Transcriptomic Changes Predict Metabolic Alterations in LC3 Associated Phagocytosis in Aged Mice

LC3b (Map1lc3b) plays an essential role in canonical autophagy and is one of several components of the autophagy machinery that mediates non-canonical autophagic functions. Phagosomes are often associated with lipidated LC3b to promote phagosome maturation in a process called LC3-associated phagocytosis (LAP). Specialized phagocytes, such as mammary epithelial cells, retinal pigment epithelial (RPE) cells, and sertoli cells, utilize LAP for optimal degradation of phagocytosed material, including debris. In the visual system, LAP is critical to maintain retinal function, lipid homeostasis, and neuroprotection. In a mouse model of retinal lipid steatosis-mice lacking LC3b (LC3b−/−), we observed increased lipid deposition, metabolic dysregulation, and enhanced inflammation. Herein, we present a non-biased approach to determine if loss of LAP mediated processes modulate the expression of various genes related to metabolic homeostasis, lipid handling, and inflammation. A comparison of the RPE transcriptome of WT and LC3b−/− mice revealed 1533 DEGs, with ~73% upregulated and 27% downregulated. Enriched gene ontology (GO) terms included inflammatory response (upregulated DEGs), fatty acid metabolism, and vascular transport (downregulated DEGs). Gene set enrichment analysis (GSEA) identified 34 pathways; 28 were upregulated (dominated by inflammation/related pathways) and 6 were downregulated (dominated by metabolic pathways). Analysis of additional gene families identified significant differences for genes in the solute carrier family, RPE signature genes, and genes with a potential role in age-related macular degeneration. These data indicate that loss of LC3b induces robust changes in the RPE transcriptome contributing to lipid dysregulation and metabolic imbalance, RPE atrophy, inflammation, and disease pathophysiology.


Introduction
Autophagy, the activity of "self-eating", is vital for cell homeostasis; this catabolic process degrades old, damaged, or abnormal organelles and other substances in the cell cytoplasm [1]. Macroautophagy (hereafter referred to as autophagy or canonical autophagy (CA)) is evolutionarily conserved across eukaryotes and involves formation of a double membrane autophagosome, a specialized structure that engulfs cellular components such as protein aggregates, lipids, and damaged organelles [2]. Upstream signals, such as nutrient deprivation or stress, trigger the activation of a multistep autophagy cascade involving several autophagy related proteins (ATGs), culminating in the lipidation of microtubule-associated protein light chain 3 (LC3/ATG8) to phosphatidylethanolamine (PE) during autophagosome biogenesis, a step necessary to target autophagosomes for Int. J. Mol. Sci. 2023, 24, 6716 2 of 23 lysosomal degradation [3][4][5][6]. Lipid homeostasis requires specialized autophagy processes to maintain mitochondrial and peroxisome health through mitophagy and pexophagy, respectively [5,7].
A sub-set of the autophagy related proteins are involved in a hybrid autophagyphagocytosis pathway, termed LC3-associated phagocytosis (LAP), considered a noncannonical autophagy pathway [8][9][10][11][12]. Professional phagocytes such as macrophages, monocytes, and dendritic cells, as well as non-professional/specialized phagocytes, such as mammary epithelial cells, retinal pigment epithelial (RPE) cells, and sertoli cells utilize LAP for optimal degradation of phagocytosed material, including dead cells/cellular debris, and to recycle nutrients, such as vitamin A derivatives [6,11,[13][14][15]. While LAP shares molecular machinery with CA, it is a distinct process; unlike CA, LAP is AMPK-mTORC1-ULK1 independent and unresponsive to nutrient status [8,11,16]. In LAP, a sub-set of autophagy machinery proteins conjugate lipidated LC3 (LC3II) to single membrane phagosome to form a LAPosome followed by degradation in the lysosome [8]. LAP and CA are regulated by RUBCN (Run domain Beclin-1 interacting and cysteine-rich containing protein); this protein promotes LAP and concomitantly suppresses CA [17,18]. Defects in macrophage LAP lead to enhanced pro-inflammatory cytokine production in chronic obstructive pulmonary disease (COPD) and in response to cigarette smoke [10,19]. In vitro knockdown of LC3b in lung epithelia sensitized these cells to bleomycin-induced apoptosis [20]. Similarly, the LC3b −/− mouse shows increased susceptibility to lung injury and fibrosis [20]. LAP mediates an anti-inflammatory and immune-suppressive response that, in the context of the tumor microenvironment, leads to evasion of immune surveillance resulting in tumor progression; LAP deficiency results in hyper-inflammation [21][22][23]. Numerous bacteria have evolved strategies to actively evade (or exploit) host cell LAP as a survival mechanism [9,24].
In the eye, LAP supports visual function and lipid homeostasis in the post-mitotic RPE cells. The RPE, residing in close apposition to photoreceptors, supplies nutrients to the retina, regulates fluid/ion balance, supports the visual cycle, and maintains the blood retinal barrier [11,25,26]. These functions rely on daily phagocytic clearance of lipid-rich photoreceptor outer segment tips over the lifetime of the retina [27]. RPE-LAP requires melanoregulin (Mreg, an LC3 interacting protein)-dependent recruitment of LC3II for subsequent phagosome maturation and phagolysosome formation, leading to degradation of the phagocytosed cargo [12,28]. Some of the degradation by-products are removed from the cell by transport to the choroid, and others are recycled; they replenish essential components and metabolites needed by the neural retina [6,11,29]. Mouse models of defective RPE-LAP (Atg5∆RPE, Mreg −/− and LC3b −/− ) exhibit increased phagosome accumulation, inadequate recovery of retinoids, and reduced visual capacity [11,12,28,30].
RPE is intimately associated with the neural retina, and RPE-LAP plays an important role in lipid regulation in the eye. Indeed, several studies have demonstrated that lipid dysregulation contributes to the pathogenesis of age-related macular degeneration (AMD) [31][32][33][34]. The RPE exploits LAP to efficiently degrade lipid and protein rich OS material daily; it metabolizes OS lipids to generate β-hydroxybutyrate (by ketogenesis), which may support the energy needs of the outer retina [35,36]. The abrogation of LAP, defective processing of lipid rich phagosomes in LC3b −/− RPE, causes phagosome accumulation, resulting in a decrease in fatty acid oxidation and ketogenesis, as well as disrupted lipid homeostasis. This results in an increase in RPE and sub-RPE lipid deposits, elevated levels of lipid peroxidation products, pro-inflammatory sterol, 7-ketocholesterol (7KCh), and decreased levels of lipid pro-survival factors Neuroprotectin D1 (NPD1) and Maresin-1 [28]. Moreover, these mice show a sub-retinal recruitment of microglia [28]. These studies point to a vital role for LAP in lipid regulation with a loss of lipid/cholesterol homeostasis, contributing to an AMD-like phenotype [11,12,28,30]. Over time in aging LC3b −/− mice, a chronic low-grade inflammatory response (para-inflammation) is manifested, which, combined with observed lipid steatosis, is reminiscent of non-alcoholic fatty liver disease [37].
Herein, we investigate how a loss of LC3b affects gene expression using an unbiased, RNAseq based comparative transcriptomics approach, thus allowing us to further uncover the mechanistic aspects of defective LAP leading to inflammation and lipid metabolic dysregulation. We provide a detailed report of the differentially expressed genes (DEGs) and perform gene ontology-based enrichment analyses and hallmark pathway analyses to decipher affected pathways. Our results show that in the LC3b −/− mice, inflammatory response, cholesterol homeostasis, and complement activation are upregulated while fatty acid metabolism and oxidative phosphorylation are downregulated. Moreover, the data provides novel insights into gene expression changes in the solute carrier family and some RPE signature genes suggesting metabolic changes resulting from defective LAP. Finally, we identify 8 DEGs that have corresponding risk alleles reported for AMD in a large genomewide association study (GWAS) [38] and experimentally validate the correlation between gene transcription (based on RNAseq) and protein expression for one of these DEGs. These results point to molecular links between defective LAP and AMD-like pathology and potentially can be extended to other age-related pathologies.

RNAseq Data Source and Quality (High Throughput RNA Sequencing Data Set Preparation for Analysis)
Aged LC3b −/− mice show retinal lipid imbalance and enhanced pro-inflammatory micro-environment. This is due to RPE and sub-RPE lipid deposits and a concomitant decrease in the synthesis of protective lipid species, NPD1, and maresin-1 [28]. To identify RPE-specific changes in the LC3b −/− mouse at the gene expression level, we prepared RNA from RPE isolated from mouse retina. The experimental design used for RPE cell isolation for subsequent RNA preparation and sequencing is illustrated in Figure 1A. Sequencing was performed using Illumina NovaSeq 6000 SP flow cell with the XP workflow (100 bp single read sequencing) to a minimum depth of 28.54 MR with a median depth of 30.4 MR (+/− 5.8 MR). Total reads for the samples ranged from 26,704,479 to 35,616,153. For each sample, over 85% of reads mapped to the transcriptome. Principal component analysis (PCA) shows the overall variation among the samples with a clear separation in the first two components ( Figure 1B). which, combined with observed lipid steatosis, is reminiscent of non-alcoholic fatty liver disease [37].
Herein, we investigate how a loss of LC3b affects gene expression using an unbiased, RNAseq based comparative transcriptomics approach, thus allowing us to further uncover the mechanistic aspects of defective LAP leading to inflammation and lipid metabolic dysregulation. We provide a detailed report of the differentially expressed genes (DEGs) and perform gene ontology-based enrichment analyses and hallmark pathway analyses to decipher affected pathways. Our results show that in the LC3b −/− mice, inflammatory response, cholesterol homeostasis, and complement activation are upregulated while fatty acid metabolism and oxidative phosphorylation are downregulated. Moreover, the data provides novel insights into gene expression changes in the solute carrier family and some RPE signature genes suggesting metabolic changes resulting from defective LAP. Finally, we identify 8 DEGs that have corresponding risk alleles reported for AMD in a large genome-wide association study (GWAS) [38] and experimentally validate the correlation between gene transcription (based on RNAseq) and protein expression for one of these DEGs. These results point to molecular links between defective LAP and AMD-like pathology and potentially can be extended to other age-related pathologies.

RNAseq Data Source and Quality (High Throughput RNA Sequencing Data Set Preparation for Analysis)
Aged LC3b −/− mice show retinal lipid imbalance and enhanced pro-inflammatory micro-environment. This is due to RPE and sub-RPE lipid deposits and a concomitant decrease in the synthesis of protective lipid species, NPD1, and maresin-1 [28]. To identify RPE-specific changes in the LC3b −/− mouse at the gene expression level, we prepared RNA from RPE isolated from mouse retina. The experimental design used for RPE cell isolation for subsequent RNA preparation and sequencing is illustrated in Figure 1A
design to isolate RPE from mouse eyes; ~24-month old WT and LC3b −/− mice were sacrificed, and the eyeballs were removed and processed to isolate RPE cells by enzymatic treatments as indicated. Total RNA was extracted and used for library preparation and RNA sequencing. (B) Principal component analysis (PCA), X-axis, and Y-axis show PC1 and PC2, respectively.

RNA-Sequencing Reveals Pathways Related to Inflammation and Metabolism
We performed Gene Ontology (GO) based enrichment analysis using the Metascape tool under 3 categories: Molecular Functions (MF), Cellular components (CC), and Biological Processes (BP). Table 1 shows a list of GO terms for top 10 DEGs under each of these categories. Closer examination of these GO terms revealed that the majority overlap with upregulated DEGs. This is not surprising since there are many more upregulated DEGs (>2.5 fold) in our list. Next, we performed a similar enrichment analysis across all upregulated and downregulated genes separately. Figure 3A shows top 5 most enriched GO terms for each of the 3 categories; more than 5 GO terms have been shown in cases where LogP values were exactly the same for more than one term. Interestingly, some of the most enriched GO terms for the upregulated genes are related to inflammation (Inflammatory response and cellular response to cytokine stimulus) and cell-cell adhesion

RNA-Sequencing Reveals Pathways Related to Inflammation and Metabolism
We performed Gene Ontology (GO) based enrichment analysis using the Metascape tool under 3 categories: Molecular Functions (MF), Cellular components (CC), and Biological Processes (BP). Table 1 shows a list of GO terms for top 10 DEGs under each of these categories. Closer examination of these GO terms revealed that the majority overlap with upregulated DEGs. This is not surprising since there are many more upregulated DEGs (>2.5 fold) in our list. Next, we performed a similar enrichment analysis across all upregulated and downregulated genes separately. Figure 3A shows top 5 most enriched GO terms for each of the 3 categories; more than 5 GO terms have been shown in cases where LogP values were exactly the same for more than one term. Interestingly, some of the most enriched GO terms for the upregulated genes are related to inflammation (Inflammatory response and cellular response to cytokine stimulus) and cell-cell adhesion ( Figure 3A), whereas the downregulated DEGs were enriched in GO terms related to metabolism, including fatty acid metabolic process, transmembrane transporters, and mitochondria ( Figure 3B). There was no significant enrichment of GO terms corresponding to phagosome maturation (GO:0090382) or phagosome lysosome fusion (GO:0090385). Reactome pathway analysis also supported these results; interleukin and cytokine signaling seen in 2 of the top 5 upregulated DEGs ( Figure 3C) while metabolism of lipids and fatty acid metabolism was seen in the downregulated DEGs ( Figure 3D).   In addition, an unfiltered list of genes ranked by their DESeq2 statistic was analyzed with GSEA against the hallmark gene collections in the Molecular Signatures Database (MSigDB). GSEA identified 34 significantly different (FDR q-val < 0.05) hallmark pathways; 28 were upregulated ( Figure 4, top), and 6 were downregulated (Figure 4, bottom). Upregulated pathways were dominated by inflammation/related pathways (eg, inflammatory response, complement, cholesterol homeostasis); those enriched by down-regulated genes were dominated by metabolic pathways (eg., oxidative phosphorylation, fatty acid metabolism, and peroxisome). We have previously established that LC3b plays a critical role in lipid-mediated homeostasis in the RPE; loss of LC3b leads to decrease in fatty acid oxidation, peroxisomal turn-over defects, age-dependent lipid accumulation, oxidative stress, and pro-inflammatory microenvironment [28,42]. LC3b −/− -RPE exhibits decrease in fatty acid oxidation and ketogenesis along with chronic low-grade inflammation [28]. To understand the contribution of gene expression changes associated with lipid metabolic imbalance and pro-inflammatory environment, we further examined some of the key hallmark pathways in relation to RPE physiology and metabolism ( Figure 4, pathways marked by arrowheads). The hallmark inflammatory response gene set is composed of 196 genes; 112 genes were part of the leading edge with a normalized enrichment score (NES) of 2.39 ( Figures 5A and S1A). The hallmark complement gene set includes 187 genes; 85 genes were part of the leading edge with an NES of 2.1 ( Figures 5B and S1B). The hallmark cholesterol homeostasis gene set includes 74 genes; 38 genes were part of the leading edge with an NES of 2.1 ( Figures 5C and S1C). LC3b −/− -RPE exhibits decrease in fatty acid oxidation and ketogenesis along with chronic low-grade inflammation [28]. To understand the contribution of gene expression changes associated with lipid metabolic imbalance and pro-inflammatory environment, we further examined some of the key hallmark pathways in relation to RPE physiology and metabolism ( Figure 4, pathways marked by arrowheads). The hallmark inflammatory response gene set is composed of 196 genes; 112 genes were part of the leading edge with a normalized enrichment score (NES) of 2.39 ( Figures 5A and S1A). The hallmark complement gene set includes 187 genes; 85 genes were part of the leading edge with an NES of 2.1 ( Figures 5B and S1B). The hallmark cholesterol homeostasis gene set includes 74 genes; 38 genes were part of the leading edge with an NES of 2.1 ( Figures 5C and S1C).
) were selected for further analysis.
LC3b −/− -RPE exhibits decrease in fatty acid oxidation and ketogenesis along with chronic low-grade inflammation [28]. To understand the contribution of gene expression changes associated with lipid metabolic imbalance and pro-inflammatory environment, we further examined some of the key hallmark pathways in relation to RPE physiology and metabolism (Figure 4, pathways marked by arrowheads). The hallmark inflammatory response gene set is composed of 196 genes; 112 genes were part of the leading edge with a normalized enrichment score (NES) of 2.39 ( Figures 5A and S1A). The hallmark complement gene set includes 187 genes; 85 genes were part of the leading edge with an NES of 2.1 ( Figures 5B and S1B). The hallmark cholesterol homeostasis gene set includes 74 genes; 38 genes were part of the leading edge with an NES of 2.1 ( Figures 5C and S1C). The hallmark oxidative phosphorylation gene set is composed of 199 genes; 138 genes were part of the leading edge with an NES of −3.02 ( Figures 5D and S2A). The hallmark fatty acid metabolism gene set includes 155 genes; 57 genes were part of the leading edge with an NES of −1.72 ( Figures 5E and S2B). The hallmark peroxisome gene set is composed of 102 genes; 35 genes were part of the leading edge with an NES of −1.72 ( Figures 5F and  S2C).

Several Solute Carrier Family Genes Are Differentially Expressed
Transporters play one of the most essential roles in metabolic homeostasis [43]. Anion transmembrane transporter activity was identified as one of the top GO terms in the pathway enrichment analysis of downregulated DEGs ( Figure 3B). As RPE is intimately connected with choroid (basal) and neural retina (apical) and has an important regulatory role in movement of metabolites and ions into and out of retina, we broadened our analysis to look specifically at solute carrier family gene expression. There were 43 DEGs (depicted in the heat map, Figure 6A top) between LC3b −/− and WT out of a total of 390 solute The hallmark oxidative phosphorylation gene set is composed of 199 genes; 138 genes were part of the leading edge with an NES of −3.02 (Figures 5D and S2A). The hallmark fatty acid metabolism gene set includes 155 genes; 57 genes were part of the leading edge with an NES of −1.72 ( Figures 5E and S2B). The hallmark peroxisome gene set is composed of 102 genes; 35 genes were part of the leading edge with an NES of −1.72 ( Figures 5F and S2C).

Several Solute Carrier Family Genes Are Differentially Expressed
Transporters play one of the most essential roles in metabolic homeostasis [43]. Anion transmembrane transporter activity was identified as one of the top GO terms in the pathway enrichment analysis of downregulated DEGs ( Figure 3B). As RPE is intimately connected with choroid (basal) and neural retina (apical) and has an important regulatory role in movement of metabolites and ions into and out of retina, we broadened our analysis to look specifically at solute carrier family gene expression. There were 43 DEGs (depicted in the heat map, Figure 6A top) between LC3b −/− and WT out of a total of 390 solute carrier genes (χ2 = 9.599; two-tailed p value < 0.05). Interestingly, Slc16a1 and Slc16a8 encoding for MCT1 and MCT3, respectively, two of the key lactate transporter genes expressed in the RPE [44] are downregulated while Slc16a11 (encoding for MCT11, an orphan transporter), expression is upregulated in LC3b −/− RPE [45]. Moreover, expression of Bsg, encoding for Basigin (CD147), a protein required for proper trafficking of MCT1 and MCT3 to the plasma membrane, was also downregulated (Table S1, Figure 6A, bottom) [46,47]. Consistent with the transcriptomic analysis, we observed~40% decrease in MCT3 protein levels in the LC3b −/− compared to the WT control ( Figure 6B). Mouse RPE/choroid flat mounts also showed a decrease in MCT3 with a mis-localization of much of the remaining MCT3 as intracellular and not plasma membrane-associated, consistent with the decrease in Bsg ( Figure 6C). carrier genes (χ2 = 9.599; two-tailed p value <0.05). Interestingly, Slc16a1 and Slc16a8 encoding for MCT1 and MCT3, respectively, two of the key lactate transporter genes expressed in the RPE [44] are downregulated while Slc16a11 (encoding for MCT11, an orphan transporter), expression is upregulated in LC3b −/− RPE [45]. Moreover, expression of Bsg, encoding for Basigin (CD147), a protein required for proper trafficking of MCT1 and MCT3 to the plasma membrane, was also downregulated (Table S1, Figure 6A, bottom) [46,47]. Consistent with the transcriptomic analysis, we observed ~40% decrease in MCT3 protein levels in the LC3b −/− compared to the WT control ( Figure 6B). Mouse RPE/choroid flat mounts also showed a decrease in MCT3 with a mis-localization of much of the remaining MCT3 as intracellular and not plasma membrane-associated, consistent with the decrease in Bsg ( Figure 6C).

Loss of LC3b Is Associated with Differential Expression of Select RPE Signature and Lipid Metabolism Related Genes
Gene expression alterations in some of the key lactate transporter genes (such as Slc16a1) prompted us to evaluate expression of RPE signature and lipid metabolism related genes. We used mouse orthologs from a list of RPE signature genes (human) reported previously [48]. Our analysis revealed significantly different expression profile between LC3b −/− and WT (χ2= 6.43; two-tailed p value < 0.05) with 19 DEGs among 149 genes in the list. Expression profile of the DEGs is depicted in the heat map ( Figure 7A). Gene and P-cadherin (red). Images are projections from 5 µm stacks captured using 60× water objective (NA1.2) on Nikon A1R laser scanning confocal microscope.

Loss of LC3b Is Associated with Differential Expression of Select RPE Signature and Lipid Metabolism Related Genes
Gene expression alterations in some of the key lactate transporter genes (such as Slc16a1) prompted us to evaluate expression of RPE signature and lipid metabolism related genes. We used mouse orthologs from a list of RPE signature genes (human) reported previously [48]. Our analysis revealed significantly different expression profile between LC3b −/− and WT (χ2 = 6.43; two-tailed p value < 0.05) with 19 DEGs among 149 genes in the list. Expression profile of the DEGs is depicted in the heat map ( Figure 7A). Gene expression of Rbp1 (retinoid binding protein 1), Enpp2 (Ectonucleotide Pyrophosphatase/Phosphodiesterase 2), and Ptgds (Prostaglandin D2 Synthase), further suggest some functional decline in the LC3b −/− RPE. Moreover, expression of Slc6a20b (Sit1, a proline and betaine transporter) is upregulated ( Figure 7A). Interestingly, expression of Best1 and Rpe65, two well studied RPE markers did not change significantly ( Figure S3A,B) [49]. expression of Rbp1 (retinoid binding protein1), Enpp2 (Ectonucleotide Pyrophosphatase/Phosphodiesterase 2), and Ptgds (Prostaglandin D2 Synthase), further suggest some functional decline in the LC3b −/− RPE. Moreover, expression of Slc6a20b (Sit1, a proline and betaine transporter) is upregulated ( Figure 7A). Interestingly, expression of Best1 and Rpe65, two well studied RPE markers did not change significantly ( Figure S3A,B) [49]. . For all the heatmaps, the variance stabilized counts were converted to colors using Clustvis software. Rows are centered; unit variance scaling is applied to rows. Rows are clustered using correlation distance and average linkage. The intensity scale ranges from low expression (dark blue) to high expression (red) relative to the median of the gene's expression across all samples. Gene names (left) and mouse ID (bottom) are indicated.
On a molecular scale, our GSEA analysis showed transcriptional changes in fatty acid metabolism and cholesterol homeostasis. We sought to further explore if LAP-mediated processes modulate the level of genes related to lipid transport/handling. Gene expression profiles for lipoprotein metabolism associated genes, including apolipoproteins (30 genes), apolipoprotein B mRNA editing enzyme catalytic polypeptides 1-4 (4 genes), Apobr, Mttp, Pltp and Lcat, Acat1, Acat2, and Lrp8 also showed significant differences between LC3b −/− and WT with 9 DEGs out of 41 genes examined (χ2 = 13.93; two-tailed p value <0.05) ( Figure 7B, left). PPI string diagram ( Figure 7B, right) depicting interaction between the differentially expressed genes Apoc4, Apoc1, Apoa2, Pltp, and Mttp suggests altered lipoprotein metabolism in the LC3b −/− RPE with Apoa2 potentially playing a central role. . For all the heatmaps, the variance stabilized counts were converted to colors using Clustvis software. Rows are centered; unit variance scaling is applied to rows. Rows are clustered using correlation distance and average linkage. The intensity scale ranges from low expression (dark blue) to high expression (red) relative to the median of the gene's expression across all samples. Gene names (left) and mouse ID (bottom) are indicated.
On a molecular scale, our GSEA analysis showed transcriptional changes in fatty acid metabolism and cholesterol homeostasis. We sought to further explore if LAP-mediated processes modulate the level of genes related to lipid transport/handling. Gene expression profiles for lipoprotein metabolism associated genes, including apolipoproteins (30 genes), apolipoprotein B mRNA editing enzyme catalytic polypeptides 1-4 (4 genes), Apobr, Mttp, Pltp and Lcat, Acat1, Acat2, and Lrp8 also showed significant differences between LC3b −/− and WT with 9 DEGs out of 41 genes examined (χ2 = 13.93; two-tailed p value < 0.05) ( Figure 7B, left). PPI string diagram ( Figure 7B, right) depicting interaction between the differentially expressed genes Apoc4, Apoc1, Apoa2, Pltp, and Mttp suggests altered lipoprotein metabolism in the LC3b −/− RPE with Apoa2 potentially playing a central role.

DEGs with Potential Role in Age-Related Retinal Diseases
Dysregulation in lipid/cholesterol homeostasis is linked to age-related diseases and loss of RPE function [55][56][57]. Several features observed in the LC3b −/− RPE, such as reduced phagocytosis, lipid deposits, recruitment of immune cells, and complement pathway activation are reminiscent of AMD-like pathophysiology [28,[58][59][60]. To gain insight into the molecular changes underlying this pathology, we analyzed expression of genes with potential role in AMD.
We started with a list of 41 genes based on a list identified as genes with top priority and statistical significance combined in a genome-wide association study (GWAS) published by the International AMD Genomics Consortium (IAMDGC). Currently, this is both the most recent and largest AMD GWAS study and reported 52 independent genetic signals distributed over 34 loci associated with AMD at genome-wide significance [38,61]. There was a significantly different expression profile between LC3b −/− and WT (χ2 = 12.02; twotailed p value = 0.0005) with 8 DEGs out of 37 mouse orthologs identified (depicted in the heat map, Figure 8), including Rdh5, Slc16a8, and Htra1. Slc16a8 (MCT3) plays an important role in regulating ionic composition of outer retina and Slc16a8 −/− mice showed reduced visual function [62]. A subset of patients with RDH5 (retinol dehydrogenase 5) mutation develop macular atrophy; reduced RDH5 activity is considered a risk factor for AMD [63,64]. Increased protein expression of HTRA1 (High Temperature Requirement A Serine Peptidase 1), a serine protease, was seen in iPSC-derived RPE cell line from subjects carrying AMD risk-associated 10q26 locus [65]. Lipoprotein metabolism is intimately linked to cholesterol biosynthesis and transport. GSEA of cholesterol homeostasis pathways show upregulation of genes for cholesterol biosynthesis (Hmgcr, Hmgcs1, Fdps, Sqle, Cyp51a1, Idl1, Nsdhl, and others), as well as genes associated with cholesterol trafficking/import (Stard4, Ldlr) [50]. Core enrichment subset in the cholesterol homeostasis pathway also includes genes coding for proteins involved in lipid metabolic disorders, some of which link lipoprotein metabolism with cholesterol homeostasis (Lpl, Ldlr) and the cholesterol biosynthetic enzymes (Dhcr7, Nsdhl, and Ebp) (depicted as heat map in Figure 7C) [51,52]. GSEA of fatty acid metabolism pathway shows downregulation of genes involved in ketogenesis (Hmgcs2), several genes critical for the TCA cycle (Sdhc, Sdha, Sdhd, Aco2, Suclg1), peroxisomal fatty acid oxidation (Acox1, Acaa1), and mitochondrial β-oxidation (Acadl, Acadvl, Cpt2, Cpt1a) (depicted as heat map in Figure 7D). Acox1 catalyzes desaturation of acyl-CoA esters, which is the initial committed step in peroxisomal fatty acid oxidation [53,54].

DEGs with Potential Role in Age-Related Retinal Diseases
Dysregulation in lipid/cholesterol homeostasis is linked to age-related diseases and loss of RPE function [55][56][57]. Several features observed in the LC3b −/− RPE, such as reduced phagocytosis, lipid deposits, recruitment of immune cells, and complement pathway activation are reminiscent of AMD-like pathophysiology [28,[58][59][60]. To gain insight into the molecular changes underlying this pathology, we analyzed expression of genes with potential role in AMD.
We started with a list of 41 genes based on a list identified as genes with top priority and statistical significance combined in a genome-wide association study (GWAS) published by the International AMD Genomics Consortium (IAMDGC). Currently, this is both the most recent and largest AMD GWAS study and reported 52 independent genetic signals distributed over 34 loci associated with AMD at genome-wide significance [38,61]. There was a significantly different expression profile between LC3b −/− and WT (χ2= 12.02; two-tailed p value = 0.0005) with 8 DEGs out of 37 mouse orthologs identified (depicted in the heat map, Figure 8), including Rdh5, Slc16a8, and Htra1. Slc16a8 (MCT3) plays an important role in regulating ionic composition of outer retina and Slc16a8 −/− mice showed reduced visual function [62]. A subset of patients with RDH5 (retinol dehydrogenase 5) mutation develop macular atrophy; reduced RDH5 activity is considered a risk factor for AMD [63,64]. Increased protein expression of HTRA1 (High Temperature Requirement A Serine Peptidase 1), a serine protease, was seen in iPSC-derived RPE cell line from subjects carrying AMD risk-associated 10q26 locus [65].

Discussion
Post-mitotic cells such as neurons, RPE cells, and cardiac myocytes utilize autophagic processes to sustain homeostasis; these cells get rid of unwanted material (waste) and recycle cellular components, thus maintaining intracellular quality control over their lifetime [6,66]. In the RPE, LAP plays a necessary role in cell homeostasis through the optimal clearance of phagocytosed OS fragments; this supports (1) the visual cycle by recycling retinoids, (2) retinal metabolism (e.g., by metabolizing lipids to generate ketones and thus providing energy to the neural retina), and by producing insulin locally to support homeostasis, (3) RPE and retinal health and function by preventing buildup of excess lipids, (4) ensuring the synthesis of anti-inflammatory lipids [11,28,30,43,67]. LAP thus serves a protective role in the retina. We have previously established that the LC3b, an LC3 isoform, plays a crucial role in RPE-LAP; it is critical for lipid homeostasis and in regulating the inflammatory state of the retina/RPE [5,28,41]. Here, using RNAseq based comparative transcriptomics, we identified 1533 DEGs with~73% of these upregulated and 27% downregulated in the LC3b −/− RPE. There was no compensatory upregulation of any other Atg8/LC3 family members-LC3a, Gabarap, Gabarapl1, and Gabarapl2, consistent with previous studies showing no upregulation of LC3a in LC3b −/− [5,28,39,41,68]. In the absence of LC3b, inflammatory and complement-related pathways are upregulated while fatty acid metabolism, oxidative phosphorylation, and peroxisome pathways are downregulated. To the best of our knowledge, this is the first report using RNAseq based expression profiling in a defective LAP model.

LC3b and Lipid Metabolic Dysregulation
Daily phagocytosis of OS presents a unique set of challenges for the RPE; estimates suggest that each RPE processes~0.08-0.15 pmoles of OS derived fatty acid per day [6,27,35]. Excessive lipid accumulation and formation of lipid-peroxidation adducts in RPE/Ch is a common feature of mouse models of defective phagosome transport and degradation [30,56,69]. The RPE utilizes LAP in the daily ingestion of lipids that it degrades to provide metabolic intermediates; the lipids provide a critical energy source for the RPE through fatty acid oxidation and metabolically couple with photoreceptors to provide β-hydroxybutyrate (β-HB) [35,36,42].
Herein, we found that the hallmark pathways for oxidative phosphorylation and fatty acid metabolism are significantly downregulated in 24-month-old old LC3b −/− mice. Leading-edge gene subsets of these pathways include Acadvl, Cpt2, Cpt1a, and Acadl (involved in mitochondrial fatty acid β-oxidation), Sdhc, Sdha, Sdhd, Aco2, Suclg1 (involved in TCA cycle), and Hmgcs2 (rate limiting enzyme in ketogenesis). Approximately 30% of OS lipids are very long chain fatty acids (VLCFAs, C ≥ 20), which are preferentially catabolized in peroxisomes via β-oxidation [53,54]. Therefore, efficient lipid catabolism in the RPE requires the coordination of β-oxidation pathways within both mitochondria and peroxisomes [42,53,54,70]. Our GSEA analysis showed significant downregulation of peroxisome pathways, both biogenesis and turnover, as indicated by alterations in Acox1 and Cat, as well as Pex6, Pex11a, and Pex2. The protein products of the Acox gene family, the acyl-CoA oxidases (ACOXs) catalyze desaturation of the AcylCoA esters in the committed step of peroxisome fatty acid oxidation [53,54]. Hydrogen peroxide generated in the ACOX reaction is rapidly degraded by the anti-oxidant enzyme catalase in a coupled reaction. Peroxisomes are the main site for catalase dependent antioxidant activity to maintain redox homeostasis of the RPE [71]. In the RPE, peroxisome turnover is an LC3b dependent process; LC3b −/− RPE have elevated peroxisome numbers with reduced antioxidant function reminiscent of aging/disease RPE [42,71,72]. Furthermore, peroxisome catalase activity is lower in primary RPE from aged donors and those with AMD [73]. While the interplay between LAP, peroxisome FAO and peroxisome turnover is a subject of future investigation; the downregulation of the hallmark peroxisome pathway in the LC3b −/− further lends support to decline in peroxisome function contributing to lipid dysregulation, higher oxidative stress, and disease pathogenesis.
Mitochondrial ketogenesis often serves as a fat disposal conduit, specifically when fatty acid oxidation exceeds mitochondrial capacity [37,74]. Such decreased capability may be due to mitochondrial dysfunction, aging, or disease [75], as well as lipid overload (for example, after a high fat meal). By disposing of acetyl-CoA, a product of FAO, ketogenesis prevents the need for terminal oxidation through the TCA cycle thus decreasing oxidative stress due to enhanced TCA flux [76]. Progressive mitochondrial dysfunction is a major contributor to the pathophysiology of age-related diseases of the outer retina [55,77,78]. Mice lacking Hmgcs2 exhibited ketogenic deficiency and develop non-alcoholic fatty liver disease spontaneously [79]. RetSat, coding for retinol saturase, an oxidoreductase with diverse functions, including retinoid metabolism, was also included in the leading edge gene subsets. In adipocytes, RetSat expression is regulated by PPAR γ and is proposed to be involved in insulin sensitivity [80,81]. Retsat −/− mice show reduced production of all-trans-13,14-dihy-droretinol from dietary vitamin A, an increased accumulation of neutral lipids, and increased adiposity [80,82].
Our transcriptomic analysis provided unique insight into the regulation of metabolic pathways at the level of gene expression in RPE-LAP deficient cells. Although fatty acid metabolism and oxidative phosphorylation pathways were down regulated, cholesterol homoeostasis pathways were upregulated. The RPE not only synthesizes cholesterol through the mevalonate pathway, but also takes up cholesterol from the circulation, as well as the daily phagocytosis of OS tips with cholesterol homeostasis regulated via cholesterol efflux [83][84][85]. In the aging macula, dysregulation of cholesterol metabolism and lipoprotein processing are considered contributors to the onset and development of age-related degeneration [56]. Non-alcoholic fatty liver disease is similarly associated with dysregulated cholesterol metabolism, and it manifests on a molecular level as cholesterol accumulation in the mitochondria [86]. Schirris and colleagues [87], using metabolic network modeling, suggest cholesterol biosynthesis as a potential compensatory pathway in mitochondrial dysfunction serving as a means by which to restore the NADH and NADPH balance. Our own transcriptomic studies (Figures 5D and S2A) and previous in vitro studies suggest mitochondrial dysfunction in LAP-deficient RPE [88]. If compensatory, cholesterol biosynthesis as a means to maintain homeostasis presents a unique challenge to the RPE, as cholesterol is also taken up through daily the LAP-dependent degradation of OS tips, the loss of which results in high levels of the pro-inflammatory sterol, 7-keto cholesterol, a component of AMD-associated drusen [89].
A critical aspect of cholesterol efflux in the RPE is cholesterol esterification and subsequent packaging and secretion in lipoprotein-like particles (LLP) by the RPE back to the circulation [56]. These lipoproteins called BrM-LLP are distinct from circulating plasma lipoproteins [90]; esterified cholesterol is the largest component of BrM-LLP, it is over 10-fold more abundant than triglyceride [91]. Our transcriptomic analyses highlights modulation of numerous genes associated with lipoprotein metabolism, including ApoO, Pltp, Mttp, ApoC4, and ApoA2 all of which are downregulated. As a group these genes are associated with plasma lipoprotein assembly, remodeling, and clearance. Of note, Mttp, encoding for microsomal transfer protein is necessary for the synthesis of β-lipoproteins, likely BrM-LLP by the RPE [55,90,92]. Interestingly, two genes, ApoBec1 and ApoBec3, involved in apolipoprotein B mRNA editing, are both upregulated. What role this gene editing activity plays in RPE-specific β-lipoproteins assembly is unknown. Further studies are necessary to understand how cholesterol biosynthesis and packaging in lipoproteins is influenced by LAP with an emphasis on therapeutics designed to restore lipid homeostasis [57].
The metabolic fate of tissues is not only regulated by the levels of rate-limiting metabolic enzymes, but also by the expression and localization of solute transporters. Glucose is the primary metabolic substrate of the neural retina and is transported from the choroidal blood supply to the outer retina by GLUT1 (Slc2al) transporters in the apical and basolateral membranes of the RPE. The lactate generated in the outer retina through aerobic glycolysis is transported out of the retina by lactate transporters expressed in the apical (MCT1) and basolateral (MCT3) membranes of the RPE [93,94]. We found the genes encod-ing these transporters (Slc2a1, Slc16a1, and Slc16a8) were all decreased in the LC3b −/− RPE ( Figure 6A). Previous studies have shown MCT3 expression is reduced by mitochondrial dysfunction and RPE atrophy [95]. Furthermore, SLC16a8 (MCT3) is a risk allele for AMD, and the loss of MCT3 expression was observed at early stages of AMD [96,97]. The reduced MCT3 expression in the LC3b −/− RPE is consistent with previous findings, showing that the expression of RPE signature genes are impacted by mitochondrial dysregulation and inflammation [98].

Defective LAP, Inflammation, and Complement System
Progress in understanding LAP in professional and specialized phagocytes has shown some interesting parallels. In macrophages and other immune cells, defective LAP leads to increased inflammatory signaling and inflammatory cytokine production [3,99,100]. RPE cells are one of the "most oxidative environments" in the body [101,102], generating reactive oxygen species resulting from an oxygen-rich environment, high metabolic activity, daily flux of polyunsaturated fatty acids due to OS phagocytosis, and bisretinoid accumulation [6,103,104]. Chronic oxidative stress contributes to numerous retinal degenerative diseases, including AMD [105,106]. In the absence of LC3b, undigested lipids act as substrates for lipid peroxidation reactions; there is an accumulation of the proinflammatory sterol, 7-ketocholesterol (7KCh), and reduced biosynthesis of anti-inflammatory molecules-NPD1 and maresin-1-thus further exacerbating inflammation [24,28,107,108]. Our GSEA analyses show that inflammatory and complement pathways are significantly upregulated in the LC3b −/− , potentially contributing to an AMD-like phenotype. Polymorphisms in complement factor genes in the context of environmental risk factors are associated with AMD [109][110][111]. Several studies point to an intimate relationship between oxidative stress, complement activation, and inflammation in models of AMD [56,109,[112][113][114][115]. Moreover, in an in vitro RPE lipid steatosis model, in which the loss of the autophagy protein, LC3b, results in defective phagosome degradation and metabolic dysregulation, we show that lipid overload results in increased gasdermin cleavage, IL-1β release, lipid accumulation, and decreased oxidative capacity [88]. The current RNAseq analyses expands these in vitro studies, as numerous cytokines and caspases (inflammatory response) are associated with the GSEA leading edge gene subsets in LC3b −/− RPE (see Figures 5A and S1A).

Lapses in LAP and AMD Associated Phenotype
Lipid metabolic imbalance is a hallmark of cardiovascular diseases, type 2 diabetes, Alzheimer's disease and AMD [116][117][118][119][120][121][122]. Moreover, several age-related disorders, including neurodegenerative disorders and AMD are associated with dysregulated autophagy [123]. Defects in autophagy and associated processes in long-lived post-mitotic cells like RPE that are challenged with lipid-rich OS throughout life makes them especially vulnerable to age-related pathologies [11,28,69]. A recent study on a cohort of Finnish patients reported several SNPs in the autophagy related genes, including rs73105013-T for MAP1LC3A that are associated with increased risk for wet AMD [124]. With a range of structural, metabolic, and functional defects, especially with regards to lipid dysregulation and inflammation, LC3b −/− RPE offers an opportunity to investigate the AMD-like phenotype in a mouse model. Using the comparative transcriptomics approach, here, we identified 8 DEGs overlapping with a list of prioritized genes associated with independent genetic signals in the largest AMD GWAS published to date [38,61]. These include: Htra1, Tnfrsf10b, Tgfbr1, Trpm3, Tspan10, Slc16a8, Adamts9, and Rdh5 (mouse orthologs). Four of these genes: Tspan10, Tnfrsf10b, Htra1, and Rdh5 also overlapped with a list of putative causal genes for AMD based on human eye eQTL (expression quantitative trait loci) and GWAS signal co-localization [125]. There is ongoing research in the field to investigate AMD pathogenesis using a range of approaches, such as gene expression studies on post-mortem eyes from AMD patients, in vitro studies with iPSC-RPE cells from subjects carrying risk allele variants, and animal models [126]. iPSCS-RPE cells derived from a patient carrying two copies of SLC16A8 risk allele (rs77968014-G) showed deficit in transepithelial lactate transport [127]. Furthermore, there was a progressive loss of MCT3 with an increasing severity of dry AMD [96]. An increase in the expression of HTRA1 protein was reported in iPSC-RPE cell line from subjects carrying a high-risk genotype at 10q26 locus [65,128], and a clinical trial evaluating the inhibition of HTRA1 is ongoing [129]. Mutations in RDH5 have been reported in a subset of patients with macular atrophy [63,130]. A variant in RDH5 has been shown to be associated with an increased skipping of exon 3, nonsense-mediated decay of the of aberrant transcript, and lower minor allele specific expression, thus providing a potential mechanistic link by which RDH5 allele contributes to AMD risk [64]. Interestingly, expression of both RDH5 and TRPM3 decreases under inflammatory conditions in vitro and has been suggested to contribute to RPE dysfunction in AMD [131,132]. TRPM3 codes for a calcium permeable cation channel activated by noxious heat, neurosteroid pregnenolone sulfate, or osmotic pressure [131][132][133][134]. TRPM3 gene hosts miR-204, a microRNA highly expressed in the RPE that plays an important role in gene regulation and retinal/RPE homeostasis [134][135][136]. Furthermore, cholesterol enrichment inhibits TRPM3 activation and has been suggested as a mechanism for proinflammatory cytokine secretion associated with atherosclerotic processes [137]. Modulation of Trpm3 levels along with cholesterol accumulation and increased inflammation observed in the LC3b −/− mouse RPE offers potential mechanistic insights into AMD pathogenesis [28]. Downregulation of Trpm3 may also reduce inflammatory thermal hyperalgesia as has been seen in Trpm3 −/− mice [138,139]. The RNAseq data using the LC3b −/− mouse model reported here is a valuable addition to bridge the gaps in our understanding of AMD pathophysiology and to identify potential therapeutic targets, and it has the potential to be extended to other lipid metabolic defects/age-related diseases as well.

Animals
LC3b −/− mice (strain name: B6;129P2-Map1Lc3btm1Mrab/J; stock # 009336 [68]) and C57BL6/J wildtype (WT) mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA). The mouse lines were confirmed to be free of the rd8 mutation by Transnetyx (Cordova, TN, USA). Maintenance of mouse colonies and all experiments involving animals were as described previously [12]. Mice were housed under standard cyclic light conditions: 12 h light/12 h dark and fed ad libitum, with both female and male mice used in these studies. All procedures involving animals were approved by the Institutional Animal Care and Use Committees (IACUC) of the University of Pennsylvania and were in accordance with the Association for Research in Vision and Ophthalmology (ARVO) guidelines for use of animals in research.

RPE Cell Isolation
RPE cells from WT (N = 4) and the LC3b −/− (N = 3) mice (age~24 months) eyes were isolated with enzymatic treatment and gentle dissociation as described [141] with minor modifications. Briefly, the mice were anesthetized by CO 2 asphyxiation, the eyes were enucleated, and muscles and connective tissue were removed in ice cold HBSS-H− (HBSS without calcium or magnesium + 10 mM HEPES). Cornea, ciliary body, and lens were removed, and the eyecups were incubated in 1mg/mL Hyaluronidase solution (in HBSS-H−) at 37 • C, 5% CO 2 for 20 min followed by incubation in ice-cold HBSS-H+ (HBSS with calcium and magnesium + 10 mM HEPES) and gently pulling away the neural retina. Next, the RPE/choroid was incubated in 0.25% trypsin at 37 • C, 5% CO 2 for 1 h, transferred to 20% fetal bovine serum (FBS) /HBSS-H+ solution, and gently shaken to detach RPE sheets, followed by centrifugation at 240× g for 5 min to collect the RPE sheets/cells.

RNA Isolation and Library Preparation
Total RNA was isolated with RNAEasy Plus Mini kit (QIAgen, Hilden, Germany) per manufacturer's instructions. RNA concentration, and sample quality was determined using an Agilent bioanalyzer. Total RNA samples with a RIN value of 7.1 (+/− 0.29) and a concentration between 1.1 and 2.7 ng/µL were used. An amount of 15 ng of total RNA were used for library preparation using NEB Ultra II stranded mRNA library kit (New England Biolabs, Ipswich, MA, USA) as per manufacturer's instructions and assessed for quality. Libraries were subjected to 100 bp single read sequencing on a NovaSeq SP flow cell using a NovaSeq 6000 sequencing system (Illumina, San Diego, CA, USA) by the Next-Generation Sequencing Core of the University of Pennsylvania. The raw RNAseq data has been deposited in the Gene Expression Omnibus database (GEO, NCBI) under the accession number GSE225344.

Enrichment Analysis
DESeq2 was used to determine the significance of differential expression between the two groups. Genes showing a fold change of ≤−1.5 or ≥1.5 (Log2-transformed ratio of ≤−0.585 or ≥0.585), and an adjusted p-value (p-value corrected for false discovery rate using Benjamini-Hochberg method) ≤ 0.05 was considered significantly differentially expressed genes (DEGs). The up-and downregulated DEGs were tested for enrichment of Gene Ontology (GO) pathways using Metascape [146]. Additionally, an unfiltered list of genes ranked by their DESeq2 statistic was analyzed with GSEA (Gene Set Enrichment Analysis, v4.3.2), where pre-ranked analyses against gene collections in the Molecular Signatures Database (MSigDB, v2022.1.Hs) were carried out [147,148]. Enrichment scores and statistics for genesets in the hallmark collection were calculated.

Gene Expression Visualization
Variance stabilized expression values (calculated with DESeq2) for DEGs were used to generate heatmaps using the ClustVis web tool [149]. Rows were clustered using correlation distance and average linkage. Volcano plot was constructed using VolcaNoseR webapp using log2Fold change on the X axis and −log10(adjusted P) on the Y axis [150,151].

Protein-Protein Interaction Network Analysis
STRING database (version 11.5) was used to predict protein-protein interactions network (PPI) with cutoff criterion score >0.4; interaction sources included text-mining, experiments, and curated databases [152,153].

Immunostaining
Immunostaining was performed on RPE/choroid flat mounts [28]. An amount of 24-month-old WT and LC3b −/− mice were anesthetized, and the eyes were enucleated and incised just below the ora serrata. RPE/choroid flat-mounts were prepared by separating the retina from the RPE followed by fixation in 4% PFA for 30 min at room temp. Flat-mounts were permeabilized and blocked in blocking solution containing 5% BSA in PBS + 0.2% Triton X-100 (PBST) at 37 • C for 1 h, incubated with primary antibody diluted in blocking solution (1:2000 for rabbit anti-MCT3 or 1:200 for goat P-cadherin) at 4 • C overnight, washed three times with PBST, incubated in appropriate secondary antibodies conjugated to Alexa Fluor dyes (Invitrogen, Donkey anti-rabbit 488 + donkey anti-goat 594; 1:1000), and Hoechst 33258 (1:10,000) at 37 • C for 1 h and washed three times. Flat-mounts were mounted in Prolong Gold (Invitrogen). Images were captured on a Nikon A1R laser scanning confocal microscope with a PLAN APO VC 60× water (NA 1.2) objective at 18 • C, and the data were analyzed using Nikon Elements AR 4.30.01 software.

Statistical Analysis
Gene level statistical tests were done using DESeq2 and pathway enrichment statistical tests were done using GSEA and Metascape. List overlaps were analyzed using the Chi squared (χ2) test of independence using GraphPad. A 2-tailed p-value was computed.