Abstract

   High myopia is a leading cause of blindness worldwide with increasing
   prevalence. Retina percepts visual information and triggers myopia
   development, but the underlying etiology is not fully understood
   because of cellular heterogeneity. In this study, single‐cell RNA
   sequencing analysis was performed on retinas of mouse highly myopic and
   control eyes to dissect the involvement of each cell type during high
   myopia progression. For highly myopic photoreceptors, Hk2 inhibition
   underlying metabolic remodeling from aerobic glycolysis toward
   oxidative phosphorylation and excessive oxidative stress was
   identified. Importantly, a novel Apoe ^+ rod subpopulation was
   specifically identified in highly myopic retina. In retinal neurons of
   highly myopic eyes, neurodegeneration was generally discovered, and the
   imbalanced ON/OFF signaling driven by cone‐bipolar cells and the
   downregulated dopamine receptors in amacrine cells were among the most
   predominant findings, indicating the aberrant light processing in
   highly myopic eyes. Besides, microglia exhibited elevated expression of
   cytokines and TGF‐β receptors, suggesting enhanced responses to
   inflammation and the growth‐promoting states involved in high myopia
   progression. Furthermore, cell–cell communication network revealed
   attenuated neuronal interactions and increased glial/vascular
   interactions in highly myopic retinas. In conclusion, this study
   outlines the transcriptional landscape of highly myopic retina,
   providing novel insights into high myopia development and prevention.

   Keywords: glucose metabolism, high myopia, microglia, ON/OFF pathways,
   retina, single‐cell RNA sequencing
     __________________________________________________________________

   Single‐cell RNA sequencing analysis revealed the metabolic remodeling
   from aerobic glycolysis toward oxidative phosphorylation and a specific
   degenerative rod subpopulation in highly myopic photoreceptors.
   Neurodegeneration was generally discovered in highly myopic retinal
   neurons with imbalanced ON/OFF signaling. Cell–cell communication
   network revealed attenuated neuronal interactions and increased
   glial/vascular interactions in highly myopic retinas, providing novel
   insights into its development and prevention.

   graphic file with name MCO2-4-e372-g004.jpg

1. INTRODUCTION

   In past decades, the prevalence of high myopia has rapidly elevated
   worldwide.[34] ^1 High myopia has become a leading cause of blindness
   with dramatically increased risks of multiple ocular pathologies,
   including chorioretinal and retinal degeneration. Therefore, it is
   important to investigate the mechanisms underlying the development of
   high myopia.

   Emerging evidence has suggested that the retina is involved in
   refractive development by sensing blurred images and triggering the
   signal transduction cascade of scleral remodeling and subsequent eye
   growth. Several genome‐wide association studies (GWAS) have also
   highlighted the essential roles of photosensitive and neuronal
   mechanisms underlying myopia progression.[35] ^2 , [36]^3 , [37]^4
   Previous studies involving pharmacological inhibition of retinal
   neurotransmitters or genetic knockdown of cell‐type‐specific genes
   (e.g., Gnat1^−/−) confirmed that a variety of retinal cell types might
   play roles in refractive development.[38] ^5 , [39]^6 However, the
   retina is a complex tissue in which multiple heterogeneous cell types
   form complicated communication networks. Studies performed on the whole
   retina may conceal critical contribution of individual cellular
   compartments to myopia.

   The advent of single‐cell RNA sequencing (sc‐RNA seq) provides an
   opportunity to comprehensively investigate the complex biological
   systems. The single‐cell atlases of various ocular structures have been
   constructed.[40] ^7 , [41]^8 , [42]^9 , [43]^10 We previously uncovered
   the single‐cell transcriptome atlas of lens epithelial cells, providing
   insight for pathogenesis and therapeutic targets for lens
   abnormalities.[44] ^9 This technique enables detailed elucidation for
   retinal transcriptional landscape in high myopia and cell‐type‐specific
   roles in refractive development at single‐cell resolutions.

   In this study, we applied sc‐RNA seq to the mouse retinas of highly
   myopic and contralateral control eyes. We reveal a high‐myopia‐specific
   degenerative rod subcluster and a decreased ratio of OFF‐cone‐bipolar
   cells (BCs) versus ON‐cone‐BCs, as well as pan‐neuronal degeneration in
   the highly myopic retinas. In addition, an enhanced response to
   inflammation is identified in the microglia of highly myopic retinas.
   These findings provide novel insights into the transcriptional
   landscape of highly myopic eyes and the retinal mechanisms underlying
   myopia development.

2. RESULTS

2.1. Identification of cell types affected by the development of high myopia

   To explore the involvement of retinal cell types in the development of
   high myopia, sc‐RNA seq was performed on retinas dissected from highly
   myopic mice models (Figure [45]1A). The models were established by
   placing a −25.00 diopter (D) lens on the right eyes for 4 weeks and the
   left eyes served as self‐controls. There was no significant difference
   in the refractive status between the two groups at the start of the
   experiments. After modeling, the right eyes exhibited a significant
   myopic shift compared with the control eyes (−1.07 D vs. 7.72 D,
   p < 0.001, paired t‐test).

FIGURE 1.

   FIGURE 1
   [46]Open in a new tab

   Single‐cell transcriptomic profiling of mice retinas in highly myopic
   and control eyes. (A) Experimental workflow for single‐cell RNA
   sequencing of mice highly myopic and control retina. (B) Uniform
   Manifold Approximation and Projection (UMAP) visualization of retinal
   transcriptional landscape, colored according to cell types. (C)
   Specific expression of cell type marker genes in each cell type. (D)
   UMAP visualization of retinal transcriptional landscape, colored
   according to groups (high myopia and control). (E) The distribution of
   differentially expressed genes (DEGs) (summarized from DEGs calculated
   within each cell type between highly myopic and control eyes);
   “Photoreceptors (PRs) and neurons” include rod photoreceptors, cone
   photoreceptors, bipolar cells, amacrine cells, and retinal ganglion
   cells; “non‐neurons” include Müller glial cells, microglial cells,
   vascular endothelial cells, and pericytes; “shared” means the genes
   upregulated or downregulated in both “PRs and neurons” and
   “non‐neurons.” (F) Gene set variation analysis (GSVA) for pathways of
   interest in PRs and neurons. (G) GSVA for pathways of interest in
   Müller glial cells and microglial cells. BC, bipolar cell; AC, amacrine
   cell; RGC, retinal ganglion cell; VEC, vascular endothelial cell; HM,
   high myopia; Ctrl, control.

   A total of 15,335 cells from highly myopic retinas and 15,991 cells
   from self‐control retinas were analyzed, respectively (Figure [47]S1A).
   The cell transcriptomes from all the samples were integrated and
   analyzed together to gain power for detecting rare cell types. Based on
   the expression profiles of the top 2000 highly variable genes, we
   initially identified 11 transcriptionally distinctive clusters and
   determined main retinal cell types (rod photoreceptors [Rho, Gnat1],
   cone photoreceptors [Opn1mw, Arr3], BCs [Trpm1, Pcp2], amacrine cells
   [ACs] [Slc6a9, Tfap2b], retinal ganglion cells [RGCs] [Pou4f1, Pou4f2],
   Müller glial cells [Slc1a3, Rlbp1], microglia [Aif1, Csf1r], vascular
   endothelial cells [Vwf, Cdh5], and pericytes [Des, Cspg4]) using
   well‐established cell type markers based on secondary clustering
   (Figure [48]1B,C and Figure [49]S1B).[50] ^11 The general distribution
   of cell types did not differ significantly between the highly myopic
   and control eyes (Figure [51]1D), suggesting that abnormal gene
   expression in each cell type might be more important to the
   pathogenesis of high myopia compared with changes in the proportions of
   cell types. Therefore, we next explored the cell‐type‐specific
   alterations in highly myopic retinas.

   We compared the transcriptional signatures of each retinal cell type
   between highly myopic and control eyes. We found that the highest
   percentage of upregulated genes was enriched in non‐neuronal cells,
   whereas the downregulated genes were mostly enriched in neuronal cells
   (Figure [52]1E). The downregulated genes in neuronal cells were
   involved in neuron development, synaptic signaling, regulation of ion
   transport, and detection of light stimulus (Figure [53]1F), consistent
   with previous large‐population GWAS analyses which revealed that visual
   perceptual and neuronal mechanisms contributed to the development of
   myopia.[54] ^2 , [55]^3 , [56]^4 By comparison, the upregulated genes
   in glia were enriched for cell activation and immune‐related processes
   (Figure [57]1G).

   We also scored each retinal cell type according to the expression of
   myopia candidate genes (Figure [58]S2),[59] ^2 and we observed
   significantly higher scores in BCs and ACs and comparable scores in
   other cell types (Figure [60]S1C). Overall, these results indicate that
   multiple mechanisms, especially neuronal mechanisms, might be involved
   in the development of high myopia. The cell‐type‐specific changes and
   mechanisms need to be further investigated.

2.2. Identification of a specific degenerative rod subcluster in highly
myopic eyes

   Several studies have indicated that rod and cone photoreceptors are
   involved in regulating ocular growth.[61] ^12 , [62]^13 , [63]^14
   Recent investigations in monkeys showed that photoreceptors play
   important roles in modulating eye growth by producing signals toward
   retinal pigment epithelial (RPE) cells and promote scleral tissue
   remodeling.[64] ^12 , [65]^13 , [66]^15 , [67]^16 However, the
   understanding of cellular heterogeneity on rod photoreceptors remained
   limited. To explore the heterogeneous changes in high myopia, we
   divided the population of rod photoreceptors into eight subpopulations
   (Figure [68]2A). Of note, subcluster 8 was specifical to highly myopic
   eyes (Figure [69]2B–D). Based on the transcriptional signature, the
   functional enrichment analysis showed that this high‐myopia‐specific
   rod subcluster was related to oxidative phosphorylation, ATP formation,
   response to stimuli, and intrinsic apoptotic signaling pathway
   (Figure [70]2E), indicating a specific mode of rod degeneration in
   highly myopic retinas. Because Apoe was the top marker gene that was
   highly expressed in subcluster 8 (Figure [71]2F), we named subcluster 8
   the Apoe ^+ rod subcluster. Immunofluorescence staining further
   confirmed the specific existence of Apoe ^+ rod photoreceptors in
   highly myopic retinas (Figure [72]2G). The morphology of Apoe ^+ cells
   in the outer nuclear layer of the retina was similar to that of common
   rod photoreceptors and the Apoe ^+ cell count was significantly higher
   in highly myopic retinas than in control eyes (Figure [73]2G).

FIGURE 2.

   FIGURE 2
   [74]Open in a new tab

   Specific subcluster of rod photoreceptors identified in highly myopic
   eyes. (A) t‐Distributed Stochastic Neighbor Embedding (TSNE)
   visualization of rod photoreceptors. (B) TSNE visualization of rod
   photoreceptors in highly myopic and control eyes, respectively. (C)
   Relative proportion of each rod subcluster in highly myopic and control
   eyes. (D) The fold change of relative proportions of each rod
   subcluster between highly myopic and control eyes. (E) Pathway
   enrichment analysis results of the marker genes for rod subcluster 8.
   (F) Marker genes of rod subcluster 8. Apoe was the top marker gene,
   therefore, the rod subcluster 8 was named as “Apoe^+ rod
   photoreceptors”. (G) Immunofluorescence staining to confirm the
   specific existence of Apoe^+ rod photoreceptors in highly myopic
   retina. (H) The degenerative trajectory of rod photoreceptors where the
   subcluster 8 appeared at the late period. (I) Heat map showing the
   dynamic expression profiles of genes along the trajectory and the
   representative pathways enriched for each pseudotime period.

   Next, we endeavored to uncover the biological significance of Apoe ^+
   photoreceptors in promoting myopia development. Using the Monocle2
   package, we constructed an unsupervised pseudotime trajectory based on
   gene signatures of all rod photoreceptors without prior clustering
   information. In this analysis, the Apoe ^+ rod subcluster was located
   at one end of the trajectory (Figure [75]2H). Because this subcluster
   was associated with stress and degeneration, we considered this
   trajectory to represent the neurodegenerative process, and the opposite
   end was defined as the starting point. Notably, genes related to normal
   photoreceptor functions and metabolic activities (photoreceptor cell
   maintenance, cilium assembly, and glycolysis) were significantly
   downregulated along the trajectory, whereas genes enriched for
   pathological states and altered metabolic characteristics (cellular
   response to hypoxia/starvation, inflammatory responses, and oxidative
   phosphorylation [OXPHOS]) were upregulated during the late period
   (Figure [76]2I).

   Notably, the expression level of Hk2, the first rate‐limiting enzyme of
   glycolysis in photoreceptors, was significantly downregulated along the
   trajectory (q value = 5.02 × 10^−80) (Figure [77]3A). To evaluate the
   metabolic outcomes of Hk2 deficiency, glycolysis and OXPHOS levels were
   measured using the Seahorse metabolic analyzer. After Hk2 knockdown
   (Figure [78]3B), the extracellular acidification rates (ECAR) exhibited
   a reduction in glycolysis and glycolytic capability (Figure [79]3C–E),
   whereas the oxygen consumption rates (OCR) exhibited an increase in
   maximal respiration compared with control cells (Figure [80]3F–G),
   indicating Hk2 deficiency caused a metabolic shift from glycolysis to
   OXPHOS, and that Hk2 is required for maintaining cellular glycolytic
   activities in photoreceptors. Hk2 knockdown also led to increased
   formation of reactive oxygen species (ROS) (Figure [81]3H). ROS is
   closely linked with myopia.[82] ^17 To evaluate the role of ROS in
   photoreceptors on myopia development, RPE cells were co‐cultured with
   photoreceptor cells pre‐exposed to H[2]O[2] and exhibited a significant
   upregulation of the myopia molecular marker, MMP2 (Figure [83]3I).
   Taken together, our results suggest that oxidative stress induced by
   glucose metabolism remodeling in photoreceptors in the context of
   blurred vision might trigger phototransduction dysfunction and
   progression of myopia.

FIGURE 3.

   FIGURE 3
   [84]Open in a new tab

   Effect of Hk2 knockdown (KD) on photoreceptor cells. (A) The decreased
   expression of Hk2 along the pseudotime trajectory. (B) Successful
   knockdown of Hk2 in a photoreceptor cell line confirmed by western
   blotting. (C–E) Seahorse metabolic analysis (glycolysis stress tests)
   of control cells and Hk2 KD cells. (F–G) Seahorse metabolic analysis
   (mitochondria stress tests) of control cells and Hk2 KD cells. The
   results are presented as the mean ± SEM. (H) Increased
   reactive oxygen species (ROS) production in the photoreceptor cells
   after Hk2 knockdown stained with a ROS detection kit. (I) Increased
   expression of MMP2 in mouse primary retinal pigment epithelial (RPE)
   cells co‐cultured with photoreceptor cells, which were pre‐exposed to
   H[2]O[2]. The control group was co‐cultured with photoreceptor cells
   without any treatment. ECAR, extracellular acidification rates; OCR,
   the oxygen consumption rates.

2.3. Decreased neuronal excitability of interneurons and imbalanced BC‐driven
ON/OFF light processing signaling in highly myopic eyes

   BCs anatomically connect the outer and inner layers of the retina and
   transmit signals from photoreceptors to RGCs. Previous studies have
   suggested that altered BC signaling transduction is involved in myopia
   development[85] ^18 , [86]^19 ; however, the underlying
   cell‐type‐specific molecular mechanisms remain largely unknown. To
   identify the role of BC signaling transduction, we first surveyed the
   differentially expressed genes between highly myopic and control BCs
   (Figure [87]4A). Surprisingly, we observed remarkable neurodegenerative
   manifestations characterized by multiple downregulated pathways related
   to synapses and neuron development (Figure [88]4B). Notably, Grik1,
   which expresses an ionotropic glutamate receptor, was the top
   downregulated gene in highly myopic BCs (Figure [89]4A). Because Grik1
   was enriched in BCs (Figure [90]S2), western blotting was performed at
   the tissue level and confirmed that GRIK1 was downregulated in highly
   myopic BCs (Figure [91]4C,D). These results were consistent with those
   of previous GWAS meta‐analyses, which identified GRIK1 as a myopia
   candidate gene in a large cohort of more than 250,000 participants.[92]
   ^2 Therefore, these results highlight the role of aberrant BC‐mediated
   light processing in the development of myopia by regulating the
   expression of glutamate receptors.

FIGURE 4.

   FIGURE 4
   [93]Open in a new tab

   Decreased neuronal excitability of interneurons and imbalanced bipolar
   cell (BC)‐driven ON/OFF light processing signaling in highly myopic
   eyes. (A) Volcano plot displaying differentially expressed genes
   detected between highly myopic BCs and control BCs. (B) Pathway
   enrichment analysis results of upregulated and downregulated genes in
   highly myopic BCs. (C and D) Downregulated protein level of GRIK1 in
   highly myopic retinas. (E) UMAP visualization of all BCs, colored
   according to cell subtypes. (F) Proportion of each BC subtype in highly
   myopic and control eyes. (G) Downregulated expression of Grik1 in
   highly myopic OFF‐cone‐BCs. (H) Volcano plot displaying differentially
   expressed genes detected between highly myopic amacrine cells (ACs) and
   control ACs. (I) Pathway enrichment analysis results of upregulated and
   downregulated genes in highly myopic ACs. HM, high myopia/highly
   myopic.

   To further investigate the transcriptional alterations of highly myopic
   BCs, we performed unsupervised subclustering of these cells and
   classified them into three subpopulations, rod‐BCs (Prkca ^+),
   ON‐cone‐BCs (Scgn ^+ Isl1 ^+), and OFF‐cone‐BCs (Scgn ^+ Isl1 ^−),
   based on the expression of established subtype markers
   (Figure [94]4E).[95] ^20 We first surveyed the proportional variance of
   each cell type and observed an increased proportion of ON‐cone‐BCs and
   a decreased proportion of OFF‐cone‐BCs in highly myopic retinas than in
   control eyes, whereas the proportion of rod‐BCs was comparable between
   highly myopic and control eyes (Figure [96]4F and Figure [97]S3A). This
   result further implied the role of imbalanced ON–OFF signaling
   (especially cone‐BC‐driven ON–OFF signaling) in promoting high myopia
   development. In addition, we investigated the expression profile of
   Grik1 in BC subpopulations and found that it was specifically expressed
   in OFF‐cone‐BCs (Figure [98]S3B), and its expression was also decreased
   in highly myopic OFF‐cone‐BCs (Figure [99]4G). These results provide
   additional support for the role of glutamate receptor regulation in the
   development of high myopia.

   The differential analyses were extended to ACs that exhibited similar
   trends toward increased responses to stress and decreased expression of
   genes related to neuronal function and synaptic signaling
   (Figure [100]4H,I). Notably, Drd4 was among the top downregulated genes
   in highly myopic ACs (Figure [101]4H). As a component of D2‐like
   dopamine receptors (D2Rs), Drd4 could actively respond to the
   environmental lighting by modulating intracellular Ca^2+ levels and
   therefore regulating Ca^2+‐dependent signaling processes.[102] ^21 This
   result supports previous observations suggesting that D2Rs, rather than
   D1‐like receptors (D1Rs), mediate the effect of dopamine for protecting
   against lens‐induced myopia.[103] ^22 , [104]^23 In summary, these
   findings strongly suggest that abnormal BC/AC‐mediated light processing
   within the retina may be the vital factor that triggers and maintains
   myopia development.

2.4. Highly myopic eyes showed enhanced responses to inflammatory and growth
factors

   Accumulated evidence supports the potential involvement of inflammation
   in high myopia progression.[105] ^24 , [106]^25 , [107]^26 For
   instance, previous studies have reported elevated levels of interleukin
   (IL)‐6 and transforming growth factor (TGF)‐β in the ocular environment
   of highly myopic patients.[108] ^24 , [109]^25 To dissect the
   inflammatory profiles of highly myopic eyes at a single‐cell
   resolution, we determined the expression levels of cytokines, growth
   factors, and the corresponding receptors in microglia, the
   retina‐resident immune cells. We found that Il1a, Il6ra, Il21r, Tgfbr1,
   and Tgfbr2 and downstream transcriptional regulators (Stat3, Nfkbr1,
   and Nfkbr2) were significantly increased in the microglia of highly
   myopic eyes (Figure [110]5A). These results suggested these eyes
   exhibited enhanced responses to cytokines and growth factors, as well
   as a possible activation of the STAT3 signaling pathway. Western
   blotting confirmed the increased STAT3 expression in highly myopic
   retinas (Figure [111]5B). Double‐label immunofluorescence staining of
   phosphorylated‐STAT3 (p‐STAT3) and IBA1 further showed the activation
   of STAT3 signaling in highly myopic retinal microglial cells
   (Figure [112]5C). Meanwhile, genes enriched for cell activation, cell
   migration, and cellular responses to stress were also upregulated in
   highly myopic microglia (Figure [113]5D). Collectively, these findings
   provide novel evidence for the potential involvement of enhanced
   microglial responses to inflammation and the growth‐promoting states in
   high myopia progression.

FIGURE 5.

   FIGURE 5
   [114]Open in a new tab

   Enhanced responses to inflammatory and growth factors in highly myopic
   eyes. (A) Violin plot displaying upregulated cytokine and growth factor
   receptor genes in highly myopic microglia. (B) Up‐regulation of STAT3
   in highly myopic retina confirmed by western blotting. (C)
   Immunofluorescence staining of IBA1 and phosphorylated‐STAT3 (p‐STAT3).
   (D) Pathway enrichment analysis for upregulated genes in highly myopic
   microglial cells.

2.5. High myopia alters retinal cell‒cell communication

   The retina is composed of multiple cell types, which form a complex
   regulatory network for light processing through synaptic transmission
   and interactions via ligand‒receptor pairs or adhesive molecules. To
   investigate the alterations of cell‒cell interactions in highly myopic
   retinas, the cell‒cell communication network was constructed in highly
   myopic and control retinas based on the expression profiles of
   ligand‒receptor pairs in each cell type (Figure [115]6A). A total of
   2651 interactions among cell types were observed in the control
   retinas, compared with 2136 were found in highly myopic retinas. Of
   note, the interactions within photoreceptors or neurons were attenuated
   or eliminated, whereas those between glial or vascular cells and other
   cell types were remarkably increased in highly myopic retinas
   (Figure [116]6B), consistent with our prior observations of
   degenerative neurons and activated glia in highly myopic eyes.

FIGURE 6.

   FIGURE 6
   [117]Open in a new tab

   High myopia alters retinal cell–cell communication. (A) Chord diagrams
   of cellular interactions among all cell types in highly myopic and
   control retinas, separately. (B) Heatmaps exhibiting the interaction
   profiles between every two cell types in highly myopic and control
   retinas, separately. (C) Dot plot of ligand‐receptor interactions
   between Apoe− or Apoe+ rod photoreceptors and other cell types. BCs,
   bipolar cells; ACs, amacrine cells; RGCs, retinal ganglion cells; VECs,
   vascular endothelial cells.

   Furthermore, because we identified a specific subcluster of rod
   photoreceptors (Apoe ^+ rods) in highly myopic eyes, we explored the
   altered interactions associated with Apoe ^+ rods (Figure [118]6C).
   Notably, these cells showed downregulated expression of brain‐derived
   neurotrophic factor (Bdnf) receptors (Ncam1 or Gpr152) and insulin‐like
   growth factor 1 (Igf1) receptor (Igf1r); these changes may hinder the
   cells’ ability to receive nutritional support from the surrounding
   neurons (Figure [119]6C). Besides, the overexpression of Sirpa in Apoe
   ^+ rods may enhance the interactions between Apoe ^+ rods and all other
   retinal cell types which expressed Cd47 (Figure [120]6C). While recent
   studies have reported the important roles of Cd47‐Sirpa ligand‐receptor
   pair in regulating synapse strengthening and neuron survival,[121] ^27
   this result implicated the retinal neuronal network undergoing
   remodeling in highly myopic eyes. Taken together, this analysis of the
   cell–cell communication network revealed attenuated neuronal
   interactions and increased glial/vascular interactions in highly myopic
   retinas, and these aberrant communication patterns provide novel
   insight for understanding the pathogenesis of myopia.

3. DISCUSSION

   High myopia is becoming increasingly prevalent, and it has been
   estimated that it will affect 10% of the world's population by
   2050.[122] ^28 There is sufficient evidence to suggest that the retina
   triggers the development of myopia by perceiving aberrant visual
   experience and initiating the signaling cascade that finally transmits
   to the myopia target tissue, the sclera. However, owing to the
   heterogeneity of retinal cell types, the contribution of individual
   retinal pathways to myopia development remains largely unknown. In this
   study, we explored the single‐cell transcriptional signatures of highly
   myopic and control retinas and identified cell‐type‐specific mechanisms
   underlying the pathogenesis of high myopia that included
   high‐myopia‐specific degenerative rod photoreceptor subpopulation,
   imbalanced ON‐BC/OFF‐BC signaling, and decreased retinal neuronal
   activity. Moreover, microglia activity was increased in high myopic
   retinas, suggesting that metabolic and inflammatory factors play roles
   in the development of high myopia.

   The retina processes light information through a neuronal signaling
   cascade and maintains its homeostasis via glia and vascular cells.
   Photoreceptors have been implicated in the pathogenesis of abnormal
   refractive development. Tin et al. reported that “abnormal
   photoreceptor inner segment morphology” was the most significant gene
   set enriched for myopia‐associated genes; other photoreceptor‐related
   gene sets include the “thin retinal outer nuclear layer”, “detection of
   light stimulus”, and “nonmotile primary cilium”.[123] ^2 Previous
   functional studies have also confirmed the attenuation of photoreceptor
   cell responses to flash electroretinography (a‐wave and d‐wave) in a
   murine model of myopia.[124] ^6 However, the underlying molecular
   mechanism was largely unknown.

   Photoreceptors consume a large amount of energy for phototransduction
   and constant renewal of their outer segments, but they preferentially
   consume glucose through aerobic glycolysis despite possessing abundant
   mitochondria and enzymes for OXPHOS,[125] ^29 , [126]^30 , [127]^31
   sharing similar metabolic characteristics to cancer cells, in a process
   known as the Warburg effect. Because photoreceptors are the most
   abundant cells in the retina, their metabolic changes would have a
   remarkable impact on the retinal microenvironment. In this study, we
   revealed a shift toward OXPHOS in highly myopic photoreceptors. Similar
   metabolic alternations in myopic eyes have also been also mentioned in
   previous studies, which used bulk RNA sequencing of retinas.[128] ^6 ,
   [129]^32

   More importantly, we identified the involvement of Hk2 deficiency in
   this metabolic remodeling process. Hexokinases (HKs) are the first
   rate‐limiting enzymes in glycolysis, and HK2 is the main subtype
   expressed in photoreceptors.[130] ^33 In this study, Hk2 was among the
   top downregulated genes along the rod degenerative trajectory in highly
   myopic eyes. Silencing Hk2 led to a decreased glycolysis and an
   increased in OXPHOS in a photoreceptor cell line using Seahorse
   metabolic analysis. This result was consistent with those performed in
   a hepatocellular carcinoma cell line by DeWaal et al. and those in Hk2
   mutant mice by Zhang et al.[131] ^33 , [132]^34 Importantly, Hk2
   knockdown contributed to ROS as possible by‐products of excessive
   OXPHOS in mitochondria, because HK2 could help to limit ROS generation
   by maintaining local ADP levels.[133] ^35 , [134]^36 ROS is closely
   linked with myopia.[135] ^17 It may induce upregulation of HIF‐1α or
   proinflammatory cytokines, which have been elaborated to promote myopia
   development.[136] ^26 , [137]^37 , [138]^38 This may explain why ROS
   induced in photoreceptors further upregulated the expression of myopic
   molecular marker (MMP2) in co‐cultured RPE cells in our study, and
   therefore indicate how HK2‐related glucose metabolism might be related
   to high myopia development. Taken together, our findings demonstrate
   that the oxidative environment derived from an aberrant metabolic shift
   toward OXPHOS in the retina may help trigger the signaling transmission
   cascade toward the sclera, providing a new perspective for
   understanding the development of high myopia and the early‐onset of
   retinal degeneration in highly myopic eyes.

   Another important finding is the discovery of a novel rod photoreceptor
   subcluster characterized by APOE expression in highly myopic retinas.
   APOE is critical for transporting cholesterol and lipids in the central
   nervous system, and it is strongly associated with neuronal
   degenerative diseases and aging.[139] ^39 Previous studies have
   demonstrated neuronal APOE production after acute stress and injury,
   including oxidative stress, traumatic brain injury, and amyloid‐β (Aβ)
   accumulation.[140] ^40 , [141]^41 , [142]^42 Therefore, these specific
   Apoe ^+ rod photoreceptors might be related to the neuropathologic
   microenvironments of highly myopic retinas, playing roles in regulating
   cholesterol redistribution for cellular repair. The specific factors
   underlying these observations remain unclear; therefore, further
   functional mechanistic studies are necessary.

   An interesting finding of this study was the imbalanced ON‒OFF
   signaling systems mediated by BCs in highly myopic eyes. Accumulative
   evidence has suggested the involvement of BC‐mediated light processing
   in the development of myopia.[143] ^18 , [144]^19 , [145]^43 Hendriks
   et al. investigated the incidence and degree of myopia in inherited
   retinal dystrophies with genetical degeneration of a specific cell type
   and found that BC dysfunction was associated with the highest risk of
   high myopia.[146] ^18 In our study, the proportion of OFF‐cone‐BCs was
   decreased. We also observed downregulation of Grik1 in highly myopic
   OFF‐cone‐BCs. In support, Jiang et al. found that OFF‐cone‐BC signaling
   was highly associated with a myopia‐susceptibility locus which
   genotyped between 186 twin volunteers,[147] ^19 providing additional
   evidence for OFF‐cone‐BC pathways in myopia development. The ON and OFF
   visual systems are established at the first retinal synapses (the
   photoreceptor‒BC synapses) and separated throughout the downstream
   retinal information processing systems (BC‒AC and BC‒RGC synapses).
   They respond to light stimuli in opposite ways because ON‐BCs
   depolarize, whereas OFF‐BCs hyperpolarize in response to light. It is
   predicted that aberrant visual experience would trigger
   glutamate‐receptors‐mediated ON/OFF‐BC signaling imbalance, manifested
   as altered contrast sensitivity and spatial resolution, directly
   leading to changes in retinal dopamine metabolism and choroidal
   thickness as eyes are attempting to reduce image defocus on the
   retina,[148] ^43 , [149]^44 and therefore promoting or maintaining high
   myopia progression. Further investigations using animal models with
   conditional knockout of interested genes (e.g. Grik1) in BCs may
   provide more evidence.

   Neuronal dysfunction in the light processing pathway also extended to
   ACs. Emerging evidence has revealed a role of ACs in ocular refractive
   development. Pan et al. reported that deregulated phosphorylation
   modification of connexin36 in aII ACs contributed to the development of
   myopia.[150] ^28 In our study, we identified significant dysfunction of
   ACs, including downregulated synaptic signaling transmission and neuron
   development, implicating impaired AC function in complex retinal image
   processing, especially in the adjustment of image brightness, which is
   strongly associated with the onset of myopia. Furthermore, because ACs
   are the main retinal neurons that possess a dopaminergic synaptic
   system, our finding that the D2‐like dopamine receptor Drd4 was
   significantly downregulated in highly myopic ACs provides further
   support for the widely acknowledged myopia development theory involving
   dopamine signaling.[151] ^45

   Microglia are the main resident immune cells in the retina.
   Neuroinflammation mediated by microglia plays vital roles in retinal
   pathologies. Accumulative evidence has suggested that inflammation is
   involved in the development of myopia. Lin et al. reported a higher
   incidence of myopia among patients with inflammatory diseases, and
   elevated expression levels of c‐Fos, nuclear factor‐κB, and IL‐6 in
   murine myopic eyes.[152] ^26 However, in our study, we found that a
   group of cytokine receptors, rather than cytokines, were significantly
   elevated. While proinflammatory cytokines could be derived from other
   cell types, such as blood‐derived macrophages that infiltrated the
   retina, our results highlight the enhanced responses of highly myopic
   eyes to proinflammatory environment. Notably, the STAT3 signaling
   pathway was activated in highly myopic microglia characterized by
   elevated p‐STAT3 expression, exhibiting an aging or neuroinflammatory
   profile.[153] ^46 , [154]^47 Because this activation might be a
   response to the pathological environment in highly myopic retinas or an
   amplifier to exacerbate myopia progression via regulating
   myopia‐related genes, further studies are needed to dissect the causal
   relationship between these two events.

   Given that RPE and choroid also play important roles in scleral
   remodeling and eye growth, one limitation of this study is the absence
   of RPE transcriptomic data as is rarely captured in retinal sc‐RNA‐seq
   samples.[155] ^48 Further RNA sequencing of RPE/choroid is expected to
   elucidate the role of retina‐RPE/choroid‐sclera system on the
   regulation of myopia development.

   In conclusion, we have provided an outline of the transcriptional
   landscape of highly myopic retinas and revealed their generalized
   retinal neurodegeneration and enhanced responses to pathogenic factors
   (Figure [156]7). Our results imply that multiple retinal cell types are
   involved in the development of high myopia development. We have also
   identified a subpopulation of high‐myopia‐specific degenerative rods
   and a loss of balance between cone‐BC‐driven ON/OFF visual pathways,
   providing novel perspectives of the retinal mechanisms and some
   potential therapeutic targets in high myopia.

FIGURE 7.

   FIGURE 7
   [157]Open in a new tab

   Schematic illustration of transcriptional landscape of retina in high
   myopia. In axially elongated highly myopic eyes,
   metabolic‐remodeling‐induced oxidative stress and a novel Apoe^+ rod
   subpopulation was discovered, and aberrant light processing
   characterized by ON/OFF signaling imbalance driven by cone‐bipolar
   cells and dopamine receptors (Drd4) downregulation in amacrine cells
   were found. In addition, highly myopic microglia exhibited enhanced
   responses to cytokines. Furthermore, cell–cell communication network
   revealed attenuated neuronal interactions and increased glial/vascular
   interactions in highly myopic eyes. These results indicated a specific
   degeneration mode characterized by metabolic aberration, neuronal
   dysfunction, and increased inflammatory responses in highly myopic
   retinas, and revealed the cell‐type specific mechanisms underlying
   disease progression.

4. MATERIALS AND METHODS

4.1. Animals

   Mice were obtained from SLAC Laboratory Animal Co. Ltd. and reared in
   specific pathogen‐free (SPF) barrier systems under a consistent 12‐h
   light/dark cycle (light on at 7:00 am and off at 7:00 pm). The humidity
   was kept at 40%−60% and the room temperature was kept at 21°C.

   A −25.00 D lens was placed to the right eye of 4‐week‐old male C57BL/6
   J mice to construct the defocus‐induced high myopia model. At the
   beginning of the study, mice with interocular baseline refraction of
   more than 1.00 D measured by a mouse infrared photorefractor (Steinbeis
   Transfer Center) were excluded. After 4 weeks of defocus induction,
   mice with the right eye showing at least 6.00 D more myopic than the
   left eye (self‐control) were considered an effective model of high
   myopia.

4.2. Retinal cells dissociation and single‐cell RNA sequencing

   A total of three highly myopic retina samples and three control samples
   were prepared separately for sc‐RNA seq, and each sample was composed
   of five biologically duplicated retinas. The retinas were dissected in
   DMEM (Thermo Fisher Scientific) on ice, and then chopped up with
   scalpels and digested in Earle's balanced salt solution containing
   papain (20 units/mL) and DNAse I (2,000 units/mL) for 20 min at 37°C.
   Dissociation was stopped with ovomucoid. Cells were filtered through a
   40 μm cell strainer (Bel‐Art), centrifuged (300 g, 5 min), and
   resuspended in Earle's balanced salt solution with 10% FBS. Trypan blue
   staining was used to evaluate cell concentration and viability. Cells
   from the six samples were then added to six respective channels on
   Chromium Single Cell Controller (10× Genomics). ScRNA‐seq libraries
   were constructed on 10× Genomics system and sequenced on NovaSeq 6000
   (Illumina) after quality tests.

4.3. Data preprocessing, quality control, and sc‐RNA seq analysis

   Raw sequencing data were mapped to the mm10 mouse reference genome
   using the 10× Genomics CellRanger software (version 5.0.0). The
   expression matrixes were processed using the Seurat package (version
   4.1.0) in R software (version 4.1.2). Cells with RNA counts ≥1000,
   expressed genes ≥200, log[10](genes per unique molecular identifier
   [UMI]) > 0.7, and mitochondrial RNA <25% were retained as high‐quality
   cells. Doublets were further detected using the DoubletFinder package
   (version 4.1.2) and removed from following analyses. After scaling and
   normalization, principal component analysis was performed using the top
   2000 highly variable genes. Cell clustering was performed with the top
   30 principal components at a resolution of 0.4 using the “FindClusters”
   function and visualized using Uniform Manifold Approximation and
   Projection (UMAP) or t‐Distributed Stochastic Neighbor Embedding
   (t‐SNE). Cell types were annotated using well‐established marker genes,
   and sub‐clustering would further be performed if one cluster was found
   to be mixed with multiple cell types.

4.4. Differentially expressed genes identification and functional enrichment
analysis

   Differentially expressed genes were generated with use of the
   “FindMarkers” function in Seurat package. Heat maps, volcano plots, and
   violin plots were generated with R software. Gene Ontology and Kyoto
   Encyclopedia of Genes and Genomes functional enrichment analyses were
   performed using the Metascape tool ([158]http://metascape.org).[159]
   ^49

4.5. Gene set variation analysis

   The gene set variation analysis (GSVA) algorithm was used to evaluate
   the enrichment scores of interested pathways/gene sets among different
   cell types and groups. Heat map was used to visualize the GSVA results.

4.6. Pseudotime analysis

   The Monocle2 package (version 2.8.0) was used to calculate the
   pseudotime trajectory of rod photoreceptors with unsupervised
   approaches.[160] ^50 The start point of the pseudotime trajectory was
   determined by cell states. DDRTree was applied to reduce dimensions,
   and the “plot_cell_trajectory” function was used to visualize the cell
   trajectories.

4.7. Cell–cell communication analysis

   To investigate the potential interactions between cell types, cell–cell
   communication analysis was performed using the CellPhoneDB Python
   package (version 1.0.0) based on expression of known interacting
   ligand‐receptor pairs.[161] ^51 Two cell types were interacted if one
   cell type expressed the ligand and the other cell type expressed the
   receptor. “Expressed” was defined as at least 10% of cells in the
   specific cell‐type having nonzero read counts for the corresponding
   receptor or ligand genes.

4.8. Cell culture and treatment paradigms

   The photoreceptor cell line, 661 W, and primary mouse RPE cells were
   both cultured in DMEM/F12 medium with 10% FBS and 1% penicillin and
   streptomycin. All cell cultures were maintained in a humidified
   environment of 95% air and 5% CO[2] at 37°C. For Hk2 knockdown, 661 W
   cells were transfected with siRNA (#siG190508023415; Guangzhou Ribo
   Company) using LipofectamineTM 3000 Reagent (Thermo Fisher Scientific)
   for 72 h. For ROS detection, cells were incubated in DCFH‐DA (#S0033;
   Beyotime) at 37°C for 30 min, washed by DPBS for three times, and then
   observed using a fluorescence microscope. For the co‐culture
   experiments, the 661 W cells were cultured in six‐well plates, and
   pre‐exposed to 0.005 mmol/L hydrogen peroxide (H[2]O[2], #349887;
   Sigma) for 24 h before the Transwell inserts seeded with RPE cells were
   placed onto the plates.[162] ^52

4.9. Metabolomic assays

   To evaluate the glycolysis and OXPHOS levels, 661 W cells were plated
   onto Seahorse 96‐well plates. Cells were transfected with Hk2 siRNAs or
   negative control siRNAs before being washed and incubated in the base
   medium (Agilent Technologies) at 37°C for 1 h. ECAR and OCR were
   measured by glycolysis stress tests and mitochondria stress tests on
   seahorse Bioscience X96 extracellular flux analyzer (Agilent
   Technologies) according to the manufacturer's instructions, serving as
   quantitative indicators for cellular glycolysis and OXPHOS levels,
   respectively.[163] ^34

4.10. Western blotting

   Protein was extracted from retinas or cultured cells, separated on 10%
   SDS‐PAGE gel, and transferred to nitrocellulose membranes. After
   blockage, the membranes were incubated with primary antibodies at 4°C
   overnight, followed by incubating with HRP‐conjugated secondary
   antibodies at room temperature for 1 h. After visualization with ECL
   Plus Western Blotting Substrate (Thermo Fisher Scientific) and ECL
   detection system (ChampChemi, SAGECREATION), the density of target
   protein bands was evaluated using the ImageJ software and normalized to
   the loading control (GAPDH) before statistical analysis. The following
   primary antibodies were used for western blotting: GRIK1 (GLUR5, 1:100,
   #sc‐393420; Santa Cruz), HK2 (1:2000, 22029‐1‐AP; ProteinTech), MMP2
   (1:500, #10373‐2‐AP; ProteinTech), STAT3 (1:1000, #9139; Cell Signaling
   Technology), p‐STAT3 (1:1000, #9145; Cell Signaling Technology), and
   GAPDH (1:3000, #5174; Cell Signaling Technology).

4.11. Immunofluorescence

   Frozen sections of mice eyes were blocked, incubated with primary
   antibodies at 4°C overnight, and then incubated with secondary
   antibodies and DAPI for 2 h. Slides were observed using a confocal
   microscope (Leica Microsystems). The primary antibodies used for
   immunofluorescence were as follows: PRKCA (1:200, #P4334; Sigma
   Aldrich), APOE (1:50, #66830‐1‐lg; ProteinTech), p‐STAT3 (1:200, #9145;
   Cell Signaling Technology), and IBA1 (1:200, #ab283319; Abcam).

4.12. Statistics

   The experimental data between highly myopic and self‐control eyes were
   compared using paired t‐test. Other experimental data between two
   groups were compared using independent sample t‐test. p Value < 0.05
   was considered statistically significant.

AUTHOR CONTRIBUTIONS

   Yunqian Yao and Xiangjia Zhu conceived the idea and wrote the initial
   and revised versions of the manuscript. Xingtao Zhou supervised the
   study and was involved in critical revision of manuscript. Yi Lu
   provided material support. Qingfeng Wu provided technical support.
   Yunqian Yao and Zhenhua Chen performed bioinformatic analyses and
   experiments. All authors reviewed the results and approved the final
   version of the manuscript.

CONFLICT OF INTEREST STATEMENT

   The authors declare no conflicts of interest.

ETHICS STATEMENT

   All the animal procedures conformed to the ARVO Statement and were
   approved by the Ethics Committee for Animal Studies of Eye & ENT
   Hospital of Fudan University (approval number, 2020068).

Supporting information

   Supporting Information
   [164]Click here for additional data file.^ (1.2MB, docx)

ACKNOWLEDGEMENTS