Editor's evaluation: Nuclear NAD+-biosynthetic enzyme NMNAT1 facilitates development and early survival of retinal neurons

Abstract

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Despite mounting evidence that the mammalian retina is exceptionally reliant on proper NAD+ homeostasis for health and function, the specific roles of subcellular NAD+ pools in retinal development, maintenance, and disease remain obscure. Here, we show that deletion of the nuclear-localized NAD+ synthase nicotinamide mononucleotide adenylyltransferase-1 (NMNAT1) in the developing murine retina causes early and severe degeneration of photoreceptors and select inner retinal neurons via multiple distinct cell death pathways. This severe phenotype is associated with disruptions to retinal central carbon metabolism, purine nucleotide synthesis, and amino acid pathways. Furthermore, transcriptomic and immunostaining approaches reveal dysregulation of a collection of photoreceptor and synapse-specific genes in NMNAT1 knockout retinas prior to detectable morphological or metabolic alterations. Collectively, our study reveals previously unrecognized complexity in NMNAT1-associated retinal degeneration and suggests a yet-undescribed role for NMNAT1 in gene regulation during photoreceptor terminal differentiation. Editor's evaluation Mutations in the gene encoding the NMNAT1 enzyme cause Leber congenital amaurosis type 9 (LCA9), a blinding disease. Using conditional inactivation of the mouse gene in the retina, this study extends previous observations on the requirement of this enzyme for photoreceptors homeostasis. The study not only shows NMNAT1 is involved in photoreceptor terminal differentiation but also provides evidence that the survival of other cell types depends on this enzyme. It also shows that NMNAT1 deficiency leads to the activation of different cell death pathways and causes metabolic defects in retinal cells. The study thus provides a better picture of the retinal defects that may underlie LCA9 in humans. https://doi.org/10.7554/eLife.71185.sa0 Decision letter eLife's review process Introduction Nicotinamide adenine dinucleotide (NAD+) is a ubiquitous cellular metabolite with a diverse palette of biological functions across all kingdoms of life. In addition to serving a central role in redox metabolism as an electron shuttle, NAD+ has well-defined roles as a substrate for a host of enzymes including sirtuins (SIRTs), mono- and poly-ADP-ribose polymerases (PARPs), and NAD+ glycohydrolases (CD38, CD157, and SARM1). Collectively, these roles implicate NAD+ metabolism in phenomena as diverse as aging, cell proliferation, immunity, neurodegeneration, differentiation, and development (Houtkooper et al., 2010; Cantó et al., 2015; Nikiforov et al., 2015; Cambronne and Kraus, 2020; Navas and Carnero, 2021). A relatively recent advance in the field is the notion of compartmentalized NAD+ metabolism—that regulation of NAD+ in distinct subcellular compartments dictates function in diverse manners (Cantó et al., 2015; Nikiforov et al., 2015; Cambronne and Kraus, 2020; Navas and Carnero, 2021). While many aspects of this compartmentalization remain to be explored, it is now known that spatiotemporal NAD+ regulation plays prominent roles in processes including axon degeneration, circadian regulation, and adipogenesis (Cambronne and Kraus, 2020). Among mammalian tissues, the retina appears particularly reliant on proper NAD+ homeostasis for survival and function. This is suggested by associations between retinal NAD+ deficiency and pathology in diverse models of retinal dysfunction (Lin et al., 2016; Williams et al., 2017) as well as multiple mutations to NAD+- or NADP+-utilizing enzymes which cause blindness in humans (Bowne et al., 2006; Aleman et al., 2018; Bennett et al., 2020). Among these enzymes is nicotinamide mononucleotide adenylyltransferase-1 (NMNAT1), a highly conserved, nuclear-localized protein which catalyzes the adenylation of nicotinamide mononucleotide (NMN) or nicotinic acid mononucleotide (NaMN) to form NAD+, the convergent step of all mammalian NAD+ biosynthetic pathways (Nikiforov et al., 2015). To date, over 30 NMNAT1 mutations have been linked to the severe blinding diseases Leber congenital amaurosis type 9 (LCA9) and related cone-rod dystrophy (Perrault et al., 2012; Falk et al., 2012; Chiang et al., 2012; Koenekoop et al., 2012; Coppieters et al., 2015; Nash et al., 2018). Although NMNAT1 is ubiquitously expressed, and many of these mutations reduce NMNAT1 catalytic activity or stress-associated stability (Falk et al., 2012; Koenekoop et al., 2012; Sasaki et al., 2015), patients with these disorders rarely report extra-ocular phenotypes, a puzzling observation which is recapitulated by two LCA-NMNAT1 mutant mouse models (Greenwald et al., 2016). Further puzzling is the existence of two other NMNAT paralogs (Golgi-associated NMNAT2 and mitochondrial NMNAT3), which are detectable in the retina (Kuribayashi et al., 2018) but have not been linked to blindness. Importantly, while a crucial role for retinal NAD+ was recently described through characterization of mice conditionally lacking the NAD+ pathway enzyme NAMPT in photoreceptors (Lin et al., 2016), the significance of nuclear-synthesized NAD+ in vision—suggested by the fact that NMNAT1 is the only NAD+-pathway enzyme to date linked to blindness—remains poorly understood. Current results point to multiple, potentially distinct roles for NMNAT1 in the retina—ex vivo studies suggest that NMNAT1 supports sirtuin function to facilitate the survival of retinal progenitor cells (Kuribayashi et al., 2018), while ablation of NMNAT1 in mature mice results in rapid death of photoreceptors mediated by the neurodegenerative NADase SARM1 (Sasaki et al., 2020b). Global deletion of NMNAT1 in mice is embryonically lethal (Conforti et al., 2011), suggesting non-redundant roles for nuclear NAD+ synthesis during development. Consistent with this notion, pan-retinal NMNAT1 deletion is shown to cause rapid and severe retinal degeneration in mice shortly after birth (Wang et al., 2017; Eblimit et al., 2018). While these studies suggest diverse functions of retinal NMNAT1 beyond its canonical role in redox metabolism, the degree to which these functions overlap—as well as the mechanistic basis for the severity of NMNAT1-associated retinal dystrophy in animal models and patients—have not been comprehensively explored. In this study, we investigate the roles of NMNAT1-mediated NAD+ metabolism in the retina by generating and characterizing a retina-specific NMNAT1 knockout mouse model. Utilizing histological and transcriptomic approaches, we demonstrate that NMNAT1 deletion causes severe and progressive retinal degeneration affecting specific retinal cell types beyond photoreceptors, and that this severe degeneration likely results from activation of multiple distinct cell death pathways. Comprehensive metabolomics analysis reveals specific metabolic defects in NMNAT1 knockout retinas and suggests impaired central carbon, purine nucleotide, and amino acid metabolism as a cause for severe degeneration. Strikingly, RNA-sequencing reveals a collection of photoreceptor and synapse-specific genes which are downregulated in knockout retinas preceding degeneration. Immunostaining of several of these genes suggests severe impairment of photoreceptor terminal differentiation in the absence of NMNAT1. Overall, our results reveal a previously unappreciated complexity in NMNAT1-associated retinal degeneration, provide possible explanations for the retina-specific manifestations of NMNAT1 deficiency, and propose a yet-undescribed role for NMNAT1 in gene regulation during late-stage retinal development. Results Generation and validation of NMNAT1 conditional knockout mouse model To establish a retina-specific NMNAT1 knockout model, we crossed mice homozygous for a loxP-targeted Nmnat1 locus (Nmnat1flox/flox) with transgenic mice expressing Cre recombinase under a Six3 promoter (Nmnat1wt/wt;Six3-Cre), which is expressed throughout the retina at embryonic day 9.5 (E9.5) and shows robust activity by E12.5 (Furuta et al., 2000). After several crosses, mice inheriting Six3-Cre and a floxed Nmnat1 locus (Nmnat1flox/flox;Six3-Cre, hereafter referred to as ‘knockouts’) exhibit Cre-mediated excision of the first two exons of Nmnat1—which contain important substrate binding domains—in the embryonic retina (Figure 1A). We determined that retinal Nmnat1 expression in postnatal day 4 (P4) knockout mice was reduced by 75.6% (95% CI 56.1–95.0%) compared to littermate controls (Figure 1B), a reduction consistent with a previously reported NMNAT1 retinal knockout model (Sasaki et al., 2020b). We further verified that retinal NMNAT1 protein levels were drastically reduced in P0 knockout mice using a custom-made polyclonal antibody against NMNAT1 (Figure 1C and Figure 1—figure supplement 1). Finally, we confirmed that embryonic Six3-Cre expression alone does not cause gross retinal abnormalities by staining for several well-characterized cell type markers in mature Nmnat1wt/wt; Six3-Cre retinas and littermate controls (Figure 2—figure supplement 1; markers discussed below). Figure 1 with 1 supplement see all Download asset Open asset Loss of NMNAT1 leads to early and severe retinal degeneration. (A) Schematic depicting retina-specific Six3-Cre mediated excision of a segment of the Nmnat1 gene. (B) Relative Nmnat1 expression in retina from P4 knockout (-/-) and littermate control (+/+) mice as assessed by RT-qPCR (grey bars represent mean, significance determined using Mann-Whitney U test, n = 3 biological replicates). (C) Representative western blot showing levels of NMNAT1 and β-tubulin loading control in retinal lysate from P0 knockout and control mice. (D–G) Representative H&E-stained retinal cross sections from knockout and control mice at indicated ages. (H–J) Spider plots depicting mean retinal thickness at P0, P4, and P10. Data are represented as mean ± SD. *p < 0.05 using Student’s t-test, n = 3 biological replicates per age. Scale bars, 30 μm. Abbreviations: LP, loxP site; E1-4, exon 1–4; P, postnatal day; GCL, ganglion cell layer; NBL, neuroblastic layer; IPL, inner plexiform layer; OPL, outer plexiform layer; INL, inner nuclear layer; ONL, outer nuclear layer; IS/OS, photoreceptor inner segment/outer segment layer. Figure 1—source data 1 Quantification of Nmnat1 mRNA levels in P0 WT and KO retinas. Numerical source data for retinal thickness quantification in P0, P4, and P10 WT and KO retinas. https://cdn.elifesciences.org/articles/71185/elife-71185-fig1-data1-v2.xlsx Download elife-71185-fig1-data1-v2.xlsx Early-onset and severe morphological defects in the NMNAT1-null retina As a first step toward characterizing the effects of NMNAT1 ablation on the retina, we performed retinal histology using hematoxylin and eosin (H&E) staining. H&E-stained retinal cross sections from P0 knockout and control mice reveal no obvious morphological differences (Figure 1D and H); however, by P4, knockout retina are markedly thinner than controls and exhibit disrupted lamination and evidence of large-scale cell death in the inner and outer nuclear layers (Figure 1E). Degeneration is most severe in the central retina, with a ~45% reduction in central retinal thickness (but unaffected peripheral retinal thickness) in P4 knockout mice (Figure 1I). By P10, knockouts show a ~62% reduction in central retinal thickness and ~27% reduction in peripheral retinal thickness compared to controls (Figure 1J). Degeneration of the entire inner and outer nuclear layers is nearly complete by P30 (Figure 1G), while remaining inner retinal structures persist until approximately P60 (data not shown). Proper segregation of inner and outer retinal neurons appears disrupted in P4 knockouts, but this segregation is established in P10 knockouts despite severe degeneration (Figure 1F). Interestingly, formation of the outer plexiform layer (containing photoreceptor and bipolar neuron synaptic structures) appears disrupted in P4 and P10 knockout retinas (Figure 1E and F). NMNAT1 loss affects survival of major inner retinal neurons Histological examination suggests severe photoreceptor degeneration in NMNAT1 knockout retinas but also indicates loss of specific inner retinal neuron populations. To further characterize these effects, we quantified populations of several major inner retinal cell types in our knockout by staining retinal sections with well-characterized antibody markers: retinal ganglion cells were identified by labelling for brain-specific homeobox/POU domain protein 3A (BRN3A), amacrine cells by labeling for calretinin (CALR), horizontal cells by labeling for calbindin (CALB), and bipolar cells by labeling for Ceh-10 homeodomain-containing homolog (CHX10) (Key Resources Table). With the exception of calbindin, we performed this analysis at P4 and P10—representing early and late stages of degeneration, respectively—revealing an interesting cell-type-dependent sensitivity to NMNAT1 loss (Figure 2). At both tested ages, relative numbers of retinal ganglion cells are not significantly different between knockout and control retinas (Figure 2A–D and M), while numbers of amacrine cells are unchanged at P4 but reduced by ~51% (95% CI 36–65%) in P10 knockout retinas (Figure 2E–H and O). Numbers of bipolar cells are similarly unchanged at P4 but reduced by ~75% (95% CI 62–85%) in P10 knockout retinas (Figure 2I–L and P). In P0 knockout retinas—an age at which ganglion, amacrine, and bipolar cell numbers are unchanged (data not shown)—horizontal cell counts are reduced by ~36% (95% CI 18–53%) (Figure 2N), and this trend persists in P4 knockout retinas (Figure 2A–B and N). These results identify retinal bipolar, horizontal, and amacrine neurons as targets of NMNAT1-associated pathology and suggest that horizontal and bipolar neurons are more sensitive to NMNAT1 loss than amacrine neurons. Interestingly, while retinal ganglion cells do eventually degenerate at timepoints past P30 (data not shown), they appear largely agnostic to NMNAT1 loss in the young postnatal retina. Figure 2 with 1 supplement see all Download asset Open asset NMNAT1 loss affects retinal bipolar, horizontal and amacrine cells. Representative retinal sections from knockout (-/-) and floxed littermate control (+/+) mice at the indicated ages labeled with antibodies against BRN3A (A-D, green), Calbindin (CALB) (A–B, magenta) Calretinin (CALR) (E–H), and CHX10 (I–L). Quantification of BRN3A (M), CALB, (N), CALR (O), and CHX10-positive cells (P) are shown. In (O), only CALR-positive cells on the outer side of the IPL (layer indicated by white arrowheads) were counted. Data is represented as mean ± SD. n = 3 biological replicates for all panels; significance determined using Student’s t-test. Scale bars, 30 μm. Figure 2—source data 1 Numerical source data for retinal cell type quantification in P4 and P10 KO and WT retinas. https://cdn.elifesciences.org/articles/71185/elife-71185-fig2-data1-v2.xlsx Download elife-71185-fig2-data1-v2.xlsx Loss of NMNAT1 impairs photoreceptor terminal differentiation Turning our attention to photoreceptors, we repeated the above approach with antibodies against the photoreceptor markers recoverin (anti-RCVRN) and rhodopsin (anti-RHO). While anti-RCVRN cleanly labels developing photoreceptor somas in P4 and P10 control retinas (Figure 3H, Figure 3—figure supplement 1F), we observe a complete lack of recoverin expression in knockout retinas at both ages (Figure 3I, Figure 3—figure supplement 1G). Barring a small amount of non-specific staining likely originating from the secondary antibody (Figure 3—figure supplement 1D,E), rhodopsin expression at P4 and P10 showed an identical trend to that of recoverin (Figure 3J and K, Figure 3—figure supplement 1H,I). Figure 3 with 5 supplements see all Download asset Open asset NMNAT1 loss impairs photoreceptor terminal differentiation. Representative retinal sections from knockout (-/-) and floxed littermate control (+/+) mice at the indicated ages labelled with antibodies against OTX2 (A–D), Phosducin (PDC) and peanut agglutinin (PNA) (E, F), M-opsin (OPN1MW) and PNA (G, H), recoverin (RCVRN) (H,I) or rhodopsin (RHO) (J, K). (L) Comparison of differentially expressed genes in E18.5 and P4 knockout retinas as assessed by RNA-sequencing. (M) GSEA of differentially expressed genes identified in E18.5 and P4 knockout retinas. (N, O) Relative expression of indicated genes in P4 and E18.5 knockout retinas as assessed by RNA-sequencing. n = 3 biological replicates for all panels, except (N), where n = 5 biological replicates. Corresponding zoom panels are indicated with dotted rectangles. Scale bars, 30 μm. Figure 3—source data 1 Gene ontology (GO) overrepresentation analysis of genes differentially expressed in KO retinas at E18.5 and P4 timepoints. Numerical source data for expression (relative to WT levels) of select photoreceptor-specific genes in E18.5 and P4 KO retinas. https://cdn.elifesciences.org/articles/71185/elife-71185-fig3-data1-v2.xlsx Download elife-71185-fig3-data1-v2.xlsx Intrigued by the magnitude of recoverin and rhodopsin loss and hypothesizing defects in the expression of other retinal proteins in our knockout, we comprehensively profiled the transcriptomes of knockout and control retinas at two timepoints— pre-degeneration (E18.5) and during degeneration (P4) and using RNA-sequencing. At P4, this analysis reveals 2976 differentially expressed genes in NMNAT1 knockout retinas (Figure 3—figure supplement 2B), several of which we validated using RT-qPCR (Figure 3—figure supplement 1J). Consistent with the lack of recoverin and rhodopsin staining at this age, gene set enrichment analysis (GSEA) of P4 differentially expressed genes (DEGs) reveals several large, highly-overrepresented clusters of downregulated photoreceptor-related genes including both recoverin and rhodopsin (Figure 3—figure supplement 2C). Strikingly, among 815 DEGs in E18.5 knockout retinas, a similar cluster of downregulated genes associated with visual perception and the photoreceptor outer segment was observed (Figure 3—figure supplement 3B,C). Combining both RNA-sequencing datasets reveals a group of 365 DEGs in knockout retinas common to both timepoints (Figure 3L). Importantly, GSEA on this gene set reveals highly overrepresented clusters of photoreceptor and synapse associated genes (Figure 3M), and further analysis identifies a core set of 21 photoreceptor-associated genes which are significantly downregulated in E18.5 and P4 NMNAT1 knockout retinas (Figure 3N and O). Notably, this set includes rod-specific (e.g. Gngt1), cone-specific (e.g. Opn1sw, Cnga3) and photoreceptor-specific (e.g. Prph2, Rcvrn, Aipl1) genes of diverse function, many of which have important roles in photoreceptor development and function. Consistent with a specific transcriptional effect on photoreceptors, we confirmed that expression of several well-known ganglion cell, amacrine/horizontal cell, and bipolar cell-specific genes was largely unchanged in NMNAT1 knockout retinas at either tested age (Figure 3—figure supplement 4). To further confirm the relevance of our RNA-sequencing results, we immunostained knockout and control retinas with several markers of developing cone photoreceptors: anti-phosducin (PDC) (Rodgers et al., 2016), anti-M-opsin (OPN1MW), and peanut agglutinin (PNA), which labels developing cone outer segments (Blanks and Johnson, 1984). While these markers showed normal cone accumulation and rudimentary outer segment formation in the OPL of P0 and P4 control retinas (Figure 3E and G), P0 knockout retinas showed markedly reduced and widely dispersed phosducin-positive cones (Figure 3F), and P4 knockouts demonstrated a near complete absence of M-opsin expression, mirroring recoverin and rhodopsin at this age (Figure 3H). Accordingly, knockouts at both ages show significant attenuation of cone outer segment formation as determined by PNA staining (Figure 3F and H). Due to the extent of photoreceptor-specific transcript and protein alterations in knockout retinas, we hypothesized that NMNAT1 is essential for photoreceptor terminal differentiation—the process by which partially committed retinal progenitors initiate photoreceptor-specific transcriptional and morphological programs (Swaroop et al., 2010; Brzezinski and Reh, 2015; Daum et al., 2017). To test this possibility, we probed knockout and control retinas with an antibody against OTX2, a well-characterized transcription factor necessary for photoreceptor terminal differentiation (Beby and Lamonerie, 2013). In P0 and P4 control retinas, OTX2 is enriched in developing photoreceptors in the ONL (Figure 3A and C, arrowheads) and in developing bipolar neurons in the INL (Figure 3C). Consistent with our hypothesis, OTX-positive photoreceptors are absent in P0 knockout retinas (Figure 3B), and P4 knockout retinas display a striking expression of OTX2 throughout the ONL/INL (Figure 3D). Finally, to determine whether impaired retinal progenitor proliferation may explain reduced photoreceptor numbers in knockouts, we assessed proliferation at P0 using the well-characterized marker phospho-histone H3 (PHH3), which labels M-phase cells (Prigent and Dimitrov, 2003). This analysis revealed a slight but not significant decrease in PHH3-positive cells in P0 knockout retinas (Figure 3—figure supplement 1A–C) Altogether, these results argue that NMNAT1 is crucial for terminal differentiation—but not early proliferation—of retinal photoreceptors and suggest photoreceptor-specific transcriptional dysregulation as a driver of the severe photoreceptor phenotype in NMNAT1-deficient retinas. Beyond affecting photoreceptor-specific gene expression, we also note downregulation of 6 synapse-associated genes (Stx3, Syngr1, Cln3, Scamp5, and Sv2b) in both E18.5 and P4 knockout retinas (Figure 3—figure supplement 5C,D), consistent with disruptions to outer plexiform layer formation in P4 knockout retinas on histology and on staining with the synapse marker synaptophysin (anti-SYPH) (Figure 3—figure supplement 5). Loss of NMNAT1 during retinal development triggers multiple cell death pathways As NMNAT1 deficiency drastically impairs the postnatal survival of photoreceptor, bipolar, horizontal, and amacrine retinal neurons, we sought to determine the mechanisms by which these cells degenerate. To this aim, we began by staining retinal sections with an antibody against activated caspase-3 (AC3). While P0 knockout retinas show little AC3 staining compared to controls—consistent with grossly normal retinal morphology at this age—P4 knockout retinas show robust AC3 immunoreactivity in the inner and outer nuclear layers (Figure 4A–D). As expected, most AC3-immunoreactive (AC3+) cells display nuclear chromatin condensation (‘pyknosis’) characteristic of dying cells; however, staining also revealed a population of pyknotic nuclei not immunoreactive to AC3 (AC3-) (Figure 4A’–D’, arrows). Interestingly, these pyknotic, AC3- nuclei were sparsely present in P0 and P4 control retinas (Figure 4A and C) and to a larger extent in P0 and P4 knockout retinas (Figure 4B and D). Quantification (Figure 4E) reveals a general trend of cell death consistent with our histology and cell-marker investigations: nuclear pyknosis is slightly elevated in P0 knockout retinas compared to controls, peaks at P4 where we observe robust retinal degeneration, and is virtually absent by P10, by which time the majority of outer and inner nuclei in the knockout are lost (Figure 1F). Interestingly, we observe roughly equal amounts of AC3+ and AC3- pyknotic cells in P0 knockout retinas, whereas by P4 AC3- pyknotic cells constitute ~30% of pyknotic cells in knockout retinas (Figure 4E). In addition to being present at all tested ages and following the same general trend as AC3+ pyknotic cells, AC3- pyknotic cells often appear in distinct clusters (Figure 4D'), distinguishable from the more evenly dispersed AC3+ pyknotic cells. These results suggest the activation of at least two distinct cell death pathways in NMNAT1 knockout retinas between P0 and P4. Figure 4 with 2 supplements see all Download asset Open asset NMNAT1 loss causes activation of multiple cell death pathways in the retina. (A–D) Representative retinal sections from knockout (-/-) and floxed littermate control (+/+) mice at the indicated ages labelled with an antibody against active Caspase-3 (AC3). Corresponding zoom panels are indicated with dotted rectangles. Arrows denote pyknotic, AC3-negative nuclei. (E) Quantification of pyknotic nuclei in sections from knockout and control mice at the indicated ages, grouped by presence (AC3+) or absence (AC3-) of active Caspase-3 labeling. Relative expression of several apoptotic (F), pyroptotic (G), and necroptotic (H) genes in P4 knockout retinas as assessed by RNA-sequencing. (I) Relative expression of Sarm1 in P4 knockout and control retinas as assessed by RNA-sequencing. (J) Relative abundance of cyclic-ADP-ribose (cADPR) in P4 knockout and control retinas as measured by mass spectrometry (grey bars represent means). Data are represented as mean ± SD. Significance determined using unpaired t-tests for (E) and (J) or DESeq2 for (F–I) (see Materials and methods). n = 3 biological replicates per condition for (A–E), n = 5 biological replicates for (F–I), n = 6 biological replicates (one outlier removed) for (J). Scale bars, 30 μm. Figure 4—source data 1 Numerical source data for quantification of Caspase-3-positive (AC3+) and Caspase-3-negative (AC3-) pyknotic cells in WT and KO retinal sections. Numerical source data for expression (relative to WT) of cell death-related genes in P4 KO retinas. Numerical source data for relative abundance of cyclic ADP-ribose (cADPR) in P4 WT and KO retinas. https://cdn.elifesciences.org/articles/71185/elife-71185-fig4-data1-v2.xlsx Download elife-71185-fig4-data1-v2.xlsx To more comprehensively characterize NMNAT1-associated cell death and identify possible caspase 3-independent cell death pathways in our knockout, we leveraged the E18.5 and P4 RNA-sequencing datasets mentioned above. This allowed us to systematically assay the expression of a collection of genes associated with several major cell death pathways (Figure 4—figure supplement 2). Consistent with AC3 staining, we observe deregulation of a collection of apoptosis-related genes in P4 knockout retina, including significant increases in Noxa and Fas, two pro-apoptotic genes previously associated with cell death in NMNAT1-deficient retinas (Kuribayashi et al., 2018; Figure 4F). Notably, two of these genes—Noxa and Chop—are also significantly deregulated at E18.5, prior to significant retinal degeneration (Figure 4—figure supplement 1A). In addition to transcriptional signatures of apoptosis, we identified upregulation of a collection of genes associated with pyroptosis in P4 NMNAT1 knockout retinas (Figure 4G). Pyroptosis is characterized by assembly of a multi-protein complex called the ‘inflammasome,’ which ultimately cleaves and activates the pore-forming members of the gasdermin family of proteins to elicit lytic cell death in response to a variety of perturbations (Swanson et al., 2019; McKenzie et al., 2020). Interestingly, we find upregulation of all three classical inflammasome components—Nlrp3, Casp1, and Pycard (ASC)—in P4 knockout retinas (Figure 4G), with Nlrp3 upregulation at E18.5 as well (Figure 4—figure supplement 1B). In addition, we observe significant increases in Irf2, a transcriptional activator of gasdermin D (Kayagaki et al., 2019), as well as pyroptosis-associated proteins Ripk1 and Tlr2 at P4 (Figure 4G). Notably, expression of Tlr2 and related protein Tlr4 is also significantly elevated in E18.5 knockout retinas (Figure 4—figure supplement 1B). Finally, we also observed dysregulation of several genes associated with necroptosis (Figure 4H) and ferroptosis (Figure 4—figure supplement 2D) in P4 knockout retinas; while none of these genes were significantly upregulated at E18.5, we do observe an early induction of necroptosis-associated protein Nox2 at this age (Figure 4—figure supplement 1C). Recently, photoreceptor cell death in a postnatally induced global NMNAT1 knockout mouse was shown to depend heavily on the activity of the pro-degenerative axonal protein SARM1 (Sasaki et al., 2020b). Reasoning SARM1 as the culprit behind the caspase 3-independent cell death in our model, we checked Sarm1 expression in our RNA-seq data and assayed SARM1 activity by measuring levels of its catalytic product cyclic ADP-ribose (cADPR) using targeted mass spectrometry in P4 and E18.5 NMNAT1 knockout and control retinas. Surprisingly, we found no significant changes in SARM1 expression (Figure 4I, Figure 4—figure supplement 1D) or activity (Figure 4J, Figure 4—figure supplement 1E) at either tested age. Overall, these data reveal that activation of multiple cell death pathways underlies the early and severe degeneration observed in NMNAT1 knockout retinas, and suggest pyroptosis, necroptosis and apoptosis as potential drivers of this degeneration. Global metabolic alterations in NMNAT1-deficient retinas To identify possible mechanisms for the severe and cell-type-specific retinal degeneration in our model, we next sought to characterize global metabolic consequences of embryonic NMNAT1 deletion in the retina. To this end, we used targeted liquid chromatography-tandem mass spectrometry (LC-MS/MS) to quantify levels of ~112 cellular metabolites spanning many essential biochemical pathways in NMNAT