhibit profound retinal degeneration linked to failed clearance of POS by the RPE 14,15 . While Mertk is highly expressed by the RPE 14,15,21 , RPE do not express BAI1 endogenously ( Figure S2A). We first confirmed that downstream components of the BAI1 signaling pathway are present within the RPE. Isolated RPE from Mertk +/− and Mertk −/− mice were analyzed by RT-PCR and immunoblotting. These analyses showed expression of the BAI1 signaling components ELMO, Dock180 and Rac1 by RT-PCR ( Fig. 2A,B). Importantly, we found that ELMO2, Dock180 and Rac1 were all expressed. Furthermore expression of BAI1 signaling components did not differ between RPE from Mertk +/− and Mertk −/− mice as assessed by both RNA and protein ( Fig. 2A,B), suggesting that BAI1 could theoretically function in the RPE.
We next assessed Bai1 Tg expression and localization within the RPE. The Bai1 Tg mouse construct contains three separate methods to detect expression in situ (Fig. 2C). First, the Bai1 Tg construct includes an IRES-driven GFP, which allows visualization of the transcriptional activity at the Bai1 Tg locus. To assess GFP expression in RPE, we prepared RPE flat mounts from Mertk +/− Bai1 Tg and Mertk −/− Bai1 Tg mice and imaged GFP by confocal microscopy (Fig. 2D). The GFP expression pattern appears uniform across the RPE layer in both Mertk +/− Bai1 Tg and Mertk −/− Bai1 Tg mice. On average, 45% of the area in each 20x field is GFP positive (data not shown). Importantly, no GFP signal was detected in mice lacking the Bai1 Tg (Fig. 2D). Second, the Bai1 Tg is derived from the human Bai1 cDNA 1 . Despite being highly homologous to mouse BAI1 at the protein level and indistinguishable in functional assays 14,18 , human Bai1 transcript can be distinguished from the murine transcript by RT-PCR. Human Bai1 was readily detected in the RPE of Mertk +/− Bai1 Tg and Mertk −/− Bai1 Tg but not littermate control mice (Fig. 2E). Third, the Bai1 Tg construct has an N-terminal HA-tag, which facilitates detection of the Bai1 Tg protein. The HA-BAI1 protein was readily detected in the extracts of isolated RPE cells by immunoblotting (Fig. 2F). RPE cells are highly polarized and the apical microvilli mediate the phagocytosis of photoreceptor outer segments. Therefore, to assess whether the Bai1 Tg protein was properly localized on the apical surface in a location similar to MerTK 21 , we stained for the HA-tag in eyecup cross sections. HA-BAI1 showed apical localization and co-localized with ezrin, a cytoskeletal protein enriched in microvilli (Fig. 2G). Furthermore, HA-BAI1 was detected on the apical RPE surface of both Mertk +/− Bai1 Tg and Mertk −/− Bai1 Tg mice. These data suggested that RPE cells in the Mertk −/− Bai1 Tg mice express the Bai1 Tg , and that the BAI1 protein localizes to the region of the RPE that mediates phagocytosis. To examine whether there are signals initiated by the BAI1 transgene within the RPE, we performed RNAseq on Mertk +/+ and Mertk +/+ Bai1 Tg mice to identify genes likely altered by the Bai1 Tg . Significant transcriptional changes were noted in 63 genes from the Bai1 Tg mice. When we focused on processes previously associated with BAI1, such as cholesterol homeostasis, axonal growth and synaptogenesis [33][34][35] , we identified 10 genes with functions that are associated with the aforementioned processes ( Figure S2B). These data suggest that BAI1 is not only expressed by RPE cells at the correct location, but can also signal to induce transcriptional changes in RPE cells.
Mertk-linked retinal degeneration is not rescued by BAI1. We next assessed the retinal degeneration in Mertk −/− Bai1 Tg mice. Photoreceptors in Mertk −/− mice begin to show signs of overgrowth at post-natal day 17 and exhibit highly disorganized POS by 35 days after birth 14,21 . Degeneration of the photoreceptor layer begins soon afterwards and previous reports have demonstrated that most photoreceptors are lost by 12 weeks of age 14,21,26 . To assess retinal degeneration, we collected sagittal sections of eyes that transect the optic cup (where the optic nerve meets the retina) (Fig. 3A,B). Inspection of central retinas from 12 week-old Mertk −/− and Mertk −/− Bai1 Tg revealed equivalent degeneration of the outer nuclear layer (ONL) consisting of photoreceptor nuclei. Importantly, Mertk +/− Bai1 Tg retinal cross sections exhibited normal retinal architecture suggesting that the retinal structure was not adversely affected by transgenic Bai1 expression ( Fig. 2A).
ONL thickness is not equivalent across the retina and degeneration is not necessarily homogeneous 26,27 . To standardize the measurement of ONL across the entire section, we adapted a previously described technique 26 and measured the ONL thickness at 20 standardized points across the 'inferior-superior' axis of the retinal section (Fig. 3B). We measured ONL thickness in mice at 8 and 12 weeks of age ( Fig. 3C). At 8 weeks of age, both Mertk −/− and Mertk −/− Bai1 Tg mice exhibited moderate degeneration across the entire inferior-superior axis (Fig. 3C). By 12 weeks of age, retinal degeneration had progressed in both Mertk −/− and Mertk −/− Bai1 Tg mice (Fig. 3C). As previously reported 26,27 , degeneration was more severe in the superior than the inferior retina ( Fig. 3D,E). Importantly, degeneration was equivalent between Mertk −/− and Mertk −/− Bai1 Tg animals. Collectively, these data indicate that overexpression of Bai1 did not rescue retinal degeneration due to loss of Mertk.

Phagocytosis defect in MerTK-null mice is not rescued by the BAI1 transgene.
Retinal degeneration in Mertk −/− mice is attributed to decreased phagocytosis of POS. Therefore, we analyzed RPE phagocytosis of POS in Mertk −/− and Mertk −/− Bai1 Tg mice. Phagocytosis by RPE in vivo is readily analyzed due to its temporal regulation (around light onset) and because the content of the phagosomes (rhodopsin) can be visualized 14,30 . The rate and amount of phagocytosis is a dual function of phagocyte efficiency and the ratio of targets to phagocytes. Since Mertk −/− mice exhibit photoreceptor degeneration, the number of POS targets decreases as degeneration progresses. Therefore, decreases in RPE phagocytosis after the onset of degeneration could be due to the inefficient uptake by the RPE or a decreased target to phagocyte ratio. To ensure that any observed changes in phagocytosis were due to RPE phagocytic efficiency, we analyzed mice prior to the onset of retinal degeneration, at P17-P21 days of age. Importantly, analysis of ONL thickness at this age confirmed that degeneration in Mertk −/− and Mertk −/− Bai1 Tg mice had not yet begun (Fig. 4A).
To quantify phagocytosis, we harvested eyes 1 h after light onset and immunostained sagittal-eye sections for rhodopsin ( Fig. 4B, white arrows). Importantly, rhodopsin puncta are degraded over time and thus are mostly absent 8 h after light onset, suggesting that these puncta are not artifacts of sectioning or staining (Fig. 4B). Quantification of puncta at 1 h post-light onset revealed a striking decrease in rhodopsin puncta in the RPE of Mertk −/− mice relative to Mertk +/− mice. However, neither Mertk +/− Bai1 Tg nor Mertk −/− Bai1 Tg mice exhibited differences in the number of puncta compared to their respective controls (Fig. 4C). Furthermore, neither the loss of Mertk nor overexpression of BAI1 affected phagosome trafficking in RPE, as indicated by basolateral localization of phagosomes ( Figure S4). BAI1 overexpression was previously shown to enhance phagocytosis in WT epithelial cells (both in vitro and in vivo) 1 . However, BAI1 was unable to enhance phagocytosis in Mertk +/+ mice and importantly, the number of puncta in Mertk +/+ mice were equivalent to the number seen in Mertk +/− ( Figure S3D). Collectively, these data suggest that Bai1 Tg expression is unable to rescue the RPE phagocytic defect observed in Mertk −/− mice.
MerTK regulates the gene expression program in the RPE. The lack of rescue by BAI1 prompted us to ask whether MerTK regulates the phagocytic capacity of RPE in ways beyond PtdSer binding. To address this question, we chose an unbiased transcriptomics approach, and performed RNAseq analysis of RPE from Mertk +/+ and Mertk −/− mice. RPE were isolated at day 14 after birth, a time when the mouse pups have opened their eyes but Mertk −/− mice do not yet exhibit photoreceptor overgrowth that could impact RPE gene expression 14,21 .
RNAseq analysis identified 60 genes that were differentially regulated in Mertk −/− RPE (Fig. 5A). Further analysis revealed that 11 of these genes have annotated functions related to either cytoskeletal rearrangement or endosomal maturation, two processes that are essential for phagocytosis (Fig. 5B). The gene fibulin 7 (Fbln7), important for extracellular matrix adhesion and cytoskeletal rearrangement [36][37][38] , has a SNP associated with age-related macular degeneration and its expression is altered in patients with retinoschisis, a form of inherited retinal dystrophy 39,40 . qRT-PCR validation of the original library and an additional cohort of Mertk +/+ and Mertk −/− mice confirmed that Fbln7 is upregulated in Mertk −/− RPE (Fig. 5C). Phagosome maturation culminates in acidification, which promotes the breakdown of the phagosome content 41 . Chloride flux across the lysosome membrane regulates lysosome acidification 42,43 . Therefore we further assessed the chloride channel anoctamin 1 (Ano1) and the bicarbonate transporter Slc4a4 for further validation. Although Ano1 has yet to be implicated in the chloride flux across lysosomal membranes, it is a well-established calcium-activated chloride transporter 44 and RPE phagocytosis is accompanied by substantial calcium accumulation in the RPE 45 . Slc4a4 is a bicarbonate transporter that is thought to regulate intracellular pH by electrogenic flux of H + and HCO 3 −46 . Interestingly, it has been proposed that these ion transporters work together in certain cellular processes 47 . We validated that Slc4a4 and Ano1 were reproducibly downregulated in both the original RPE samples as well as in an additional cohort of RPE (Fig. 5D). These data suggest that the presence of MerTK can regulate the expression of various components of the phagocytic machinery, either directly or indirectly.
Phagocytosis is a metabolically demanding process for the phagocyte that requires significant ATP generation to mediate cytoskeletal rearrangement and digestion of cargo 48,49 . In addition to annotating multiple genes involved in phagocytosis, we identified an additional 11 genes related to cell metabolism (Fig. 5B). Although these genes we identified have yet to be directly linked to phagocytic processes, the number of genes that were differentially regulated was intriguing. We validated the fatty acid elongase, Elovl1 and the acetyl CoA synthetase, Acss2, and found that the expression changes were reproducible, in both the original library and a fresh cohort of Mertk +/+ and Mertk −/− mice (Fig. 5E). Overall, these data suggest that MerTK might coordinate the phagocytic process at multiple levels from PtdSer binding, to cytoskeletal reorganization and phagosome maturation, as well as coordinating metabolic changes during phagocytosis (and potentially the visual cycle).
In addition, 9 genes identified from the RNAseq screen were linked to human retinal disease ( Table 1). Five of these genes were also annotated under the metabolic subset (marked with*) and three were annotated in the phagocytosis subset (marked with #) (Fig. 5B, Table 1). Two genes, Gja1 and Cyp27a1 have a striking monogenic association with oculodentodigital dysplasia and cerebrotendinous xanthomatosis respectively, both of which are associated with retinal abnormalities 50,51 . Three genes had SNP associations to retinal disease: Ldlrad3, Fads2, and Fbln7 ( Table 1). Five of the genes we annotated as being linked to retinal disease, were identified in the comparative toxicogenomics database, which infers gene-disease associations by examining curated associations between pharmacological agents, known diseases and the gene of interest 52,53 . These associations were further validated by experimental links to retinal disease or RPE function. Slc4a4 has also been linked to retinitis pigmentosa (RP) type 17 and Elovl1 is an endogenous inhibitor of the visual cycle enzyme RPE65, defects in which cause RP type Rhodopsin puncta in the RPE, referred to as phagosomes, were quantified in ImageJ by an automated particle count. Particle size was constrained from 0.5 μm to ∞ and minimum circularity was restricted to 0.2. Symbols within the bars represent the average particle count from the left and right eyes of one mouse. Mertk +/− (n = 5) Mertk +/− Bai1 Tg (n = 8) Mertk −/− (n = 8) Mertk −/− Bai1 Tg (n = 6). Error bars represent SEM. Significance was determined by one-way ANOVA. Multiple comparison analysis was corrected with a post-hoc Tukey's test.
MerTK −/− mice exhibit diminished retinyl ester accumulation prior to the onset of degeneration. In addition to promoting photoreceptor survival, RPE contribute indirectly to phototransduction by participating in the visual cycle, which supplies photoreceptors with the chromophore 11-cis-retinal. The visual or retinoid cycle is the recycling of vitamin A (all-trans-retinol) to the light-reactive chromophore 11-cis-retinal (Fig. 6A) 32 . Interestingly, one of the first characterizations of Mertk −/− mice revealed that vision is affected in mice as early as P20, prior to the onset of retinal degeneration 14 . Given the broader changes in gene expression that we observed in Mertk −/− RPE, we investigated whether Mertk expression influenced the visual cycle.
High-performance liquid chromatography was used in a blinded assay to quantify all-trans-retinyl esters and 11-cis-retinal in the eyes of dark-adapted mice at P21 (prior to degeneration but after significant retinal maturation) 14,21,57,58 . Retinyl esters are notoriously difficult to extract and quantify, thus we developed a novel procedure that allows extraction and quantification of the esters with more than 90 percent accuracy 59 . Interestingly, we found that all-trans-retinyl ester levels were significantly lower in Mertk −/− relative to WT mice (Fig. 6B). However, there was no difference in the levels of the chromophore 11-cis-retinal at this time point (Fig. 6C). Retinyl esters are a form of stored vitamin A; esterification to fatty acids prevents their release in to the extracellular space 60,61 . The shifts we observed in metabolic gene expression could influence the availability of fatty acids for esterification, possibly affecting retinyl ester accumulation. Diminished retinyl ester levels in Mertk −/− mice prior to the loss of photoreceptors support the hypothesis that Mertk can influence critical processes beyond phagocytosis, and that MerTK is uniquely critical for RPE function.

Discussion
Phagocytes express multiple PtdSer receptors and it has been unclear whether these receptors are functionally redundant or if they have tissue and context specific functions. In an attempt to determine if PtdSer receptors are functionally unique, we asked if PtdSer receptors are capable of compensating for one another in specialized phagocytes in vivo. Our analysis of the Mertk −/− Bai1 Tg mice led to two major findings. First, Bai1 Tg expression appears to rescue phagocytic deficits in Mertk −/− Sertoli cells. Second, Bai1 Tg expression was not able to prevent retinal degeneration or rescue phagocytic deficits in the Mertk −/− RPE. These findings provide the first evidence for functional compensation between two different PtdSer phagocytosis receptors from two distinct receptor families: The adhesion family GPCR, BAI1 and the receptor tyrosine kinase MerTK. Our results also suggest that the role of a given PtdSer receptor in cell clearance may differ depending on the tissue context. Overall, this indicates that PtdSer receptors are not fully interchangeable, which could, in part, explain why phagocytes express a variety of PtdSer receptors on their cell surface. It also suggests that expression of MerTK is non-redundant in the eye/retina.
To understand what unique role(s) MerTK might have in the RPE, we performed RNAseq on Mertk −/− and Mertk +/+ RPE and identified 60 genes that were differentially expressed in Mertk −/− RPE cells. From those genes, three critical subsets were dysregulated: genes linked to retinal disease, phagocytosis, and metabolism. Genes linked to human retinal disease have associations that are either monogenic, SNP, or inferred based on curated experimental and pharmacological data. Interestingly, 8 of the 9 genes associated with retinal disease were linked to either metabolic or phagocytic functions. These perturbations could explain why Mertk −/− mice exhibit visual defects (as measured by electroretinogram recordings) at P20, prior to the onset of retinal degeneration 14 . Furthermore, the observed transcriptional changes may help to explain the decreased retinyl ester accumulation in Mertk −/− mice as our data revealed significant changes in metabolic genes that might perturb the availability of fatty acids for esterification.
The finding that Bai1 overexpression can reduce apoptotic corpse accumulation in the testes is intriguing. Though RPE phagocytosis of POS is PtdSer dependent, it is not 'traditional apoptotic corpse clearance' and could instead be analogous to the cell pruning observed in the nervous system. It is possible that the mechanism of action of BAI1 in phagocytosis is not conducive to pruning-type events, but rather that BAI1 is more suited for larger corpse uptake. Alternatively, BAI1-mediated phagocytosis could require an as-of-yet unidentified co-factor that RPE cells do not express. This might explain why the Bai1 Tg can enhance phagocytosis in Mertk −/− Sertoli cells, where it is known to function endogenously 1,62 . Overall, this work suggests that the variety of PtdSer receptors did not evolve as a simple redundancy mechanism. Rather, these receptors likely play unique roles beyond PtdSer recognition, including the regulation of  genes crucial to the process of phagocytosis and biological processes essential for general homeostasis and even specialized tissue functions. Overall, this suggests that PtdSer receptors should not be regarded as having homogenous functionality. Further research into the downstream functions of PtdSer receptors is critical to understanding how the phagocytic process is coordinated in different tissues and contexts.
Experimental Procedures. Mice. Mertk −/− mice were purchased from Jackson Laboratories (stock no: 011122 -B6;129-Mertk tm1Grl /J). Bai1 Tg mice were previously generated by our lab on a C57Bl/6 N background and then backcrossed with C57Bl/6 J mice 1 . The Bai1 Tg mice were screened for the RD8 mutation and were confirmed to be RD8 negative 63  RPE isolation for gene expression analyses. The RPE isolation protocol was previously described 64 . Eyes were enucleated from P14 neonatal mice 2 hours after light onset. Globes were incubated in serum free DMEM (Corning) with 2% dispase (Worthington). Following incubation, eyes were washed 3 times in DMEM supplemented with 10% fetal bovine serum (FBS) (Gemini). Cornea, iris and lens were removed from each eyecup. The eyecups were incubated for 20 minutes at 37 °C in DMEM supplemented with 10% FBS. Following incubation, the neural retina was removed from each eyecup. The eyecup was tugged at opposite ends to release the RPE layer. RPE sheets were washed in Ca 2+ /Mg 2+ free HBSS (Gibco). RPE to be used for RT-PCR or RNAseq were lysed in 350 μL of RLT buffer (Qiagen). RPE to be used for protein analysis were lysed in 40 μL of RIPA buffer.
Eyecup dissection and ONL analysis. Central corneas of enucleated eyes were punctured with a 25-gauge needle. Eyes were submerged in Hartman's fixative (Sigma) and incubated for 1-3 h at room temperature. Following initial fixation, the cornea, iris and lens were removed. Eyecups were returned to fixative and incubated overnight at 4 °C. Eyecups were moved to 30% sucrose for cryo-protection and incubated at 4 °C until they sank. Eyecups were embedded in O.C.T. (Tissue-Tek) and flash frozen in an isobutane dry ice bath. Consistent orientation of the 'nasal-notch' during embedding was maintained to control eyecup orientation. Sagittal eyecup sections were cut at 10 μm thickness and sections transecting the optic cup were collected. Sections for ONL analysis were stained with Mayer's haematoxylin (Sigma) and eosin (Fisher) and tiled at 20x magnification on an Axio Imager. z1 (Zeiss) with Stereo Investigator software (MBF Biosciences). An image mask with 20 fixed measurement points was applied to eyecup images in Photoshop (Adobe). The ONL was measured at points indicated by the mask in ImageJ software (NIH).
In situ rhodopsin analysis. P17-P21 mice were euthanized 1 h after light onset. Eyes were dissected, fixed and sectioned as described above. Eye sections were blocked in PBS (Corning) with 10% normal goat serum (Jackson Immunoresearch) for 1 h at room temperature. Sections were stained overnight with rhodopsin antibody diluted 1:500 (Abcam clone Rho 4D2). AF647 conjugated secondary antibody was used to detect rhodopsin. Tiled images were acquired at 40x magnification on an Axio Observer.z1 microscope (Zeiss). The RPE layer was isolated in Photoshop (Adobe). Quantification of puncta was performed by automated particle count in ImageJ (NIH).
Flat mount preparation and staining. Mice were euthanized between 4 and 6 weeks of age. Eyes were enucleated and the cornea, iris, lens and neural retina were removed. Eyes were cut at four points for flat mount 'clover' preparation and fixed for 1 h in PBS 4% paraformaldehyde (PFA) at room temperature. Flat mounts were blocked and permeabilized in PBS 10% normal horse serum (Hyclone) and 0.1% Tween 20 then stained over night at 4 °C with anti-GFP antibody at 1:50 (Abcam #6673). Secondary antibody conjugated to AF488 was used to detect GFP staining. Flat mounts were then stained with anti-ZO-1 at 1:250 (Thermo Fisher #61-7300) for 1 h at room temperature. Secondary antibody conjugated to AF647 was used to detect ZO-1 staining. Images were acquired at 20x magnification on a confocal Axio Observer.z1 microscope (Zeiss).
Eyecup staining for HA. Mice were euthanized between 4 and 6 weeks of age. Eyes were enucleated and the cornea, iris, lens, and neural retinas were removed. Eyecups were fixed for 1 h in PBS 4% PFA at room temperature then moved to 30% sucrose for cryo-protection and incubated at 4 °C until the eyecups sank. Eyes were embedded in O.C.T. Eyecups were sectioned at 10 μm and sections bisecting the optic cup were collected. Sections were blocked in PBS with 10% normal goat serum for 1 h at room temperature. Sections were stained overnight at 4 °C with anti-HA antibody 1:100 (Cell Signaling Technology clone C29F4). AF647 conjugated secondary antibody was used to detect staining (ThermoFisher). Sections were then stained with biotin conjugated anti-ezrin antibody 1:100 (ThermoFisher clone 3C12) overnight at 4 °C. AF488 conjugated streptavidin was used to detect ezrin staining (ThermoFisher). Images were acquired at 40x magnification on an Axio Imager.z2 microscope with Apotome (Zeiss).
Testicular torsion. Torsion surgery was performed as previously described 8 . The studies were performed in accordance with the 'Guiding Principles of the Care and Use of Research Animals' promulgated by the Society for the Study of Reproduction. The male mice were anesthetized with an intraperitoneal (IP) injection of a mixture of 6 mg /100 g of ketamine and 0.5 mg/100 g of xylazine. After the testes were exteriorized through a low ventral midline incision, the testes were released from the epididymo-testicular membrane through incising the gubernaculums. For torsion, the testis of right side was rotated 720° for 2 h, during which time it remained in the abdomen with a closed incision. For sham control, the testis was freed of the epididymo-testicular membrane and left in the abdomen. After 2 h, the incision was reopened, the testis was counter-rotated to the natural position, the gubernaculum was rejoined, the testes were reinserted into the scrotum, and then the incision was closed. 24 h after operation, mice were euthanized. The testes were removed and fixed for 6 h in Bouin's fixative for paraffin embedding. Testicular cross sections were stained for cleaved caspase-3 at the University of Virginia Biorepository and Tissue Research Facility. Testicular cross sections were imaged at 20x with an Aperio Scanscope at the University of Virginia Biomolecular Analysis Facility and Shared Instrumentation Core. The number of apoptotic cells per tubule was determined for the entire cross section.
Sertoli cell isolation and staining. Testes from P11-P21 day-old mice were isolated and decapsulated. Tubules were dispersed in a solution of HBSS, 0.0625% Trypsin (Corning) 10 μg/mL DNase (Sigma) for 20 min. at 37 °C then 0.00625% Soybean trypsin inhibitor (SBTI) (Sigma) was added and the supernatants were decanted. Tubules were re-suspended in HBSS with 1 M glycine, 2 mM EDTA, 0.0625% SBTI, and 10 μg/mL DNase and incubated for 10 min. at room temperature. Suspensions were spun at 1000 RPM. Tubules were minced and re-suspended in HBSS with 1 mg/mL collagenase and 10 μg/mL DNase in a shaking water bath at 37 °C for 30 minutes. Tubule fragments were allowed to sediment at room temperature and re-suspended in HBSS with 15 U/mL hyaluronidase (Sigma) and 10 μg/mL DNase and incubated in a shaking water bath at 37 °C for 30 min. Tubule fragments were centrifuged at 1000 RPM. Pellets are washed and re-suspended in F12/DMEM (Corning) with 10% FBS, 1% PSQ 1% (Corning), sodium pyruvate (Corning), 10 mM HEPES (Corning), 5 μg/mL Transferrin (Sigma), 2.5 ng/mL epidermal growth factor (Gibco), 10 μg/mL insulin (Gibco). Media was changed after 24 h of culture to remove floating germ cells. Sertoli cell cultures were passaged after 3-5 days to chamber slides for staining. Sertoli cells were fixed in PBS with 4% PFA for 30 min. at room temperature, blocked in PBS with 10% normal goat serum (Jackson Immunoresearch) and stained overnight in anti-BAI1 antibody 1:100 (R&D #AF4969).
Retinoid analysis. P20-P21 mice were dark adapted overnight. Eyes were collected under dim red light and stored at −80 °C until they were analyzed. Whole globes were then homogenized in 10 mM sodium phosphate buffer, pH 8.0, containing 50% methanol (v/v) and 100 mM hydroxylamine. The resulting mixture was extracted twice with 4 ml of ethyl acetate. The combined organic layers were dried in vacuo, reconstituted in 400 μl of hexanes, and 100 μl of the extract was injected on to a normal-phase high-performance liquid chromatography (HPLC) (Agilent Sil, 5 μm, 4.6 × 250 mm; Agilent Technologies, Santa Clara, CA) in a stepwise gradient of ethyl acetate in hexanes (0-17 min, 0.6%; 17.01-42 min. 10%) at a flow rate of 1.4 ml·min −1 . Retinoids were detected by monitoring absorbance at 325 nm and quantified based on a standard curve representing the relationship between the amount of synthetic retinoid standard and the area under the corresponding chromatographic peak.
RNAseq. RPE were isolated from littermate P14 mice 2 h after light onset. RNA was isolated with an RNeasy kit (Qiagen). An mRNA library was prepared using an Illumina TruSeq platform. The transcriptome was sequenced on a NextSeq. 500 cartridge. The statistical software package R (version 3.3.2) was used for all analyses. The Bioconductor package DESeq. 2 was used for differential gene expression analysis of RNA-seq data. Heatmaps were created using the R package gplots via the heatmap.2 package. The R code used for bioinformatics analysis and heatmap generation is available upon request.
Statistics. Statistics analysis was performed in GraphPad Prism 7.0. Statistical test used as indicated in the figure legends. A p-value < 0.05 was considered statistically significant.
Data availability. All data is available upon request. RNAseq datasets are publicly accessible on GEO.