Molecular Vision 2019; 25:625-635 <>
Received 19 September 2018 | Accepted 18 October 2019 | Published 21 October 2019

Gremlin-1: An endogenous BMP antagonist induces epithelial-mesenchymal transition and interferes with redifferentiation in fetal RPE cells with repeated wounds

Duo Li, Dongqing Yuan, Han Shen, Xiying Mao, Songtao Yuan, Qinghuai Liu

These authors contributed equally to this work.

Department of Ophthalmology, the First Affiliated Hospital of Nan Jing Medical University, Nanjing, China

Correspondence to: Qinghuai Liu, Department of Ophthalmology, the First Affiliated Hospital of Nanjing Medical University, 300 Guangzhou Road, Nanjing, 210029, P.R. China; email:


Purpose: To investigate the role of Gremlin-1, which is an endogenous antagonist of the bone morphogenetic protein (BMP) signaling pathway, in inducing epithelium-mesenchymal transition (EMT) in fetal RPE cells after repeated wounds.

Methods: Subconfluent repetitive passages in fetal RPE cells were regarded as a model of repeated wounds. A phase contrast microscope was used to observe the morphology and pigment formation in cells. The expression of GREM1 (Gene ID: 26585; OMIM 603054) and EMT- or RPE-related genes in cells was evaluated with quantitative PCR (qPCR). Recombinant human protein Gremlin-1 (0.1 μg/ml) was added every day to investigate the molecular effects of Gremlin-1 on fetal RPE cells. The cell migration rate was investigated using a cell wound scratch assay, and western blotting was used to analyze the representative proteins (P-cadherin, ZO-1, vimentin, Smad4, and phosphorylated-Smads). In addition, transfection of siRNA was used to explore the rescue effects on EMT cells through the downregulation of GREM1. Finally, LDN193189, which is a type of pan-inhibitor of BMP receptors, was used to verify whether complete blocking of the BMP pathway interferes with the redifferentiation in low-passage fetal cells, even if the cells were treated with transforming growth factor beta 1 (TGF-β) inhibitors.

Results: In fetal RPE cells, the expression of GREM1 were gradually upregulated with repetitive passages, and at the same time, the function-specific genes in fetal RPE cells (TJP1, PMEL, BEST1, RPE65, and MERTK) were downregulated while the EMT-specific genes were upregulated. In addition, GREM1 had a similar expression pattern as SNAI1, which is a key transcription factor to trigger EMT. Recombinant human Gremlin-1 promoted EMT with the upregulation of SNAI1 and elevated the cell migration rate in a cell scratch assay, as well as decreased the expression of two key transcription factors of RPE embryonic development (MITF and OTX2) and the RPE marker, RPE65. Furthermore, the EMT marker, vimentin, and the TGF-β pathway downstream transcription factor phosphorylated-Smad2 (p-Smad2) increased, but the epithelial marker, ZO-1, was reduced. Additionally, Smad4, which plays a role as a Snail1 cooperator by binding Smad3, was also increased. In contrast, GREM1 silencing increased the expression of MITF and OTX2, which means there was better redifferentiation in subconfluent fetal RPE cells, but it had little influence on p-Smad2 compared to the negative control group. Finally, by adding LDN193189, the BMP signaling pathway was blocked, and this block led to poor redifferentiation in low-passage cells, although the cells were treated with TGF-β inhibitors. In addition, as positive feedback to block the BMP pathway, GREM1 was subsequently upregulated.

Conclusions: In fetal RPE cells, Gremlin-1 induces EMT and inhibits redifferentiation by promoting the TGF-β pathway and inhibiting the BMP pathway. GREM1 silencing alleviates EMT and increases the redifferentiation of cells by relieving the blockade of the BMP pathway. However, GREM1 silencing has no effects on the TGF-β pathway. Thus, Gremlin-1 may serve as a novel target to treat proliferative vitreoretinopathy (PVR) and inhibit subretinal fibrosis, which is a risk factor for influencing the therapeutic effects of anti-vascular endothelial growth factor (anti-VEGF) on neovascular age-related macular degeneration (nAMD).


RPE is the pigmented cell layer located between the neurosensory retina and the vascular choroid. Fibrosis in RPE causes diseases such as proliferative vitreoretinopathy (PVR) and neovascular age-related macular degeneration (nAMD) [1]. In fibrosis, epithelial-mesenchymal transition (EMT) has been identified as a major driver, and in this process, epithelial cells such as RPE lose their polarity and tight junction. These changes result in an increase in migration and invasive properties [2]. In PVR, RPE cells become more invasive after EMT. These cells migrate into the vitreous cavity and form a contractile epiretinal membrane (ERM) that causes tractional retinal detachment [3]. In AMD, repetitive damage in RPE cells is generally considered to be the major pathogenesis that leads to the loss of central vision and choroidal neovascularization (CNV). Although intravitreal injection of anti-vascular endothelial growth factor (VEGF) medicine has become a standard therapeutic method for addressing CNV, subretinal fibrosis and formation of scars after injection threaten the therapeutic effect and result in unexpected visual acuity loss [4,5]. According to some studies, approximately half of the eyes after treatment could develop scars after 2 years, and in untreated CNV, scar formation is also an important morphological feature that influences the prognosis of disease [6,7]. EMT occurring in RPE is regarded as a major cause of this phenomenon [8].

The molecular mechanism of EMT is complex. Some transcription factors, such as Snail1, are conventionally known as keys to trigger the process, and some signaling pathways, such as the transforming growth factor beta 1 (TGF-β) pathway, are also important factors for promoting EMT [9]. In addition, bone morphogenetic protein (BMP), Notch, and the wingless (Wnt) pathways regulate this process [10]. These signaling pathways have extensive crosstalk, but the specific relationship in EMT is largely unknown [11]. Radekeet et al. confirmed that repetitively passaged RPE cells could induce EMT, and after treatment with A83–01, which is a TGF-β inhibitor, mesenchymal cells could be restored. However, when the cells were continuously passaged to passage 7, RPE cells still lost their functions and entered the mesenchymal state [12]. Therefore, a TGF-β inhibitor has great potential to prevent EMT of fetal RPE cells, but other factors that exist in the long run undermine the effectiveness of the inhibitor in an unknown way.

Gremlin-1 is one of the endogenous BMP antagonists that preferentially binds to BMP-2 or BMP-4 but secondarily binds to BMP-7 [13].In some research studies, BMP-4 and BMP-7 had inhibitory effects on EMT [14,15]. Therefore, Gremlin-1 likely promotes EMT by inhibiting BMP signaling. In some studies of pancreatitis and chronic kidney disease, Gremlin-1 was a key profibrotic factor for promoting fibrosis by inhibiting the BMP pathway and activating the TGF-β pathway. These studies also showed that GREM1 (Gene ID: 26585; OMIM 603054) knockdown or knockout in cells or mice inhibit EMT [16,17]. In some cancers, such as mesothelioma, Gremlin-1 has been reported to promote cell migration and results in cancer cells that are more invasive by activating the TGF-β pathway and changing the extracellular matrix (ECM) [18]. Therefore, in this study, we investigated the effects of Gremlin-1 on inducing fetal RPE cells in the epithelial-mesenchymal transition and interfering with their redifferentiation.


Fetal RPE cultures and treatments

RPE cells were isolated from three different aborted fetuses and cultured using a previously published protocol [19]. Plating density was 10,000 cells/cm2, and the cultured medium was exchanged every 2 days. Cells that were needed to maintain differentiation were treated with SB431542 (Sigma, St. Louis, MO, 10 μM) every day. On day 32, cells were harvested by using trypsin and plated at 10,000 cells/cm2 for routine serial passage. In addition, recombinant human Gremlin-1 (0.1 μg/ml, R&D, Shanghai, China) was added to the cells every day to explore the molecular functions of Gremlin-1 in RPE cells. The fetal RPE cells used in this study were obtained from the Center of Reproductive Medicine at the First Affiliated Hospital of Nanjing Medical University with informed consent and in accordance with the Declaration of Helsinki and the study adhered to the ARVO statement on human subjects.

An observation about the morphology and pigments of fetal RPE

The morphology of the cells cultured in vitro was observed with a phase contrast microscope (Nikon TS-100, Tokyo, Japan). And The pigments in cells were analyzed using bright-field micrographs [12].

Quantitative PCR (qPCR)

Total RNA from cells was extracted using TRIzol (Invitrogen Life Technologies, Shanghai, China). cDNA was synthesized with a Revert Aid First Strand cDNA Synthesis Kit (Thermo, Shanghai, China). FastStart Universal SYBR Green Master (ROX; Roche, Basel, Switzerland) was used to amplify cDNA, and the StepOnePlus real-time PCR system (ABI, Carlsbad, CA) was used to detect the real-time change in cDNA. qPCR was conducted with the following protocol: 1 cycle of 95 °C for 3 min, followed by 45 cycles of 95 °C for 10 s, 58 °C for 20 s, 72 °C for 10 s. ΔΔCT was used to calculate the differences in relative gene expression, and the results represented the mRNA expression level. ACTIN (Gene ID:60; OMIM 102630) was used as an internal reference.

Cell scratch wound assay

Fetal RPE cells at passage 2 were grown to 100% confluency and divided into two groups (the control and Gremlin-1 groups). A marker pen was used to mark the edges of the scratches on the underside of the culture plates. A single scratch was made through the center of the cell sheets with a sterile pipette tip. The scratched areas were photographed immediately after wounding. After 2 days with treatment, the scratched areas were re-photographed using a phase contrast microscope, and the cell migration rate was assessed. Assays were performed in triplicate.

siRNA and transfection

GREM1 siRNA (Silencer®Select, Ambion, Boston, MA) was used to knock down GREM1, and negative control siRNA (Silencer™ FAM-labeled Negative Control No. 1 siRNA, Invitrogen, Boston, MA) was used as negative control. Lipofectamine 3000 (Invitrogen) was used to transfect siRNA into cells, and the method of transfection was conducted according to the reagent protocol (Invitrogen). After transfection, the cells were harvested on day 4 for measurement.

Western blotting

Fetal RPE cells were scraped off with cell scrapers and collected in a 1.5 ml centrifuge tube. Radioimmunoprecipitation assay (RIPA; Servicebio, Wuhan, China), which contained a protease inhibitor (Servicebio), was added and triturated with a pipette to ensure complete cell lysis in an ice bath for 30 min. The supernatant was collected after centrifuging for 10 min (20000 ×g, 4 °C). The protein samples were separated with sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) gels and then transferred to the polyvinylidene difluoride (PVDF) membrane (Millipore, Shanghai, China). The membranes were incubated in the blocking buffer (5% nonfat dry milk or 5% bovine serum albumin [BSA] in 0.5% Tris-buffered saline/Tween 20 [TBST]) for 1 h at room temperature and hybridized with specific primary antibodies for P-cadherin (Abcam), Smad4, vimentin (Servicebio), phosphorylated-Smad2 (p-Smad2; BIOSS, Beijing, China), Smad1 (Ruiying Biology, Suzhou, China), and ZO-1 (Proteintech Group, Wuhan, China) at a dilution of 1:1,000 overnight at 4 °C. The membranes were washed three times before they were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies at an appropriate dilution (1:3,000) for 1 h at room temperature. The bands were visualized with an enhanced chemiluminescence (ECL) HRP substrate (Millipore) using a chemiluminescence imaging system (Syngene G:BOX Chemi HR16; Syngene, Frederick, MD). β-actin served as an internal reference.

Statistical analysis

All experiments were performed at least three times. The data are shown as the mean ± standard error of the mean (SEM) and were analyzed with GraphPad Prism 6 (GraphPad Software, San Diego, CA). Differences among the two groups were analyzed with a t test or a paired t test. One-way ANOVA (ANOVA) was used to analyze the samples that had more than two groups, and Dunnett’s multiple comparisons test was performed with the control groups. A p value of less than 0.05 was considered statistically significant.


The expression of GREM1 was upregulated in the EMT of fetal RPE cells

Fetal RPE is one of the in vitro experimental models that resembles human RPE tissue more than ARPE-19. Generally, in early passages (from passage 1 to passage 3), most of the fetal RPE cells have a stable proliferation capacity and maintain some primal characteristic morphologies and functions, such as having a cobblestone-like appearance and pigments with abundant expression of some RPE function-specific genes, such as MERTK (Gene ID:10461, OMIM 604705), BEST1 (Gene ID:7439,OMIM: 607854), RPE65 (Gene ID:6121, OMIM 180069), CDH1 (Gene ID:999, OMIM 192090), and PMEL (Gene ID: 6490, OMIM:155550). However, in this study, when cells were passaged to passage 4, the fetal RPE cells lost their characteristic shapes and pigments, and the cells gradually gained a fibroblast-like appearance (Figure 1A). The function-specific genes in fetal RPE cells decreased at passage 4 compared to the cells at passage 2 (Figure 1B). Moreover, passage 2 cells expressed more epithelial markers, especially CDH1, and the specific RPE epithelial marker P-cadherin also had more expression compared to the passage 4 cells (Figure 1C). However, compared to passage 2 cells, the EMT marker, VIM (Gene ID: 7431, OMIM: 193060), had higher expression in passage 4 cells (Figure 1D). In addition, the expression of GREM1, which encodes Gremlin-1, was statistically significantly upregulated in the passage 4 cells, which were in an EMT state (Figure 1E).

To investigate the expression level of GREM1 in high passage cells when a TGF-β inhibitor prevented the rescue effects, fetal RPE cells were treated daily with SB431542 (10 μM). The cells treated with SB431542 had a more cobblestone-like appearance and pigments. However, when the cells were passaged to passage 11, the rescue effects of SB431542 disappeared because most of the cells became fibroblast-like, and the pigments were statistically significantly reduced (Figure 2A). Furthermore, the expression of RPE function-specific genes was decreased with repetitive passages (Figure 2B), and the expression of GREM1 in SB431542-treated cells also had passage-related upregulation, especially in passage 11 cells (Figure 2C).

GREM1 and SNAI1 had a similar expression trend in the progress of the EMT

To explore the dynamic expression changes in GREM1 in fetal RPE cells and EMT cells, the relation between GREM1 and SNAI1 (Gene ID: 6615 OMIM: 604238) was analyzed. Passage 4 cells, which were supposed to enter EMT, were seeded at 10,000 cells/cm2 and were divided into two groups: one that was treated daily with SB431542 (the 10 μM SB431542 group), and the control group without any treatment (the control group). Cells were observed under microscopy and were harvested to evaluate the gene expression level and trend by qPCR on day 2, 5, 10, 15, and 30 after seeding. In the SB431542 group, the cells differentiated into RPE-like cells, while in the control group, the cells gradually developed a fibroblast-like shape (Figure 3A). In addition, the expression of GREM1 and SNAI1 had a similar trend in both groups. In the control group, SNAI1 was approximately upregulated along with the upregulation of GREM1. However, in the SB431542 group, SNAI1 had low expression similar to GREM1. Meanwhile, RPE65 was expressed in a gradually increasing manner while GREM1 and SNAI1 had low expression but maintained low expression when GREM1 and SNAI1 had high expression (Figure 3B).

Exogenous gremlin-1 induced EMT by upregulating SNAI1 and downregulating the expression of key transcription factors in fetal RPE cells

To explore the exact effects of Gremlin-1 fetal RPE cells, recombinant Gremlin-1 proteins (0.1 μg/ml, once per day) were added to passage 2 cells that were seeded at 10,000 cells/cm2 until day 15. Obviously, the cells in the Gremlin-1 group appeared mesenchymal-like, but the cells in the control group appeared more cobblestone-like (Figure 4A). A cell wound scratch assay showed that the cells in the Gremlin-1 group had a higher cell migration rate compared to the control group cells on the 2nd day after treatment (Figure 4B). Additionally, fetal RPE cells treated with recombinant Gremlin-1 proteins expressed more SNAI1 and VIM; however, there was less MITF (Gene ID:4286 OMIM: 156845) and OTX2 compared to the control group, and RPE65 was also downregulated in the Gremlin-1 group cells (Figure 4C). Furthermore, in terms of the protein expression, ZO-1 had lower expression than the control group while vimentin had higher expression in the Gremlin-1 group (Figure 4D). Furthermore, a western blot showed that the active form of Smad2, p-Smad2, was enhanced, and at the same time, Smad4 also increased compared to the control group (Figure 4D).

GREM1 knockdown alleviated EMT and promoted redifferentiation in high passage fetal RPE cells by enhancing the BMP pathway

To investigate whether GREM1 knockdown prevents EMT of fetal RPE cells, siRNA was transfected into cells, and the transfection efficiency was confirmed with qPCR (Figure 5A). On the 4th day after transfection, the siGREM1 group cells appeared to have a more cobblestone-like shape, but all of the negative control (NC) group cells maintained a fibroblast-like shape (Figure 5B). Meanwhile, in the siGREM1 group cells, the expression level of MITF, OTX2, and RPE65 increased and that of VIM decreased. However, SNAI1 had no statistically significant change compared to the NC group cells (Figure 5C). For the changes in the representative protein expression after transfection, ZO-1 and p-Smad1 increased, vimentin was reduced, and p-Smad2 between the siGREM1 and NC group cells had no statistically significant differences (Figure 5D).

Completely blocking the BMP pathway inhibited redifferentiation in low passage fetal RPE cells and upregulated GREM1

BMPs play an important role in the embryonic development of RPE, and they have some effects on inhibiting EMT [14,15]. Although Gremlin-1 is one of the endogenous BMP antagonists that binds BMPs and is upregulated in mesenchymal RPE, binding some kinds of BMPs did not provide abundant evidence to prove that completely blocking the BMP pathway would inhibit mostly sub-confluent low passage fetal RPE cells from redifferentiating into fetal RPE cells. Therefore, LDN193189 was used as a pan-inhibitor to verify whether inhibition of the BMP pathway interferes with redifferentiation. LDN193189 is a selective chemical inhibitor of BMP type I receptor kinases, which interrupts the phosphorylation activity of BMP type I receptors ALK2 and ALK3 (IC50=5 nM and 30 nM, respectively). Therefore, the molecular concentration was kept to 100 nM to ensure the suppressive efficiency of the BMP pathway and to avert other redundant concentration-related effects, such as inhibition of TGF-β type I receptors (ALK4, ALK5, and ALK7, IC50≥500 nM). Passage 2 cells were divided into four groups: the control, SB431542 (SB group), SB431542+LDN193189 (S+L group), and LDN193189 (LDN group) groups. After treatment, cells in the LDN and S+L groups had a poor appearance with abnormal shapes and lacked pigment compared to the control group and SB group cells. The control group cells and the SB group cells had no statistically significant differences in their morphology (Figure 6A). In addition, RPE function-specific genes in the LDN group and S+L group cells had much lower expression than the other two groups (Figure 6B), and GREM1 was upregulated after LDN193189 was used, especially in the LDN group (Figure 6C).


EMT is one of the most important pathological changes in RPE wound responses. EMT is not only the key process in the pathogenesis of PVR but also causes fibrosis in the subretinal space that decreases the therapeutic effects of anti-VEGF on retinal neovascular diseases, such as CNV. Thus, the need to inhibit EMT for treating retinal fibrosis is extremely urgent. Although TGF-β inhibitors can rescue mesenchymal RPE, they always have a smaller effect on cells with repetitive wounds, such as cells that are repetitively passaged in vitro. Therefore, other factors must promote EMT.

Gremlin-1 is an endogenous BMP antagonist. Gremlin-1 serves as a negative regulator in the self-feedback system in the BMP signaling pathway and interacts with neighbor cells or itself through paracrine or autocrine mechanisms [20]. Gremlin-1 predominantly binds to BMP-2 and BMP-4, and interferes with their binding to receptors. In this study, mesenchymal RPE cells expressed more GREM1, although they were treated with SB431542 to inhibit EMT after several passages. The expression process of GREM1 in the EMT cells showed an approximately similar trend to SNAI1; however, GREM1 was already shown to be upregulated before SNAI1 expression. According to studies in other diseases, Gremlin-1 promotes EMT in renal fibrosis and chronic pancreatitis, as well as enhances cancer cell invasiveness, such as mesothelioma cells. The results of these studies indicated that Gremlin-1 could promote SNAI1 expression, but the exact mechanism of this promotion is unknown. The same results were also verified in the present study, and the fetal RPE cells treated with recombinant Gremlin-1 protein expressed more SNAI1. Therefore, it is not clear whether Gremlin-1 is one of the determining factors for SNAI1 expression, but it is likely one of the stimulators. Therefore, to detect the real effects on cells, recombinant human Gremlin-1 was used, and it obviously facilitated RPE cell migration, which is one of the mechanisms underlying retinal detachment in PVR. According to the western blotting results, Gremlin-1 not only inhibited the BMP pathway but also promoted the TGF-β pathway because of an increase in p-Smad2, which is an activated transcription factor downstream of the TGF-β pathway. At the same time, Smad4, which binds Smad3 and cooperates with Snail1, was also increased [9]. The TGF-β pathway is one of the most important pathways that triggers EMT, but in some ways, the BMP pathway plays a negative role in the occurrence of EMT [15]. Therefore, Gremlin-1 exacerbates EMT by inhibiting BMP signaling and promotes TGF-β signaling to induce EMT.

In the embryonic development of eyes, BMPs secreted by the surface ectoderm can specify RPE development in chick embryos. In the optic cup without a surface ectoderm, exogenous BMPs can rescue the function of these cells and initiate MITF expression so that it triggers embryonic development of RPE [21,22]. Generally, low passage fetal cells express MITF and OTX2 and regain a hexagon shape and pigments [23,24]. However, in the present study, exogenous Gremlin-1 downregulated the expression of MITF and OTX2 in low passage cells, which indicates that blockade of the BMP pathway leads to poor differentiation in fetal RPE cells. In contrast, GREM1 knockdown by transfecting siRNA relieved the downregulation effects of Gremlin-1 on MITF and OTX2 and rescued high passage mesenchymal RPE cells. However, SNAI1 maintained a similar expression level compared with its expression in the NC group cells, but VIM decreased in a novel way. The possible reasons for this phenomenon are that p-Smad1 increased, but p-Smad2 had no statistically significant change, and these results indicate that GREM1 knockdown can alleviate the suppression of the BMP signaling pathway but may lack an ability to inhibit the TGF-β pathway. Therefore, although stable TGF-β signals induce continuous SNAI1 expression, GREM1 knockdown facilitates activation of the BMP pathway to induce RPE differentiation by upregulating MITF and OTX2. These results counteract the effects of TGF-β [9]. In addition, although some cells were treated with SB431542, LDN193189 (as an inhibitor of the BMP type I receptor) still interfered with the redifferentiation in low passage fetal RPE cells, and the expression of GREM1 was simultaneously increased. These results reveal that complete inhibition of ALK2 and ALK3 can promote EMT and interrupt fetal RPE cells redifferentiation, and the upregulation of GREM1 further induces EMT as positive feedback. The results also reveal that the BMP pathway is necessary for redifferentiation in fetal RPE cells, but Gremlin-1 has a similar negative effect as LDN193189. In contrast, inhibiting TGF-β with SB431542 had no influence on low passage cells.

Until recently, there had been no specific chemical molecule to inhibit Gremlin-1; therefore, there has been no opportunity to confirm whether persistently inhibiting Gremlin-1 restores mesenchymal RPE instead of transient downregulation using siRNA. Meanwhile, stable transfection should also be used to make long-term observations. However, the transient transfection results reveal that there is a potential chance to inhibit EMT when TGF-β inhibitors lose their effects. In addition, Gremlin-1 has other novel biologic functions. In bones and intestines, the expression of GREM1 defines osteochondral reticular (OCR) stem cells and intestinal reticular stem cells (iRSCs) [25]. Researchers have also identified that Gremlin-1 is a proangiogenic regulator like VEGF that binds vascular endothelial factor receptor 2 (VEGFR2) and can increase angiogenesis in vitro or in vivo [16] [26]. The results suggest that Gremlin-1 may be not only a promoter of fibrosis but also an inducer of neovascular disease in the retina. Therefore, Gremlin-1 is likely to be a potential therapeutic target for some intraocular diseases, like PVR, proliferative diabetic retinopathy (PDR), and neovascular AMD, through antifibrosis and anti-neovascularization mechanisms.

Gremlin-1 is not the only endogenous BMP inhibitor. Several endogenous inhibitors, such as Noggin and chordin (chordin-like 1 and chordin-like 2), also have the ability to block the BMP signaling pathway, but the relation and exact functions of these factors for EMT are unknown [14]. However, as a hypothesis, these factors could have some regulating effects on RPE cells in embryonic development, differentiation, and EMT. This possibility needs to be studied further.


Prof. Qinghuai Liu and Prof. Songtao Yuan are co-corresponding authors for this article. This research was supported by grants from the National Key Research and Development Project of China (2017YFA0104101), the National Natural Science Funds of China (81,870,694) and the Jiangsu Natural Science Foundation (BK20151586). We thank the Center of Reproductive Medicine at the First Affiliated Hospital of Nanjing Medical University for providing the fetal RPE cells.


  1. Friedlander M. Fibrosis and diseases of the eye. J Clin Invest. 2007; 117:576-86.
  2. Nieto MA, Huang RYJ, Jackson R, Thiery JP. EMT: 2016 Cell. 2016; 166:21-45.
  3. Tamiya S, Kaplan HJ. Role of epithelial–mesenchymal transition in proliferative vitreoretinopathy Exp Eye Res. 2016; 142: :26-31.
  4. Nowak JZ. AMD–the retinal disease with an unprecised etiopathogenesis: in search of effective therapeutics. Acta Pol Pharm. 2014; 71:900-16.
  5. Salomon Y. Cohen, Hassiba Oubraham, Joel Uzzan, Lise Dubois, Ramin Tadayoni. Causes of unsuccessful ranibizumab treatment in exudative age-related macular degeneration in clinical settings. Retina. 2012; 32:1480-85
  6. Daniel E, Cynthia A. Toth, Juan E Grunwald, Glenn J Jaffe, Daniel F Martin, Stuart L Fine, Jiayan Huang, Guishuang Ying, Stephanie A Hagstrom, Katrina P Winter, Maureen G Maguire. Risk of Scar in the Comparison of Age-related Macular Degeneration Treatments Trials Ophthalmology. 2014; 121:656-66.
  7. Daniel E, Pan W, Ying G, Benjamin J. Kim, Juan E Grunwald, Frederick L Ferris, Glenn J Jaffe, Cynthia A Toth, Daniel F Martin, Stuart L Fine, Maureen G Maguire. Development and Course of Scars in the Comparison of Age-related Macular Degeneration Treatments Trials Ophthalmology. 2018; 125:1037-1046.
  8. Ishikawa K, Kannan R, Hinton DR. Molecular mechanisms of subretinal fibrosis in age-related macular degeneration Exp Eye Res. 2015; 142:19-25.
  9. Lamouille S, Xu J, Derynck R. Molecular mechanisms of epithelial–mesenchymal transition Nat Rev Mol Cell Biol. 2014; 15:178-96.
  10. Gonzalez DM, Medici D. Signaling mechanisms of the epithelial-mesenchymal transition Sci Signal. 2014; 7:re8
  11. Guo X, Wang X. Signaling cross-talk between TGF-β/BMP and other pathways Cell Res. 2009; 19:71-88.
  12. Monte J. Radeke, Carolyn M Radeke, Yinghsuan Shih, Jane Hu, Dean Bok, Lincoln V Johnson, Peter J Coffey. Restoration of mesenchymal retinal pigmented epithelial cells by TGFβ pathway inhibitors: implications for age-related macular degeneration Genome Med. 2015; 7:58-58.
  13. Derek P. Brazil, Rachel H Church, Satnam Surae, Catherine Godson, Finian Martin. BMP signalling: agony and antagony in the family Trends Cell Biol. 2015; 25:249-64.
  14. Zeisberg M, Shah A, Kalluri R. Bone Morphogenic Protein-7 Induces Mesenchymal to Epithelial Transition in Adult Renal Fibroblasts and Facilitates Regeneration of Injured Kidney J Biol Chem. 2005; 280:8094-100.
  15. Yao H, Li H, Yang S, Li M, Zhao C, Zhang J, Xu G, Wang F. Inhibitory Effect of Bone Morphogenetic Protein 4 in Retinal Pigment Epithelial-Mesenchymal Transition. Sci Rep. 2016; 6
  16. Staloch D, Gao X, Liu K, Xu M, Feng X, Judith F Aronson MF, George H. Greeley, Cristiana Rastellini, Celia Chao, Mark R Hellmich, Yanna Cao, Tien C Ko. Gremlin is a key pro-fibrogenic factor in chronic pancreatitis J Mol Med. 2015; 93:1085-93.
  17. Rodriguesdiez R, Lavoz C, Carvajal G, Rayegomateos S, Diez RR, Ortiz A, Egido J, Mezzano S, Ruizortega M. Gremlin is a downstream profibrotic mediator of transforming growth factor-beta in cultured renal cells Nephron, Exp Nephrol. 2012; 122:62
  18. Yin M. Mira Tissari, Jenni Tamminen, Irene Ylivinkka, Mikko Ronty, Pernilla Von Nandelstadh, Kaisa Lehti, Marko Hyytiainen, Marjukka Myllarniemi, Katri Koli. Gremlin-1 is a key regulator of the invasive cell phenotype in mesothelioma Oncotarget. 2017; 8:98280-97.
  19. Gamm D. J Nicholas Melvan, R L Shearer, I Pinilla, Grzegorz Sabat, Clive N Svendsen, Lynda S Wright. A Novel Serum-Free Method for Culturing Human Prenatal Retinal Pigment Epithelial Cells Invest Ophthalmol Vis Sci. 2008; 49:788-99.
  20. Scheel C, Eaton EN, Li S, Christine L. Chaffer, Ferenc Reinhardt, Kong Jie Kah, George W Bell, Wenjun Guo, Jeffrey S Rubin, Andrea L Richardson, Robert A Weinberg. Paracrine and autocrine signals induce and maintain mesenchymal and stem cell states in the breast. Cell. 2011; 145:926-40.
  21. Muller F, Rohrer H, Vogelhopker A. Bone morphogenetic proteins specify the retinal pigment epithelium in the chick embryo Development. 2007; 134:3483-93.
  22. Steinfeld J, Steinfeld I, Coronato N, Hampel M, Paul G. Layer, Masasuke Araki, Astrid Vogelhopker. RPE specification in the chick is mediated by surface ectoderm-derived BMP and Wnt signalling Development. 2013; 140:4959-69.
  23. Singh R. M Joseph Phillips, David Kuai, Jackelyn Meyer, Jessica M Martin, Molly A Smith, Enio T Perez, Wei Shen, Kyle Wallace, Elizabeth E Capowski, Lynda S Wright, David M Gamm. Functional Analysis of Serially Expanded Human iPS Cell-Derived RPE Cultures Invest Ophthalmol Vis Sci. 2013; 54:6767-78.
  24. Shih Y, Monte J. Radeke, Carolyn M Radeke, Peter J Coffey. Restoration of Mesenchymal RPE by Transcription Factor-Mediated Reprogramming. Invest Ophthalmol Vis Sci. 2017; 58:430-430.
  25. Daniel L. Worthley, Michael Churchill, Jocelyn T Compton, Yagnesh Tailor, Meenakshi Rao, Yiling Si, Daniel E Levin, Matthew G Schwartz, Aysu Uygur, Yoku Hayakawa, Stefanie Gross, Bernhard W Renz, Wanda Setlik, Ashley N Martinez, Xiaowei Chen, Saqib Nizami, Heon Goo Lee, H Paco Kang, Jonmichael Caldwell, Samuel Asfaha, C Benedikt Westphalen, Trevor A Graham, Guangchun Jin, Karan Nagar, Hongshan Wang, Mazen A Kheirbek, Alka Kolhe, Jared Carpenter, Mark A Glaire, Abhinav Nair, Simon Renders, Nicholas A Manieri, Sureshkumar Muthupalani, James G Fox, Maximilian Reichert, Andrew S Giraud, Robert F Schwabe, Jeanphillipe Pradere, Katherine D Walton, Ajay Prakash, Deborah L Gumucio, Anil K Rustgi, Thaddeus S Stappenbeck, Richard A Friedman, Michael D Gershon, Peter A Sims, Tracy C Grikscheit, Francis Y Lee, Gerard Karsenty, Siddhartha Mukherjee, Timothy C Wang. Gremlin 1 Identifies a Skeletal Stem Cell with Bone, Cartilage, and Reticular Stromal Potential Cell. 2015; 160:269-84.
  26. Mitola S, Ravelli C, Moroni E, Salvi V, Leali D, Ballmerhofer K, Zammataro L, Presta M. Gremlin is a novel agonist of the major proangiogenic receptor VEGFR2. Blood. 2010; 116:3677-80.