Serum Level of Polyubiquitinated PTEN and Loss of Kidney Function in American Indians With Type 2 Diabetes

Abstract

Rationale & Objective:

Fibrosis is a major driver of chronic kidney disease and epithelial-mesenchymal transition (EMT) may contribute to its development. A polyubiquitinated form of phosphatase and tensin homolog (PTEN ^K27polyub) promotes EMT in vitro. Thus, it is a potentially useful biomarker of progressive kidney fibrosis and may predict loss of kidney function.

Study Design:

Observational cohort study.

Setting & Participants:

Southwest USA. American Indians (154 women, 80 men) with or at high risk for diabetic kidney disease (DKD).

Predictors:

Serum PTEN^K27polyub.

Outcomes:

≥40% loss of glomerular filtration rate (GFR) or onset of kidney failure. Kidney structural measures in a sub-set of study participants who underwent research kidney biopsies (n=77).

Analytical approach:

Cox proportional hazards models adjusted for age, sex, diabetes duration, HbA1c, blood pressure, use of renin angiotensin system (RAS) blockers, measured GFR and albuminuria. Spearman correlations for associations with structural measures.

Results:

At baseline, mean age was 42.8 ± 10.5 (SD) years, diabetes duration 11.5 ± 7.1years, mean arterial pressure 90.5 ± 9.5 mmHg, HbA1c 9.3 ± 2.4%, GFR 152 ± 45 ml/min and median albumin:creatinine ratio 38 mg/g (interquartile range 14–215). 64 (27.4%) participants were using RAS blockers. Higher PTEN^K27polyUb was associated with greater risk of ≥40% loss of GFR during a median follow-up of 6.3 years (Hazard ratio [HR] for the 4^th vs. 1^st quartile = 3.95, 95% CI 2.23–6.98, p<0.001). Serum PTEN^K27polyUb was associated with an increased risk of kidney failure over a median follow-up of 15.8 years (HR quartile 4 vs. 1 = 5.66, 95% CI 1.99–16.13, p=0.001). Baseline serum PTEN^K27polyUb in the biopsy subset correlated with structural measures including glomerular basement membrane width (rho =0.370, p<0.001) and mesangial fractional volume (rho =0.392, p<0.001).

Limitations:

Small study in single population.

Conclusion:

Higher serum PTEN^K27polyUb is associated with increased risk for GFR decline and kidney failure in American Indians with type 2 diabetes.

PLAIN-LANGUAGE SUMMARY

A recently identified modified form of the protein PTEN has been implicated in kidney fibrosis in animals and in fibrotic mechanisms in human cellular studies. We wanted to see if circulating levels of this form of PTEN were associated with progression of kidney disease in American Indians with type 2 diabetes. We found that the modified form of PTEN was associated with increased risk of decline in kidney function and the onset of kidney failure. In a subgroup of participants, serum modified PTEN levels were also associated with the severity of lesions in kidney structure, which signify early manifestations of diabetic kidney disease.

Introduction

Fibrosis is a major element of progressive chronic kidney disease (CKD) regardless of etiology. As such, the mechanisms underpinning the development of fibrosis are of great importance for understanding the pathophysiology of CKD and, markers of fibrosis could also be clinically informative biomarkers of CKD progression.¹ One process that may contribute to renal fibrosis in CKD is epithelial-mesenchymal transition (EMT), whereby epithelial cells transform into myofibroblasts. ²

Phosphatase and tensin homolog (PTEN), a lipid and protein phosphatase, ^3,4 plays a role in fibrosis and PTEN expression is reduced in fibrotic conditions.⁵ Low intracellular PTEN in kidneys is associated with the development of fibrosis following acute kidney injury ⁶ as well as the development of fibrosis and “mesangial hypertrophy” in cellular and animal models of diabetic kidney disease (DKD). ^7,8 Intracellular PTEN levels can be regulated through changes in expression of the PTEN gene, but also by post-translational modifications. ⁹ PTEN^K27polyUb is a recently described K27-linked polyubiquitinated form of PTEN which, unlike unmodified PTEN, exhibits reduced lipid phosphatase activities but gains serine and threonine protein phosphatase activities. As such, it dephosphorylates and stabilizes key regulators of EMT, both in vitro and in animal studies. ^10,11 In view of the in-vitro findings, we considered this form of PTEN a potential biomarker for progression of CKD. PTEN^K27polyUb is detectable in multiple tissues in humans, including kidney tubules, ¹⁰ and is measurable in serum and urine. ¹¹ Serum concentrations of PTEN^K27polyUb are higher in people with diabetes than in those without, with the highest concentrations seen in people with DKD. ¹¹ Here, we measured serum PTEN^K27polyUb concentrations in a cohort of American Indians with type 2 diabetes and obtained longitudinal measures of kidney function to determine how serum PTEN^K27polyUb relates to loss of kidney function. We also examined the association of serum PTEN^K27polyUb in a subset of participants who had research kidney biopsies to explore whether serum PTEN^K27polyUb is also associated with the structural lesions of DKD.

Methods

Population

Participants in this study were selected from prior kidney studies conducted in American Indians with type 2 diabetes from the Gila River Indian Community in Arizona that used the same methods for evaluating kidney function. ^12,13 Study visits included annual measurement of measured glomerular filtration rate (mGFR) and collection of biological samples for storage. For this study, we included participants who had an available stored serum sample and ≥2 measures of mGFR (n=234). The timeline of participants baseline samples and follow-up is detailed in Figure S1. This study was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases. Each participant signed an informed consent document.

Clinical measures

GFR was measured after an overnight fast by the urinary clearance of iothalamate (described in detail elsewhere ¹⁴). Due to the level of obesity in participants, absolute mGFR was reported, uncorrected for body surface area. ¹⁵ We measured albumin:creatinine ratio (ACR) in the baseline urine sample collected at the clearance study. ACR was categorized as normal (<30 mg/g), microalbuminuria (≥30 −299 mg/g), or macroalbuminuria (≥300 mg/g). We also measured height and weight for calculation of body mass index (BMI), and blood pressure with the participant resting in the seated position. Mean arterial pressure (MAP) was calculated as (2 x diastolic blood pressure + systolic blood pressure)/3. High-performance liquid chromatography was used to measure iothalamate ¹⁴ and HbA1c. ¹⁶ Urine albumin was measured by nephelometry, and urine creatinine by a modified Jaffe reaction. ^17,18 All serum samples were stored at −80° C prior to measurement of PTEN^K27polyUb. Medication use at baseline was obtained by self-report.

PTEN measurements

PTEN^K27polyUb was measured in stored samples at the MD Anderson Cancer Center, University of Texas using a Ubiquant ELISA kit and an antibody specific to the ubiquitinated form of PTEN, as previously described.¹¹

Outcomes

Participants were followed until December 31, 2018, onset of kidney failure requiring renal replacement therapy (RRT), or death, whichever came first. Vital status and initiation of RRT were ascertained independent of study visits. Outcomes for the study were an ≥40% decline in mGFR, kidney failure and all-cause mortality. To avoid the bias associated with informative censoring for the ≥40% mGFR decline outcome, we imputed GFR slope, as previously described ¹⁹ and detailed in Item S1 and Figure S2. In brief, for participants who did not meet the 40% decline in mGFR threshold, we calculated a slope based on either the final two mGFRs or the final mGFR and the onset of kidney failure (taken as an mGFR of 15ml/min). This slope was applied to the final mGFR to calculate imputed values of mGFR for up to 2-years. If the imputed mGFR passed the 40% threshold, we calculated the time it crossed the threshold and used that as the outcome date. For participants with an mGFR that met the 40% decline threshold, we used that and the immediately prior mGFR to calculate the slope to estimate the date they crossed the 40% decline threshold. Follow-up was truncated at the earliest date of reaching the 40% decline threshold, death, 2 years after final mGFR or December 31, 2018. As a sensitivity analysis, we repeated analyses using the unimputed data. Cause of death data are complete through the end of 2016 and are based on review of death certificates and medical records.

Kidney morphometry

One hundred and four of the current study participants had a research kidney biopsy. ¹⁵ All biopsies took place after the baseline exam (median time between serum PTEN^K27polyUb measure and biopsy = 5.5 years, interquartile range (IQR) 5.0 – 6.1 years). We present data for the 77 biopsies obtained within <6.0 years of the PTEN^K27polyUb measurement. Kidney biopsy tissues were embedded in Epon and prepared for light and electron microscopy according to standard procedures. ^20–22 The following glomerular structural parameters were measured by unbiased morphometry on electron microscopy images, as described elsewhere ^{20, 21, 23}: glomerular basement membrane width, mesangial fractional volume, ²⁴ glomerular filtration surface density, ^{24, 25} mean foot process width, ²⁶ percentage of endothelial fenestrations, ²⁶ fractional podocyte volume per glomerulus,²⁷ and the number density of podocytes per glomerulus. ²⁷ Cortical interstitial fractional volume, ²² and mean glomerular volume ²⁸ were estimated using light microscopy.

Statistical analysis

Cross sectional analyses:

For associations with serum PTEN^K27polyUb, we used analysis of variance for continuous variables and logistic regression for categorical variables.

Survival analyses:

We used PROC LIFETEST to calculate the log rank test for equality and tests for linear trend for univariate survival curves. We used Cox proportional hazards models to assess the association between quartiles of PTEN^K27polyUb and the outcomes of interest, and hazard ratios (HR) are presented relative to the lowest quartile as well as for overall trend for an ordinal measure of quartiles of PTEN^K27polyUb. We tested the proportionality assumption for each variable, as described by Kalbfleisch and Prentice.²⁹ When a variable violated this assumption, follow-up was truncated to a point when the proportionality assumption still applied. As sensitivity analyses for the kidney failure outcome, we also examined sub-distribution HRs using the competing risk model of Fine and Gray, ³⁰ and interval-specific HRs based on three follow-up intervals (<7.5, 7.5->12.5 and ≥12 years). For all outcomes we also examined serum PTEN^K27polyUb as a continuous measure, using Cox models including cubic splines to evaluate non-linear associations using the LGTPHCURV9 SAS macro. ³¹

Kidney structural analyses:

We used Spearman correlation coefficients for the associations between structure and serum PTEN^K27polyUb.

Genetic analyses:

We explored potential causes for the distribution of serum PTEN^K27polyUb by examining genetic variation in the genes involved in the production of PTEN^K27polyUb - PTEN, Ubiquitin B, Ubiquitin C and MEX3C. We used data from 207 study participants previously genotyped using a custom Pima Indian Axiom genome-wide array (Affymetrix,Santa Clara, CA) ³² and/or an exome chip by Regeneron Genetics Center (Tarrytown, NY). ³³ We compared genotypes for variants in the genes between participants with serum PTEN^K27polyUb above and below the median value using Fisher’s exact tests.

All analyses were carried out using SAS 9.4 (Cary, NC).

Results

PTEN^K27polyUb levels were measured in baseline serum samples collected from 234 study participants (154 women/80 men). Samples were stored at −80 ⁰ C for a median of 19.3 years (IQR 17.0 – 26.1) before measurement of PTEN^K27polyUb with a range from 10.9 to 27.3 years. There was no evidence of an association between serum PTEN^K27polyUb levels and storage time (Spearman correlation coefficient = −0.005, P=0.9). The median serum PTEN^K27polyUb was 1.38 ng/ml (IQR 0.16 – 6.69). Serum PTEN^K27polyUb levels had a bimodal distribution (Figure 1). We tested SNPs within the PTEN, Ubiquitin B, Ubiquitin C and the MEX3C genes against a dichotomized serum PTEN^K27polyUb variable (divided at the median) but found no significant associations (Table S1).

Figure 1:

Distribution of K27 polyubiquitinated PTEN

A. untransformed

B. log transformed (x-axis units are anti-logged to indicate actual values of serum PTEN^K27polyUb)

Baseline participant characteristics are shown in Table 1. Kidney function at baseline was mostly preserved or elevated among study participants with a mean mGFR of 152 ± 45 ml/min; only 4 participants had a mGFR <60 ml/min (CKD stage 3). However, 53.8% of participants had some degree of albuminuria (33.3% microalbuminuria and 20.5% macroalbuminuria). Cross-tabulation of GFR and ACR shows that 108 participants (46.2%) had no albuminuria and a GFR >90 ml/min (Table 2). In univariate analyses higher serum PTEN^K27polyUb was associated with older age (rho =0.177, P=0.007), longer diabetes duration (rho =0.219, P=0.001), higher MAP (rho = 0.194, P=0.003), higher HbA1c (rho =0.153, P=0.02), and greater use of insulin (rho =0.293, P<0.001), lower mGFR (rho = −0.321, P <0.001) and greater albuminuria (rho = 0.349, P <0.001). The 4 participants in the study with CKD stage 3 had serum PTEN^K27polyUb levels in the upper quartile, and of the 15 participants with a baseline mGFR <90 ml/min, only one had a serum PTEN^K27polyUb level below the upper quartile (Figure S3).

Table 1:

Baseline clinical characteristics by quartile of K27 polyubiquitinated PTEN concentration

		Quartile of PTENK27polyUb
	All (N=234)	Q1 (N=59)	Q2 (N=58)	Q3 (N=59)	Q4 (N=58)	P-value trend
Serum PTENK27polyUb (ng/ml)	1.38 (0.16–6.69)	0.14 (0.12–0.15)	0.21 (0.18–0.23)	4.15 (2.82–5.51)	10.95 (7.70–12.86)	-
Age (years)	42.5 ± 10.6	42.2 ± 9.3	38.3 ± 11.6	43.2 ± 10.3	46.5 ± 9.5	0.004
Male sex	80 (34.2%)	24 (40.7%)	21 (36.2%)	14 (23.7%)	21 (36.2%)	0.3
Diabetes duration (years)	11.3 ± 7.1	8.9 ± 6.2	10.7 ± 7.8	13.6 ± 6.3	12.1 ± 7.1	0.001
MAP (mmHg)	90 ± 10	89 ± 10	89 ± 9	91 ± 11	93 ± 9	0.005
GFR (ml/min)	152 ± 45 (Range 34–284)	162 ± 40 (Range 100–284)	168 ± 40 (Range 74–265)	157 ± 41 (Range 94–265)	122 ± 45 (Range 34–233)	<0.001
ACR (mg/g)	38 (14–215)	18 (9–51)	22 (8–69)	119 (28–366)	83 (21–486)	<0.001
Albuminuria						<0.001
Normo	108 (46.2)	40 (67.8)	33 (56.9)	15 (25.4)	20 (34.5)
Micro	78 (33.3)	15 (25.4)	18 (31.0)	26 (44.1)	19 (32.8)
Macro	48 (20.5)	4 (6.8)	7 (12.1)	18 (30.5)	19 (32.8)
HbA1c (%) †	9.3 ± 2.4	8.5 ± 2.4	9.3 ± 2.5	10.0 ± 2.0	9.4 ± 2.4	0.01
BMI (kg/m2)	35.1 ± 8.3	36.0 ± 8.1	35.0 ± 8.6	34.3 ± 8.0	35.1 ± 8.6	0.5
Antihypertensive treatment (N/%)	94 (40.2)	21 (36.6%)	23 (39.7)	27 (45.8)	23 (39.7)	0.5
Renin Angiotensin System blocker treatment (N/%)	64 (27.4)	15 (25.4)	19 (32.8)	17 (28.8)	13 (22.4)	0.6
Diabetes treatment						<0.001
No drugs	81 (34.6%)	29 (49.2%)	22 (37.9%)	18 (30.5%)	12 (20.7%)
Orals only	98 (41.9%)	26 (44.1%)	29 (50.0%)	18 (30.5%)	25 (43.1%)
Insulin ± orals	55 (23.5%)	4 (6.8%)	7 (12.1%)	23 (39.0%)	21 (36.2%)

MAP = mean arterial pressure; GFR = glomerular filtration rate; ACR = albumin:creatinine ratio; Normo = normoalbuminuria (ACR <30mg/g); Micro = microalbuminuria (ACR 30–300 mg/g); Macro = macroalbuminuria (ACR >300 mg/g); BMI = body mass index. p-values are from linear regression models for continuous measures and from logistic regression models for categorical measures.

^†N=228 and for quartile 1 N=58, quartile 2 N=57, quartile 3 N=57 and quartile 4 N=56

Conversion of units: HbA1c to mmol/mol (HbA1c in % – 2.152)/0.09148

Table 2:

Cross-tabulation of glomerular filtration rate and albumin:creatinine ratio at baseline

	ACR <30 mg/g	ACR 30–300 mg/g	ACR > 300 mg/g
GFR <60 ml/min		2	2
GFR 60–90 ml/min	2	4	5
GFR >90 ml/min	106	72	41

ACR = albumin to creatinine ratio; GFR = glomerular filtration rate

PTEN^K27polyUb and 40% decline in mGFR

After imputation, 168 participants (72%) had a ≥40% fall of mGFR from baseline over a median follow-up of 6.3 years (IQR 3.4–11.6 years). Higher PTEN^K27polyUb was associated with greater risk of a ≥40% decline in univariate survival analysis (P <0.0001 log rank test for equality of survival curves and for log rank test of trend) (Figure 2a). We report results from a series of Cox models with sequential additions of covariates (Table 3). Quartiles of serum PTEN^K27polyUb were significantly associated with 40% decline in mGFR (p<0.001 for all models). In the fully adjusted model (adjusted for age, sex, diabetes duration, HbA1c, BMI, MAP, RAS blocker use, mGFR and ACR) the HR for the 4^th vs. 1^st quartile was 3.95, 95% CI 2.23–6.98, P <0.001. Findings were similar when repeated using unimputed data (Table S2).

Figure 2:

Survival plots for kidney outcomes by quartile of serum PTEN^K27polyUb

A. ≥40% decline in GFR

B. Kidney failure

C. End stage renal disease or death from any cause

Black line – quartile 1 (lowest) serum PTEN^K27polyUb

Blue line – quartile 2 serum PTEN^K27polyUb

Green line– quartile 3 serum PTEN^K27polyUb

Red line– quartile 4 (highest) serum PTEN^K27polyUb

Number of participants in each quartile shown at 2.5 year (panel A) or 5 year (panels B & C) are shown below each survival plot. All curves are truncated when group <10 participants

Table 3:

Association of K27 polyubiquitinated PTEN concentration with kidney outcomes and all-cause mortality

	Cases	Follow-up (years)	IR (/1000 pyrs)	Univariate model*		Model 1		Model 2		Model 3
40 % loss of GFR				HR (95%CI)	P	HR (95%CI)	p	HR (95%CI)	P	HR (95%CI)	p
Q1	23	7.9 (4.1–14.8)	4.3	Reference	-	Reference	-	Reference	-	Reference	-
Q2	37	6.2 (3.1–13.7)	7.6	1.80 (1.07–3.03)	0.03	1.39 (0.81–2.37)	0.2	1.58 (0.92–2.73)	0.1	1.48 (0.86–2.55)	0.2
Q3	56	7.3 (4.0–11.6)	11.4	2.75 (1.69–4.48)	<0.001	2.43 (1.48–3.98)	<0.001	2.12 (1.28–3.50)	0.004	1.44 (0.85–2.44)	0.2
Q4	49	4.9 (2.8–8.4)	12.7	3.19 (1.94–5.27)	<0.001	3.56 (2.10–6.02)	<0.001	3.90 (2.27–6.70)	<0.001	3.95 (2.23–6.98)	<0.001
All	165	6.3 (3.4–11.6)	8.7		<0.001		<0.001		<0.001		<0.001
Kidney failure
Q1	5	17.5 (9.6–21.1)	0.5	Reference	-	Reference	-	Reference	-	Reference	-
Q2	12	17.5 (10.5–20.3)	1.2	2.63 (0.84–8.26)	0.1	1.45 (0.50–4.20)	0.49	1.63 (0.57–4.68)	0.4	1.92 (0.67–5.54)	0.2
Q3	28	15.1 (9.7–19.6)	3.1	3.12 (1.02–9.58)	0.05	5.45 (2.09–14.18)	0.001	4.39 (1.67–11.50)	0.003	3.19 (1.19–8.53)	0.02
Q4	28	14.0 (6.9–18.3)	3.7	6.14 (2.11–17.88)	0.001	8.75 (3.39–22.94)	<0.001	8.09 (3.03–21.62)	<0.001	5.66 (1.99–16.13)	0.001
All	73	15.8 (9.4–19.8)	2.0		<0.001		<0.001		<0.001		0.001
Death
Q1	34	17.6 (10.2–21.6)	3.5	Reference	-	Reference	-	Reference	-	Reference	-
Q2	29	18.5 (14.8–21.3)	2.8	0.79 (0.48–1.30)	0.3	0.77 (0.46–1.29)	0.3	0.79 (0.47–1.32)	0.4	0.84 (0.49–1.42)	0.5
Q3	32	17.8 (14.5–21.4)	3.0	0.86 (0.53–1.40)	0.5	0.81 (0.50–1.33)	0.4	0.82 (0.49–1.35)	0.4	0.74 (0.44–1.25)	0.3
Q4	38	17.6 (12.3–19.7)	4.1	1.24 (0.78–1.97)	0.4	1.04 (0.65–1.67)	0.9	1.03 (0.64–1.67)	0.9	0.97 (0.56–1.68)	0.9
All	133	17.9 (13.1–21.1)	3.3		0.3		0.8		0.9		0.7

IR = incident rate; pyrs = person years; IQR = interquartile range; HR = hazards rate; 95% CI = 95% confidence interval; P = p-value;

Q1 = lowest quartile of serum PTEN^K27polyUb; Q2 = second quartile of serum PTEN^K27polyUb; Q3 = third quartile of serum PTEN^K27polyUb; Q4 = highest quartile of serum PTEN^K27polyUb; P_trend shows the p-value over quartiles of serum PTEN^K27polyUb; pyrs = person years;

Data shown in bold represents statistically significant results (p<0.05)

Model 1 – adjusted for age, sex and diabetes duration; Model 2 = model 1 + hba1c, BMI, mean arterial pressure, renin angiotensin system blocker use; Model 3 – model 2 + glomerular filtration rate and albumin-to-creatinine ratio.

^*follow-up truncated at 13 years (49 cases) for kidney failure

PTEN^K27polyUb and kidney failure

Over a median follow-up of 15.8 years (IQR 9.4–19.8) there were 74 cases of kidney failure (32%) with the majority (64/74) also included in the ≥40% decline outcome prior to reaching the kidney failure outcome. Higher quartiles of PTEN^K27polyUb were significantly associated with risk of kidney failure in univariate survival plots (P<0.001 log rank test for equality of survival curves and for log rank test of trend) (Figure 2b). Quartiles of PTEN^K27polyUb were significantly associated with kidney failure in all the Cox survival models (Table 3). However, we had limited power for the Cox model which was censored at 13 years (49 cases) due to violation of the proportionality assumption for the 3^rd quartile of serum PTEN^K27polyUb. In the fully adjusted Cox models, in which follow up was not censored (74 cases) and which was adjusted for age, sex, diabetes duration, HbA1c, BMI, RAS blocker use, mGFR and albuminuria the highest quartile of serum PTEN^K27polyUb was associated with a greater risk of kidney failure compared to quartile 1 (HR =5.66, 95% CI 1.99–16.13, P =0.001).

In kidney failure models which included death as a competing risk, HRs for kidney failure were modestly potentiated (kidney failure sub-distribution HR for the 4^th vs. 1^st quartile in the fully adjusted model =6.25, 95% CI 2.17–17.99, p=0.001) (Table S3). The interval-specific models are consistent with our main findings. After 12.5 years, the highest risk in a univariate model was for the 3^rd quartile (HR Q3 vs Q1= 9.38; HR Q4 vs Q1 =5.30, however both had wide confidence intervals (Table S4).

PTEN^K27polyUb and all-cause mortality

During a median follow-up of 17.9 years (IQR 13.1–21.1) there were 133 deaths (53 following progression to kidney failure). Cause of death was available for all deaths prior to 2016 (n=124) (Table S5). Cardiovascular disease (n=31) was the leading cause of death. There was no association between quartiles of PTEN^K27polyUb and all-cause mortality in either the univariate (Figure 2c) or multivariate analyses (Table 3).

Serum PTEN^K27polyUb sensitivity analyses

As a sensitivity analysis, we used cubic splines to model serum PTEN^K27polyUb as a continuous measure. For each model, we included 3 knots and used the median of serum PTEN^K27polyUb as the reference value. For ≥40% decline in mGFR and kidney failure, the univariate models showed marked non-linearity. With greater adjustment, the observed associations became more linear. In the fully adjusted models, the best fit for both the 40 % decline in mGFR and kidney failure outcomes was linear rather than curved (Figure S4). These models showed no significant association with serum PTEN^K27polyUb and death (Figure S5).

PTEN^K27polyUb and kidney structure

Seventy-seven participants had a research kidney biopsy within the 6 years of the serum PTEN^K27polyUb measurement. These participants were younger, with shorter diabetes duration, lower MAP, HbA1c and insulin use than the whole cohort (Table S6). They also had higher mean mGFR (161 ± 40 ml/min) and lower median ACR (28, IQR 9–63 mg/g). Serum PTEN^K27polyUb correlated with a number of morphometric measures (Table 4), including glomerular basement membrane width (rho =0.370, P=0.001), mesangial fractional volume (rho =0.392, P<0.001), glomerular filtration surface density (rho = −0.266, P=0.02), podocyte fractional volume (rho = −0.382, P=0.001) and the number density of podocytes per glomerulus (rho = −0.396, P<0.001). These findings were essentially unchanged in models adjusted for age, sex and time between serum measurement and kidney biopsy.

Table 4:

Kidney morphometry and association with serum PTEN^K27polyUb

			Univariate		Partial correlation coefficient controlling for age, sex and time from biopsy
Morphometric measure	N	Mean (SD)/median (IQR)	Spearman correlation coefficient	P-value	Spearman correlation coefficient	P-value
GBM width (nm)	77	477 ± 105	0.370	0.001	0.402	<0.001
Mesangial fractional volume	77	0.27 ± 0.08	0.392	<0.001	0.379	0.001
Glomerular filtration surface density (μm2/μm3)	77	0.10 ± 0.04	−0.266	0.02	−0.259	0.03
Cortical interstitial fractional volume	73	0.18 ± 0.04	0.205	0.08	0.221	0.07
Glomerular volume (106 μm3)	70	2.36 ± 0.80	−0.057	0.6	−0.010	0.9
Foot process width (nm)	76	608 (468 – 773)	0.089	0.4	0.046	0.7
Percentage of endothelial fenestrations (%)	76	45 ± 19	−0.210	0.07	−0.190	0.1
Podocyte fractional volume per glomerulus	77	0.16 ± 0.05	−0.382	0.001	−0.437	<0.001
Numeric density of podocytes per glomerular volume (x106)	77	133 ± 83	−0.396	<0.001	−0.401	<0.001

SD = standard deviation; IQR = interquartile range

Discussion

Higher serum PTEN^K27polyUb levels are associated with increased risk of 40 % decline in mGFR and incident kidney failure. Serum PTEN^K27polyUb was also positively correlated with key kidney morphometric measurements including glomerular basement membrane width and mesangial fractional volume which are themselves strong correlates of kidney function ²³ and early markers of DKD progression. ^34,35 Higher serum PTEN^K27polyUb levels were correlated with podocyte measures, including lower fractional podocyte volume and number density of podocytes per glomerular volume, which are also associated with DKD progression.³⁶ Our findings are in line with prior work in animals and in vitro studies that demonstrated a role for PTEN^K27polyUb in EMT, ^10,11 which is one mechanism for kidney fibrosis.

The PTEN gene is a tumor suppressor gene ^{5, 6} which is expressed throughout the body. ³⁷. PTEN is a lipid and protein phosphatase which principally dephosphorylates phosphatidylinositol 3,4,5 trisphosphate (PIP₃), thus, inhibiting the phosphoinositol 3-kinase (PI3K)/protein kinase B (Akt) pathway which is vital to cell growth and cell cycle regulation. ³⁸ Intracellular PTEN is regulated by changes in gene expression and protein modifications. ⁹ We focused on a K27-linked polyubiquitinated form of PTEN, PTEN^K27polyUb, first studied by Lin and colleagues. ^10,11 PTEN^K27polyUb is formed by polyubiquitination at the lysine 80 position, catalyzed by the E3 ubiquitin-protein ligase MEX3C. This form of PTEN exhibits reduced enzymatic activities in dephosphorylating PIP₃ and is primarily a serine and threonine phosphatase. ¹⁰ As such it dephosphorylates substrates with serine/threonine phosphorylation, including some with a role in EMT. ¹⁰ Mice genetically predisposed to kidney disease and homozygous for a genetic variant of PTEN resistant to the formation of PTEN^K27polyUb (PTEN^K80R/K80R) had less severe kidney disease than mice that were heterozygous for the variant (PTEN^WT/K80R). The kidneys of the homozygotes also showed less glomerulosclerosis and fibrosis and lower expression of EMT markers such as alpha-smooth muscle actin and vimentin, than the PTEN^WT/K80R mice. ¹¹ Exposing human kidney cells in vitro to high glucose suppresses PTEN expression ^7,39,40 and increases PTEN^K27polyUb. ¹¹ PTEN^K27polyUb expression is also increased by exposure to cytokines including TGF-β¹¹ which is an important instigator of EMT.⁴¹

While human kidney cells can undergo EMT in vitro it is not clear whether they do so in vivo ^41–43 and, if they do, to what extent this contributes to progression of CKD.⁴⁴ Partial EMT, whereby epithelial cells express markers of mesenchymal cells and form fibrillar collagen without fully becoming myofibroblasts, may be more common than complete EMT. ^43,45

Studies of PTEN and PTEN^K27polyUb in kidney disease have primarily assessed intracellular levels whereas we measured serum PTEN^K27polyUb. We do not know whether serum PTEN^K27polyUb reflects kidney PTEN^K27polyUb. Serum PTEN^K27polyUb concentrations are higher in people with diabetes compared to people without diabetes and higher still in people with DKD. ¹¹ We found that serum PTEN^K27polyUb was strongly correlated with kidney function as assessed by mGFR and ACR, which is in keeping with an earlier study in a Caucasian population. ¹¹ Of note the correlation we report with GFR is weaker than previously reported ¹¹ which may reflect that our cohort included fewer participants with advanced CKD.

We found significant correlations between serum PTEN^K27polyUb and a variety of clinically relevant morphometric measures including glomerular basement membrane width and mesangial fractional volume, both of which are associated with progression of DKD. ^34,46 This was observed despite the limited numbers of research kidney biopsies and the interval between measurement of serum PTEN^K27polyUb and the kidney biopsy. However, this provides important supportive data for the prospective associations between serum PTEN^K27polyUb and declining kidney function. Ideally, these findings will be replicated in a larger study with measurements of serum PTEN^K27polyUb at the time of biopsy.

Serum PTEN^K27polyUb was negatively correlated with podocyte fractional volume and the number density of podocytes per glomerulus. To date, there has not been any work on the actions of PTEN^K27polyUb in podocytes. However, podocytes can undergo EMT ⁴⁷ and so it may be that PTEN^K27polyUb has a similar effect in podocytes to its effects observed in kidney epithelial cells.¹¹ Inhibition of PTEN in mice resulted in actin cytoskeleton rearrangement within podocytes and the onset of albuminuria, with similar changes seen in cultured human podocytes. ³⁹ Insulin resistance in podocytes leads to albuminuria, ⁴⁸ and PTEN is a key mediator of insulin signaling in podoctyes via the PI3k/Akt pathway. Suppressing PTEN expression in mice podocytes improves podocyte insulin sensitivity.⁴⁹

Strengths of the study include the use of a well-characterized cohort of carefully phenotyped individuals with repeated measures of mGFR and long-term follow-up for kidney events. This allowed us to not only document the range of PTEN^K27polyUb in people with type 2 diabetes but to assess PTEN^K27polyUb as a possible biomarker for clinical outcomes, including the hard endpoints of kidney failure and death as well as the accepted surrogate endpoint for kidney failure of GFR loss ≥40%. The findings were consistent over a number of modelling strategies. We also extended our observations to include data on the association with morphometric measures of kidney, though in a limited subset with up to 6 years between serum measurement and research kidney biopsy. We report data for nine structural measures, albeit raising issues of multiple testing and viewed as exploratory, nonetheless provide supportive evidence for the primary findings relating to the development and progression of DKD. The study also has limitations. With only a single measure we could make no assessment regarding PTEN^K27polyUb level stability over time. Serum was stored at −80°C over 10.9 years prior to measurement and we lack data on the impact of storage on the measurements. However, we found no correlation between storage time and serum PTEN^K27polyUb over the 15-year range in sample storage times in this study. There are also a number of features of this study which may limit its generalizability. It is restricted to one ethnicity, includes more women than men, and most participants had normal or elevated mGFR at baseline. However, our cross-sectional findings are in keeping with previously published work in a Caucasian population with more advanced CKD, ¹¹ and more generally we have shown on multiple occasions that data regarding DKD from this population of American Indians is in keeping with findings in other populations ^{34, 50}. We are unable to explain the bimodal distribution of serum PTEN^K27polyUb which was not due to changes in assay methods, any of the demographic or clinical measures we examined or any variations in the genes coding for PTEN or ubiquitin.

In conclusion, we demonstrated that serum levels of a K27 polyubiquitinated form of PTEN are positively associated with the risk of kidney function decline and kidney failure. Prior experimental studies linking PTEN^K27polyUb with EMT suggest a potential mechanism for our observations and justify further investigation.

Supplementary Material

1

Figure S1: Schematic of participants by timing of baseline sample and length of follow-up

Figure S2: Illustrations of imputation for determining case status for ≥40% decline in GFR outcome and kidney failure outcome

Figure S3: Serum K27 polyubiquitinated PTEN and hazard rates for outcomes from models incorporating cubic splines

A. ≥40 decline in GFR outcome and univariate model

B. ≥40 decline in GFR outcome and fully adjusted model

C. Kidney failure univariate model

D. Kidney failure fully adjusted model

Figure S4: Association of serum K27 polyubiquitinated PTEN with kidney function

A. with glomerular filtration rate

B. with albumin:creatinine ratio

Figure S4: Serum K27 polyubiquitinated PTEN and hazard rate for all-cause mortality from models incorporating cubic splines

A. Univariate

B. Adjusted for age, sex and diabetes duration

C. Fully adjusted model

Item S1: Imputation of decline in measured GFR

Table S1: Genotypes for variants in genes relating to K27 polyubiquitinated PTEN production by high or low serum K27 polyubiquitinated PTEN

Table S2: Sensitivity analysis of association of K27 polyubiquitinated PTEN concentration with GFR loss without imputation

Table S3: Cause specific hazard rates for kidney failure in models with death as a competing risk

Table S4: Interval specific hazard rates for kidney failure

Table S5: Causes of death for participants who died before 2017

Table S6: Clinical characteristics for participants with kidney biopsy tissue within 6 years of serum PTEN^K27polyUb measurement