PRELP Regulates Cell–Cell Adhesion and EMT and Inhibits Retinoblastoma Progression

Simple Summary Mutation of the RB1 tumor suppressor gene is fundamental in retinoblastoma initiation and progression although its downstream mechanism has not been well elucidated. Here, we found that expression of proline/arginine-rich end leucine-rich repeat protein (PRELP) is strongly downregulated in human retinoblastoma and highly expressed in Müller glial cells in normal retina. Deletion of PRELP in mice resulted in retinal dysplasia associated with enhanced proliferation. mRNA expression profiling revealed that cancer pathways were strongly activated in PRELP−/− retina. Additionally, cell–cell adhesion was inhibited while epithelial mesenchymal transition (EMT) and inflammation were activated. On the other hand, application of PRELP protein to retinoblastoma cell lines enhances cell–cell and cell–substrate adhesion and inhibits anchorage independent growth by reversing EMT. These observations indicate that PRELP downregulation in human retinoblastoma can contribute cancer progression through regulation of cell adhesion and EMT suggested that PRELP application might be a novel strategy for retinoblastoma treatment. Abstract Retinoblastoma (RB) is the most common intraocular pediatric cancer. Nearly all cases of RB are associated with mutations compromising the function of the RB1 tumor suppressor gene. We previously demonstrated that PRELP is widely downregulated in various cancers and our in vivo and in vitro analysis revealed PRELP as a novel tumor suppressor and regulator of EMT. In addition, PRELP is located at chromosome 1q31.1, around a region hypothesized to be associated with the initiation of malignancy in RB. Therefore, in this study, we investigated the role of PRELP in RB through in vitro analysis and next-generation sequencing. Immunostaining revealed that PRELP is expressed in Müller glial cells in the retina. mRNA expression profiling of PRELP−/− mouse retina and PRELP-treated RB cells found that PRELP contributes to RB progression via regulation of the cancer microenvironment, in which loss of PRELP reduces cell–cell adhesion and facilitates EMT. Our observations suggest that PRELP may have potential as a new strategy for RB treatment.


Introduction
RB is the most common intraocular pediatric cancer with an estimated frequency of 1 in 15,000 to 20,000 live births. RB is aggressive and if left untreated is invariably fatal. Highly specialized clinical interventions including chemotherapy, cryotherapy localized radiotherapy, and enucleation (eye removal) are required to treat RB [1]. These therapies are effective at prohibiting death but often come at the cost of loss of vision in one or both eyes. Moreover, these therapies exert deleterious physical, cognitive, and psychological damage that persists throughout life. Therefore, there is an urgent need to develop more effective and less invasive treatments.
Nearly all cases of RB are associated with mutations compromising the function of the RB1 tumor suppressor gene. RB1 was one of the first tumor suppressor genes identified following its cloning in 1986-1987 [2]. MYCN hyper amplification without RB1 mutation has also been reported in a small subset of RB cases [3,4]. Previously, our whole genome sequencing identified at least one RB1 allele in all 19 RB samples, with several samples also harboring MYCN amplification. However, we could not identify any mutation other than RB1 and MYCN [5]. Interestingly, it was thought that loss of RB1 function results in the initial development of a benign retinoma that later develops into aggressive RB [6]. Furthermore, familial RB patients with RB1 mutation do not always initiate cancer in both eyes. These observations of incomplete penetrance and phenotypic varaibility of the RB1 gene indicate that other factors are also involved in RB progression. Indeed, nonmutational epigenetic reprogramming and microenvironmental cues have been shown to drive tumorigenesis in many cancers [7].
PRELP is a class II member of the small leucine rich proteoglycan (SLRP) family. SLRPs are highly expressed in ocular tissues and associated with many ocular diseases and cancer [8,9]. These proteins generally reside in the extracellular matrix (ECM) after being secreted, where they were initially recognized to organize and maintain the structural integrity of the ECM via collagen fibrillogenesis. Indeed, PRELP is a molecular anchor between constituents of the ECM and the basement membrane [10]. More recent studies have suggested that SLRPs regulate signaling pathways such as TGF-β and Wnt through direct interaction with either ligands or receptors [8]. This can have a profound effect on cancer-specific mechanisms, Recently, we reported that PRELP expression was profoundly suppressed in the majority of epithelial cancers [11]. We found that PRELP −/− mice spontaneously generated benign urothelial papillary cancer, accompanied by a reduction in tight junctions between urothelial epithelial cells that is associated with EGF and TGF-β pathway activation. These changes are quintessential to the EMT phenotype. As PRELP is located on chromosome 1q31.1, a region hypothesized to be associated with the initiation of malignancy in RB [6], we investigated the potential role of PRELP on the progression of RB through EMT and cell adhesion in this study.

Materials
Antibodies used are listed in Supplemental Table S1. Y79 cells and WERI-RB1 cells were obtained from ATCC and cultured in RPMI-1640 with 10% FBS and 5% L-glutathione. Purified recombinant PRELP were previously reported [12]. Anti-PRELP antibodies were a gift from the Tsumoto laboratory. Conditioned media of PRELP was obtained from tetracycline-inducible PRELP expressed in HEK293T cells.

Human RB Data
Human RB expression profiling data were obtained in our previous paper [5].

Mouse Experiments
Construction and analysis of the PRELP −/− mice is described previously [11]. All experiments were covered by Home Office Animal Project Licence and performed in accordance with University College London guidelines for animal handling.

Viability Assays
First, 0.5 × 10 5 cells were seeded under either suspension or attached conditions in 96-well plates and incubated at 37 • C for 24 h. Cells were then resuspended in a 1:1 ratio with 0.4% trypan blue. Counting of live and dead cells was performed using a Countess II Automated Cell Counter (Thermo Fisher Scientific, Waltham, MA, USA).

Proliferation and Adhesion Assays
For proliferation assays, 2.0 × 10 4 cells were cultured under either suspension or attached conditions in 96-well plates and incubated at 37 • C for the indicated time. Then, the CCK-8 assay (ab228554, Abcam, Cambridge, UK) was performed according to the manufacturer's instructions. A measure of 10 µL of WST-8 reagent was added per 100 µL cell media and incubated for 2 h at 37 • C. Absorbance was then measured at 460 nm. For adhesion assays, 5.0 × 10 4 cells were seeded in non-treated or 1.0 µg/mL laminin (Sigma, Burlington, MA, USA, L2020) coated plates and incubated for 24 h. Cells were washed twice in PBS and the CCK-8 assay was performed as above.

Anchorage-Independent Growth Assays
First, 0.6% 2-hydroxyethyl agarose was added as a base to 6-well plates and allowed to solidify. In addition, 2.0 × 10 4 cells were dissociated and suspended in 0.3% 2-hydroxyethyl agarose. Cells were incubated for up to 14 d to form colonies which were imaged at regular intervals using a phase-contrast microscopy. At the end of the experiment, 0.05% crystal violet was used to stain colonies.

Immunostaining
PRELP −/− retina: Perfusion through direct cardiac approach was used. To analyze dysplasia and other phenotypes, we isolated eye from PRELP −/− mice and wild type mice at ages of 18-25 weeks as indicated. The isolated eyes were fixed with 4% PFA at room temperature for 1 h and embedded in OCT compound (Tissue Tek, Sakura Finetek, Alphen aan den Rijn, The Netherlands). Then, 10 µm sections were used for X-gal staining and immunohistochemistry as reported [11].
Cultured cells: Cells were fixed for 10 min in methanol for ZO-1 antibody, or in 4% PFA against E-cadherin, N-cadherin, paxillin, and β-catenin and incubated with the primary antibody. Then, cells were stained with the AlexaFluor 488 secondary antibody and visualized by Zeiss LSM 710 confocal microscope.
Single cell expression profiling of WERI-RB1 cells: 0.1 × 10 6 cells were seeded under suspension conditions in 96-well plates and treated with 50 µg/mL PRELP for 16 h. Cells were then dissociated. A total of 12,000 cells from each sample were processed through the 10× Genomics Chromium platform pipeline according to the manufacturer's instruction. Gel beads in Emulsion were barcoded by oligonucleotides. After RNA quantification with a Qubit 2.0 Fluorometer, sequencing was performed by Illumina NextSeq 2000. Single cell RNA-sequencing was done in duplicate for both control and PRELP-treated cells and single cell data were combined to produce a pooled mRNA expression profile to obtain the robust result [13,14].
DESeq2 was used to identify differentially expressed genes which was uploaded to Qiagen Ingenuity Pathway Analysis (IPA) for ontological analysis.

Quantification and Statistical Analysis
Error bars in all graphs indicate standard error of mean. Student t-test was used to perform statistical analysis unless stated otherwise. Student t-test vales are * = p-value < 0.05, ** = p-value < 0.005, and *** = p-value < 0.001. For in vitro and in vivo experiments, n > 3. At least three fields of view were analyzed for each immunofluorescence assay and five fields of view for anchorage-independent growth assay. Statistical analysis was performed using SPSS 27 (version 27.0.1; IBM, New York, NY, USA).

PRELP Expression Is Strongly Suppressed in RB
Recently, we have performed whole genome and mRNA expression profiling of human RB [5] and investigated the expression of PRELP and other members of the SLRP family in human RB (Table 1). Almost all members of the SLRP family are suppressed in RB compared with the normal retinal tissue. In particular, PRELP expression was reduced 163-fold (p-value of 1.46 × 10 −22 ). We also confirmed low expression of PRELP in RB cell lines of Y79 and WERI-RB1 as observed in other cancer cell lines [11]. Our whole genome sequencing analysis confirmed that gene mutations or alternations are largely limited to RB1 and NMYC and SLRP family genes are not mutated [5], indicating that PRELP suppression is a consequence of RB1 mutation in RB.

PRELP Is Highly Expressed by Glial Cells in the Retina
After confirmation of PRELP suppression in RB, we examined expression patterns of PRELP in healthy tissues using PRELP −/+ mice [11]. We first examined endogenous expression of PRELP by X-gal staining using a Lac reporter gene cloned into PRELP −/+ mice. Expression patterns between the LacZ reporter gene in PRELP −/+ mouse and the PRELP transcript using in situ hybridization in their wild type equivalents was almost identical [11]. In adult retina, X-gal staining was mainly observed in the retinal ganglion cell layer, the outer plexiform layer, and the basal side of the outer nuclear layer around inner limiting and outer limiting membranes as stripe pattern from apical to basal retina, mimicking Müller glial cell processes ( Figure 1A).
The X-gal staining pattern was further confirmed by β-galactosidase antibody staining ( Figure 1B). In the peripheral retina, PRELP is also expressed at non-pigmented layer of the ciliary body ( Figure 1C). To confirm cell types that express PRELP, double staining of anti-β-galactosidase monoclonal antibody with Müller glial cell-specific marker GFAP was performed ( Figure 1D-I). To determine the Müller glial cell subtypes expressing SLRP family members in the retina, we performed single cell mRNA expression profiling analysis of mouse retina ( Figure 1J). Clustering of cell markers identified four subtypes of Müller glial cells. All Müller glial cell clusters expressed PRELP ( Figure 1K), while other SLRP members were expressed in specific clusters ( Figure 1L-N). This is consistent with the literature [15]. PRELP localization was determined using monoclonal PRELP antibodies [12]. We found that PRELP antibody stained the whole retina with the strongest expression in the RGC layer and outer side of photoreceptor cells, which is consistent with our mRNA expression ( Figure 1O,P).
Other cell types were also examined, using markers for micoglia (IBA-1), pericytes (NG-2), and vascular endothelial cells (isolectin or green-tomato lectin). We found that in addition to Müller glial cells, a small population of microglia (Figure 2A-F) and pericytes ( Figure 2G-L) expressed PRELP while no expression was detected in endothelial cells.

PRELP −/− Mice Exhibit Dysplasia in the Retinal Cell Layers
We examined the effects of PRELP inactivation in the developing retina by comparing retinal morphology in wild type and PRELP −/− mice. Wild type mice presented discrete borders between each retinal layers throughout the entire retina ( Figure 3A,B). However, architecture in the PRELP −/− mouse retina was disrupted ( Figure 3C-G). All PRELP −/− mice had localized regions of severe dysplasia. This occurred in the outer nuclear layer ( Figure 3C-E), inner nuclear layer ( Figure 3F), and ganglion cell layer ( Figure 3G). Although dysplasia was observed in PRELP −/− retina ( Figure 3H), most prominently in the outer nuclear layer. We then quantified cell number in these cell layers ( Figure 3I-K). No significant differences in cell number at non-dysplastic sites were found. On the other hand, cell number was significantly increased in dysplastic areas, associated with the widening of plexiform layers in some cases. These observations suggest that decreasing cell-cell integrity may enhance proliferative potential, inducing dysplasia. The X-gal staining pattern was further confirmed by -galactosidase antibody staining ( Figure 1B). In the peripheral retina, PRELP is also expressed at non-pigmented layer of the ciliary body ( Figure 1C). To confirm cell types that express PRELP, double staining of anti-β-galactosidase monoclonal antibody with Müller glial cell-specific marker GFAP was performed ( Figure 1D-I). To determine the Müller glial cell subtypes expressing SLRP family members in the retina, we performed single cell mRNA expression profiling analysis of mouse retina ( Figure 1J). Clustering of cell markers identified four subtypes of Müller glial cells. All Müller glial cell clusters expressed PRELP ( Figure 1K), while other SLRP members were expressed in specific clusters ( Figure 1L-N). This is consistent with the literature [15]. PRELP localization was determined using monoclonal PRELP antibodies [12]. We found that PRELP antibody stained the whole retina with the strongest expression in the RGC layer and outer side of photoreceptor cells, which is consistent with our mRNA expression ( Figure 1O,P).
Other cell types were also examined, using markers for micoglia (IBA-1), pericytes (NG-2), and vascular endothelial cells (isolectin or green-tomato lectin). We found that in addition to Müller glial cells, a small population of microglia (   Abnormalities were also seen in the structure around both inner limiting membrane and outer limiting membrane in PRELP −/− mice. In the wild type retina, the structure of the inner limiting membrane was uniform. In the PRELP −/− retina, this structure was thinner, more cystic, either hypoplastic or hyperplastic, and often detached ( Figure 3C,D). Detachment of the idiopathic epiretinal membrane, which is located at the apical end feet of Müller glial cells and caused by Müller glial cell-mesenchymal transition, was observed in many cases ( Figure 3C-E,L) [16]. We stained mouse retina with Laminin A to elucidate ECM structure. Laminin A staining, corresponding to basal membrane, was strongly disrupted in PRELP −/− retina ( Figure 3M  Abnormalities were also seen in the structure around both inner limiting membrane and outer limiting membrane in PRELP −/− mice. In the wild type retina, the structure of
Our recent studies showed that PRELP directly binds multiple extracellular proteins such as TGF-β, EGF receptor, IGFI receptor, and p75NTR components [11,12]. These pathways are known to regulate cell cycle, EMT, cell adhesion, and inflammation. Our expression profiling of PRELP −/− retina shows that these pathways are significantly affected ( Figure 4D), supporting that PRELP-mediated regulation of these pathways are involved in the mouse retina as observed in bladder. To estimate the contribution of PRELP downregulation in human cancer, we biostatistically compared expression profiling data of PRELP −/− mouse retina with publicly available data using the Analysis Match function of IPA software (QIAGEN). Figure 4E shows that PRELP −/− mouse retina shows very high similarity with publicly available deposited cancer data in all four analysis methods of Canonical Pathways (CP), Upstream Regulations (UR), Causal Networks (CN), and Diseases and Function (DE), indicating that PRELP suppression in almost all cancers has significant contribution to the cancer initiation or progression.

PRELP Application Reduces RB Cancer Cell Viability in Association with Enhanced Cell Adhesion, Inhibition of Anchorage Independent Growth, and Facilitation of MET
Our observations suggest that suppression of PRELP potentiates RB progression through EMT mechanisms, especially those associated with cell adhesion, as observed in bladder cancer. Thus, we hypothesized that application of PRELP protein to RB cells is likely to inhibit RB progression by reversing EMT via mesenchymal epithelial transition (MET). To elucidate the roles of PRELP in RB progression, we examined the effect of purified recombinant PRELP protein [12] on Y79 and WERI-RB1 RB cell lines. RB cells are known to grow as adhesive or suspension cultures depending on cell culture substrates. We used two substrates: Nunclon Delta-treated plates (attached culture) and non-treated plates (suspension culture). Approximately 18.1% or 25.5% of either Y79 or WERI-RB1 cells attached to non-treated plates, respectively. Almost all RB cells attach to Delta-treated plates.

PRELP Application Reduces RB Cancer Cell Viability in Association with Enhanced Cell Adhesion, Inhibition of Anchorage Independent Growth, and Facilitation of MET
Our observations suggest that suppression of PRELP potentiates RB progression through EMT mechanisms, especially those associated with cell adhesion, as observed in bladder cancer. Thus, we hypothesized that application of PRELP protein to RB cells is likely to inhibit RB progression by reversing EMT via mesenchymal epithelial transition (MET). To elucidate the roles of PRELP in RB progression, we examined the effect of purified recombinant PRELP protein [12] on Y79 and WERI-RB1 RB cell lines. RB cells are We examined the effect of PRELP on live RB cell numbers on non-treated plates or Delta-treated plates. Figure 6A,B shows that, on non-treated plates, PRELP treatment significantly reduced the number of live cells in both Y79 and WERI-RB1. On the other hand, on Delta-treated plates, effect of PRELP on live cell number was weak ( Figure 6C,D), suggesting that the effect of PRELP depends on cell environment or adhesion status. Next, we examined the effect of PRELP on the ratio of live cells to dead cells by trypan blue exclusion. PRELP application significantly reduced the ratio of live cells on both non-treated plates and Delta-treated plates ( Figure 6E,F), indicating that PRELP has negative effect on RB cell survival. Then, we examined the effect of PRELP on cell-substrate adhesion. PRELP application to Y79 or WERI-RB1 cells strongly enhanced cell adhesion to non-treated plates. A 1 µg/mL laminin treatment enhanced cell adhesion to the substrate and PRELP application further enhanced adhesion ( Figure 6G,H). The enhancement of adhesion by PRELP must be related by its biological function because addition of BSA (50 µg/mL) did not alter the adhesion. The observation that PRELP can enhance adhesion of Y79 and WERI-RB1 cells to non-treated or laminin-treated plates suggests that PRELP treatment may inhibit cancer cell growth through inhibition of anchorage-independent growth. Therefore, we examined the effect of PRELP under anchorage-independent growth conditions. In the absence of PRELP, Y79 cells and WERI-RB1 cells developed many large colonies. Addition of PRELP strongly inhibited colony formation ( Figure 6I,J). We counted the number of colonies and single cells ( Figure 6K,L). These observations indicate that PRELP inhibits anchorage independent growth of solo RB cells. Furthermore, we quantified relative size of colonies ( Figure 6M,N), indicating that PRELP reduces the colony size. These corroborate our recent study in bladder cancer using EJ28 bladder cell lines [11]. Our results suggest that PRELP has tumor suppressor-like properties through regulation of EMT and cell adhesion in many types of cancer.
Anchorage-independent growth has been previously shown to produce distinct colony types, which correspond to EMT states and contribute to metastatic potential. As observed in breast cancer [17], there were three distinct colony types of RB cells in our anchorageindependent growth assays ( Figure 6O). Round colonies were highly organized, with a discrete boundary. Mass colonies displayed a lack of organization while clustered colonies were disseminated. These phenotypes have been previously demonstrated to be spaced along the EMT spectrum, with round colonies presenting epithelial-like qualities and clustered colonies demonstrating mesenchymal-like qualities. Application of PRELP to RB cell lines increased the proportion of round colonies ( Figure 6P,Q) whilst decreasing the number of clustered colonies ( Figure 6R,S), indicating that PRELP application resulted in MET. Next, we examined the effect of PRELP application on expression of adherens junctions (β-catenin, E-cadherin, N-cadherin) and tight-junction (ZO-1). PRELP application enhanced β-catenin, E-cadherin, and N-cadherin staining at the plasma membrane ( Figure 7A-U). ZO-1 staining was also enhanced by PRELP application (Figure 7V-BB), indicating that PRELP indeed activate cell-cell adhesions. We also examined the effect of PRELP on paxillin and actin. PRELP treatment resulted in diffused paxillin staining ( Figure 7CC-II). In control, long and clear actin fibers were observed, which were reduced with PRELP treatment ( Figure 7JJ-PP) PRELP application clearly enhanced epithelial type staining and weakens mesenchymal type staining.

Expression Profiling after Application of PRELP Revealed That PRELP Suppresses Various Tumor Related Pathways and Enhances Cell-Cell Adhesion
To elucidate comprehensive action mechanism of PRELP on RB, we applied PRELP protein to Y79 or WERI-RB1 cells and performed mRNA expression profiling. In the case of WERI-RB1, the cells were cultured as suspension condition and then the effect of PRELP (50 µg/mL) was examined by single-cell based bulk mRNA expression profiling as indicated in Experimental Procedures (Supplemental Data S2). Firstly, we examined expression of EMT-related genes and cancer-related genes. Many of EMT-related genes and cancerrelated genes were strongly suppressed after PRELP application ( Figure 8A,B). Then, we performed ontological analysis using IPA software. Cell adhesion pathways are strongly activated, while EMT related pathways are inhibited, supporting our in vitro observations ( Figure 8C). Cancer pathways are strongly inhibited by PRELP ( Figure 8D).  with a discrete boundary. Mass colonies displayed a lack of organization while clustered colonies were disseminated. These phenotypes have been previously demonstrated to be spaced along the EMT spectrum, with round colonies presenting epithelial-like qualities and clustered colonies demonstrating mesenchymal-like qualities. Application of PRELP to RB cell lines increased the proportion of round colonies ( Figure 6P,Q) whilst decreasing the number of clustered colonies ( Figure 6R,S), indicating that PRELP application resulted in MET. Next, we examined the effect of PRELP application on expression of adherens junctions (β-catenin, E-cadherin, N-cadherin) and tight-junction (ZO-1). PRELP application enhanced β-catenin, E-cadherin, and N-cadherin staining at the plasma membrane ( Figure 7A-U). ZO-1 staining was also enhanced by PRELP application (Figure 7V-BB), indicating that PRELP indeed activate cell-cell adhesions. We also examined the effect of PRELP on paxillin and actin. PRELP treatment resulted in diffused paxillin staining (Figure 7CC-II). In control, long and clear actin fibers were observed, which were reduced with PRELP treatment (Figure 7JJ-PP) PRELP application clearly enhanced epithelial type staining and weakens mesenchymal type staining.  (50 μg/mL) was examined by single-cell based bulk mRNA expression profiling as indicated in Experimental Procedures (Supplemental Data S2). Firstly, we examined expression of EMT-related genes and cancer-related genes. Many of EMT-related genes and cancer-related genes were strongly suppressed after PRELP application ( Figure 8A,B). Then, we performed ontological analysis using IPA software. Cell adhesion pathways are strongly activated, while EMT related pathways are inhibited, supporting our in vitro observations ( Figure 8C). Cancer pathways are strongly inhibited by PRELP ( Figure 8D).  Figure 9 shows a drawing of "Molecular Mechanism of Cancer", indicating that many oncogenic pathways such as RAS, NFB, and AKT are negatively regulated by PRELP. Furthermore, cell cycle is inhibited, while apoptosis is activated. TGF- pathway  Figure 9 shows a drawing of "Molecular Mechanism of Cancer", indicating that many oncogenic pathways such as RAS, NFκB, and AKT are negatively regulated by PRELP. Furthermore, cell cycle is inhibited, while apoptosis is activated. TGF-β pathway and growth factor pathways are inhibited by PRELP as observed in biochemical analysis on bladder cancer cells [11]. Figure 9 shows strong contrast to that of PRELP −/− mouse retina analysis (Figure 5), confirming the role of PRELP in cancer progression. Figure 8E shows that PRELP inhibits cell cycle and potentiates cell death. PRELP also inhibits inflammation-related pathways ( Figure 8F). We also performed standard bulk expression profiling analysis using Y79 cells as indicated in Experimental Procedures (Supplemental Data S3). As expected, PRELP application to Y79 cells affected cell adhesion, EMT-related pathways, cancer pathways, cell cycle, and inflammation as observed in WERI-RB1 cells. We compared similarity between WERI-RB1 cell analysis and Y79 cell analysis by Analysis Match. Figure 8G clearly indicates high similarity of these two analyses.
shows that PRELP inhibits cell cycle and potentiates cell death. PRELP also inhibits inflammation-related pathways ( Figure 8F). We also performed standard bulk expression profiling analysis using Y79 cells as indicated in Experimental Procedures (Supplemental Data S3). As expected, PRELP application to Y79 cells affected cell adhesion, EMT-related pathways, cancer pathways, cell cycle, and inflammation as observed in WERI-RB1 cells. We compared similarity between WERI-RB1 cell analysis and Y79 cell analysis by Analysis Match. Figure 8G clearly indicates high similarity of these two analyses. Figure 9. Molecular mechanism of cancer: Schematic drawing of "Molecular Mechanism of Cancer" obtained by ingenuity canonical pathway analysis of PRELP application to WERI-RB1 cells. The majority of cancer-associated pathways were downregulated in PRELP-treated WERI-RB1 cells. Red Figure 9. Molecular mechanism of cancer: Schematic drawing of "Molecular Mechanism of Cancer" obtained by ingenuity canonical pathway analysis of PRELP application to WERI-RB1 cells. The majority of cancer-associated pathways were downregulated in PRELP-treated WERI-RB1 cells. Red = upregulated; green = downregulated; orange = predicted upregulation; blue = predicted downregulation.

Discussion
Enhanced proliferative capacity and remodeling of cell adhesive property are major characteristics of cancers. Here, our analysis of human RB expression profiling and subsequent analysis of PRELP −/− mouse retina identified that suppression of PRELP contributes to weakening cell adhesive property and facilitating EMT in RB, which enhances progression of RB. Furthermore, we demonstrated that PRELP application activates cell-cell adhesion and MET and suppresses viability of RB cell lines. These observations indicate the importance of regulation of cell adhesive property by controlling microenvironment in cancer. PRELP may have a fundamental role in the microenvironment and seems to function as a novel type of a tumor suppressor against RB.

PRELP Secreted from Müller Glial Cells Has Function to Maintain Rigid Retinal Structure
The retina is an immune privileged tissue and is against tumor development even when exposed to UV radiation in daily life, which is a strong carcinogen. The retina is rigidly packed into distinct layers, which are important for the highly coordinated retinal functions. This rigid layered structure of the retina is maintained by a wide range of mechanisms. In particular, Müller glial cells have important roles through providing structural frameworks in the retina and also providing secreted proteins that contribute to cell-cell adhesion [18][19][20]. Our results indicate that PRELP is selectively expressed by Müller glial cells. Furthermore, deletion of PRELP in mice clearly leads to retinal dysplasia and down regulation of many adhesion-related biological events. Moreover, we observed that detachment of inner limiting membrane as observed in idiopathic epiretinal membrane in PRELP −/− mouse retina. Idiopathic epiretinal membrane is located around the apical end feet of Müller glial cells and is caused by TGF-β-Snail pathway-mediated Müller glial cell-mesenchymal transition [16], suggesting that PRELP deletion induced EMT in Müller glial cells themselves. Interestingly, idiopathic epiretinal membrane is associated with myopia [16]. Since PRELP is highly expressed in sclera [21] and regulated EMT [11], PRELP might be aassociated with myopia [22]. These observations indicate that PRELP is one of the molecules that contribute to maintenance of retinal structure.
How does the dysplasia of retina with localized activation of proliferation occur? There are two potential explanations; (1) the damage of cell-cell adhesion firstly initiated dysplasia and then localized proliferation was activated around the dysplasia or (2) abnormal proliferation resulted in dysplasia. We observed that the increased number of cells is only observed in dysplasia regions, suggesting that defect of cell adhesion damage is required for activation of proliferation. Our analysis of molecular mechanism indicates that PRELP can regulate at least three signaling pathways of TGF-β, EGF, and the canonical Wnt pathways [11]. Our expression profiling analysis indicate that all these signaling seems to be affected in PRELP −/− retina and human RB. These pathways are major pathways in the regulation of EMT [23,24]. Indeed, EMT is strongly affected in PRELP −/− retina. In addition, these pathways are well known as major oncogenic pathways [25][26][27][28]. These observations indicate that the defect of PRELP mediated activation of cell-cell adhesion and MET through regulation of TGF-β, EGF, and the canonical Wnt pathways is the major cause of dysplasia in PRELP −/− retina.

Role of PRELP-Depleted Microenvironment in RB Progression
We have found that PRELP expression is strongly downregulated in RB (Table 1) and that downregulation of PRELP revealed important roles in progression of RB as we demonstrated in this study. We previously discovered the downregulation of PRELP in many different cancers, including retinoblastoma [11]. We revealed that loss of PRELP is an important prerequisite for bladder cancer onset using PRELP −/− mouse. In this study, only four SLRPs had a fold change less than −100. Of these, NYX and OGN expression levels were 80.0 and 1584.3, respectively. PRELP and OPTC expression levels were 4102.7 and 5104.3 respectively, indicating that these proteins could be more important in normal retina function. Additionally, the PRELP gene is located on chromosome 1q31.1, which is a site frequently perturbed in cancer. The comparison of expression profiling data of PRELP −/− retina and human cancers has revealed that influenced mechanisms and genes in PRELP −/− retina are highly conserved with cancer progression mechanisms in other cancers ( Figure 4E), indicating that the PRELP downregulation in human RB contributes to the progression through regulation of the pathways regulated by PRELP. In addition, expression profiling analysis of PRELP −/− mouse retina indicated that PRELP itself has ability to regulate cancer-related pathways, cell adhesion, and EMT without RB1 abnormality. Regulation of cell-cell adhesion in cancer is fundamental. Anchorage independent growth is a hallmark of cancer cells [29,30]. EMT, which contains loss of cell-cell adhesion, is required for cancer initiation and progression [31]. In general, less adhesive cancer is more malignant. RB is a very malignant type of cancer. Reflecting this, cancer cell lines originated from RB are not adhesive, requiring suspension culture. The importance of regulation of cell-cell adhesion in RB progression is clearly demonstrated in this study, in which PRELP has fundamental roles in the regulation of cell-cell adhesion and inhibiting RB cell progression.
In RB, the importance of cell-cell adhesion was previously demonstrated by CDH11 study [32]. CDH11 is a type II classical cadherins family, which regulate cell-cell adhesion. Marchong et al. showed that CDH11 gene copy number and expression are frequently lost in human RB. Crossing of Cdh11 null mice with TAg-RB, RB model mice, resulted in faster tumor growth than Tag-RB alone through downregulation of tumor cell death, proposing that CDH11 works as a tumor suppressor through activation of cell-cell interaction. CDH11 is reported to inhibit the canonical Wnt pathway through downregulation of β-catenin [33]. However, our expression profiling data of human RB shows that CDH11 is not significantly suppressed in RB [5]. Furthermore, in seven RB that we analyzed, we did not identify mutation or amplification of CDH11 [5]. In addition, our expression profiling analyses of PRELP −/− mice and PRELP overexpression show that the expression of CDH11 was not significantly affected, indicating that PRELP and CDH11 are important for cell-cell adhesion through interaction with the canonical Wnt pathways but work under distinct mechanisms in tumor suppression.
Moreover, the molecular function of PRELP as a regulator of major cancer-related pathways of TGF-β, EGF, and Wnt also strongly indicates that PRELP suppression is not just consequence of RB development. PRELP is located at 1q31.1, where it is associated with transition from retinoma to RB [6]. Our study strongly suggests that PRELP downregulation may have an important role in RB progression. We found that the upregulation of both N-cadherin and E-cadherin in PRELP-treated cells indicated that the two cadherins are regulated by PRELP. Although the upregulation of N-cadherin is associated with EMT, our in vitro assays demonstrated that PRELP application enhanced cell adhesion and suppressed anchorage-independent growth of RB cells. These findings suggest that PRELP altered oncogenic and tumor suppressive cadherin expression coincidentally, but tumor suppressive roles are dominant.
To link these processes to RB1 mutation, genes directly under expression control of RB1 or E2Fs (de-repressed by RB1 mutation) were analyzed. RB1 loss-of-function directly reduced expression of a cadherin family member, CELSR1, linked to adherens junction stability [34]. E2F1 de-repression also induced ECT2 expression; an oncogene encoding a RhoA-activating GEF linked to NSCLC cell proliferation and invasion [35]. RB1 mutation may directly impair immune cell recruitment and contribute to reduced inflammation, as several immune cell adhesion molecules were reduced by E2F1 de-repression (SELE, ICAM1, TAPBP). Several additional regulators of chromatin structure and transcription were directly enhanced by RB1 loss-of-function and/or E2F activation, including RBBP4, a ubiquitous component of histone deacetylase complexes that was upregulated in an RB1-mutation-driven embryonic brain tumor model in zebrafish [36] and is known to enhance mesenchymal marker expression and invasion in human cervical and colon cancer cell lines [37,38].

Difference between PRELP −/− Mouse Retina and Human RB
Our analysis of PRELP −/− mouse retina showed dysplasia-associated activation of proliferation. However, mice did not develop malignancies, as observed with human RB. This observation indicates that RB1 mutation and/or MYCN amplification can regulate PRELP-independent mechanisms that are important for proper development of malignant RB.

Generation of PRELP-Depleted Microenvironment in RB
Our results indicate that PRELP suppression is important for RB progression. However, there are two possible explanations of how the low expression is established. (1) PRELP expression is indeed suppressed in the cancer cells during RB progression. If so, this suppression must be under transcriptional control because our previously published whole genome sequencing did not identify any mutations or deletion of the PRELP gene [5]. Interestingly, RBBP4, a downstream regulator of RB1, is upregulated by RB1 mutation [36] and enhances mesenchymal marker expression in human cervical and colon cancer cell lines [37,38]. (2) If RB is not formed from Müller glial cells, exclusion of Müller glial cells from cancer tissue because of expansion of cancer cells might be the mechanism as reduction of PRELP expression in RB. To identify which mechanism is correct, the cell of origin for RB is important. However, the cell of origin has not been clearly determined [39]. Amacrine neurons and horizontal neurons are reported to initiate RB-like cancer in RB1/p107 mouse model and RB1/p130 mouse model, respectively [40][41][42]. PRELP is expressed largely in Müller glial cells in mice. In RB1/p107 and RB1/p130 models, some cancer cells were positive to Müller glial cell markers. Müller glial cells have proliferative potential properties [43,44]. Cone photoreceptor cells are reported as a candidate of cell of origin in human iPS cell-based model [7]. Recently, Norrie et al. demonstrated that iPS cells originated from RB patients can initiate RB after xenograft into mouse eyes. Single cell analysis of RB development process shows that the RB had mixed cell type properties [45]. This is consistent with expression profiling of human RB. Furthermore, our PRELP −/− retina caused dysplasia in both the INL and ONL. Further study is required to determine RB cell of origin. Although cell of origin is important for understanding RB, understanding of the autonomous PRELP suppression mechanisms may not be very important because PRELP is a secreted protein and influences cancer progression via microenvironment.

ECM in Cancer
Secreted PRELP protein distributes largely around the apical side of Müller glial cell end-feet but widely defuses over the retina. ECM proteins have fundamental roles in various aspects in cancer initiation and progression [46,47]. In particular, they create a complex microenvironment for cancer progression [48]. The importance of ECM proteins was clearly observed when overexpression of MMP3 in mice induced mammary carcinoma, despite having cancer-associated gene mutation [49]. Some SLRP family members are known substrates of MMPs [50,51], and Type-I SLRPs, such as Decorin and Biglycan have been reported to affect cancer progression [52]. This paper demonstrates that the Type-II SLRP member, PRELP, is involved in RB progression. ECM proteins, including PRELP are important in regulating ocular tissues This paper and our previous paper [11] propose that PRELP is an important ECM protein in the initiation and progression of various cancers through control of cell-cell. PRELP-mediated improvement of the cancer microenvironment might be a better strategy for treatment of RB.

Conclusions
In this paper, we demonstrated the functional roles of PRELP, which are the association with the inhibitory effects for RB through regulating cell-cell adhesion, EMT, and inflammatory signaling. Moreover, exogenously supplied PRELP protein to RB cells showed enhancement of cell-cell and cell-substrate adhesion and the inhibition of anchorage-independent growth by reversing EMT. Therefore, PRELP may have potential as an anti-cancer drug to treat RB patient in the future.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers14194926/s1, Table S1: List of antibodies; Data S1: Signaling pathways and biological functions regulated genes in PRELP −/− retina; Data S2: Signaling pathways and biological functions significantly affected after application of PRELP protein to WERI-RB1 cells; Data S3: Signaling pathways and biological functions significantly affected after application of PRELP protein to Y79 cells.

Data Availability Statement:
The data presented in this study are available in the supplementary materials and are available from the corresponding author upon reasonable request.