Loss of Glomerular Permselectivity in Type 2 Diabetes Associates With Progression to Kidney Failure Free

Alterations in glomerular hemodynamic function are proposed to contribute to the onset and progression of kidney disease in people with type 1 and type 2 diabetes (T2D). The impact of T2D on glomerular hemodynamic function has been characterized extensively in an American Indian population from the southwestern U.S., which has a high incidence of kidney failure ascribed to diabetic kidney disease (DKD) (1). In this population, whole-kidney glomerular filtration rate (GFR), measured by the urinary clearance of iothalamate is, on average, 14% higher in individuals with impaired glucose tolerance than in those with normal glucose regulation, and GFR increases further with the onset of T2D, typically reaching a plateau in those with moderate albuminuria (urine albumin-to-creatinine ratio [ACR] 30–299 mg/g; previously referred to as microalbuminuria) (2,3). The GFR generally remains high, even with long-standing diabetes, but declines with the appearance of severe albuminuria (urine ACR ≥300 mg/g; previously referred to as macroalbuminuria) (2), with nearly one-half of patients developing severe albuminuria progressing to kidney failure within 10 years (4). Specific intraglomerular hemodynamic parameters, namely the balance in vascular resistance between the afferent (R_A) and efferent arterioles (R_E) and intraglomerular hydraulic pressure (P_GLO), also associate strongly with incident kidney failure in this population (5).

Functional changes in the size-selective properties of the glomerular permeability barrier in diabetes may also play a role in the initiation and progression of DKD. The fractional clearances of uncharged dextrans of broad size distribution are elevated in American Indians with T2D of <3 years duration compared with those who have normal glucose regulation. The greatest increases in fractional clearance are seen for the larger, more impermeant dextrans with molecular radii between 48 and 60 Å (6). Among those with diabetes, the sieving coefficients for larger dextrans increase further, relative to normal urine albumin excretion, in severe albuminuria, especially when urine ACR exceeds 3,000 mg/g, but not in moderate albuminuria, suggesting a significant size defect or shunt in the glomerular barrier that appears in parallel with major changes in hemodynamic function and glomerular structure in DKD (7,8).

In this study, we examined the contribution of early defects in the size-selective properties of the glomerular capillary to the risk of progression to kidney failure in an American Indian cohort with T2D. We used a hydrodynamic model of hindered solute transport to characterize the size-selective properties of the glomerular barrier. In this model, the glomerular capillary wall was represented as a heteroporous membrane perforated predominantly by small restrictive cylindrical pores with a log-normal distribution of pore radii and a parallel population of larger nondiscriminatory pores characterized by a parameter, ω₀, that estimates the fraction of total filtrate volume passing through this shunt (7,9). We hypothesized that a greater shunt magnitude associates with a higher risk of progressive DKD.

Between 1965 and 2007, American Indians in Arizona participated in a longitudinal study of the natural history of diabetes and its complications. Each member of this community who was at least 5 years old was invited to undergo a research examination approximately every 2 years. Diabetes was diagnosed by a 2-h postload plasma glucose concentration ≥200 mg/dL (11.1 mmol/L) at these biennial examinations or when the diagnosis was documented in the medical record. The date of diabetes diagnosis for each participant was ascertained from this longitudinal population study. In 1988, informative subsets of individuals from this population were selected to undergo more detailed longitudinal studies of kidney function to better understand the natural course of DKD and the factors associated with its development and progression (2,6,7,10,11). The current study included 185 adult participants with T2D who underwent these detailed studies of kidney function, which included measurements of GFR, effective renal plasma flow (RPF), and dextran sieving characteristics of the glomerular capillary. The first study at which all these measures were available was considered the baseline examination. These baseline examinations were conducted between September 1988 and February 1994.

Participants were followed until kidney failure, death, or 4 August 2020. Both health outcomes were ascertained independently of the research examinations. Kidney failure was defined as the initiation of kidney replacement therapy or death from DKD if the participant refused kidney replacement therapy. Two participants refused kidney replacement therapy and their deaths were classified as being due to kidney failure. The vital status of all study participants was confirmed through 4 August 2020. The study adhered to the Declaration of Helsinki and was approved by the institutional review board of the National Institute of Diabetes and Digestive and Kidney Diseases. Each participant provided written informed consent.

BMI was defined as weight divided by the square of height (kg/m²). Blood pressure was measured twice with the participant resting in the seated position, and the results were averaged. Mean arterial pressure (MAP) was calculated as (2 × diastolic blood pressure + systolic blood pressure)/3. HbA_1c was measured by high-performance liquid chromatography (HPLC). Urine albumin concentration was measured in untimed urine samples by nephelometric immunoassay and urine creatinine by a modified Jaffé reaction. Urine albumin concentrations below the detection limit of the assay (≤6.8 mg/L) were set to 6.8 mg/L in the analyses. Albumin excretion was assessed by the urine ACR.

GFR and RPF were measured by the urinary clearance of iothalamate and p-aminohippurate (PAH), respectively. Loading doses of each clearance marker based on the participant’s weight were given intravenously following an oral water load. Each clearance marker was then delivered continuously by an infusion pump. The infusion rates of iothalamate and PAH were calculated to maintain respective serum concentrations constant at 1.5 and 2.0 mg/dL (6). After an equilibration period, the bladder was emptied by voiding, and four timed urine collections were made at 20-min intervals. Blood was drawn to bracket each urine collection, and results from each of these clearance periods were averaged. An HPLC system with an ultraviolet light detector was used to measure the concentrations of iothalamate and PAH in urine and serum at 236 nm (#LC-6A; Shimadzu Instruments, Columbia, MD). RPF was computed by dividing the clearance of PAH by estimates of the extraction ratio of PAH in the kidney at different levels of GFR. Indirect intraglomerular hemodynamic parameters, including P_GLO and the balance in vascular resistance between R_A and R_E (R_A/R_E ratio), which are not directly measurable in humans, were estimated with the use of equations described previously (5).

To investigate underlying alterations in glomerular capillary wall function, dextrans measured in molecular size intervals of 2 Å between a 26- and 60-Å radius were infused continuously at half the rate of iothalamate following a loading dose of dextran at the initiation of the clearance study (6). The effective Stokes-Einstein radius of albumin is 35 Å. Dextran concentrations in urine and serum were measured using a refractive index detector (#RID-6A; Shimadzu Instruments) as described previously (7). Size separation of dextrans in ZnSO₄-precipitated urine and serum was achieved by HPLC using an automated injector (WISP 710; Waters Corporation, Milford, MA) and two size exclusion columns in series (Ultrahydrogel 500 and 250; Waters Corporation). The molecular radius (r_S) measured in Å of each dextran fraction was calculated from its molecular weight (MW) in daltons using the following equation: r_S = 0.33 × (MW)^0.463.

The fractional clearance of dextran (θ) for each size interval was calculated for the first timed collection period by dividing the urine-to-serum dextran concentration ratio by the urine-to-serum iothalamate concentration ratio. The interday coefficient of variation for urine-to-serum dextran concentration ratios for molecular size intervals between 32 and 60 Å was reported previously to vary between 7.7 and 12.4% (7).

We used a hydrodynamic model of hindered transport through water-filled cylindrical pores to describe the observed transport of dextran across the glomerular membrane (9). We selected a log-normal + shunt model to represent the glomerular capillary, as it best fits dextran sieving data for patients with and without proteinuria (12). The log-normal + shunt model contains three parameters: u, the mean pore radius; s, a measure of the breadth of the log-normal distribution (ln s is the SD of the distribution of ln r, where r is the pore radius); and ω₀, the shunt parameter (7,9,12). A schematic illustration of this model is provided in Fig. 1. We also examined the relationship between ω₀ and kidney failure using a simpler isoporous + shunt model that represents the glomerular capillary as an isoporous membrane perforated predominantly by small restrictive cylindrical pores of identical radii and a parallel population of larger nondiscriminatory pores characterized by the shunt parameter ω₀ (9). The computation of the hydrodynamic models was programmed in MATLAB (MathWorks, Natick, MA), and the programming code is available upon request from Dr. Layton ([email protected]).

Research kidney biopsies were performed in 51 of the study participants. The morphometric measurements are described in detail elsewhere (13). Measured structural parameters included glomerular sclerosis, glomerular volume, glomerular basement membrane (GBM) thickness, cortical interstitial fractional volume, mesangial fractional volume, peripheral capillary surface area, podocyte foot process width, podocyte number per glomerulus, podocyte density, area per podocyte, podocyte cell volume, nonepithelial cell number, and nonepithelial cell density, which includes mesangial, interstitial, and endothelial cells.

Patient characteristics are expressed as mean ± SD, median (25th–75th percentile [interquartile range (IQR)]) for skewed distributions, and frequencies and percentages for qualitative variables. Groups were compared using a parametric (ANOVA) test for ω₀. Positively skewed parameters were log-transformed as appropriate.

The Kaplan-Meier survival curve for the outcome of kidney failure was plotted by tertiles of ω₀ and compared using the log-rank test. The category boundaries for the tertiles of ω₀ were at 0.00102 between the low and middle tertiles and 0.00166 between the middle and high tertiles. The association of ω₀ with this health outcome was further examined by Cox proportional hazards regression with adjustments for baseline age, sex, MAP, HbA_1c, GFR, and ACR, since these represent traditional confounders. We limited the number of covariates to adequately fit to the number of kidney failure events. The hazard ratio (HR) and 95% CI was expressed per 1-SD increment of ω₀. The model was tested for log-linearity and proportionality assumptions.

Pearson correlations were used to assess the relationships between ω₀ and the morphometric variables derived from the kidney biopsies. Correlations were standardized (mean 0; SD 1) prior to analysis. Partial correlation analysis was used to study these relationships after adjustment for the effects of age, sex, diabetes duration, MAP, and HbA_1c. Associations are illustrated graphically by partial residual regression plots, where residuals were computed by regressing each of the morphometric variables on the clinical covariates.

We performed four sensitivity analyses. First, we further adjusted our final model for diabetes duration and then for the intraglomerular parameters P_GLO and R_A/R_E ratio, since we previously demonstrated that these parameters represent risk factors for kidney failure in this population (4,5). To determine the prognostic value of ω₀ for kidney failure, while accounting for death, we performed a competing risk analysis using the proportional subhazard (sub-HR) method (14). Finally, we examined the relationship between ω₀ and kidney failure using the simpler isoporous + shunt model described above.

Statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC) and R version 4.0.4 or later (The R Foundation for Statistical Computing, https://www.R-project.org/) software. P < 0.05 was considered statistically significant.

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Baseline characteristics of the 185 participants in this study (74 men, 111 women) are summarized in Table 1. Mean age at baseline was 43 ± 10 years, and mean GFR was 147 ± 46 mL/min. Median duration of diabetes was 9.7 (IQR 3.0–15.0) years, and median urine ACR was 41 (IQR 11–230) mg/g. Nineteen (10%) participants were receiving antihypertensive treatment at the baseline examination, and only one of these participants was receiving an ACE inhibitor. Five (3%) participants had a baseline GFR of <60 mL/min, and 86 (46%), 58 (31%), and 41 (22%) had normal, moderate, and severe albuminuria, respectively. Participants were followed for a median of 17.7 (IQR 10.0–27.3) years. During follow-up, 67 (36%) participants developed kidney failure and 123 (74%) died, 57 after developing kidney failure. Only two participants died from kidney failure without first receiving kidney replacement therapy. The relationships among the three parameters that characterize the log-normal + shunt model at baseline (shunt parameter [ω₀], breadth of the log-normal distribution of the restrictive pores [s], and mean radius of the restrictive pores [u]) and kidney failure are shown in Fig. 2. On average, participants experiencing kidney failure had slightly but not statistically significantly higher dextran sieving coefficients for pore radii of 56–60 Å and significantly lower dextran sieving coefficients for pore radii of 26–38 Å than those who did not progress to kidney failure (Fig. 3). The differences in sieving coefficients between the groups led to significantly higher shunt magnitudes in the hydrodynamic model, on average, among those who progressed to kidney failure (Table 2). Neither of the parameters characterizing the restrictive pores in the model was associated with kidney failure.

The Kaplan-Meier survival plot for kidney failure by tertiles of ω₀ is shown in Fig. 4. The probability of developing kidney failure (log-rank P = 0.0001) was significantly higher among participants in the higher tertiles of ω₀. In an unadjusted Cox proportional hazards model, the risk of kidney failure was significantly higher with each 1-SD increment in ω₀ in the log-normal + shunt model (HR 1.96 [95% CI 1.52, 2.52]). After adjustment for age, sex, MAP, HbA_1c, GFR, and ACR at baseline, the risk of kidney failure (HR 1.55 [95% CI 1.17, 2.05]) was modestly attenuated but remained statistically significant (Table 3). When either diabetes duration (HR 1.46 [95% CI 1.09, 1.95]) or P_GLO and R_A/R_E ratio (HR 1.55 [95% CI 1.13, 2.10]) were added to the Cox proportional hazards model, the effect of ω₀ on the risk of kidney failure was not attenuated and remained statistically significant.

After accounting for death as a competing event in the multivariable model, ω₀ remained associated with kidney failure (sub-HR 1.53 [95% CI 1.20, 1.94]). The association between ω₀ and kidney failure was also significant in each of the simpler isoporous + shunt models of solute transport (Table 3).

Morphometric parameters from research kidney biopsies were available in a subset of 51 participants. The kidney biopsy was performed a median of 3.0 (IQR 1.5–3.4) years after the dextran sieving study and a maximum of 4.6 years. Characteristics of the participants who underwent a kidney biopsy and those who did not are summarized and compared in Supplementary Table 1. Participants who had a kidney biopsy had lower BMI and higher HbA_1c, GFR, and RPF, on average, than those who did not undergo a biopsy. Pearson correlation coefficients between ω₀ and structural parameters from the kidney biopsy are shown in Supplementary Table 2. Higher ω₀ correlated positively with mesangial fractional volume and nonepithelial cell density after adjustment for age, sex, diabetes duration, MAP, and HbA_1c (Fig. 5).

We demonstrated that enhanced transglomerular passage of macromolecules in the glomerular capillaries of American Indians with T2D and early DKD via a nonselective shunt pathway was associated with kidney failure later in life, independent of albuminuria and GFR. This finding suggests that mechanisms responsible for impaired glomerular barrier permselectivity are important and potentially independent determinants of DKD progression in T2D. The reduced sieving of dextrans of smaller radius among participants who progressed to kidney failure may be attributable to local hemodynamic changes that deleteriously effect permeation by reptation in these linear macromolecules, leading to a greater effective hydrodynamic radius than in those who did not progress. We also showed that the shunt parameter correlated strongly with some of the glomerular structural lesions that best predict loss of GFR and progression to kidney failure (15). It is therefore crucial to understand the molecular and metabolic factors that contribute to the enhanced transglomerular passage of macromolecules in order to better understand possibly targetable mechanisms of disease progression.

In the kidneys, the physiologic importance of permeability in glomerular function and in maintaining a barrier to plasma proteins demands a high degree of specialization in the glomerular capillary wall. While much research has focused on the role of the GBM and the podocyte and its slit diaphragm in maintaining the glomerular filtration barrier, recent work has demonstrated the importance of other components of this barrier, including the endothelium and the endothelial glycocalyx, and emphasizes the importance of crosstalk among various components in maintaining an effective barrier (16,17). The proximal tubule may also play an important role as a cellular retrieval mechanism for proteins crossing the glomerular capillary into Bowman space (18). The extent to which filtration within the glomerulus and cellular uptake downstream (18) affect the appearance of circulating proteins in the urine remains an area of intense investigation and controversy, although we presume that significant postglomerular reabsorption of dextrans does not occur in the kidney.

Hydrodynamic models of hindered solute transport through water-filled cylindrical pores are often used to describe the observed transport of dextran macromolecules across the glomerular capillary wall (9). The model that best replicates observed values for glomerular sieving of dextran in people with normal urine albumin excretion or nephrotic range proteinuria represents the glomerular capillary wall as a heteroporous membrane perforated by a series of restrictive pores with a log-normal distribution of pore radii and a parallel distribution of nonrestrictive shunt-like pores (12). In the current study, we selected this model to assess the association between shunt magnitude and progression of DKD to kidney failure. A simpler model in which the restrictive pores were assumed to be of uniform radius also demonstrated a statistically significant association between shunt magnitude and progression to kidney failure.

Detailed morphometric studies conducted in research kidney biopsies in these American Indians have demonstrated strong associations between the cells that make up the glomerular filtration barrier, namely the podocytes and endothelial cells, and progressive DKD (8,13,15,19,20). Increased thickness of the GBM, the extracellular matrix component of the selectively permeable glomerular filtration barrier, is also associated with DKD progression in this population. The early appearance in American Indians with T2D of defects in barrier size selectivity, based on dextran sieving studies, raises the possibility that an early increase in the filtered protein load across the glomerular capillary may contribute to the pathogenesis of DKD (6). This hypothesis is supported by a study in people with type 1 diabetes that found that transglomerular passage of dextran macromolecules of a broad size distribution (radii 28–60 Å) was lowered significantly by treatment with the ACE inhibitor enalapril. The improvements in sieving characteristics observed in the study were accompanied by reductions in fractional albumin and IgG clearances, which increased after withdrawal of the drug. Since neither GFR nor RPF were affected by treatment with enalapril in the study participants, the primary action of enalapril was thought to be the modulation of intrinsic membrane properties of the glomerular filtration barrier (21). On the other hand, a study in people with T2D and biopsy-confirmed DKD failed to demonstrate an effect of either the ACE inhibitor perindopril or the calcium channel blocker nitrendipine on the size-selective properties of the glomerular barrier. Neither drug modified the fractional dextran clearance over the entire range of molecular radii investigated (26–66 Å), even though the fractional clearances of large dextran macromolecules in those with T2D were significantly elevated over control values (22). Regardless of the precise nature of the underlying disturbance in the glomerular filtration barrier that leads to the early loss of size selectivity observed in American Indians with T2D, the magnitude of this defect was strongly associated with progression to kidney failure in this study, suggesting that mechanisms underlying early loss of glomerular permselectivity may be important therapeutic targets in DKD.

Increasing R_A/R_E ratio associates strongly with incident kidney failure independent of GFR and traditional risk factors. Likewise, higher P_GLO followed by a progressive decline in this parameter confers greater risk of this outcome (5). Each of these intraglomerular hemodynamic parameters correlates strongly with the glomerular structural lesions that best predict loss of GFR and progression to kidney failure. When each of these covariates was added to our Cox proportional hazards model along with the other covariates, ω₀ remained significantly associated with progression to kidney failure.

This study has important strengths and limitations worth discussing. Strengths include the comprehensive assessment of glomerular physiology using gold standard measures of hemodynamic and barrier function in a well-characterized and relatively large cohort of adults with T2D with decades of follow-up for major health outcomes. Moreover, nearly all the kidney failure that develops in this population is ascribed to diabetes (23), and other causes of kidney disease were specifically excluded from this study. Only one participant in our study was receiving an ACE inhibitor at the time of the dextran sieving measurement, so we were unable to assess its effect on the sieving characteristics of the glomerular barrier. Although the study was conducted in an American Indian population, raising concerns about its generalizability, risk factors for kidney failure identified in American Indians, including long diabetes duration, hyperglycemia, elevated blood pressure, albuminuria, and the presence of diabetic retinopathy, as well as the clinical course of the disease, have been consistently replicated in other populations, arguing that findings in this population are likely generalizable (1). Furthermore, the permselectivity defects observed in American Indians with T2D have also been observed in other patients with diabetes as described above (21,22). The longer duration of follow-up in our study permitted us to examine the impact of these permselective defects on kidney failure, the pivotal hard outcome. A limitation of this study is that estimates of barrier function are based on mathematical modeling that may not fully reflect membrane properties and performance. Additionally, dextran sieving measurements are not feasible in the outpatient setting because of the requirement of intravenous access and serial collection of blood and urine samples, which limits its clinical application to research. However, it is possible that some panels of urine permselectivity markers, such as the selectivity index (24–26), could capture much of the predictive power of the shunt parameter. Despite adjustment for traditional kidney risk factors and renal hemodynamics parameters, residual confounding may still affect our findings. GFR and albuminuria were measured over many years in this study, but variation in attendance at these examinations could lead to informative censoring bias since loss to follow-up was sometimes related to these study outcomes. Accordingly, we elected to focus our analysis on the relationship of glomerular permselectivity with the end-stage complication of kidney failure, which was identified independently of the research examinations.

In conclusion, loss of size selectivity of the glomerular barrier is an early marker of glomerular injury in diabetes that associates with progression to kidney failure, independent of albuminuria and GFR. Revealing the molecular mechanisms underlying this early alteration in barrier function may provide new therapeutic targets for DKD in T2D. Identifying a simpler analog of the shunt parameter could also provide an early marker of risk.

Acknowledgments. The authors acknowledge the work of Lois I. Jones, Enrique Diaz, Bernadine Waseta, and Camille Waseta from the National Institute of Diabetes and Digestive and Kidney Diseases in Phoenix.

Funding. This work was supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. P.B. receives salary and research support from the National Institute of Diabetes and Digestive and Kidney Diseases (grants DK132399, DK129211, DK129720, DK116720), National Heart, Lung, and Blood Institute (grant HL165433), JDRF (grant 3-SRA-2022-1097-M-B), American Heart Association (grant 20IPA35260142), Boettcher Foundation Center for Women’s Health at the University of Colorado, and the Department of Pediatrics, Section of Endocrinology, and Barbara Davis Center for Diabetes at the University of Colorado School of Medicine.

Duality of Interest. P.B. has acted as a consultant for AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Eli Lilly, LG Chem, Sanofi, Novo Nordisk, and Horizon Pharma and serves on the advisory boards or steering committees for AstraZeneca, Bayer, Boehringer Ingelheim, Novo Nordisk, and XORTX. P.J.S. has acted as a consultant for AstraZeneca and serves on the advisory board for Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. P.J.S., H.C.L., R.G.N., and P.B. wrote the manuscript and researched data. A.L. and K.V.L. assisted in the analyses, contributed to the discussion, and reviewed and edited the manuscript. R.G.N. designed the study. R.G.N. 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.

Prior Presentation. Parts of this study were presented in abstract form at the American Society of Nephrology Kidney Week 2022, Orlando, FL, 3–6 November 2022.