Postnatal eye size in mice is controlled by SREBP2-mediated transcriptional repression of Lrp2 and Bmp2

How organs achieve a reproducible size is a central question in biology. The eye is by far the most important sensory organ, and its size and dimension closely relate to its functional properties. Eye axial diameter plays a key role in determining retinal image size (Hughes, 1977). Moreover, the size of an eye has to match its optic parameters to perceive clear vision. For the vertebrate camera-like eye composed of multiple structures, it requires a sophisticated control system that coordinates individual tissues to ensure correct size and function of the organ. Despite the staggering differences in eye size across mammalian species, the final eye size difference between adult animals within a species is insignificant (Howland et al., 2004). In humans, for example, the average axial length of adult eyes is 23.6 mm with a standard deviation of ±0.7 mm (Gordon and Donzis, 1985). These findings suggest that there is a strong genetic basis to eye size control. However, in comparison with other organs, the mechanisms underlying eye size control remain poorly understood.

In a study aiming to understand the role of lipid synthesis in retina development and diseases such as retinitis pigmentosa (as photoreceptors shed 10% of their outer segment daily and need to synthesize membrane discs rapidly; Young, 1967), we made an unexpected discovery that the lipogenic transcription factor sterol regulatory element binding protein 2 (SREBP2; also known as SREBF2) has a function in eye size regulation. SREBP2 is a master transcription factor that regulates cholesterol synthesis and metabolism in all cells (Brown and Goldstein, 1997). Full-length SREBP2 (flSREBP2) is the precursor protein tethered in the membranes of the endoplasmic reticulum. In cells with low levels of sterols, SREBP2 is cleaved to leave just the N-terminal domain (nSREBP2), which translocates to the nucleus and functions as a transcription activator. nSREBP2 binds to specific sterol regulatory element (SRE) sequences or E-box motifs and activates the transcription of the enzymes involved in cholesterol synthesis as well as enzymes involved in generating NADPH (Athanikar and Osborne, 1998; Shimomura et al., 1998). SREBP2 is expressed in the neural retina and retinal pigment epithelium (RPE) cells (Zheng et al., 2012, 2015), but its function in the eye development remains elusive.

In this study, we investigate the role of SREBP2 and its downstream signaling pathway in regulating eye size in mice. We find that overexpression of nSREBP2 in the RPE cells of postnatal mice leads to extremely enlarged eye globes. Taking this observation as a starting point, we reveal that Lrp2, a gene that is known to restrict eye overgrowth, is transcriptionally repressed by SREBP2. Transcriptome analysis and functional assays identified that BMP2 is the downstream effector of both Srebp2 and Lrp2, and the level of Bmp2 in the RPE is the key determinant of eye size. As the upstream regulator, SREBP2 transcriptionally represses Bmp2. Over postnatal development, the levels of Lrp2 and Bmp2 transcripts increase and the SREBP2 protein level decreases, in accordance with their functions to restrict and promote eye growth, respectively, as the eye growth rate slows down. Notably, RPE-specific Bmp2 overexpression by adeno-associated virus (AAV) can effectively prevent the eye enlargement and retinal thinning caused by Lrp2 loss. Together, our study shows that rapid postnatal eye size increase is governed by an RPE-derived signaling pathway, which consists of both positive and negative regulators of eye growth. Overall, this study unveils an essential role of the SREBP2-LRP2-BMP2 signaling in the RPE in determining eye size.

To study the function of SREBP2 in postnatal eye development, we overexpressed Srebp2 in neonatal mouse eyes by subretinal injection of serotype 8 adeno-associated virus (AAV8). Viral transgene expression driven by the ubiquitous CMV promoter first started in the RPE as early as postnatal day (P) 1 (Fig. 1A-C), and strong transgene expression could be observed in both the RPE and photoreceptors later at P7 and P14 (Fig. S1). Whereas the control eyes that overexpressed GFP appeared normal, a striking eye enlargement phenotype induced by Srebp2 overexpression was observed at mouse eye opening (Fig. 1D-G). Overexpression of the truncated N terminus of SREBP2 (nSREBP2) or full-length SREBP2 (flSREBP2) induced eye enlargement, but the phenotype of nSREBP2 overexpression was much more prominent than that of flSREBP2 (Figs 1F,G and 2A), possibly owing to the constitutive activity of the nuclear-located nSREBP2.

In wild-type mice, both axial length (AL) and equatorial diameter (ED) of the eye globes increased rapidly in the first two postnatal weeks (Fig. 1H). After eye opening, eye size increase greatly slowed down (Fig. 1H). To examine the effects of nSREBP2 on eye size growth, we injected AAV8-CMV-nSrebp2 vectors into the right (R) eye and normalized its AL and ED lengths to the uninjected left (L) eye and assessed the R/L ratio. nSREBP2 overexpression led to significant eye overgrowth (∼20% increases in both dimensions) in the first month (Fig. 1I,J), after which the phenotype stabilized and persisted throughout adulthood (Fig. 1I,J). These results suggest that SREBP2 promotes eye size during early postnatal development in mice.

Next, we investigated which cell type(s) is responsible for the eye enlargement phenotype. The RPE and photoreceptors had the highest infection and transgene expression level by AAV8 viruses with the CMV promoter (Fig. S1). Targeted gene expression in the RPE was driven by the bestrophin 1 (Best1) promoter in the AAV8 vector (Fig. S1) (Esumi et al., 2004; Xiong et al., 2019). RPE-specific nSREBP2 overexpression was sufficient to induce eye enlargement, and the phenotype was comparable to that induced by broad nSREBP2 overexpression (Fig. 2A,B). By contrast, robust photoreceptor-specific nSREBP2 expression was driven by the human rhodopsin kinase (RK; also known as RHOK and GRK1) promoter (Khani et al., 2007) (Fig. S1), but it did not cause any change of eye size (Fig. 2C). In summary, we conclude that SREBP2 functions in the RPE to promote mouse eye size.

What is the potential downstream molecule of SREBP2 that mediates its effects on eye size? It was recently reported that low density lipoprotein receptor-related protein 2 (Lrp2) is also required in the RPE to regulate eye size (Storm et al., 2019). We confirmed the eye enlargement phenotypes of Lrp2 loss by conditional knockout (cko) as well as by shRNA-mediated knockdown in the RPE (Fig. 3A-C) (Cases et al., 2015; Storm et al., 2019). The enlarged eyes caused by nSREBP2 overexpression are characterized by the expansion of the posterior segment and retinal thinning, with essentially normal anterior segment and intraocular pressure (Fig. S2A-D), resembling the phenotype of Lrp2 cko mice (Fig. S2E-H). The highly similar phenotypes induced by nSREBP2 overexpression and Lrp2 deficiency imply that these two factors may function in the same pathway with opposing functions.

We hypothesized that Lrp2, which is a member of the low-density lipoprotein receptor (LDLR) family of proteins, is a transcriptional target of SREBP2, given the well-known function of SREBP2 as a transcription factor for lipogenic genes (Horton et al., 1998). To test this, we examined the mRNA levels of Lrp2 in response to AAV-mediated Srebp2 overexpression or knockdown in vivo. Overexpression of nSREBP2 in the RPE increased the mRNA level of Hmgcr and Ldlr, two known SREBP2 transcriptional target genes (Horton et al., 2002), but it significantly reduced the level of Lrp2 mRNA (Fig. 3D). Conversely, downregulation of endogenous Srebp2 by shRNA increased the Lrp2 mRNA level by nearly twofold (Fig. 3E). We further tested a boron-containing small molecule, BF175, which can directly suppress SREBP2 transcriptional activity by blocking the binding of SREBP2 to its transcriptional co-factor mediator complex (Zhao et al., 2014). In the mouse RPE explant model, adding BF175 to the culture medium reduced the level of Hmgcr and Ldlr mRNA but significantly increased the level of Lrp2 mRNA (Fig. 3F), mirroring the effects of Srebp2 knockdown. Moreover, suppression of Srebp2, either by co-injection of AAV-Srebp2 shRNA (sh) or BF175, effectively suppressed Lrp2 shRNA1 (sh1)-induced eye size increase (Fig. 3G,H). These results suggest that SREBP2 has a physiological role in eye size regulation by suppressing the expression of Lrp2.

Does SREBP2 directly regulate transcription at the Lrp2 promoter? Within 350 bp upstream of the transcription start site of the human LRP2 promoter sequence, there are three putative SREBP2-binding motifs, including a binding site (TGGTGTGAC) predicted by the JASPAR dataset, an SRE-like sequence (GTGGGG) and an E-box motif (CACGTG) (Fig. 3I) (Amemiya-Kudo et al., 2002; Fornes et al., 2020; Shimano et al., 1999). To examine whether SREBP2 functionally regulates the Lrp2 promoter, we measured the transcriptional activity of the Lrp2 promoter in response to Srebp2 overexpression in a luciferase reporter assay. The results showed that the activity of the Ldlr promoter (−335 to +3 bp) was greatly enhanced whereas the activity of the Lrp2 promoter (−505 to −13 bp) was significantly repressed by nSREBP2 co-transfection (Fig. 3I). This finding, together with the qPCR results, strongly suggests that SREBP2 acts as a transcriptional repressor of Lrp2 rather than its usual role as a transcriptional activator. We further excluded the possibility that Srebp2 is also downstream of and regulated by Lrp2, as neither Srebp2 mRNA nor protein level changed as a result of Lrp2 knockdown (Fig. S3). Hence, we propose a model in which SREBP2 is an upstream regulator of eye size, and it promotes mouse eye size by repressing Lrp2 (Fig. 3J).

To investigate which downstream pathways of Srebp2 and Lrp2 are responsible for regulating eye growth, we performed RNA-sequencing (RNA-seq) analysis with mouse RPE tissues. P0 C57BL/6 pups were injected with AAV8-Best1-GFP/nSrebp2 or AAV8-Best1-ctrl sh, Lrp2 sh1 or Lrp2 sh2 viruses. At P14, RPE cells were carefully dissociated for RNA extraction. Differential gene expression (DGE) was determined between the three pairs of datasets (nSREBP2 versus GFP, Lrp2 sh1 versus ctrl sh, Lrp2 sh2 versus ctrl sh) (Fig. 4A,B). We reasoned that any key downstream effector or pathway responsible for eye growth control should be similarly regulated by nSREBP2 overexpression or Lrp2 knockdown. This approach allowed the number of the genes/pathways identified by RNA-seq to be narrowed down to a shortlist. Gene set enrichment analysis (GSEA) of all canonical pathways (total 181 gene sets) identified five common pathways that were significantly differentially expressed (P<0.05) in all three enlarged eye groups in comparison with their control groups (Fig. 4C, Fig. S4). The BMP pathway caught our attention because of its possible involvement in the regulation of eye growth and development of myopia (Cheng et al., 2013; Li et al., 2015; Liu et al., 2009; Nixon et al., 2019; Verhoeven et al., 2013). BMP pathway target genes Smad6, 7, 9 and Id1-4 were clearly downregulated, suggesting an overall attenuation of BMP signaling in the enlarged eyes (Fig. 4D). This is consistent with a negative enrichment score of the pathway (Fig. S4). Several Bmp ligands (Bmp2, 4, 6, 7 and 11), which are highly expressed in the wild-type RPE (Fig. S5), were downregulated by nSREBP2 overexpression or Lrp2 knockdown (Fig. 4D).

To determine whether any Bmp ligand is the downstream effector of the Srebp2-Lrp2 pathway, we knocked down each of the five highly expressed Bmp ligand genes in the neonatal mouse RPE (Fig. 4E). Two shRNAs with high knockdown efficiency were tested for each gene (Fig. S6). We found that Bmp2 knockdown induced eye enlargement, whereas Bmp4, 6, 7 or 11 knockdown did not cause any significant change in mouse eye size (Fig. 4E). These results suggest that BMP2 is the downstream effector of Srebp2 and Lrp2.

Next, we investigated whether Bmp2, similar to Lrp2, is directly regulated by SREBP2. qPCR results confirmed that Bmp2 mRNA level is decreased by nSREBP2 overexpression in the RPE in vivo (Fig. 5A). By analyzing a previously published ChIP-seq dataset that profiled genome-wide SREBP2 binding in the HCC70 human carcinoma epithelial cell line (Cai et al., 2019a), we found that SREBP2 binding is enriched at the promoter and the intron 1 of the BMP2 gene as well as in the promoter of the LRP2 gene (Fig. S7). Although no putative SREBP2-binding site in the BMP2 promoter region can be identified, there are two E-box motifs in BMP2 intron 1 (Fig. 5B). To verify SREBP2 binding in the RPE cells, ChIP-qPCR was performed using three primer sets, with one pair (P1) in the promoter region and the other two pairs (P2 and P3) flanking each of the E-box motifs in intron 1. ChIP-qPCR results showed that endogenous SREBP2 protein is enriched at the promoter as well as at the first E-box motif in intron 1 (Fig. 5B). We further performed a luciferase reporter assay to examine whether SREBP2 activates or represses the expression of BMP2. When nSREBP2 was co-transfected, the activity of the BMP2 promoter (−500 to −1 bp) was greatly suppressed (Fig. 5C). The intron 1 sequence (+1271 to +1778 bp) was cloned in front of a minimal promoter (MinP), and its activity was also suppressed at the presence of nSREBP2 (Fig. 5C). Lrp2 knockdown led to the decrease of Bmp2 mRNA level in the RPE in vivo (Fig. 5D), suggesting that LRP2 is a positive regulator of the Bmp2 gene. However, the mechanism by which LRP2 promotes the expression of Bmp2 warrants further investigation. Together, our data suggest that SREBP2 represses the transcription of Bmp2 both directly and indirectly by suppressing Lrp2.

If SREBP2-LRP2-BMP2 is a key signaling pathway that controls eye size, one would expect that its activity changes along with the eye growth rate during postnatal development. We first examined the mRNA and protein levels of the three genes in the wild-type mouse RPE at three time points: P0, P14 and P30. Bmp2 and Lrp2, the two genes which inhibit eye growth, showed a clear upregulation of both mRNA and protein levels from P0 to P30, during which period the eye growth rate slows down (Fig. 5E,F). However, Srebp2 mRNA levels did not show any significant change over time (Fig. 5G). As Srebp2 has been shown to be regulated post-transcriptionally (Brown and Goldstein, 1997), we further examined the SREBP2 protein level in the RPE. The levels of both full-length and mature truncated SREBP2 proteins declined from P0 to P30 (Fig. 5G), in concordance with its function to suppress Lrp2 and Bmp2 expression. Together, our data suggest that the dynamic and opposite changes of the eye growth promoting and inhibiting genes in the SREBP2-LRP2-BMP2 pathway govern eye size growth in mice. In our model, the relatively high Srebp2 and low Lrp2/Bmp2 in neonatal mice promote the rapid eye size increase, whereas low Srebp2 and high Lrp2/Bmp2 in adult mice ensure that eye growth stops at the proper size (Fig. 5H).

BMP2 is a key signaling molecule that functions in the downstream part of the SREBP2-LRP2-BMP2 pathway. BMP2 has been proposed as an eye growth ‘STOP’ signal previously. Decreased expression of Bmp2 in myopic eyes in various animal models has been previously reported (He et al., 2018; Wang et al., 2015; Zhang et al., 2012, 2016, 2019). The human BMP2 single nucleotide polymorphism (SNP) rs235770 is associated with myopia in multi-ethnic cohorts (Verhoeven et al., 2013). In mice, Bmp2 is expressed mainly in the RPE from embryonic day 11-11.5 (Dudley and Robertson, 1997), and the Bmp2 level is much higher in the RPE than in the retina in adult eyes (Fig. S5). However, loss-of-function phenotypes of Bmp2 in the RPE have not been examined. We used both shRNA-mediated and CRISPR/Cas9-mediated knockdown to suppress Bmp2 expression specifically in the RPE in the neonatal mice (Fig. S6). Insufficient Bmp2 in the RPE led to eye enlargement, and eye size was inversely correlated with the dose of Bmp2 (Fig. 6A,B,J). Histological analysis showed that the enlarged eye globe resulting from Bmp2 knockdown is caused by expansion of the posterior chamber without other gross ocular morphological defects, which is highly comparable to the histology of nSREBP2 overexpression and Lrp2 cko eyes (Fig. 6D,E, Fig. S2). Retina structure appeared normal except for a uniform thinning of all layers (Fig. 6G,H,L), which is likely due to the expansion of the posterior eye globe. In fact, retinal thinning is a major complication of high myopia, which may increase the risks of retinal detachment and tears (Curtin and Karlin, 1970; Ohno-Matsui and Jonas, 2020; Vongphanit et al., 2002).

Interestingly, excessive BMP2 by RPE-specific Bmp2 overexpression resulted in the opposite effect, which is smaller eyes (Fig. 6C). The smaller eye is characterized by reduced posterior globe size and thickening of the posterior ocular layers (Fig. 6F,I,K,L). The severity of the phenotype was also correlated with the dose of BMP2 overexpressed (Fig. 6K, Fig. S6). Therefore, mouse eye size is inversely correlated with the Bmp2 level in the RPE, suggesting that RPE-derived BMP2 level is a key determinant of eye size.

Congenital high myopia with enlarged eye globes and retinal dystrophy are the main ocular phenotypes of the Donnai–Barrow syndrome caused by LRP2 mutations (Kantarci et al., 2007; Longoni et al., 2008; Pober et al., 2009). Although gene therapy has emerged as a promising approach to treat inherited eye diseases, it is difficult to rescue Lrp2 loss-of-function phenotypes by gene augmentation therapy given the large molecular weight of LRP2 (∼522 kDa). Because our data suggest that LRP2 functions via Bmp2 to restrict eye growth, we hypothesized that forced Bmp2 expression could rescue the ocular phenotypes caused by Lrp2 loss. To address this, we produced Lrp2 cko in the RPE by subretinally injecting AAV8-Best1-Cre virus to Lrp2^fl/fl mice at P0 together with the AAV8-Best1-Bmp2 or GFP virus (Fig. 7A). At P30, the axial length and retinal thickness were measured by optical coherence tomography (OCT). Lrp2 cko eyes with the control GFP virus injection showed obvious AL elongation and retinal thinning, but these phenotypes were largely rescued by AAV-mediated Bmp2 overexpression (Fig. 7B-D, Fig. S8). These results suggest that BMP2 acts downstream of Lrp2 and that targeted Bmp2 expression in the RPE may be an effective therapeutic intervention for eye enlargement and associated complications caused by Lrp2 loss.

Srebp2 is a key gene in cholesterol synthesis and lipid metabolism, and it is highly expressed in both the retina and RPE (Zheng et al., 2012, 2015). Although a Srebp2 hypomorphic mutation has been linked to cataract formation in the lens in mice (Merath et al., 2011), the function of Srebp2 in the posterior eye has not been examined. Here, we propose that SREBP2 has an important function in eye development, which is eye size control. We showed that the postnatal increase in eye size is controlled by SREBP2-mediated transcriptional repression of Lrp2 and Bmp2, which are two suppressors of eye size. SREBP2 normally functions as a transcriptional activator (Horton et al., 2002), and it activates the transcription of two lipogenic genes, Ldlr and Hmgcr, in the RPE (Figs 3D-F,I and 5C). Interestingly, in the same cell type SREBP2 represses Lrp2 and Bmp2 transcription to control eye size (Figs 3D-F,I and 5A-C). Our results suggest that SREBP2 has more diverse functions in the eye and displays distinct transcriptional activities towards different downstream targets.

The function of SREBP2 in eye size regulation is also consistent with changes in its expression level in the RPE during mouse development. In neonates, Srebp2 level is high whereas Lrp2 and Bmp2 levels are relatively low, promoting the rapid eye size increase. Over postnatal development, there is a decrease of SREBP2 protein expression in the RPE but the mRNA level of Srebp2 remains constant (Fig. 5G). Downregulation of SREBP2, the eye size-promoting protein, and simultaneous upregulation of Lrp2 and Bmp2, the two eye size-inhibiting genes, may be the mechanism that ensures that the eye grows to and stops at the proper size (Fig. 5E,F). How SREBP2 is regulated at the protein level during the key postnatal period of eye size determination and whether SREBP2 recruits any transcriptional co-repressor to suppress Lrp2 and Bmp2 transcription require further investigation.

Given the known functions of Srebp2 and Lrp2 in lipid metabolism, it is natural to question whether lipid metabolism also plays a role in eye growth and axial length determination. SREBP2 is a prominent protein that activates cellular cholesterol synthesis and uptake. Our transcriptome analysis identified the lipid metabolic process as one of the significantly changed biological pathways in the nSREBP2 overexpression group, and all the known SREBP2 transcriptional target genes, including HMGCR, LDLR, SCD, ACACA and FASN, were upregulated. LRP2 belongs to the LDLR family and may play a role in lipoprotein uptake. However, recent studies underscored the function of Lrp2 in internalizing and processing signaling molecules (Christ et al., 2012, 2015; Gajera et al., 2010), but relatively less is reported about the function of Lrp2 in lipid metabolism. Our RNA-seq analysis did not identify any lipid-related pathway in the RPE with Lrp2 knockdown, but this could be because RNA-seq only reveals gene expression and not lipid molecule changes. Future studies using proteomic profiling would be helpful to identify any common lipid metabolism pathway downstream of Srebp2 and Lrp2. Interestingly, SREBP2 is regulated by post-translational mechanisms, and its activation is controlled by the cellular cholesterol level and other nutrient sensors, such as mTORC1 (Brown and Goldstein, 1997; Eid et al., 2017). In the future, it would be interesting to examine whether lipid and other nutrient signals impinge on eye size control through Srebp2.

We provide direct evidence showing that eye size regulation by Srebp2 and Lrp2 occurs through Bmp2. Previous studies have demonstrated a direct link between Bmp4 and Lrp2 (Gajera et al., 2010). LRP2 is a clearing receptor of BMP4 in the subependymal zone in the adult mouse brain (Gajera et al., 2010). Loss of Lrp2 results in increased Bmp4 expression and activation of SMAD1/5/8 in the stem cell niche (Gajera et al., 2010). One recent study in zebrafish also reported that the bmp4 pathway was changed in Lrp2^−/− eyes (Collery and Link, 2018 preprint). Zebrafish Bmp4 protein binds to the extracellular domain of Lrp2, and its signaling can be facilitated as well as reduced by Lrp2 via different mechanisms (Collery and Link, 2018 preprint). However, our study showed that BMP2, but not other BMP ligands, is the key signaling molecule in the context of mouse eye size regulation (Fig. 4E). LRP2 promotes Bmp2 expression (Fig. 5D), although the detailed mechanism remains to be determined. As LRP2 is an endocytic receptor, it may promote Bmp2 expression indirectly via a third pathway, such as sonic hedgehog, which has been shown to be directly regulated by LRP2 and further regulates BMP signaling in other developmental contexts (Christ et al., 2012, 2016; Huycke et al., 2019; McCarthy et al., 2002).

It is worth noting that BMP2 signaling may not be the only signaling pathway downstream of Srebp2 and Lrp2 that is responsible for eye size regulation, as the Bmp2 knockdown phenotype is less significant than that of nSREBP2 overexpression or Lrp2 knockdown, which cannot be simply explained by Bmp2 levels (Figs 2, 3, 5, and Fig. S6). Additional candidate pathways, such as the Jak/Stat and sonic hedgehog pathways, are highly differentially expressed in the RPE with Srebp2 overexpression and Lrp2 knockdown in the RNA-Seq dataset; therefore, the involvement of these pathways in eye size regulation should be further investigated.

There are several possible mechanisms leading to eye enlargement. Buphthalmos is most commonly found in congenital or infantile glaucoma patients. The increased eye globe size in the congenital or infantile glaucomatous cases is secondary to the stretching of the globe by high intraocular pressure (IOP), given the elasticity of the sclera at this young age (Aziz et al., 2015). However, we did not detect increased IOP in this case (Fig. S2), excluding high IOP being the cause of eye enlargement. Another common mechanism responsible for organ size increase is cell overproliferation. A prominent example is that liver size is controlled by the Hippo pathway via its regulation of cell proliferation and survival (Dong et al., 2007). However, eye enlargement induced by Bmp2 knockdown in postnatal mice is unlikely to be caused by increased cell proliferation in the neuroretina, as retinal cell proliferation rate was not affected by the RPE Bmp2 level using a 5-ethynyl-2′-deoxyuridine flow cytometry assay (Fig. S9). Therefore, the driving force of eye enlargement may not originate from the retina.

The sclera provides structural support to the eye globe (Watson and Young, 2004). The ‘mechanical’ theory of myopia development suggests that scleral extracellular matrix remodeling and thinning leads to exaggerated eye growth and axial elongation (Metlapally and Wildsoet, 2015). Similar mechanisms may underlie the eye enlargement in early postnatal development observed in this study and during myopia development. One hypothesis, which is to be further tested, is that early postnatal sclera development may be under the influence of RPE-derived BMP2. Our preliminary data showed that scleral cell proliferation rate is controlled by the level of Bmp2 in the RPE in vivo (Fig. S9), which supports the hypothesis. A previous in vitro study also showed that BMP2 promoted scleral cell proliferation and changed the expression levels of genes related to extracellular matrix remodeling (e.g. MMP2 and TIMP2) in cultured human scleral fibroblasts (Hu et al., 2008), but the exact functions of RPE-derived BMP2 in regulating scleral development require further in vivo studies. Moreover, choroidal development may be also controlled by BMP2 from the RPE and further contribute eye size regulation directly or indirectly.

Eye size control is of great biomedical relevance, as refractive error occurs when the axial length of the eye does not match its refractive power. Myopia, which is the most common type of refractive error, is caused by abnormal enlargement or elongation of the eye globe (Chakraborty et al., 2020; Siegwart and Norton, 2011). High myopia, which is defined by the World Health Organization as a refractive error ≤−5.00 diopter (D) or an axial length ≥26 mm, can lead to secondary complications such as retinal detachment and myopic macular degeneration that cause irreversible vision impairment (Cai et al., 2019b; Ohno-Matsui and Jonas, 2020). Despite the alarming prevalence of myopia worldwide and increasing evidence of genetic predisposition (Cai et al., 2019c; Holden et al., 2016), there are few effective therapeutic treatments to prevent myopia and especially high myopia, which is in part owing to our poor understanding of the genes and molecular mechanisms underlying eye growth and eye size control.

Our study demonstrates that high myopia caused by Lrp2 insufficiency is prevented by targeting the downstream effector Bmp2. We showed that a low dose of the AAV-Best1-Bmp2 vector can completely prevent the development of high myopia and secondary retinal thinning (Fig. 7), which could be a potential early intervention of inherited high myopia caused by Lrp2 genetic defects. Compared with drug treatment, the advantages of using AAV vectors include long-term effects and cell-type specificity. The RPE is the target cell type of the successful Leber's Congenital Amaurosis gene therapy, which has been proven to be safe and effective (Bainbridge et al., 2008; Hauswirth et al., 2008; Maguire et al., 2008). As our research highlighted the role of the RPE as a signaling center in controlling postnatal eye growth, RPE cells would be ideal target cells for treating high myopia by gene therapy as well. Together, our findings suggest that therapeutic strategies targeting SREBP2-LRP2-BMP2 signaling to control eye growth could have significant clinical implications.

All animal procedures performed were approved by Hong Kong Department of Health under Animals Ordinance Chapter 340 [Ref: (20-130) in DH/HT&A/8/2/5 Pt.2] and by City University of Hong Kong Animal ethics committee (Ref: A-0475).

CD1 mice were purchased from the Chinese University of Hong Kong, and C57BL/6J mice were purchased from The Jackson Laboratory. Lrp2^fl/fl mice were obtained as a gift from Prof. Thomas Willnow (Max Delbrück Center for Molecular Medicine, Berlin, Germany) (Christ et al., 2016; Leheste et al., 2003). Mice were kept on a 12 h light/12 h dark cycle in City University of Hong Kong Laboratory Animal Research Unit.

pAAV2/8 and pAdDeltaF6 plasmids were obtained from the Penn Vector Core (University of Pennsylvania). pAAV-Best1-GFP-WPRE was made and published previously (Xiong et al., 2019). The mouse Srebp2 coding sequence was cloned from a pLKO-puro FLAG Srebp2 plasmid (Addgene plasmid #32018). Full-length or N-terminal (1-1371 bp) sequences were cloned into AAV plasmids by Gibson ligation. AAV-shRNA vectors were cloned by replacing the GFP sequence with mCherry-shRNA in the pAAV-Best1-GFP vector. See Supplementary Materials and Methods for further details of plasmid cloning.

pAAV, Rep/Cap 2/8 and pAdDeltaF6 plasmids were mixed with polyethylenimine and added to HEK293T cells; 24 h after transfection, the cell medium was changed to DMEM only; 72 h after transfection, supernatant was collected, and cell debris was spun down and discarded. AAV8 in the supernatant were precipitated by PEG-8000 (8.5% wt/vol PEG-8000 and 0.4 M NaCl for 1.5 h at 4°C), centrifuged at 7000 g for 10 min, and resuspended in virus buffer (150 mM NaCl and 20 mM Tris, pH 8.0). The resuspension was run on an iodixanol gradient, and viruses in a 40% fraction were collected. Recovered AAV virus particles were washed three times with cold PBS using Amicon 100K columns (EMD Millipore). Protein gels were run to determine virus titers.

Subretinal injection into P0 neonate eyes was performed as previously described (Wang et al., 2014; Xiong et al., 2015). Briefly, 0.25 μl of viruses in PBS was injected into the subretinal space using a pulled angled glass pipette controlled by a FemtoJet (Eppendorf). AAV8-CMV-GFP/nSrebp2/ flSrebp2, AAV8-Best1-GFP/nSrebp2/flSrebp2/Ctrl sh/Srebp2 sh1/Srebp2 sh2/Lrp2 sh1/Lrp2 sh2/Bmp2/4/6/7/11 sh1/sh2, and AAV8-RK-ZsGreen/nSrebp2 were injected at a dose of 1E9 vg/eye. AAV8-Best1-Cre was injected at a final dose of 1E7 vg/eye in all groups, and the doses of AAV8-Best1-Bmp2 virus were 2E6 vg/eye and 1E7 vg/eye (low and high dose, respectively). AAV8-Best1-saCas9 and AAV8-Best1-Bmp2 g1 or g2 were mixed at a 1:1 ratio and injected at a total dose of 2E9 vg/eye. For animals used for qPCR and RNA-seq, both left and right eyes were injected and used for RNA extraction. For animals used for eye size measurement or other phenotype characterizations, only the right eye of the animal was injected, and the left eye was left uninjected as within-animal controls.

Mice were sacrificed at the indicated ages. Eyes were enucleated, and connective tissues and muscles were carefully removed using tweezers and scissors. Eyes were immersed in PBS in a 6 cm Petri dish and imaged under a Nikon SMZ800N dissection scope with 2× magnification. ED and AL were measured using ImageJ and converted to ml or ratios.

OCT images of mouse eyes were taken using a SD-OCT (Bioptigen Envisu R4310 SD-OCT, Germany). Mice were anaesthetized by intraperitoneal injection of a 100 mg/kg ketamine and 10 mg/kg xylazine mixture dosed by weight. A drop of 0.5% proxymetacaine hydrochloride (Provain-POS, Germany), and a drop of 0.5% tropicamide and 0.5% phenylephrine hydrochloride (Mydrin-P, Santen Pharmaceutical Co., Japan) solution were separately instilled on the ocular surface for corneal anesthesia and dilation of the pupil, respectively. Lubricating eye drops (Systane Ultra, Alcon) was applied to prevent desiccation of the cornea during imaging. Then, the anesthetized mouse was put onto a stereotaxic platform for alignment with the imaging lens. Whole-eye biometry and retinal OCT images were separately measured and captured along the horizontal meridian centered at a point one optic disk diameter away from the outer optic disc margin using an SD-OCT (Bioptigen Envisu R4310 SD-OCT, Germany). Axial resolution was 2.6 μm and scanning speed was 20,000 lines per second. SD-OCT imaging was conducted at P30 on the same cohort of mice. Dimensions of individual ocular components were quantified using ImageJ. Axial length was defined as the distance between the anterior cornea and the outer boundary of the RPE layer.

For noninvasive measurement of IOP, an Icare TonoLab tonometer (Colonial Medical Supply) was used. Mice were anesthetized using 2% isoflurane, and IOP measurements were acquired from each eye within 3 min of induction of anesthesia. Each instrument-generated average was derived from six individual measurements. All measurements were performed at the same time during daylight for three consecutive days.

Eyeballs were quickly removed from the euthanized mouse and dipped in 70% ethanol for decontamination. Under a dissecting stereomicroscope, connective tissues and muscles were carefully removed. After washing twice in PBS, eyeballs were immersed in warm culture medium (DMEM:F12+10% fetal bovine serum). the cornea was cut off using curved scissors, and the lens was pulled out gently with tweezers. The ora serrate was cut off to remove the iris and cornea. The retina and optic nerve were carefully and completely removed from the eye cups. Four radial cuts were made to enable the eye cups to be flat-mounted. Each eye cup was transferred onto the center of a floating polycarbonate nucleopore filter membrane (Whatman 110406, 0.2 μm) placed in 6-well plates with the RPE side facing down. Freshly prepared BF175 stock solution was added to the full culture medium to a final BF175 concentration of 12.5 μM. See Supplementary Materials and Methods for description of BF175 synthesis procedures. Half of the medium was replaced with fresh medium on the second day. RPE flat-mounts were harvested at 48 h in explant and processed for RPE isolation and RNA extraction.

Eyecups without retina and optic nerve tissues were dissected as described in the RPE explant section. Two eyes of the same mouse were pooled in one tube and processed together. RPE cells were incubated in papain solution (Worthington Biochemical Corporation) for 15 min. After washing twice in warm medium, RPE samples were triturated with a 600 μl pipette tip gently to dissociate the pigmented RPE cells from the sclera. The resuspended cell solution was transferred to a clean tube and spun down at 600 g. RNAs were extracted from mouse RPE using Trizol (Thermo Fisher Scientific) followed by Quick RNA MicroPrep Kit (Zymo Research) and were used for qPCR or RNA-seq.

EdU (100mg/kg, Abcam, ab146186) was subcutaneously injected daily from P3 to P5 to mark cells in S phase. Animals were harvested at P6 and their eyes were removed and cryosectioned at 20 μm thickness. EdU staining was performed using the Click-iT™ EdU Alexa Fluor™ 488 Imaging Kit (Thermo Fisher Scientific, C10337). EdU-positive (EdU⁺) cells in the choroid and sclera across the whole retinal section were counted manually. Three mid-central retinal sections of each eyeball were selected for quantification, and the number of EdU⁺ cells per eye was averaged for statistical analysis. For quantification the number of EdU⁺ cells in the retina, fluorescence-activated cell sorting was used. Retinas were incubated in papain solution (Worthington Biochemical Corporation) for 10 min. After washing twice in warm medium, retinas were triturated gently with a 600 μl pipette tip and the resuspended cell solution was transferred to a clean tube and spun down at 600 g. The percentage of EdU⁺ cells in the retina was quantified using a Click-iT™ EdU Alexa Fluor™

488 Flow Cytometry Assay Kit (Thermo Fisher Scientific, C10420) according to the manufacturer's instructions.

RNAs were converted to cDNA using a PrimeScript RT reagent kit with gDNA Eraser (Takara Bio). qPCR was performed using the PowerUp Sybr Green Master Mix (Thermo Fisher Scientific) on QuantStudio 3 Real-Time PCR stems (Applied Biosystems). Gapdh was used as the normalizing control. qPCR primers are listed in Supplementary Materials and Methods.

ARPE19 cells were obtained from ATCC and cultured in standard complete growth medium. Cells were crosslinked with 0.5% formaldehyde for 2.5 min at room temperature. See Supplementary Materials and Methods for detailed sample processing procedures. ChIP-qPCR reactions were performed on a QuantStudio 3 Real-Time PCR System (Applied Biosystems) using TB Green Premix Ex Taq Master Mix (Takara Bio, RR036A) using 2 μl of input DNA or ChIP DNA for each 10 μl reaction. ChIP-qPCR data were normalized relative to input.

RNAs were extracted from mouse RPE or retina using Trizol (Thermo Fisher Scientific) followed by the Quick-RNA MicroPrep Kit (Zymo Research). The quality of RNA samples was first assessed using an Agilent Bioanalyzer RNA 6000 Nano Chip, and samples with RIN≥9 were used for further processing. The NEBNext rRNA Depletion kit (NEB, E6350) was used to remove ribosomal RNA and the NEBNext Ultra II Directional RNA Library Prep Kit (NEB, E7760) was to generate the cDNA library. Genewiz (Suzhou, China) performed 150 bp paired-end sequencing using an Illumina HiSeq System. See Supplementary Materials and Methods for detailed procedures on making RNA-seq libraries. Raw RNA-seq reads were mapped to the mouse reference genome (GRCm38), followed by calculation of gene counts using SATR with the default parameter settings (version 2.7.3a) (Dobin et al., 2013). Differential expression analysis was performed between different experimental conditions (nSREBP2 versus GFP, Lrp2 sh1 versus ctrl sh, Lrp2 sh2 versus ctrl sh) using the ‘DESeq2’ package (Love et al., 2014). Differentially expressed genes were selected based on |log2FC|>1 and Benjamini–Hochberg (BH)-adjusted P<0.05. GSEA was performed using HTSanalyzeR (Wang et al., 2011) with 5000 permutations on 181 canonical pathways gene sets (≥15 genes) from MsigDB v6.1.

Enucleated eyes were fixed in Hartman's fixative (Sigma-Aldrich, H0290) for 24 h at room temperature. The fixed samples were dehydrated through graded ethanol (50%, 70%, 80%, 85%, 90%, 95%, 100%; 30 min in ambient temperature for every step) and then cleared in xylene (three changes, 8 min for each change). The samples were further processed through paraffin (three changes, 1 h for each change, 60°C) before they were embedded with a Thermo HistoStar tissue-embedding workstation. Paraffin sections were then cut at 6 μm using a Thermo HM325 manual rotary microtome and mounted on Superfrost Plus microscope slides. For deparaffinization, prepared sections were heated at 62°C for 3 h and washed in xylene (three changes, 15 min for each change). For Hematoxylin & Eosin (H&E) staining, deparaffinized sections were rehydrated in graded ethanol (100%, 95%, 80%; 5 min for each change) and rinsed once in distilled water (5 min). The sections then went through a standard H&E staining protocol using a H&E staining kit (Abcam, ab245880) according to the manufacturer's instructions. Stained sections were mounted with Richard-Allan mounting medium (Thermo Fisher Scientific, 22-050-102). Slides were observed using an Olympus CX23 light microscope.

Mouse RPE cells were isolated as described in the previous section. Four eyes were pooled in one tube and processed together. Cell lysates were prepared using a Minute™ total protein extraction kit (for animal cultured cells and tissues) (Invent Biotechnologies, SD-001/SN-002) in accordance with the manufacturer's protocol. The extracted total proteins were quantified using a Pierce™ BCA protein assay kit (Thermo Fisher Scientific, 23225) and boiled with 4× Laemmli sample buffer (Bio-Rad, 1610747) for 5 min. Equal amounts of proteins were resolved by 7.5% SDS–polyacrylamide gel electrophoresis and then transferred to PVDF Membranes (Bio-Rad, 1620177). The membranes were blocked with 5% skimmed milk in TBS with 0.1% Tween 20 (TBST) for 1 h then probed with rabbit polyclonal anti-SREBP2 (Cayman Chemical, 10007663, 1:1000), mouse monoclonal anti-LRP2 (Santa Cruz Biotechnology, sc-515772, 1:200), mouse polyclonal anti-BMP2 (Proteintech, 18933-1-AP, 1:1000) and mouse monoclonal anti-GAPDH (Santa Cruz Biotechnology, sc-32233, 1:5000) at 4°C overnight. Membranes were then rinsed four times with TBST and then incubated with HRP-conjugated secondary antibody (Jackson ImmunoResearch, anti-rabbit 111-035-144 and anti-mouse 115-035-003;1:2000) for 2 h at room temperature. After rewashing four times with TBST, signal was visualized with ECL Plus WB Reagents (Bio-Rad, 1705060).

pLDLR-Luc was purchased from Addgene (plasmid #14940). Other reporter plasmids containing the LRP2 promoter region (−505/−13 bp), BMP2 promoter region (−500/−1 bp) or BMP2 intron (+1271/+1778 bp) were amplified from mouse genomic DNA and cloned into a pGL2 vector. HEK293 cells were seeded in 96-well plates and cultured until 60-70% confluence. Next, 450 ng pCAG-human nSREBP2 or a control plasmid, pCAG-Cre, was co-transfected with 500 ng reporter plasmids and 100 ng pRL-TK (Promega, E2241). After being cultured for 48 h, cells were lysed with reporter lysis buffer (Promega). Luciferase activity was determined in the cell lysates using the Promega luciferase detection kit (Promega).

Data are represented as mean±s.e.m. in all figures. Sample sizes and statistical analysis are indicated for each experiment in figure legend. All data sets were normally distributed, as confirmed by the Shapiro–Wilk normality test. ANOVA analysis with Tukey test was performed to compare multiple groups, and two-tailed Student's t-test was performed to compare two groups. A value of P<0.05 was considered statistically significant. GraphPad Prism was used to perform statistical analysis and generate figures.

We thank Dr Constance Cepko for valuable discussion and support for this project; Dr Thomas Willnow for kindly providing us with the Lrp2^fl/fl mice; Dr Krish Kizhatil and Dr Simon Johns for their generous help and advice on IOP measurement; Dr Ming Chang, Dr C.C. Amy Fong, Mr Eric K.W. Shum for their technical support; and Dr Zi-Bing Jin, Dr Fenghua Hu, Dr Jiahai Shi, Dr Jinyoung Kim, Dr Rebecca Chin, Dr Xin Deng, Dr Liang Zhang and Dr Geoffrey Lau for their valuable comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200633.