Urinary IgG4 and Smad1 Are Specific Biomarkers for Renal Structural and Functional Changes in Early Stages of Diabetic Nephropathy

Diabetic nephropathy (DN) is the major global cause of chronic kidney disease (CKD) and end-stage renal disease (ESRD). The natural history of DN is characterized by lesion development and progression during a long period of clinical silence, and lesions may be far advanced before increased albumin excretion rate (AER) and/or reduced glomerular filtration rate (GFR) are detectable (1). Once overt nephropathy manifests, it may be difficult to prevent progression to ESRD. Glomerular structural changes including glomerular basement membrane (GBM) thickening, expansion of the mesangium, and a consequent reduction in the filtration surface correlate with increasing AER and decreasing GFR (1). Of these, the fraction of the volume of the glomerulus occupied by the mesangium is the single strongest correlate of functional changes (2). In normoalbuminuric patients with type 1 diabetes (T1D), however, increased GBM width was the best early structural predictor of progression of overt nephropathy and ESRD (3). Although a persistent increase in levels of AER in the range representing microalbuminuria (MA) is currently the best early biomarker of DN risk, histological lesions in T1D and type 2 diabetes (T2D) may be quite advanced before MA develops (4,5), and GFR decline may antedate albuminuria. Moreover, only ∼35–45% of patients with MA and T1D or T2D will progress to overt nephropathy over the subsequent 6–10 years, whereas approximately one-third will spontaneously revert to normoalbuminuria (6). It is clear that measures of AER need to be supplemented with additional biomarkers in order to provide an earlier and more precise reflection of underlying DN lesions during the clinically silent period and to better predict the risk of DN progression.

We established that Suppressor of Mothers against Decapentaplegic Transcription Factor (Smad) 1, one of the Smad proteins functioning downstream of the transforming growth factor-β superfamily receptor kinases, is an important transcription factor for the α1 and 2 chains of type IV collagen, which are major components of mesangial matrix expansion observed during DN (7). Smad1 also strongly correlated with the development of mesangial expansion but not with albuminuria in streptozotocin-induced diabetic rats (8). Moreover, early increases in urinary Smad1 level predicted the later development of mesangial expansion in both streptozotocin-induced DN rats and db/db mice (9). In vivo evidence of Smad1 signaling was also linked to the development of DN in animal models (9–13).

The loss of charge selectivity in the glomerulus preceded and accompanied the increase in large pore area seen in patients with T1D with urinary albumin excretion >100 mg/24 h (14). IgG4 has a negative charge and, along with other anionic plasma proteins, was observed to selectively bind in a linear pattern to the GBM and to tubular basement membranes in patients with early and late-stage DN (15). We speculated that the increased urinary excretion of IgG4 in patients with diabetes may be related to biochemical, functional, and morphological GBM changes. We hypothesized that urinary IgG4 is a useful biomarker for predicting underlying GBM lesions, and thus we developed a highly sensitive and specific method for detecting urinary IgG4 for this study.

Here we report that in patients with T2D both urinary IgG4 and urinary Smad1 are novel reliable biomarkers for each of the classic pathological findings in the early stages of DN and are predictors of progressive GFR decline.

After providing appropriate informed written consent, 871 patients with T2D were recruited from Ikeda Hospital (Amagasaki, Hyogo, Japan). The study protocol was in accordance with the Declaration of Helsinki and was approved by the ethics committee at Ikeda Hospital. Patients were excluded if they had gastrectomy, severe hepatitis (aspartate aminotransferase or alanine aminotransferase >100 IU/L, or icterus), or nondiabetic renal disease (severe dysmorphic hematuria or any other renal diseases suggested by laboratory data and clinical symptoms). All patients underwent fundoscopy through dilated pupils to determine the presence or absence of DN. A total of 815 patients were enrolled for the follow-up study and 261 were excluded because of either macroalbuminuria or estimated GFR (eGFR) <60 mL/min at baseline. Of these patients, 554 were followed for 5 years with laboratory and clinical data (Fig. 1).

As shown in Supplementary Table 1, 17 normotensive patients (blood pressure <130/85 mmHg; 3 women, 14 men; mean age 49.5 ± 11.8 years; mean duration of T2D 13.3 ± 7.8 years) without macroalbuminuria or hematuria and with serum creatinine (Cr) <1.2 mg/dL volunteered for research kidney biopsies at Kitasato University. None had evidence of nondiabetic changes. The protocol for this study was approved by the Research Ethics Committee of Kitasato University School of Medicine, and all patients provided written informed consent (16,17). The normal renal structural reference was obtained from light microscopy of normal-appearing renal tissue from nephrectomy specimens from 10 patients with renal cell carcinoma (three women, seven men; mean age 66.8 ± 13.1 years). None of these control subjects had diabetes based on an oral glucose tolerance test, macroalbuminuria based on a dipstick test, decreased eGFR (all were >90 mL/min), or hypertension. Nine 1-h renal biopsy specimens obtained after renal transplantation were used as controls in electron microscopy analyses. The donor information is shown in Supplementary Fig. 1 and Supplementary Table 2. The protocol for this component of the study was approved by the Research Ethnics Committee of Tokushima University.

Urine samples were centrifuged for 15 min. Supernatants stored at −80°C were rapidly thawed and centrifuged to remove any urates or phosphates before use in the assays. An enzymatic method was performed to determine Cr in serum and urine (Shino-Test, Kanagawa, Japan).

eGFR/1.73 m² was estimated using the following equation applied for Japanese population (18), and the value was adjusted by body surface area at each point in order to estimate actual individual change:

Urinary albumin was measured through the use of a turbidimetric immunoassay. Macroalbuminuria was defined as an albumin-to-Cr ratio (ACR) >300 mg/g urine Cr, and MA was defined as 30–300 mg/g Cr in two consecutive measurements.

Renal biopsy tissues obtained by Kitasato University and Tokushima University were fixed in 10% buffered formalin and stained with periodic acid Schiff, periodic acid silver methenamine, and Masson trichrome. Mesangial area was estimated by measuring the mesangial periodic acid silver methenamine–positive area using an image analyzer with a light microscope (Image Processor of Analytical Pathology; Sumitomo Chemical Co., Tokyo, Japan) (11,12). The mesangial area fraction was expressed as the percentage of total glomerular area occupied by mesangial area. In addition, the percentage of global glomerular sclerosis was measured as described previously (19). Interstitial volume fraction (Vv [interstitium/cortex]) was determined on the light microscopy sections (approximate magnification ×300) by point-counting images projected onto a white surface by a projection microscope (20). The glomerular surface was outlined by tracing the outer margins of the tufts. The mean glomerular surface area was calculated (11,12).

Tissues for electron microscopy were fixed in 2.5% glutaraldehyde, postfixed in osmium tetroxide, dehydrated, and embedded. Ultrathin sections were examined using a JEOL CX 100 transmission electron microscope (JEOL, Tokyo, Japan) in the Kitasato Bio-Imaging Center. At least two, but usually three, glomeruli were examined per patient. GBM width and surface density of peripheral GBM (Sv [peripheral GBM(PGBM)/glomerulus(Glom)]) were measured using an orthogonal intercept method and a line intercept method, respectively, as previously detailed (2,17,21–23). One patient lacked sufficient biopsy material for analysis.

Recombinant human Smad1 (rhSmad1) was purified using a glutathione S-transferase system: 6-week-old mice were immunized with rhSmad1, and monoclonal antibodies were generated using established procedures. Approximately 3,000 hybridoma clones were screened, then characterized by Western blotting and ELISA. SDS-PAGE analysis revealed that the purified protein corresponds with the calculated molecular weight of Smad1. A specific hybridoma clone (Sp125) was then selected.

The F(ab′)₂ fraction of the Sp125 was coated onto microtiter plates (Nunc). The urine samples or standard (rhSmad1; 0.2–5 ng/mL) were applied, followed by goat anti-Smad1 polyclonal antibody (AF2039; R&D Systems), and then reacted with horseradish peroxidase–conjugated antigoat IgG antibody (605–4302; Rockland Immunochemicals Inc.). A 3,3′,5,5′ tetramethyl benzidine (TMB) peroxidase substrate solution system was introduced and absorbance was measured at 450 nm. Urinary Smad1 ELISA was done in duplicate, and the mean of three measurements was applied.

Adding two different concentrations of rhSmad1 to the urine samples (estimated values of 0.5 and 2 ng/mL) revealed recovery rates of 111% and 112%, respectively, indicating specificity for the assay system. The minimum detectable sensitivity, estimated using the three SDs method, was 0.2 ng/mL. The coefficient of variation of the ELISA system ranged from 3.3% to 6.0% in 10 independent assays, thus indicating simultaneous reproducibility. In addition, after mixing rhSmad1 into the urine samples with different pH values (5.0–8.0), recovery rates were obtained from 102.2% to 108.2%. The final urinary Smad1 concentration was normalized to the urine Cr concentration. Five positive urine samples were confirmed by Western blotting with a specific antibody for Smad1 (Sp125), which showed the corresponding single band of Smad1. The band disappeared when the blot was preincubated with excess rhSmad1. In addition, the presence of Smad1 in urine samples was confirmed with mass spectrometry.

The mouse monoclonal antihuman IgG heavy chain (Ab9243; Abcam) was coated on microtiter plates (Nunc). The urine samples or standard (full-length human IgG4 protein [ab158747; Abcam]; 1.0–2,000 ng/mL) was applied, followed by horseradish peroxidase–labeled mouse monoclonal antihuman IgG4 (Fc) (ab99823; Abcam). A TMB peroxidase substrate solution system was introduced and absorbance was measured at 450 nm. Urinary IgG4 ELISA was done in duplicate and the mean of three measurements was applied.

The urinary IgG4 detection system did not cross-react with human IgG1 (ab15874; Abcam), IgG2 (ab90284; Abcam) or IgG3 (ab158746; Abcam). Adding two different concentrations of human IgG4 (1,000 and 20 ng/mL) to the urine samples provided recovery rates of 104% and 118%. The minimum detectable sensitivity was 1.0 ng/mL and the coefficient of variation of the ELISA system ranged from 1.0% to 2.1%. The final urinary IgG4 concentration was normalized to the urine Cr concentration.

The data were expressed as mean ± SD. Trends were analyzed by the Jonckheere-Terpstra test based on a priori ordering of quantile analysis. The values were analyzed by t test, Mann-Whitney U test, one-way ANOVA, Kruskal-Wallis test, or χ² test in intergroups. The correlation of two variables including pathological parameters and biomarkers were compared using the Pearson product-moment correlation coefficient or Spearman rank correlation coefficient as appropriate on the basis of data distribution. The rates of eGFR decline compared with each baseline value based on four groups (reference, urinary Smad1 <0.15 μg/g Cr and urinary IgG4 <39.2 μg/g Cr; only high urinary Smad1 group, >15 μg/g Cr [U-Smad1]; only high urinary IgG4 group, >39.2 μg/g Cr [U-IgG4]; both high U-Smad1 and U-IgG4 group) were analyzed with the Wilcoxon signed rank test in intragroups and the Steel multiple comparison test in intergroups in the follow-up study. The odds ratio (OR) for a 10% decline of eGFR at the 5-year follow-up was calculated by multivariate logistic regression for patients fulfilling the baseline criteria (both eGFR >60 mL/min and no macroalbuminuria). The following covariates were used as putative risk factors: diabetes duration (<5 vs. ≥5 and <10 years, <5 vs. ≥10 years), retinopathy status, mean blood pressure, HbA_1c, U-Smad1/Cr, U-IgG4/Cr, and ACR (normoalbuminuria vs. MA). In the regression model, U-Smad1/Cr, U-IgG4/Cr, and ACR were forcibly included. For other factors, variables were selected through the use of a stepwise method, with selection criteria (P < 0.05). Two-sided P < 0.05 was considered to be statistically significant. GraphPad Prism 5 software (for Mac OS X) and SAS version 9.4 (SAS Institute) were used for analyses.

Higher baseline ACR was observed in subjects with higher baseline U-IgG4 using quantile analysis of U-IgG4 in the patients with T2D (P < 0.0001) (Supplementary Fig. 2A). A statistically significant association existed between increased baseline urinary albumin and higher baseline U-Smad1 (P < 0.0001) (Supplementary Fig. 2B).

Urinary IgG4 values were not uniformly distributed in the study population. By analyzing the correlation between albuminuria and log urinary IgG4, we were able to use a hockey stick regression test to establish the cutoff value for urinary IgG4 in order to separate the two distributions observed. The threshold value was 1.6 and the cutoff value for urinary IgG4 was 39.2 μg/g Cr (Fig. 2A). To confirm the clinical significance of this cutoff value, we analyzed the parameters of DN. The group with urinary IgG4 >39.2 μg/g Cr showed markedly lower eGFR than the group with urinary IgG4 <39.2 μg/g Cr (Fig. 2B). HbA_1c values did not differ between the two groups (Fig. 2C). The higher group of urinary IgG4 showed statistically significant differences in diabetes duration, presence of retinopathy, and U-Smad1 compared with the lower group, but no difference was found in BMI and mean blood pressure between the two groups (Supplementary Table 3).

Urinary IgG4 levels significantly and positively correlated with Sv (PGBM/Glom) (Fig. 3A), and urinary Smad1 levels significantly and positively correlated with mesangial fractional area (Fig. 3B). These urinary values did not correlate with the other pathological parameters (Supplementary Table 4). Urinary albumin levels in this small cohort correlated with no pathological parameters including mesangial matrix fraction, GBM thickness, Sv (PGBM/Glom), global sclerosis, interstitial fibrosis, and glomerular surface area.

The mesangial area fraction of the control kidney specimens was 5.88 ± 1.30%. Mesangial area fractions >8.5% (normal mean + 2 SDs) were considered to represent increased mesangial matrix. The cutoff for urinary Smad1 was defined as this mesangial area fraction. Based on this, 0.15 μg urinary Smad1/g Cr was used as the cutoff value (Fig. 3B).

We compared the follow-up data from four different groups according to each cutoff value as mentioned above: reference, U-Smad1 (only higher values for urinary Smad1), U-IgG4 (only higher values for urinary IgG4), and both (higher values for both parameters). Table 1 summarizes baseline data used in the 5-year follow-up study. Significant group differences were observed for age, diabetes duration, urinary albumin excretion, and HbA_1c, but not for eGFR or the use of medications including antihypertensive drugs, RAS inhibitors, statins, and uric acid–lowering drugs. No differences were found in the presence of retinopathy, BMI, and mean blood pressure between these groups. Baseline data from the patients who dropped out of the follow-up study showed no significant difference in albuminuria between these groups.

Multivariate logistic analysis was performed to determine a 10% decline in eGFR at the 5-year follow-up from the baseline value in the patients fulfilling both criteria: eGFR >60 mL/min and no macroalbuminuria. Duration of diabetes, presence of retinopathy, high blood pressure, and HbA_1c were not statistically significant predictors of eGFR loss. As a result of the selection of variables, none of these factors were incorporated into the model. U-Smad1, higher values for both U-Smad1 and U-IgG4), and MA were significant predictors of eGFR loss (OR 2.052 [P = 0.0034], 1.853 [P = 0.0324], and 2.331 [P = 0.0121], respectively) (Supplementary Fig. 3). Quantile analyses of urinary IgG4 and urinary Smad1 indicated that patients with higher U-Smad1 had a larger eGFR decline at the 5-year follow-up (Supplementary Fig. 4A and B). On the other hand, urinary IgG4 and urinary Smad1 were not related to albuminuria progression (Supplementary Fig. 4C and D).

Because urinary IgG4 and Smad1 correlate with specific renal structural parameters in the early stages of T2D, as mentioned above, we analyzed the patients without macroalbuminuria at baseline. The U-Smad1 group with eGFR >60 mL/min and no macroalbuminuria showed statistically significantly decreased eGFR at the 5-year follow-up. Moreover, the subset with elevations in both these analytes also showed significantly greater eGFR decline (Fig. 4A). Patients with MA showed a significantly higher decline in eGFR at the 5-year follow-up when compared with patients with normoalbuminuria in the both-positive group without macroalbuminuria (Fig. 4B). These findings were not observed in the other groups (reference, U-Smad1, and U-IgG4). The both-positive group and U-IgG4 group without macroalbuminuria showed increased urinary albumin at the 5-year follow-up compared with that at baseline (Fig. 4C).

Urinary albumin is broadly used and is a significant, albeit imprecise, predictor of DN risk (24). No specific markers reflect renal pathological changes at early stages of DN (6). This study demonstrates—to our knowledge for the first time—that both urinary IgG4 and urinary Smad1 reflect specific pathological findings at the early stages of DN in T2D, although albuminuria is not precisely correlated with renal structure at the early stages (6). In addition, in patients with eGFR >60 mL/min without macroalbuminuria at baseline, elevated urinary IgG4 and urinary Smad1 levels were associated with a greater risk for an increase of albuminuria and decline of eGFR over 5 years. Multivariate analysis and quantile analysis showed that urinary Smad1 was a risk factor for eGFR decline in the follow-up study, which corresponds to the previous study indicating a correlation between mesangial expansion and renal function (2). Therefore, both urinary IgG4 and urinary Smad1 predict the underlying renal biopsy findings and may be useful novel early prognostic markers for DN risk.

Diabetes is the leading cause of CKD in the developed world, occurring in ∼20–30% of patients with diabetes. DN has become more prevalent in spite of the increasing use of glucose-lowering medications and renin-angiotensin inhibitors (25). One possible explanation is the difficulty of correctly diagnosing the earlier stages of DN, as serious DN lesions may develop before clinical indicators such as albuminuria manifest (17,22). In fact, many patients with longstanding diabetes with normoalbuminuria show histological changes (5,6), and the degree of both mesangial expansion and GBM thickening did not differ between normoalbuminuric and microalbuminuric Japanese patients with T2D (21). Therefore, we urgently need to find biomarkers that can indicate renal histological changes at earlier stages of DN than are reflected by abnormalities in urinary albumin (26).

A heterogeneous distribution exists between the urinary IgG4–positive group and the urinary Smad1–positive group in T2D; however, the severity of GBM thickening and mesangial expansion also substantially overlap at the different stages (6). In addition, one report showed that GBM thickening was a risk factor for the development of macroalbuminuria or ESRD in T1D (3). GBM thickening and mesangial expansion predicted increased albuminuria after 6 years of follow-up of T2D (27), although renal structural-functional relationships were not present at the early stage of DN in T2D (17). The association of both positive urinary IgG4 and positive urinary Smad1 with decreased GFR may reflect the association of macroalbuminuria with worse lesions. In a similar way, the findings of both positive urinary IgG4 and positive urinary Smad1 in relation to the duration of diabetes and/or retinopathy may also reflect their connection to overt nephropathy.

It is generally accepted that about 33% to 45% of patients with MA ultimately develop overt nephropathy and that some who are normoalbuminuric despite a long diabetes duration will also eventually develop overt nephropathy in T2D (26). During 7.8 years of follow-up for MA in T2D, changes in urinary albumin are closely related to changes in GFR (28), which seem to be associated with renal structural changes (29). These findings suggest that urinary albumin per se could promote DN progression through the histological changes of DN. This study argues that patients with high urinary IgG4 and urinary Smad1 without macroalbuminuria may have progressive GFR loss, which is associated with increased urinary IgG4 and urinary Smad1 and renal structural changes, especially in patients with MA in whom both were increased. The results are also supported by a previous study showing that structural change is an indicator for later development of overt nephropathy in normo- and microalbuminuric patients with T2D (27). On the other hand, the reference group (negative for both urinary Smad1 and urinary IgG4) had an eGFR decline comparable to that of the general population and showed no renal structural changes. In this group, MA was not a predictor of eGFR decline.

The strengths of this study are that these biomarkers, including urinary IgG4 and urinary Smad1, reflect diabetic renal structural changes and predict later progression of GFR loss. In addition, Smad1 is a critical molecule implicated in mesangial matrix expansion, and phosphorylated Smad1 expression is associated with renal structural changes with angiotensin receptor blockade therapy or antibone morphogenetic protein 4 antibody treatment in diabetic rats (9–11,13).

One of the limitations of our study is that insufficient data exist to precisely determine the cutoff values of these biomarkers. From the analysis of a small number of control specimens from 1-h biopsy, Sv (PGBM/Glom) showed two different peaks of distribution (shown in Supplementary Fig. 1). The value of 39.2 μg/g Cr after analysis with hockey stick regression corresponded to Sv of 0.1243, indicating the approximately median value of Sv. However, both the cutoff value of urinary IgG4 after analysis with hockey stick regression and that of urinary Smad1 determined by their association with mesangial matrix expansion indicate their significance in DN. The exact mechanism of urinary Smad1 excretion is not completely clear, but preliminary study shows that the exosome fractions from urine samples from patients with DN contain significant amounts of Smad1. Another limitation is that we analyzed renal biopsies from only a small number of patients. In Japan, renal biopsies can almost never be performed at the early stage of DN. In this study, however, we provide evidence of early renal structural-functional relationships, which were not shown previously in patients with T2D. Further follow-up studies that examine a larger cohort over a longer period of time are warranted.

In summary, we provide evidence that Smad1 and IgG4 are molecules that can reflect and are relevant in DN development and progression (7–13) (Fig. 5). This study may help to translate previous basic research discoveries into clinical practice. Further studies to investigate the time course change of urinary IgG4 and urinary Smad1 in relation to DN structural and functional outcomes and validation in other cohorts are needed.

Funding. This study was supported by Grants-in-Aid from the Japan Science and Technology Agency (AS2115050G).

Duality of Interest. M.M. is involved in investigator-initiated research funded by Boehringer Ingelheim. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.D. conceived and designed this study and wrote the manuscript. T.Mo., and M.M. conceived and designed the study. T.Mo. and M.O. analyzed renal biopsies. N.M. and G.I. performed statistical analyses. T.D., Y.F., T.S., H.A., S.K., T.Mu., K.N., and T.T. performed immunological and urinary analyses. N.M., M.U., G.I., H.I., Y.K., N.T., and K.S. enrolled patients and acquired clinical data. K.Y. estimated the costs of the study. M.M. wrote and edited the manuscript. All authors contributed to the critical review and comments and approved the final draft. T.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.