Endothelial FGFR1 (Fibroblast Growth Factor Receptor 1) Deficiency Contributes Differential Fibrogenic Effects in Kidney and Heart of Diabetic Mice

Supplemental Digital Content is available in the text. Endothelial-to-mesenchymal transition (EndMT) has been shown to contribute to organ fibrogenesis. We have reported that N-acetyl-seryl-aspartyl- lysyl-proline (AcSDKP) restored levels of diabetes mellitus-suppressed FGFR1 (fibroblast growth factor receptor 1), the endothelial receptor essential for combating EndMT. However, the molecular regulation and biological/pathological significance of the AcSDKP-FGFR1 relationship has not been elucidated yet. Here, we demonstrated that endothelial FGFR1 deficiency led to AcSDKP-resistant EndMT and severe fibrosis associated with EndMT-stimulated fibrogenic programming in neighboring cells. Diabetes mellitus induced severe kidney fibrosis in endothelial FGFR1-deficient mice (FGFR1fl/fl; VE-cadherin-Cre: FGFR1EKO) but not in control mice (FGFR1fl/fl); AcSDKP completely or partially suppressed kidney fibrosis in control or FGFR1EKO mice. Severe fibrosis was also induced in hearts of diabetic FGFR1EKO mice; however, AcSDKP had no effect on heart fibrosis in FGFR1EKO mice. AcSDKP also had no effect on EndMT in either kidney or heart but partially suppressed epithelial-to-mesenchymal transition in kidneys of diabetic FGFR1EKO mice. The medium from FGFR1-deficient endothelial cells stimulated TGFβ (transforming growth factor β)/Smad-dependent epithelial-to-mesenchymal transition in cultured human proximal tubule epithelial cell line, AcSDKP inhibited such epithelial-to-mesenchymal transition. These data demonstrated that endothelial FGFR1 is essential as an antifibrotic core molecule as the target of AcSDKP.

D iabetic kidney disease is a major cause of end-stage kidney disease worldwide. [1][2][3][4] Regardless of its cause, kidney fibrosis is a universal pathological feature of kidney disease. Kidney fibrosis is characterized by extracellular matrix accumulation and is accompanied by a series of renal malfunctions. 5 Although controversial hypotheses have been proposed, epithelial or endothelial cells via epithelial-to-mesenchymal transition (EMT) or endothelial-to-mesenchymal transition (EndMT) programs, characterized by the expression of mesenchymal markers (α-smooth muscle actin; smooth muscle protein 22α, and vimentin), has been shown to contribute in organ fibrogenesis. [6][7][8] FGF (fibroblast growth factor) signaling plays a key role in maintaining endothelial barrier function and endothelial cell survival. [9][10][11][12][13][14] FGF exerts its biological effects through binding to related FGFRs. The FGFR family belongs to a subfamily of receptor tyrosine kinases and comprises 4 family members (FGFR1, FGFR2, FGFR3, and FGFR4). 15 FGF/FGFR1 signaling has been confirmed to suppress TGFβ (transforming growth factor β) signaling in endothelial cells 16 and TGFβinduced EndMT through inducing miRNA let-7. 17 Fibroblast growth factor receptor substrate 2, a key adaptor of FGFR signaling, deficiency in the mesoderm induced cardiac outflow tract misalignment and hypoplasia. 18 These reports indicated the significant roles of endothelial FGFR1 in organ protection, yet the roles in organ injuries have not been analyzed in diabetes mellitus.
In our study, we hypothesize that endothelial FGFR1 is essential for the antifibrotic and anti-EndMT effects of AcSDKP in diabetic mice.

Methods
Data that support the findings of this study are available from the corresponding author upon reasonable request. A detailed section of methods and materials is provided in the Data Supplement.

Summary
The detailed in vivo and in vitro experimental protocols are shown in Data Supplement. All animal experiments were approved by the IACUC of Kanazawa Medical University (protocol numbers 2015-59 & 2017-73). All experiments are performed according to Japanese guidelines and institutional ethics committee guidelines. Endothelial-specific FGFR1 knockout mice (FGFR1 EKO ; C57BL/6J background) were generated by crossing VE-cadherin-Cre recombinase-positive (Cdh5-Cre) mice (NIBIOHN, Osaka, Japan) and FGFR1 loxP/loxP floxed (FGFR1 fl/fl ) mice with genes flanked by the loxP sites of exon 4. The control group was VE-cadherin Cre-negative FGFR1 fl/fl mice. The lineage tracing experiment was validated by crossing with VE-cadherin-Cre mice and ROSA26-dtTOMATO mice. Animal experiment was conducted with 8-week old male FGFR1fl/fl (control) and endothelial-specific FGFR1 knockout mice (FGFR1fl/fl; Cdh5-Cre FGFR1 EKO ). Four or 5 number of mice was analyzed in each group. Streptozotocin (50 mg/kg×5 days body weight [BW]) was us2ed for the induction of diabetes mellitus. 34 The induction of diabetes mellitus was confirmed as a blood glucose level >16 mmol/L 2 weeks after streptozotocin injection by glucose strips. At 12 weeks after diabetes mellitus induction, the randomized male diabetic mice were divided into 2 groups (placebo [water]; AcSDKP treatment [500 μg/kg BW per day using an osmotic mini-pump for 4 weeks]). Some of nondiabetic control mice divided into 2 groups (untreated control and AcSDKP treatment [500 μg/kg BW per day using an osmotic mini-pump for 4 weeks]). Mice were anesthetized using inhalation machine NARCOBIT(II; Tokyo, Japan) with Isoflurane Inhalation Solution (Pfizer). An osmotic mini-pump was subcutaneously implanted on the back of each mouse. All mice were euthanized at 16 weeks after induction of diabetes mellitus. Blood pressure was monitored by tail cuff method. 35 Kidney and heart sections were used for morphological analyze and mesenchymal program detection. Urine albumin and plasma cystatin C level were evaluated for kidney function. Glomerular ultrastructure was analyzed by electron microscopy. Human microvascular endothelial cells, human proximal tubule epithelial cell line, AC16 human cardiac myocytes were employed for in vitro analyze and conditioned medium experiment to detect cellcell interaction.

Statistical Analysis
The data are expressed as the mean±SEM. ANOVA followed by multiple comparison Tukey analysis and nonparametric statistics ( Figure 4A) were used to determine significance, which was defined as P<0.05, if not specifically indicated.
GraphPad Prism software (ver 7.0; La Jolla, CA, RRID: SCR_002798) was used for the statistical analysis. Data analysis was blinded.

Experimental Animal Characteristics
To determine the pathological significance of endothelial FGFR1 in the diabetic kidney and heart, we first generated conditional endothelial knockout mice by crossing mice with FGFR1 genes flanked by loxP sites (FGFR1 fl/fl ) with VE-cadherin-Cre transgenic mice (FGFR1 EKO : FGFR1 fl/fl ; VE-cadherinCre; Figure S1 in the Data Supplement). Lineage tracing analysis demonstrated that both glomerular and interstitial endothelial cells expressed Cre under the control of the VE-cadherin promoter ( Figure S2). FGFR1 levels in endothelial cells were diminished in FGFR1 EKO mice ( Figure S3). Blood glucose levels were elevated to a similar degree upon streptozotocin-induced diabetes mellitus in both FGFR1 fl/fl mice without Cre (control) and FGFR1 EKO mice, and AcSDKP did not affect the diabetic mice ( Figure 1A). Compared with control mice, FGFR1 EKO mice displayed lower basal BW and blood pressure ( Figure 1B and 1D). In diabetic mice, BW was decreased, whereas kidney/BW and heart/BW ratios were increased in control and FGFR1 EKO mice ( Figure 1B, 1C, and 1D). However, in diabetic mice there was no change in absolute kidney and liver weight whereas heart weight was found decreased when compared with control mice ( Figure  S4). Heart rate and systolic blood pressure were all reduced in diabetic mice; AcSDKP normalized all of these values in diabetic control mice ( Figure 1E and 1F). However, AcSDKP had no effect on diabetes mellitus-induced heart rate and systolic blood pressure suppression in FGFR1 EKO mice ( Figure 1E and 1F).

AcSDKP Partially Ameliorated Kidney Fibrosis in Diabetic FGFR1 EKO Mice But Had No Effect on Heart Fibrosis
Masson Trichrome staining and Sirius red staining first revealed that AcSDKP normalized kidney fibrosis in diabetic control mice, as described in a previous study 6 (Figure 2A and 2C and Figure S5A and S5C). Furthermore, when compared with control mice, FGFR1 EKO mice exhibited mild kidney fibrosis under nondiabetic conditions ( Figure 2A and 2C and Figure S5A and S5C). Diabetic FGFR1 EKO mice displayed severe kidney fibrosis compared with diabetic control mice (Figure 2A and 2C and Figure S5A and S5C), but AcSDKP partially ameliorated this kidney fibrosis in FGFR1 EKO mice ( Figure 2A and 2C and Figure S5A and S5C). AcSDKP also significantly ameliorated heart fibrosis in diabetic control mice ( Figure 2B and 2D and Figure S5B and S5D). Similar to the kidney results, diabetic FGFR1 EKO mice displayed severe heart fibrosis when compared with diabetic control mice ( Figure 2B and 2D and Figure S5B and S5D); AcSDKP had no effect on heart fibrosis in diabetic FGFR1 EKO mice ( Figure 2B and 2D and Figure S5B and S5D). In the control mice, AcSDKP treatment did not cause significant alterations either on kidney fibrosis or in fibrogenic markers ( Figure S6A and S6B).
Unexpectedly, when the kidney glomerulus ultrastructure was analyzed by transmission electron microscopy, there were no significant differences between control and FGFR1 EKO mice under basal conditions ( Figure 3A and 3C). In diabetic mice, both control and FGFR1 EKO mice, some portion of the podocyte foot processes displayed thickening and effacement ( Figure 3A and 3C). Endothelial cells in diabetic control mice were not significantly altered ( Figure 3A and 3C), but endothelial cells in diabetic FGFR1 EKO mice exhibited damage ( Figure 3A and 3C). AcSDKP treatment ameliorated both podocyte and endothelial appearance abnormalities in control mice and, unexpectedly, in FGFR1 EKO mice ( Figure 3A and 3C). Periodic acid-Schiff staining revealed that diabetic FGFR1 EKO mice displayed severe glomerular pathological alterations when compared with control mice ( Figure 3B and 3D); AcSDKP completely ameliorated these alterations in diabetic control mice but only partially in diabetic FGFR1 EKO mice ( Figure 3B and 3D). Supporting these observations, increased levels of plasma cystatin C and urine albumin in diabetic control mice were restored to normal by AcSDKP ( Figure 3E and 3F). Diabetic FGFR1 EKO mice displayed higher levels of plasma cystatin C and urine albumin than diabetic control mice, but AcSDKP partially ameliorated these trends ( Figure 3E and 3F).
We then evaluated mesenchymal protein levels and found that AcSDKP suppressed the increases in vimentin and α-SMA levels in the kidneys and hearts of diabetic control mice ( Figure 4A). Nondiabetic FGFR1 EKO mice displayed increased vimentin and α-SMA expression ( Figure 4A), and these trends were further increased in both the kidneys and hearts of diabetic FGFR1 EKO mice. AcSDKP partially inhibited these mesenchymal markers in the kidneys but not in the hearts of FGFR1 EKO mice ( Figure 4A). Immunohistochemistry analysis of the kidneys and hearts also confirmed these results ( Figure 4B and 4C and Figure S7A and S7B).

FGFR1 EKO Displayed AcSDKP-Resistant EndMT in Diabetic Mice
EndMT and EMT play key roles in the development of organ fibrosis. 36,37 To investigate whether endothelial FGFR1 deficiency could influence EndMT in the heart or EndMT/EMT in the kidney, we examined CD31/α-SMA, CD31/vimentin (EndMT), and E-cadherin/α-SMA (EMT) co-labeling in the kidney and VE-cadherin/α-SMA co-labeling in the heart (endothelial markers: CD31, VE-cadherin; epithelial marker: E-cadherin, mesenchymal markers: α-SMA, vimentin). AcSDKP treatment suppressed EndMT and EMT induction in the kidneys and EndMT in hearts of diabetic control mice ( Figure 5A through 5D). When compared with nondiabetic control mice, nondiabetic FGFR1 EKO mice exhibited EndMT and EMT ( Figure 5A through 5D). Diabetic FGFR1 EKO mice displayed significantly higher levels of both EndMT and EMT than diabetic control mice ( Figure 5A through 5D); AcSDKP partially inhibited EMT in the kidneys of diabetic FGFR1 EKO mice ( Figure 5C) but did not influence EndMT in the kidneys and hearts of diabetic FGFR1 EKO mice ( Figure 5A, 5B, and 5D). TGFβ1, a key inducer of EMT, was higher in nondiabetic FGFR1 EKO mice than in nondiabetic control mice ( Figure  S8). TGF-β1 levels in kidneys were higher in diabetic control mice, and FGFR1 EKO mice displayed much higher levels of kidney TGFβ1. AcSDKP completely suppressed TGF-β1 levels in diabetic control mice; however, the anti-TGF-β1 effects of AcSDKP were only partial in diabetic FGFR1 EKO mice ( Figure S8).
AC16 human cardiac myocyte cells 39 were cultured with conditional medium, the medium from FRS2 siRNA-transfected HMVECs had no effect on the mesenchymal markers (α-SMA, SM22α, Vimentin) and TGFβR1 expression ( Figure  S13A through S13D). In addition, both N-TGFβ1 and SIS3 display no influence on the phenotype ( Figure S13A through Pathological characteristics of the organs. A, Sirius red staining for kidney fibrosis. C, Relative fibrosis areas were calculated using ImageJ software. Six independent images were analyzed. N=4-5 in each group. Scale bar: 60 µmol/L. B, Sirius red staining for cardiac fibrosis. D, Relative fibrosis areas were calculated using ImageJ software. AcSDKP indicates N-acetyl-seryl-aspartyl-lysyl-proline; FGFR1, fibroblast growth factor receptor 1; and STZ, streptozotocin. S13D), suggesting the minor fibrogenic interaction between EndMT and AC16 human cardiac myocyte.

Discussion
FGFR1 signaling has shown to be a key EndMT inhibitor and that anti-EndMT effect of AcSDKP is dependent on endothelial FGFR1 signaling in vitro. 6 Despite these important biological roles of both FGFR1 and AcSDKP, little is known about the molecular regulation and biological/pathological significance of the AcSDKP-FGFR1 relationship. Here, we demonstrated that the anti-EndMT and antifibrotic effects of AcSDKP in diabetic mice were mediated through endothelial FGFR1. In brief, (1) diabetic FGFR1 EKO mice displayed severe kidney and heart fibrosis compared with diabetic control mice and (2) AcSDKP partially ameliorated kidney fibrosis in diabetic FGFR1 EKO mice but had no effect on heart fibrosis in FGFR1 EKO mice. (3) AcSDKP partially suppressed EMT in kidney but had no effect on EndMT in either the kidneys or hearts of diabetic FGFR1 EKO mice. (4) The conditioned media from FGFR1-deficient HMVEC led to TGFβ/Smad signaling-dependent EMT in HK2 cells; AcSDKP inhibited such EMT. These data clearly demonstrated that endothelial . N-acetyl-seryl-aspartyl-lysyl-proline (AcSDKP) partially ameliorated severe kidney fibrosis in diabetic FGFR1 EKO mice. A, Electron microscopy was performed to determine glomerular pathological alterations. Representative electron microscopy (EM) images are presented in the figure. N=2. Scale bar 1 μm. C, Relative podocyte foot processes and glomerular basement membrane (GBM) thickness were calculated using ImageJ software. Six independent images of the staining were analyzed. B, Images of periodic acid-Schiff staining for glomerular pathological alterations. D, Relative pathological areas were calculated using ImageJ software. Six independent images were analyzed. N=4-5 in each group. Scale bar: 40 µmol/L. E and F, Kidney function was analyzed according to the albumin-creatinine ratio (E) and plasma cystatin C levels (F). N=4-5 in each group. The data are expressed as the mean±SEM in the graph. FGFR1 indicates fibroblast growth factor receptor 1; and STZ, streptozotocin. FGFR1 signaling is essential for the anti-EndMT and antifibrotic effects of AcSDKP.
The FGF/FGFR signaling pathway plays a critical role in normal vascular homeostasis, and endothelial FRS2 blockage induces EndMT and is associated with cardiac dysfunction in mice. 17,18 In cultured endothelial cells, inhibition of FGFR1 signaling by either a neutralizing antibody or FRS2 knockdown resulted in AcSDKP-resistant EndMT. 6 Here, we generated streptozotocin-induced type I diabetic FGFR1 EKO mice and found that endothelial FGFR1 deficiency significantly diminished the anti-EndMT and antifibrotic effects of AcSDKP in the heart. Even though some contributions of blood pressure fluctuation on the tissue injury in diabetic FGFR1 EKO mice cannot be ruled out. In the kidney, AcSDKP displayed certain antifibrotic and tissue protective effects, even in diabetic FGFR1 EKO mice, without ameliorating EndMT in the kidney and also blood pressure restoration. Cardiomyocytes, the main cell type of the heart, did not express mesenchymal markers in diabetic FGFR1 EKO mice; most likely, the cardiomyocyte phenotype did not contribute substantially to heart fibrosis, or it was not converted into a mesenchymal phenotype in the current models, even in diabetic FGFR1 EKO mice. Therefore, AcSDKP solely affected the fibrotic phenotype in the heart through the endothelial FGFR1-dependent suppression of EndMT in control mice but not in FGFR1 EKO mice. Accordingly, the EMT program in the kidney tubules of diabetic FGFR1 EKO mice was also suppressed by AcSDKP. Regard with this, AcSDKP inhibits EMT program induced by the conditioned media from FRS2 deficient endothelial cells. When analyzed by electron microscopy, AcSDKP unexpectedly ameliorated endothelial damage in the glomeruli of diabetic FGFR1 EKO mice. This result is perhaps due to the amelioration of either the mesangial or podocyte phenotype by AcSDKP because the microenvironment mediated through such cell types directly affects the endothelial phenotype in glomeruli, [40][41][42] subsequently resulting in the recovery of endothelial homeostasis, even in FGFR1 EKO mice.
Interestingly, conditioned media from FRS2-deficient endothelial cells induced TGF-β-dependent EMT. Recently, we also confirmed that the EMT phenotype induced . FGFR1 EKO mice had increased diabetes mellitus-induced mesenchymal markers, and N-acetyl-seryl-aspartyl-lysyl-proline (AcSDKP) partially inhibited these markers in the kidney but not in the heart. A, Western blotting analysis to detect α-SMA (α-smooth muscle actin) and vimentin expression in both kidney and heart tissues. Representative Western blot images from 4 independent analyses are shown. The data were normalized to β-actin and are shown as the mean±SEM. Nonparametric Mann-Whitney U test has been applied for analysis of statistical significance. The immunohistochemical analysis of vimentin levels in both kidney (B) and heart (C) tissues. Representative immunohistochemical (IHC) images are shown from 4 independent experiments. Scale bar: 60 µmol/L. FGFR1 indicates fibroblast growth factor receptor 1; and STZ, streptozotocin.
TGF-β-dependent EndMT in a paracrine manner. 43 In the kidney, any cell type could contribute to a fibrotic phenotype via the interactions between each cell type. 6,[43][44][45][46][47][48][49][50] There has been controversial discussion about the diversity of matrixproducing mesenchymal phenotypes, especially regarding the presence of EndMT or EMT programs; our analysis clearly demonstrated the significance of EndMT in organ fibrosis and that EndMT influences EMT processes in the kidney.
In conclusion, our research revealed that endothelial FGFR1 deficiency in diabetic mice resulted in severe organ fibrosis in both the kidney and heart via the induction of AcSDKP-resistant EndMT.

Perspectives
We examined the cellular and molecular bases of the antifibrotic actions of AcSDKP in diabetes mellitus. AcSDKP suppressed fibrogenic phenotypes by inhibiting the EndMT processes in the diabetic kidneys and heart. Our study discovered that the antifibrotic actions of AcSDKP is dependent on endothelial FGFR1. Moreover, FGFR1 is found as a key anti-EndMT molecule and the loss of FGFR1 led to fibrogenic phenotypes in diabetic kidneys and heart. Furthermore, our data suggests that FGFR1 is critical regulator of EndMTassociated EMT activation in diabetic kidneys. AcSDKP inhibits EMT, either FGFR1 dependent or independent mechanism, therefore AcSDKP can inhibit some fibrogenesis mechanisms in diabetic kidney even in endothelial FGFR1-deficient mice. The crosstalk between endothelium and tubular epithelial cell in the kidneys emerges as a critical mechanism in the pathobiology of fibrotic kidneys. However, in the heart, likely EndMT would display predominant roles in fibrogenesis, therefore AcSDKP cannot inhibit the heart fibrosis in diabetic endothelial FGFR1-deficient mice. Unexpectedly, AcSDKP treatment in diabetic control mice, restored the systolic blood pressure and improved kidney fibrosis. Some contributions of blood pressure alterations on the tissue injury in diabetic mice cannot be ignored. Further studies are needed to study the effect of AcSDKP in the blood pressure regulation and the role of FGFR1 in AcSDKP mediated blood pressure regulation. These data shed new light on the biology of the AcSDKP-endothelial FGFR1 interaction in organ fibrogenesis and provide a possible novel therapy for combating organ fibrosis via ameliorating endothelial FGFR1 signaling.  . Endothelial FGFR1 (fibroblast growth factor receptor 1) deficiency-induced endothelial-to-mesenchymal transition (EndMT) leads to epithelial-tomesenchymal transition (EMT) dependent on the TGFβ (transforming growth factor β) signaling pathway. A, Conditioned media experiment design. HMVECs (human microvascular endothelial cells) were transfected with scrambled or FRS2 (fibroblast growth factor receptor substrate 2) siRNA; after 6 h, the medium was changed to fresh and incubated for 48 h. The subsequently harvested media was transferred to HK2 cell cultures. B, Representative Western blotting analysis of E-cadherin, SM22α (smooth muscle protein 22α) and α-SMA (α-smooth muscle actin) expressions. Six independent experiments are shown. Densitometric analysis of the levels relative to β-actin is shown. C, ELISA analysis of TGFβ1 levels from conditioned medium. D, HK2 cells cultured in conditioned medium were treated with N-TGFβ1. α-SMA, SM22α, and TGFβR1 levels were analyzed by Western blotting. E, Densitometric analysis of the levels relative to β-actin levels (n=6) in each group was performed. F, Immunofluorescence microscopy analysis of E-cadherin, vimentin, FSP1, and P-Smad3 expression in conditioned medium with or without N-TGFβ1 treatment. For each slide, images of 6 different fields at ×400 magnification were evaluated. Scale bar 30 μm.