Allelic Depletion of grem1 Attenuates Diabetic Kidney Disease

All animal procedures were licensed by the Irish Department of Health and Children and approved by the local animal research ethics committee at University College Dublin. grem1 heterozygote knockout mice (grem1^+/−) were generated by Richard Harland, University of California at Berkeley (Berkeley, CA) (10). Experimental animals were generated by crossing wild-type C57BL/6J and grem1^+/− mice, with male offspring used in the study. Mice were maintained in a conventional animal facility in standard caging, with free access to water and standard rodent chow. Genotyping was performed using DNA extracted from ear punches as described previously (10).

Seven- to 10-week-old male mice (both wild-type and grem1^+/−), weighing ∼19 g at the onset of the experimental protocol were genotyped and then randomly divided into two groups: A, treated with streptozotocin (STZ) (Sigma) dissolved in 100 mmol/l citrate buffer, pH 4.5; or B, treated with citrate buffer alone (http://www.amdcc.org). STZ was dissolved in sterile citrate buffer and injected intraperitoneally (50 mg/kg) within 10 minutes of preparation on 5 consecutive days. Fasting blood glucose levels were assayed after a 6-h fast between 8:00 a.m. and 2:00 p.m. with tail venipuncture at 2:00 p.m. Diabetes was confirmed by two consecutive daily measurements of fasting blood glucose >15 mmol/l 2 weeks after STZ injection. To maintain weight and prevent ketoacidosis, long-acting insulin (Insulotard, Nova Nordisk) subcutaneously at a dose of ⅓ IU three times weekly was started in all diabetic mice at week 18 of diabetes.

Mice were housed individually in mouse metabolic cages (Technoplast) for 24 h to collect urine 5 days before sacrifice. Urine volumes were recorded, and urinary glucose, ketone, and erythrocyte levels were monitored semiquantitatively with Multistix reagent strips (Bayer). Cardiac puncture was performed at the time of sacrifice. The left renal artery was clamped and the left kidney was removed, weighed, and dissected. The inferior renal pole was promptly processed for electron microscopy, and the superior renal pole was snap-frozen and stored at −80°C for RNA and protein extraction. Renal perfusion was performed via cannulation of the left ventricle with an 18-gauge needle. Initial perfusion using gravity was performed with sterile normal saline (pH 7.4) for 5 min, followed by 4% (wt/vol) paraformaldehyde (pH 7.4) for 5 min. The perfused right kidney was then removed and incubated in 4% paraformaldehyde for 24 h at room temperature. Kidneys were processed, cut at 3-μm thickness on a rotary microtome, and stained with hematoxylin/eosin, periodic acid Schiff, or picrosirius red. Stained sections were scored independently (single-blinded) by pathologists using normal light microscopy. The amount of mesangial matrix was scored with periodic acid Schiff staining (score 1–4). Collagen distribution in the cortex was scored with Sirius red staining using a normal light microscope (score 1–5).

Urinary albumin was measured using an Albuwell M kit, and urinary creatinine was measured using a Creatinine Companion murine ELISA kit (Exocell, Philadelphia, PA). Urine was collected in metabolic cages, urine volume was recorded, and 24-h urinary albumin excretion levels were assayed with the Albuwell M assay kit.

Plasma creatinine, serum lipids, and whole blood A1C were measured at the Mouse Metabolic Phenotyping Core, Vanderbilt Medical Center (Nashville, TN). Plasma creatinine was measured on a high-performance liquid chromatography Zorbax SCX strong cation exchange column (Agilent, Wilmington, DE) as described previously (http://www.amdcc.org). A1C was measured using a DCA 2000 analyzer and cartridges (Bayer). Total plasma cholesterol and triglycerides were measured by standard enzymatic assays (Raichem). HDL cholesterol was measured after precipitation of VLDL and LDL using dextran sulfate and magnesium.

Creatinine clearance was calculated using the equation (urine volume [microliters] × urine creatinine [milligrams per deciliter])/(serum creatinine [milligrams per deciliter] × 1,440 [minutes]) and then corrected to total body weight at sacrifice (grams) to give creatinine clearance in microliters per minute per gram body weight.

The left inferior kidney pole was removed, diced into 1-mm cubes, and fixed in 2.5% (v/v) glutaraldehyde in 0.1 mol/l cacodylate buffer. The samples were washed in 0.1 mol/l cacodylate buffer and then postfixed in 1% (wt/vol) aqueous osmium tetroxide. The tissues were washed, dehydrated through a graded series of ethanols, and embedded in Spurr resin. Thick sections (0.5 μm) were cut, affixed to glass slides, stained with toluidine blue, and viewed by light microscopy. Thin (100 nm) sections were cut from selected areas and viewed with an FEI CM-12 transmission electron microscope operated at 80 keV. Glomerular basement membrane (GBM) thickness measurements were assessed as follows: two to four glomeruli were randomly selected in each slide, and serial measurements were taken at intervals from the margins of the lamina rara interna to lamina rara externa. Images were taken and analyzed with an AMT XR41 digital TEM camera system. Up to 60 measurements were taken per kidney.

Total RNA was isolated from snap-frozen renal poles by homogenization in 1 ml TRIzol (GibcoBRL, Life Technologies) using a Polytron (Kinematica) and a Qiagen RNeasy kit. Reverse transcription was performed using SuperScript II (Invitrogen), followed by quantitative real-time PCR on an ABI Prism 7700 sequence detection system. Mouse grem1 (Mm00488615 S1), BMP-7 (Mm00432102 m1), fibronectin (Mm00482221), vimentin (Mm00432102 m1), and CTGF (Mm00515790 g1) real-time oligo probes were purchased from Applied Biosystems.

Portions of the renal pole were lysed in radioimmunoprecipitation assay buffer containing 50 mmol/l Tris-HCl (pH 7.4), 1% (vol/vol) Nonidet P-40, 0.25% (vol/vol) sodium deoxycholate, 150 mmol/l NaCl, and 1 mmol/l EDTA, supplemented with fresh 1 mmol/l phenylmethylsulfonyl fluoride, 1× protease inhibitor cocktail (Sigma), 1 mmol/l NaF, 40 mmol/l β-glycerophosphate, 2 μmol/l Microcystin, and 1 mmol/l sodium vanadate. Protein extracts (25 μg) were then separated by 10% (vol/vol) SDS-PAGE and blotted using phospho-Smad1/5/8 (95115; Cell Signaling), total Smad1/5/8 (sc-6031R; Santa Cruz), BMP-7 (ab27569; Abcam), or β-actin (Sigma) antibodies exactly as described previously (4).

All data were plotted as mean ± SE. Student's two-tailed t tests or one-way ANOVA with Tukey-Kramer multiple comparison post hoc tests were calculated using the Instat software package.

Using the replacement β-galactosidase “knock-in” gene activity as a marker of grem1 promoter activity (10), grem1 gene expression was detected in heterozygous grem1^+/− knockout cells (Fig. 1 A). Homozygous grem1^−/− knockout mouse embryonic fibroblasts displayed approximately double the activity of β-galactosidase, suggesting that maximal grem1 expression required both copies of the grem1 gene. We therefore hypothesized that grem1 expression would be reduced in the kidneys of grem1^+/− mice in response to hyperglycemia.

Type 1 diabetes was induced in grem1^+/− and wild-type mice using STZ, with time zero defined as the day of initial STZ injection. Mice were maintained in the diabetic state for 18 or 27 weeks. Metabolic parameters, body weight, renal weight, and serum glucose levels are presented in Table 1. Nondiabetic mice gained significant weight compared with diabetic mice, as assessed at study end (wild-type control mice body weight 32.0 ± 4.47 g; wild-type diabetic mice body weight 22.82 ± 3.22 g). No significant difference in total body weight between wild-type and grem1^+/− mice in either the control or diabetic cohorts was detected (Table 1). Both diabetic groups developed marked fasting hyperglycemia that peaked at ∼33 mmol/l and was maintained up to the 27-week time point (Fig. 1,B). The onset and severity of hyperglycemia were similar between wild-type and grem1^+/− mice (Fig. 1,B). Marked and significant increases in A1C levels in diabetic mice were also detected at 18 and 27 weeks of hyperglycemia (Fig. 1 C), in both wild-type and grem1^+/− groups. Thus, the onset, severity, and progression of diabetes were similar in both wild-type and grem1^+/− mice.

We assessed the effects of type 1 diabetes on grem1 expression in kidneys from our experimental animals. Minimal grem1 mRNA was detected in wild-type and grem1^+/− control kidneys at both time points examined but the amount was dramatically increased in wild-type diabetic mice (Fig. 2,A). Approximately 62-fold increased grem1 upregulation was detected at 18 weeks compared with 42-fold at 27 weeks in wild-type diabetic mice (Fig. 2,B). In contrast, grem1^+/− mice manifested a 53-fold increase at 18 weeks and an 8-fold increase at 27 weeks in diabetic kidney (Fig. 2,B). The fold change in grem1 mRNA upregulation at 27 weeks in diabetic grem1^+/− mice was significantly lower than that seen in diabetic wild-type mice at the same time point (Fig. 2,B). Increased grem1 protein was also detected in diabetic wild-type kidney sections compared with that in controls via Western blotting and immunohistochemistry (Fig. 2,C and D and supplementary Fig. 1, available in an online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db08-1365/DC1). In contrast, the increase in grem1 protein was blunted in diabetic grem1^+/− mice (Fig. 2 C and D). These data suggest that deletion of one copy of the grem1 gene dramatically reduced grem1 induction in the diabetic mouse kidney.

Diabetic mice developed marked polyuria compared with that in age-matched controls, with no significant difference between wild-type and grem1^+/− cohorts observed at 27 weeks (wild-type control urine volume 3.35 ± 2.79 ml, wild-type diabetic urine volume 15.18 ± 9.35 ml, P < 0.01; grem1^+/− control urine volume 4.32 ± 3.15 ml, grem1^+/− diabetic urine volume 12.31 ± 10.6, P < 0.05) (Table 1). A significant increase in the ratio of left kidney weight–to–total body weight occurred in all diabetic animals compared with that in age-matched control animals after 27 weeks of diabetes, suggestive of renal hypertrophy (wild-type control vs. diabetic 5.409 ± 0.19 vs. 9.077 ± 0.38, P < 0.001; grem1^+/− control vs. diabetic 5.536 ± 0.16 vs. 8.816 ± 0.28, P < 0.001) (Table 1). The fold change in left kidney weight–to–total body weight was not significantly different between wild-type and grem1^+/− mice (data not shown).

All diabetic mice developed a significant increase in GBM thickness compared with that in nondiabetic age-matched controls at 27 weeks of hyperglycemia (Fig. 3,A and B). Interestingly, this increase was significantly greater in diabetic wild-type compared with that in diabetic grem1^+/− mice. GBM thickness increased 47% from baseline mean in age-matched control versus diabetic wild-type mice (Fig. 3,B, □). In contrast, a more modest 14% increase in GBM thickening was detected in control versus diabetic grem1^+/− mice at 27 weeks (Fig. 3,B, ■). The fold change in GBM thickening was significantly lower in grem1^+/− mice compared with that in wild-type mice (Fig. 3 C), suggesting that allelic depletion of grem1 prevented diabetes-induced early structural changes in the kidney.

Moderate increases in glomerular matrix secretion and interstitial collagen deposition were observed in diabetic wild-type and grem1^+/− mice compared with those in controls (Fig. 4,A and C). No significant tubulointerstitial fibrosis was detected in either genotype up to 27 weeks of diabetes (data not shown). Scoring of these sections revealed that the degree of increased staining of these markers of renal damage was not significantly different between diabetic wild-type and grem1^+/− mice (Fig. 4 B and D). These data suggest that our model of type 1 diabetes in mice on a C57BL/6J background manifest early diabetic nephropathy–like changes in kidney but do not develop more advanced renal disease.

Urinary microalbumin was increased in diabetic wild-type mice compared with that in controls (control 34.02 ± 4.53 μg, diabetic 27 weeks 129.32 ± 16.05 μg, P < 0.001) (Fig. 5,A). In contrast, microalbuminuria was less severe in diabetic grem1^+/− mice (control 33.09 ± 3.00 μg, diabetic 27 weeks 71.47 ± 17.54 μg, P < 0.05) (Fig. 5,A). Fold change in microalbuminuria was significantly lower in grem1^+/− diabetic mice compared with that in wild-type mice at 27 weeks of diabetes (Fig. 5,B). These data suggest that depletion of grem1 expression reduced the microalbuminuria associated with early renal damage in diabetic nephropathy. Serum creatinine levels decreased significantly in wild-type diabetic mice compared with those in age-matched controls (Table 1). This decrease in serum creatinine did not occur in grem1^+/− mice at either 18 or 27 weeks of diabetes (Table 1). Albumin-to-creatinine ratios (ACRs) increased in diabetic wild-type mice at both 18 and 27 weeks of hyperglycemia (control 63.51 ± 15.25 μg/mg, diabetic 27 weeks 594.97 ± 271.09 μg/mg) (Fig. 5,C). This increased ACR was greatly attenuated in diabetic grem1^+/− mice (control 104.13 ± 13.68 μg/mg, diabetic 27 weeks 287.10 ± 68.75 μg/mg) (Fig. 5,C). The smaller increase in grem1^+/− mice was highlighted when fold change was calculated, showing a significantly lower increase in the ACR in grem1^+/− mice compared with that in wild-type mice at 27 weeks (Fig. 5 D).

Creatinine clearance increased in diabetic wild-type mice compared with that in controls at 18 weeks of diabetes, suggesting that glomerular hyperfiltration was occurring at this time point (Fig. 5,E). In contrast, diabetic grem1^+/− mice did not develop significant increases in creatinine clearance until the 27-week time point (Fig. 5,E). A significantly higher fold change in creatinine clearance was detected in grem1^+/− mice compared with that in wild-type mice at 27 weeks, suggesting that grem1 depletion delayed the onset of peak hyperfiltration in diabetic mice (Fig. 5 F).

Serum lipids were elevated in diabetic wild-type mice, showing a significant increase in triglycerides and decrease in HDL typical of diabetic hyperlipidemia (Table 1). Diabetic grem1^+/− mice also manifested a significant elevation in serum triglycerides but demonstrated a significantly lower HDL at baseline than wild-type controls, in which HDL failed to drop significantly in the setting of diabetes (wild-type control 77 ± 3.69 g/dl, grem1^+/− control 58.3 ± 2.9 g/dl) (Table 1).

A significant correlation was detected between ACR and GBM thickening in both control and diabetic mice in our study (Fig. 6,A). To assess whether grem1 mRNA correlated with indexes of early diabetic kidney disease, grem1 levels were compared with changes in microalbuminuria and GBM thickening. The degree of grem1 expression correlated significantly with both ACR and GBM thickening in wild-type control and diabetic mice across the entire experiment (Fig. 6 B and C). These data suggested that increased grem1 gene expression occurred in parallel with cellular damage during early diabetic kidney disease.

To explore the mechanism of partial protection from diabetes-induced damage in grem1^+/− mice, we measured levels of genes implicated in kidney damage during diabetic nephropathy. Increased levels of fibronectin, vimentin, and CTGF were detected in diabetic wild-type kidney at 27 weeks (Fig. 7). In contrast, no significant increase was detected in grem1^+/− diabetic kidney, suggesting that decreased grem1 levels reduce diabetes-mediated upregulation of genes involved in mediating glomerular and renal damage. Because grem1 is a negative regulator of BMP-7, a molecule that has been shown to mediate repair processes in the damaged kidney, we examined the effect of grem1 deletion on this protein. No significant changes in BMP-7 mRNA or protein levels were detected at either 18 or 27 weeks of diabetes in either wild-type or grem1^+/− mice (supplementary Fig. 2, available in an online appendix, and data not shown). Similar to previous reports (24), levels of phospho-Smad1/5/8, a downstream target of BMP-7 signaling, were reduced in wild-type diabetic kidney at 27 weeks (Fig. 8 C and D). In contrast, no significant decrease in pSmad1/5/8 levels was detected in grem1^+/− mice, possible suggesting that BMP signaling is maintained in the diabetic kidney when levels of grem1 are reduced.

The BMP antagonist, grem1, regulates critical processes controlling limb-bud outgrowth and kidney development (6,–8,10,11). Increased grem1 levels correlated with the pathogenesis of diabetic nephropathy and/or progressive renal fibrosis (19,20). To address whether grem1 depletion would protect against diabetes-induced renal disease, we evaluated renal damage in type 1 diabetic mice lacking one copy of the grem1 gene (grem1^+/−). We provide the first evidence that allelic depletion of grem1 causes a reduction in grem1 expression levels and protects against early diabetic nephropathy–like changes in the kidney.

grem1^+/− mice on a C57BL/6J background were used in this study. Previous data compared the susceptibility of different genetic strains of mice and showed that C57BL/6J mice were “high responders” in terms of STZ-induced hyperglycemia (25). Consistent with these findings, severe and sustained hyperglycemia developed in both wild-type and grem1^+/− mice (Fig. 1,B). We demonstrate here that type 1 diabetes induced grem1 kidney mRNA and protein in wild-type mice. This reactivation of grem1 expression may be linked to a tissue repair mechanism in response to hyperglycemia and other insults that become maladaptive, leading to further kidney injury (23,26). Factors such as TGF-β, AGEs, and hemodynamic stress are features of diabetic nephropathy that have previously been shown to increase grem1 expression (17,20). Because grem1^+/− mice displayed a marked reduction in grem1 upregulation in diabetes (Fig. 2), we propose that the recapitulation of grem1 gene activation in the diabetic kidney involves both grem1 alleles.

A similar degree of renal hypertrophy was observed in both wild-type and grem1^+/− diabetic mice (Table 1). These data suggested that reduction of grem1 expression prevented diabetes-induced increases in GBM thickness during early-stage diabetic nephropathy. Staining for markers of more advanced kidney damage such as mesangial matrix secretion and interstitial collagen showed modest increases in diabetic mice (Fig. 3,D–F). Previous data had indicated that C57BL/6J mice were somewhat resistant to diabetic nephropathy–like changes in kidney, manifesting modest increases in albuminuria and glomerular damage (25). Our study also detected modest increases in glomerular damage (Fig. 3), together with a significant increase in microalbuminuria (∼4.8-fold) (Fig. 5,A and B). Mice in our study developed severe hyperglycemia (∼590 mg/dl) at 18 weeks (Fig. 1,B). In contrast, C57BL/6 mice in the study of Gurley et al. (25) manifested lower levels of hyperglycemia at 16 weeks (388 mg/dl). Thus, the more robust hyperglycemia in our model may have been sufficient to induce microalbuminuria but not gross structural changes in glomeruli or kidney tubules in our mice. We conclude, based on the early structural changes evident, that our diabetic mice develop mild but significant damage in the kidney and that this early damage is attenuated when levels of grem1 are reduced by gene deletion. Microalbuminuria developed in wild-type diabetic mice compared with control mice (Fig. 5) but was attenuated in age-matched grem1^+/− mice (Fig. 5,A and B). Diabetes-induced increases in the ACR were also reduced in grem1^+/− mice compared with wild-type mice (Fig. 5,C). Values for creatinine clearance suggest that diabetes-induced hyperfiltration peaked in wild-type mice at 18 weeks (Fig. 5,E). In contrast, creatinine clearance values for grem1^+/− diabetic mice were lower at 18 weeks compared with those in wild-type mice and were higher at 27 weeks, suggesting that the peak of hyperfiltration in diabetic grem1^+/− mice may occur at a later time than that in wild-type mice (Fig. 5 E). Together with previous data identifying grem1 upregulation in high-glucose cell culture models and human diabetic nephropathy biopsy specimens, these data add to the growing body of evidence implicating grem1 in the pathogenesis of diabetic nephropathy. Furthermore, our data provide the first evidence that a reduction in grem1 expression reduces early impairment of renal structure and function in the diabetic kidney. We are currently examining whether other vascular complications of diabetes are also altered in grem1^+/− mice. Preliminary evidence from our group suggests that grem1 deletion may also protect against aortic thickening in the diabetic state, potentially due to altered serum lipid profiles (27).

Several markers of renal damage in diabetes have been identified, and levels of several of these genes were elevated in the wild-type but not grem1^+/− diabetic kidney (Fig. 7). Thus, in the hyperglycemic state, the triggers that increase the expression of genes contributing to glomerulosclerosis and tubular damage seem to be attenuated in the grem1^+/− kidney. grem1 is an antagonist of BMP-2, -4, and -7, binding to these proteins and preventing their interaction with their cognate receptors (6,8,14). Of these, BMP-7 has attracted recent attention, as up to 90% loss of this protein has been reported in the kidneys of diabetic rats (28). Loss of BMP-7 was accompanied by an increase in grem1 expression, with both changes probably being driven by increased TGF-β1 in the diabetic kidney (28). Administration of recombinant BMP-7 or transgenic overexpression of BMP-7 in podocytes ameliorates renal injury in mouse models of diabetes (29) and lupus (30). Although significant changes in cellular BMP-7 mRNA or protein were not detected in our diabetic mice (supplementary Fig. 2), our data show that levels of phospho-Smad1/5/8, a downstream target of BMP signaling, were reduced in wild-type but not in grem1^+/− diabetic kidney compared with control kidneys (Fig. 8), suggesting that loss of grem1 expression facilitates sustained BMP-7 signaling in the diabetic kidney, extending the notion of a carefully coordinated balance between BMP signaling and grem1 in the disease state.

Previous reports have identified other genes such as TGF-β type II receptor, CTGF, and p27^Kip1 whose deletion confers protection against the sequelae of diabetic kidney disease (31,24,32). p27^Kip1+/− mice displayed an intermediate degree of protection compared with that of p27^Kip1−/− mice, suggesting that p27^Kip1 is haploinsufficient in terms of its role in diabetic nephropathy (32). Our data suggest that grem1 may also be haploinsufficient, as deletion of one allele of grem1 reduced grem1 expression and conferred a moderate degree of protection from structural renal damage induced by diabetes.

Because grem1^+/− mice developed less severe GBM thickening (Fig. 3), together with lower fold increases in ACR and estimated glomerular filtration rate (Fig. 5,C–F) compared with those in wild-type controls, a reduction in grem1 levels using gene deletion may reduce both the onset and severity of renal disease in the diabetic kidney. Increased grem1 levels correlated tightly with both GBM thickening and ACR (Fig. 6), suggesting that reactivation of grem1 in the diabetic kidney may occur in parallel with early pathological changes in renal structure and function. Because grem1^+/− mice manifest less severe microalbuminuria and GBM thickening, these data suggest that grem1 upregulation contributes to glomerular damage in response to diabetes in the kidney.

This work was supported by a Health Research Board of Ireland Clinical Fellowship (CRT/2004/41) to S.A.R. Work in the laboratory of D.P.B. is supported by Science Foundation Ireland and the Health Research Board Ireland. C.G. is supported by Science Foundation Ireland Award 06/IN.1/B114.

No potential conflicts of interest relevant to this article were reported.

We thank Richard Harland for the generous gift of grem1^+/− knockout mice. Alison Murphy, Laura Connole, George Keating, Joe Mooney, Alfie Redmond (University College Dublin), and Brendan Tobin (Mater Hospital) are gratefully acknowledged for excellent technical assistance. We thank Dionne van der Giezen (Utrecht) for histology analysis and quantitation. We acknowledge Jay Jerome and the excellent collaboration with the Vanderbilt University Medical Center (VUMC) Research Electron Microscopy Core (sponsored by National Institutes of Health Grants DK20539, CA68485, and DK58404) as well as the VUMC Mouse Metabolic Phenotyping Core (Grant DK59637).

Parts of this study were presented as a poster at the annual meeting of the American Society of Nephrology, San Francisco, California, 31 October–5 November 2007.