Abstract
Somatic reprogramming of glia into neurons is a potentially promising approach for the replacement of neurons lost to injury or neurodegenerative disorders. Knockdown of the polypyrimidine tract-binding protein Ptbp1 has been recently reported to induce efficient conversion of retinal Mϋller glia and brain astrocytes into functional neurons. However, genetic analysis of Ptbp1 function in adult glia has not been conducted. Here, we use a combination of genetic lineage tracing, scRNA-Seq, and electrophysiological analysis to show that specific deletion of Ptbp1 in adult retinal Mϋller glia and brain astrocytes does not lead to any detectable level of glia-to-neuron conversion. Few changes in gene expression are observed in glia following Ptbp1 deletion, but glial identity is maintained. These findings highlight the importance of using genetic manipulation and lineage tracing methods in studying cell type conversion.
Introduction
Neurodegenerative disorders are clinically diverse and represent a huge public health burden. To address this, considerable effort has been focused on directed transdifferentiation of endogenous glial cells into functional neurons that were lost to diseases. Most strategies to achieve this have included glial-specific viral overexpression of genes that promote neuronal identity, including transcription factors and miRNAs (Jorstad et al. 2017; Berninger et al. 2007; Cervo et al. 2017; Y. Liu et al. 2015; Caiazzo et al. 2011), or genetic or small molecule-based manipulation of extracellular signaling pathways (Zhang et al. 2015; Zamboni et al. 2020). These approaches have met with variable rates of success, with many treatments working only in vitro or in immature glia, inducing only partial reprogramming, or not generating desired neuronal subtypes (Vignoles et al. 2019). Furthermore, studies that reported highly efficient reprogramming using these approaches have been criticised for lacking data that unambiguously demonstrate a lineage relationship between reprogrammed glia and neurons (Blackshaw and Sanes 2021; C. Qian et al. 2021).
The ability to reliably and efficiently induce glial reprogramming in vivo by manipulation of a single factor would be a major advance towards effective cell-based therapy for neurodegenerative disorders. Several recent reports have claimed to achieve exactly this outcome through knockdown of Ptbp1 expression in retinal Mϋller glia and brain astrocytes (Zhou et al. 2020; H. Qian et al. 2020; Maimon et al. 2021; Fu et al. 2020). Ptbp1 is an RNA binding protein and splicing regulator that is broadly expressed in non-neuronal and neuronal progenitor cells, and which represses neuronal-specific alternative splicing (Boutz et al. 2007; Makeyev et al. 2007). Neural progenitor-specific deletion of Ptbp1 leads to precocious neurogenesis (Shibasaki et al. 2013), and knockdown of Ptbp1 has been reported to be sufficient to convert both fibroblast and N2a cells into neurons in vitro (Xue et al. 2013). Furthermore, several recent papers reported that knockdown of Ptbp1 is sufficient to induce glial conversion into neurons. One study reported that Ptbp1 knockdown in retinal Mϋller glia using AAV-mediated CasRx led to rapid and efficient transdifferentiation of Mϋller glia into retinal ganglion cells, which were then able to efficiently innervate targets in the brain following excitotoxic inner retinal injury (Zhou et al. 2020). This same study also reported that Ptbp1 knockdown was sufficient for efficient conversion of brain astrocytes into dopaminergic neurons in the striatum and rescued function in a 6-OHDA-induced mouse model of Parkinson’s disease (PD). A second study showed that lentiviral-mediated shRNA and antisense oligonucleotides (ASO)-mediated knockdown of Ptbp1 in astrocytes in the cortex, striatum, and substantia nigra all induced efficient reprogramming of astrocytes into functional neurons, which in turn also rescued neurological defects in this same PD model (H. Qian et al. 2020). Other studies have reported that Ptbp1 knockdown can convert retinal Mϋller glia to photoreceptors (Fu et al. 2020), and restore neurogenic competence in neural progenitor cells in the dentate gyrus of aged mice (Maimon et al. 2021).
This simple and elegant loss of function approach potentially addresses many of the difficulties associated with previous efforts towards directed glial reprogramming, particularly in a clinical setting, such as low reprogramming efficiencies and the use of complex overexpression constructs. However, several major concerns remain to be addressed. First, none of these approaches convincingly demonstrated a reduction of Ptbp1 expression in glial cells in situ. Second, lineage relationships between glia and neurons were inferred through the use of GFAP promoter-based AAV constructs or transgene, which are known to show neuronal expression in some contexts (Su et al. 2004; Fujita et al. 2014). Third, convincing evidence for direct glia-to-neuron conversion using reliable genetic lineage analysis and/or scRNA-Seq-based trajectory analysis is lacking. These concerns need to be addressed before the Ptbp1 knockdown approach can further advance toward clinical applications.
In this study, we address the question of knockdown specificity and efficiency of glia-to-neuron conversion upon Ptbp1 reduction, through the use of glial-specific conditional mutants of Ptbp1, combining both genetic lineage and scRNA-Seq analysis of adult wildtype Mϋller glia and astrocytes, as well as adult glia carrying heterozygous or homozygous mutants of Ptbp1. Although we observe efficient and cell-specific disruption of Ptbp1, we observe no evidence for conversion of Mϋller glia or astrocytes into neurons in either heterozygous or homozygous Ptbp1 mutants. ScRNA-Seq analysis reveals only subtle changes in gene expression in mutant glia, but no evidence for induction of neuronal-specific genes and no presence of neuronal physiology in mutant glia. Our data indicate that the glia-to-neuron conversion reported in previous studies following Ptbp1 knockdown does not reflect the effects of Ptbp1 loss of function.
Results
Genetic loss of function of Ptbp1 in retinal Mϋller glia did not lead to glia-to-neuron conversion
To simultaneously disrupt Ptbp1 in adult retinal Mϋller glia and irreversibly label these cells with a visible marker, we used three transgenic lines: GlastCreERT2, which efficiently and selectively induces Cre-dependent recombination in Mϋller glia following tamoxifen treatment (de Melo et al. 2012; Hoang et al. 2020); Sun1-GFPlox/lox, which expresses GFP targeted to the nuclear envelope under the ubiquitous CAG promoter following Cre activation (Mo et al. 2015); and Ptbp1lox/lox, in which loxP sites flank the promoter and the first coding exon of Ptbp1, and Cre activation disrupts transcription (Shibasaki et al. 2013). We then generated wildtype (GlastCreERT2;Sun1-GFPlox/lox; Ptbp1+/+), heterozygous (GlastCreERT2;Sun1-GFPlox/lox; Ptbp1lox/+), and homozygous (GlastCreERT2;Sun1-GFPlox/lox; Ptbp1lox/lox) mutant mice. Starting at ∼5 weeks old, we induced Cre activation by using 4 doses of daily injections of 4-Hydroxytamoxifen (4-OHT) and conducted immunohistochemical analysis 2 and 4 weeks later (Fig. 1A).
Immunohistochemistry data shows that PTBP1 protein expression is enriched in Mϋller glia and some other non-glial cells in adult wildtype retinas (Fig. 1B). We observe that Cre activation led to a ∼ 90% reduction in the number of PTBP1-positive Mϋller glia cells at both 2 and 4 weeks following 4-OHT induction in the homozygous retinas (Fig. 1B, C). GFP-positive nuclei of Mϋller glia remain confined to the inner nuclear layer, implying that they do not generate either retinal ganglion cells or photoreceptors (Fig. 1B). To confirm this finding, we then conducted immunostaining for the ganglion cell-specific markers RBPMS (Fig.1D, Fig. S1C) and BRN3B (Fig. S1D-E). In contrast to a recent report (Zhou et al. 2020), we did not observe any colocalization of these markers with GFP following Ptbp1 deletion. We likewise observed no colocalization of the cone-specific marker arrestin (Fig. 1E) or the photoreceptor and bipolar marker OTX2 (Fig. S1F) with GFP, in contrast to another recent study (Fu et al. 2020). No induction of GFAP expression was observed in Ptbp1-deficient Mϋller glia (Fig. S1G), indicating that Ptbp1 depletion did not initiate reactive gliosis. It had also been previously reported the robust conversion of Mϋller glia to retinal ganglion cells induced by Ptbp1 depletion resulted in restoration of visual function following excitotoxic NMDA damage (Zhou et al. 2020). Although NMDA treatment led to extensive loss of RBPMS-positive retinal ganglion cells, we likewise did not observe any GFP/RBPMS or GFP/BRN3B-positive cells nor the recovery of retinal ganglion cells after 4 weeks following NMDA injury (Fig. 1F-G, S1H). Together, our data demonstrate that loss of Ptbp1 does not convert Mϋller glia into retinal neurons either in healthy or damaged retinas.
Genetic loss of function of Ptbp1 in brain astrocytes did not lead to glia-to-neuron conversion in the brain
To genetically disrupt Ptbp1 function in astrocytes, we employed a similar approach as in the retina, using astrocyte-specific tamoxifen-inducible Aldh1l1CreERT2 mice (Srinivasan et al. 2016), as well as the Sun1-GFPlox/lox and Ptbp1lox/lox lines. We used the same breeding and tamoxifen induction strategy as in the retina to generate mice that were wildtype, heterozygous and homozygous for Ptbp1 loss of function in astrocytes, and at the same time we irreversibly labeled astrocytes with Sun1-GFP for lineage tracing. Immunohistochemical analysis was then performed at 2 and 8 weeks following 4-OHT i.p induction (Fig. 2A).
As previous studies have reported glia-to-neuron conversion following Ptbp1 knockdown in cortex, striatum, and substantia nigra (Zhou et al. 2020; H. Qian et al. 2020), we focused our analysis on these 3 brain regions. We observed a dramatic reduction in the percentage of PTBP1/GFP-double positive cells relative to total GFP-positive cells in homozygous mice in all three regions compared to wildtype, and a modest reduction in the heterozygotes relative to wildtype mice in cortex and striatum (Fig. 2B-C, Fig. S2). However, we did not observe any GFP-positive cells co-labeled with the neuronal markers NeuN and HuC/D in these regions in any genotypes, at either 2 or 8 weeks following 4-OHT i.p induction (Fig. 2D-F, Fig. S3A-C). Since previous studies had reported that Ptbp1 knockdown induced markers of dopaminergic neurons in striatum and substantia nigra (Zhou et al. 2020; H. Qian et al. 2020), we then tested whether Ptbp1 deletion induced either tyrosine hydroxylase (TH) or the dopaminergic transporter (DAT). However, we did not observe any colocalization of GFP with either marker in either region, regardless of the genotype (Fig. 2G-H, Fig. S3D-E).
We next examined if there is any potential effect of Ptbp1 deletion on astrocyte physiology. No GFP-positive astrocytes in the cortex, in either Ptbp1 heterozygotes and homozygotes, fired action potentials in response to the current steps in whole-cell patch-clamp recordings, whereas neighboring GFP-negative wildtype neurons readily fired action potentials (Fig. 3A). This indicates that Ptbp1 deletion did not change the electrophysiological functions of astrocytes to make them fire action potentials like neurons. The astrocytes displayed highly negative resting membrane potentials (Ptbp1-Het, −82.2 ± 1.1 mV; Ptbp1-KO, −81.4 ± 1.9 mV) as previously reported (McNeill et al. 2021), and the Ptbp1 heterozygous and homozygous cells did not show significant difference in resting membrane potential (Fig. 3B, p=0.7795, Mann-Whitney test). These results suggest that Ptbp1 deletion in astrocytes does not induce any specific neuronal-like electrophysiological changes.
Genetic loss of function of Ptbp1 leads to only subtle changes in the gene expression profile of Mϋller glia and astrocytes
Previous studies did not characterize the gene expression profile of glial cells after Ptbp1 knockdown (Zhou et al. 2020; H. Qian et al. 2020). To comprehensively profile the cellular phenotype induced by Ptbp1 deletion in Mϋller glia and astrocytes, we performed scRNA-Seq analysis of retina, cortex, striatum, and substantia nigra 2 weeks following tamoxifen treatment of wildtype, heterozygous, and homozygous mice (Fig. 3C, G). Mϋller glia and astrocytes were then subsetted for further analysis (Fig. 3D, 3I, S4A, B, S5A).
In all tissues, we see a consistent reduction in Ptbp1 expression levels in Mϋller glia or astrocytes in heterozygous and homozygous mice (Fig. 3E, Fig.S5E). The relative fraction of either Mϋller glia or any major subtype of retinal neurons did not obviously change among any of the genotypes (Fig. S4B). We observe few significant changes in gene expression in any glial population profiled. In Mϋller glia, we observe increased expression of the Ptbp1 paralogue Ptbp2 expression level following Ptbp1 disruption (Fig. 3F). Ptbp2 upregulation has previously been reported in neural progenitors following Ptbp1 loss of function (Boutz et al. 2007), and since Ptbp1 and Ptbp2 show partially redundant functions (Vuong et al. 2016), this may provide some functional compensation for Ptbp1 loss of function. We also observe increased expression of Id3, Trf, Eno1, Mt3, Lgals3, with reduced expression of Msi2, Dio2, Rnf121, Atp1a2 (Fig. 3F). In all cases, the changes in gene expression were quite modest (Table ST1); we did not observe either significant reductions of expression of Mϋller glia-specific markers, including Sox9, Glul, Rlbp1, Slc1a3, Apoe, Aqp4, Mlc1, Kcnj10, and Tcf7l2, or induction of genes specific to neural progenitors or mature neurons (Fig. S4D, Table ST1). Immunostaining confirmed that Ptbp1-deficient Mϋller glia retained expression of the glial marker SOX9 (Fig. S4E). Together, our data demonstrate that loss of function of Ptbp1 does not dramatically alter gene expression in Mϋller glia, or cause these cells to lose their identity.
We likewise did not observe any clear changes in the astrocytes across the 3 genotypes in any of the brain regions (Fig. 3H-I, Table ST2-ST4). A correlation plot showed that differences in gene expression in astrocytes among brain regions (cortex and striatum to substantia nigra) were higher than the differences among genotypes (Fig. 3H). We also observed S100b-expressing astrocytes in the striatum as previously reported (Gokce et al. 2016) (Fig. S5D). As with Mϋller glia, we observed only very modest changes in gene expression following Ptbp1 loss of function (Fig. 3J, Table ST2-ST4). Upregulated genes include mt-Nd4, Son, Hes5, Mt3; downregulated genes include mt-Nd3, Lars2, Ivd. Most astrocyte markers did not show altered expression after Ptbp1 deletion (Fig. S5F). Immunostaining confirmed that Ptbp1-deficient, GFP-positive astrocytes retained expression of glial marker SOX9 (Fig. S5G). We also do not observe induction of either neural progenitor or mature neuron-specific genes (Table ST2-ST4). We conclude that Ptbp1 does not function in mature glial cells to repress the expression of neuronal genes.
Discussion
Using the genetic loss of function and cell lineage analysis, in combination with scRNA-Seq analysis, we observed no evidence that either partial or complete loss of function of Ptbp1 induces glia-to-neuron conversion in retina or brain. Our data contrasts sharply with several recent studies that analyze the effects of Ptbp1 knockdown using ASO, shRNA, and/or CasRx (Zhou et al. 2020; H. Qian et al. 2020; Fu et al. 2020; Maimon et al. 2021), but is in agreement with a recent study the re-examined previous studies reporting astrocyte-to-neuron conversion following Neurod1 overexpression or either shRNA- or CasRx-mediated Ptbp1 knockdown (Wang et al. 2021). This latter study concluded that reports of glia-to-neuron conversion in these two models represented a leaky neuronal expression of the GFAP-based AAV and transgene lines used to label astrocytes. Leaky neuronal expression of GFAP-based reagents has been previously reported in other contexts (Su et al. 2004; Fujita et al. 2014), and caution should be used when interpreting results obtained using these methods without corroborating data obtained using more strongly glial-specific minipromoters and Cre lines, such as the extensively validated GlastCreERT2 and Aldh1l1CreERT2 lines used here.
Our data show that the previous reports of glia-to-neuron conversion is unlikely to have resulted from Ptbp1 loss of function in glial cells. Then what could have accounted for the recovery of visual function following NMDA excitotoxicity, or behavioral recovery following 6-OHDA-induced PD (Zhou et al. 2020; H. Qian et al. 2020)? Ectopic neuronal expression of GFAP-based reagents in endogenous native neurons represents one possible explanation. For example, inclusion of the Ptbp1-dependent control of splicing of proapoptotic gene Bak1 is essential for neuronal and animal survival (Lin et al. 2020). Knockdown of residual PTBP1 expression in neurons may therefore further promote neuronal survival. Another possibility is unexpected beneficial off-target effects of reagents targeting Ptbp1 in the previous reports. In either case, genetic methods should be used to validate future reports of glia-to-neuron conversion.
Supplemental figure legends
Supplemental Tables
Supplemental Table 1 - Differential gene expression as measured by scRNA-Seq in retinal Mϋller glia across Ptbp1 genotypes.
Supplemental Table 2 - Differential gene expression as measured by scRNA-Seq in cortical astrocytes across Ptbp1 genotypes.
Supplemental Table 3 - Differential gene expression as measured by scRNA-Seq in striatal astrocytes across Ptbp1 genotypes.
Supplemental Table 4 - Differential gene expression as measured by scRNA-Seq in substantia nigra astrocytes across Ptbp1 genotypes.
Materials and Methods
Mice
GlastCreERT2 and Sun1-GFPlox/lox transgenic mice were provided by Dr. Jeremy Nathans (de Melo et al. 2012; Mo et al. 2015). Aldh1l1CreERT2 transgenic mouse line, which allows for specific inducible Cre-recombination in astrocytes after TAM injection (Srinivasan et al. 2016), was purchased from JAX (JAX#029655). Ptbp1lox/lox mice carrying loxP sites that flank the promoter and 1st exon of Ptbp1 were generated as described previously (Shibayama et al. 2009). To induce specific Cre activation in adult Mϋller and astrocytes, 4 consecutive doses of 4-Hydroxytamoxifen (4-OHT) intraperitoneal (i.p) injection (30 mg/kg) were performed in adult wildtype, heterozygous and homozygous mice at ∼5 weeks old. Mice were sacrificed at the indicated time for analysis.
Intravitreal NMDA injection
We followed a previously described protocol (Zhou et al. 2020). Briefly, adult mice were anesthetized with 4% isoflurane inhalation. 1.5 μl of 200 mM NMDA in PBS was injected intravenously into the retinas. 4 consecutive doses of 4-OHT i.p injection (30 mg/kg) were performed at 2 weeks after NMDA injection. Mice were sacrificed, and retinas were collected at 4 weeks after 4-OHT injection.
Immunohistochemical and imaging analysis
Retinal collection and immunohistochemical analysis were performed as described previously (Hoang et al. 2020). Briefly, mice were anesthetized by CO2, and eye globes were fixed in 4% paraformaldehyde in PBS for 4 hr at room temperature. Retinas were dissected and placed into 30% sucrose in PBS at 4°C overnight. Retinas were embedded in OCT, sectioned at 16 µm thickness, and stored at −20C. Retinal sections were air-dried at 37°C for 20 min, washed 3 × 5 mins with PBS times, and incubated in a blocking buffer (0.4% TritonX-100, 10% horse serum in PBS) for 1 hr at room temperature. Primary antibodies were incubated in the blocking buffer at the indicated concentration overnight at 4°C. Primary antibodies used in this study were: rabbit anti-Ptbp1(1:200, Proteintech, #125821-AP), rabbit Rbpms (1:200, Proteintech, #151871-AP), goat anti-Brn3b (1:200, Santa Cruz, #sc6026), rabbit anti-cone arrestin (1:200, Millipore Sigma, #AB15282), goat anti-Otx2 (1:200, R&D systems, #AF1979), rabbit anti-Gfap (1;300, Dako, #z0334), rabbit anti-Sox9 (1:200, Millipore Sigma, #AB5535), rabbit anti-GFP (1:400, Life technologies, #A6455) and chicken anti-GFP (1:400, Thermo Fisher Scientific, #A10262). Secondary antibodies were incubated at 1:400 dilution in the blocking buffer. Retinal sections were then incubated with DAPI, washed 4 × 5 min in PBS, mounted with prolong gold mounting media (ThermoFisher Scientific, #P36935), air-dried, and stored in 4°C.
Brain collection and immunohistochemical analysis were performed as described previously (Hoang et al. 2020; Kim et al. 2020; Yoo et al. 2019). Briefly, mice were anesthetized by avertin and cardially perfused with 4% paraformaldehyde. Brains were then dissected and post-fixed in 4% paraformaldehyde for 12 hr at room temperature, and then placed into 30% sucrose in PBS at 4°C overnight. Brains were sectioned at 30 µm thickness and processed for immunohistochemistry as described above. Additional primary antibodies used in this study for the brain were: mouse anti-HuC/D (1:200, Thermo Fisher Scientific, #A21271), mouse anti-NeuN (1:200, Millipore Sigma, #MAB377), rat anti-Dopamine Transporter Antibody (DAT, 1:200, Millipore Sigma, #MAB369), rabbit anti-Tyrosine Hydroxylase (TH, 1:200, Millipore Sigma, #AB152).
Images were acquired using Zeiss LSM700 confocal microscope at > 6 random regions for each retina or brain. Images were processed using ImageJ.
Electroretinogram
Slice electrophysiology of cortical astrocytes
The acute brain slices were generated as previously described (K. Liu et al. 2017; Yoo et al. 2021). Ptbp1-Het or Ptbp1-KO (P54 to P73, male) were anesthetized with isoflurane and decapitated, and the brains were rapidly removed and chilled in ice-cold sucrose solution containing 76 mM NaCl, 25 mM NaHCO3, 25 mM glucose, 75 mM sucrose, 2.5 mM KCl, 1.25 mM NaH2PO4, 0.5 mM CaCl2, and 7 mM MgSO4 (pH 7.3). Acute brain slices (300 µm) including the motor cortex were prepared using a vibratome (VT 1200 s, Leica) and transferred to warm (32° to 35°C) sucrose solution for 30 min for recovery. The slices were transferred to warm (32° to 34°C) aCSF composed of 125 mM NaCl, 26 mM NaHCO3, 2.5 mM KCl, 1.25 mM NaH2PO4, 1 mM MgSO4, 20 mM glucose, 2 mM CaCl2, 0.4 mM ascorbic acid, 2 mM pyruvic acid, and 4 mM l-(+)-lactic acid (pH 7.3, 315 mOsm) and allowed to cool to room temperature. All solutions were continuously bubbled with 95% O2/5% CO2.
For whole-cell patch-clamp recordings, slices were transferred to a submersion chamber on an upright microscope [Zeiss Axio Examiner, objective lens: 5×, 0.16 numerical aperture (NA) and 40×, 1.0 NA] fitted for infrared differential interference contrast (IR-DIC) and fluorescence microscopy. Slices were continuously superfused (2 to 4 ml/min) with warm, oxygenated aCSF (32° to 34°C). Sun1-GFP(+) cells in the layer 2/3 of the motor cortex were identified under a digital camera (Sensicam QE, Cooke) using transmitted light and green fluorescence. For the whole-cell patch-clamp recordings, borosilicate glass pipettes (2 to 4 MΩ) were filled with an internal solution containing 2.7 mM KCl, 120 mM KMeSO4, 9 mM Hepes, 0.18 mM EGTA, 4 mM Mg-adenosine 5′-triphosphate, 0.3 mM Na-guanosine 5′-triphosphate, and 20 mM phosphocreatine(Na) (pH 7.3, 295 mOsm). Whole-cell patch-clamp recordings were conducted through a MultiClamp 700B amplifier (Molecular Devices) and an ITC-18 (InstruTECH), which were controlled by customized routines written in Igor Pro (WaveMetrics).
To determine the action potential firing capacity of the recorded cells, a series of depolarizing current steps (1-s long, 0 to 4000 pA) was given to the cells, and their voltage responses were recorded. During the current step test, cells were held at their resting membrane potentials, and different intensities of currents were tested until the cell membrane potential went over 0 mV without firing action potentials or the membrane potential did not hold from 4000 pA current injections. As a control, three Sun1-GFP(-) neurons from two Ptbp1-KO mice were targeted and their voltage responses to a series of depolarizing current steps (1-s long, 0 to 300 pA) were recorded by using the same internal and external solutions for comparing action potential firing properties with Sun1-GFP(+) cells from the same mice. All signals were low-pass filtered at 10 kHz and sampled at 20 kHz. Electrophysiology data was analyzed in Igor Pro (WaveMetrics).
Cell dissociation and scRNA-Seq
Retinal cell dissociation was performed as described previously (Hoang et al. 2020). Briefly, one female mouse per genotype was euthanized by CO2, and eye globes were removed and placed in ice-cold PBS. Retinas were dissected, cells were dissociated using Papain Dissociation System (LK003150, Worthington). Dissociated cells were resuspended in a buffer containing 9.8ml Hibernate A, 200μl B27, 20 μl GlutaMAX and 0.5U/ul RNAse inhibitor. Cells were filtered through a 50μm filter. Cell count and viability were determined by using 0.4% Trypan blue.
Brain cell dissociation was performed as described previously (Bell et al. 2021; Kim et al. 2021) using one female mouse per genotype. Mouse brain matrix (0.5 mm) was used to slice brains into coronal sections, and relative brain regions (cerebral cortex, striatum, midbrain) were dissected into Hibernate-A media with a 2% B-27 and GlutaMAX supplement (0.5 mM final). Tissues were dissociated in papain (Worthington).
Cells were then loaded into the 10x Genomics Chromium Single Cell System (10x Genomics) and libraries were generated using v3.1 chemistry following the manufacturer’s instructions. Libraries were sequenced on the Illumina NovaSeq platform (500 million reads per library).
ScRNA-Seq data analysis
Sequencing data were first processed through the Cell Ranger (v.6.0.1, 10x Genomics) with default parameters, aligned to the mm10 genome (refdata-gex-mm10-2020-A), and matrix files were used for subsequent bioinformatic analysis.
Matrix data were further processed using Seurat 3.4 version (Stuart et al. 2019). Low quality cells with < 500 genes, < 2000 UMI and > 30% mitochondrial genes were removed. Datasets were normalized using Seurat ‘scTransform’ function. Cells were clustered and visualized using UMAP. Retinal Mϋller glia (Slc1a3+ & Rlbp1+) and brain astrocytes (Slc1a3+ & Aldh1l1+) clusters were subsetted for further analysis to compare differential gene expression across genotypes. Differential gene expression between different genotypes was calculated using the Seurat ‘FindMarker’ function.
Author contributions
TH, DWK, and SB conceived and supervised the study. TH generated and analyzed immunohistochemistry and scRNA-Seq data from retina, assisted by HA, NAP, and TK. DWK generated and analyzed immunohistochemistry and scRNA-Seq data from brain, assisted by TH and HA. MO and SZ provided reagents. JK performed slice electrophysiology of cortical astrocytes. MY and NSP conducted ERG analysis. TH, DWK, and SB drafted the manuscript. All authors contributed to the writing of the manuscript.
Data availability
All scRNA-Seq data have been deposited to GEO as GSE184933.
Acknowledgments
We thank F. Zhou, A. Fischer, J. Ling, and W. Yap for their comments on the manuscript. We thank the Single Cell & Transcriptomics Core (Johns Hopkins) for sequencing of scRNA-Seq libraries. This work was supported by the NIH National Eye Institute grants R01EY020560 and U01EY027267 to SB and the Maryland Stem Cell Research Fund (2019-MSCRFF-5124) to DWK. SB is supported by a Stein Innovation Award from Research to Prevent Blindness.