T-Cell Expression and Release of Kidney Injury Molecule-1 in Response to Glucose Variations Initiates Kidney Injury in Early Diabetes

The presence of diabetic kidney disease (DKD) is the strongest predictor of both cardiovascular disease (CVD) and premature mortality in all individuals with diabetes (1–3). In particular, recent evidence highlights that premature mortality seen in individuals less than ∼45 years of age who develop diabetes at a younger age is not being alleviated by modern treatment regimens (4,5). In type 1 diabetes, a greater future risk of DKD and CVD may be evident as early as childhood and adolescence (6,7), detected as high-normal urinary albumin-to-creatinine ratio (uACR) preceding micro- and macroalbuminuria. This rise in uACR is seen in adolescents with type 1 diabetes, predicting some 85% of future case subjects with DKD (6).

The underlying reasons for early elevation in uACR in those at risk for kidney and CVD remain unclear, but dramatically influence mortality in adults <45 years of age with type 1 diabetes (8). Recent longitudinal studies from childhood suggest that uACR increases do not directly relate to traditional risk markers of vascular disease, such as HbA_1c in children and adolescents (6,7). HbA_1c does not necessarily capture fluctuations in blood glucose concentrations over the course of each day (glycemic variability), postulated as more potent activators of pathological pathways leading to DKD (9), although this remains hotly debated. Continuous blood glucose monitoring does show that individuals with diabetes average ∼50% of their time with glycemic variations outside ranges seen in healthy subjects (10), and those with greatest variability have increased atherosclerotic plaque formation (11). Indeed, there is some evidence that in the long term, glycemic variability is an independent risk factor for DKD progression (12–14).

This initial elevation in uACR also commonly precedes changes in other associated risk factors, including hypertension and dyslipidemia (15), but these become evident as kidney disease progresses. Further, early use of ACE inhibition and/or statin therapy, which are best practice in adults, have failed to significantly lower uACR in children and adolescents “at future risk” in a recent phase III clinical trial, the Adolescent Type 1 Diabetes Cardio-Renal Trial (AdDIT) (16). Another study has also shown that lowering blood pressure using renin-angiotensin system blockade in normotensive, normoalbuminuric individuals did not influence the progression of early kidney lesions in young adults with glomerular filtration rates >90 mL/min/1.73 m² (17).

Taken together, these previous findings suggest that other pathological factors may be at play early during the development of DKD. Circulating markers of inflammation such as hs-CRP in youth (7) and kidney injury molecule-1 (KIM-1 [18]) are elevated early in DKD progression. KIM-1, also known as hepatitis A virus cellular receptor 1 (HAVCR1) and T-cell Ig mucin 1 (TIM-1), is a transmembrane glycoprotein receptor primarily found on T cells (19) and significantly injured proximal tubules of the kidney (20,21). Activation of KIM-1 activates potent stimulatory signals for T-cell proliferation and cytokine production (22), and renal T-cell infiltration is reported in established DKD (23), sparking speculation that activation of T-cell activation, at least in murine models and type 2 diabetes, may play a major role in the initiation of DKD (24,25). Cleaved KIM-1 is also released into urine (26) and used as a biomarker for established chronic kidney disease (CKD) (18), but one study has shown that increases in circulating KIM-1 precede elevations in urinary KIM-1 and predict early progressive decline in kidney function in type 1 (18) and potentially type 2 diabetes (27). The source of circulating KIM-1 early in DKD remains to be elucidated.

Hence, the objective of this study was to examine if there was an early link among glucose variation, T-cell KIM-1 expression, and increased risk of kidney disease, which is examined both in individuals with type 1 diabetes and by using kidney biopsies from individuals with DKD.

A cross-sectional cohort of healthy adolescents/young adults (aged 15–25 years; N = 100) with diagnosed type 1 diabetes of >2 years’ duration were recruited during routine visits to the transitional diabetes clinic at the Mater Young Adult Health Centre. Exclusion criteria were: 1) previous diagnosis of micro/macroalbuminuria by uACR >2.5 mg/mmol in males and >3.5 mg/mmol in females (28) or DKD as defined by the Kidney Health Australia–Caring for Australasians with Renal Impairment algorithm for the early detection of CKD (29); 2) uncontrolled diabetes (defined as HbA_1c >9.5% [80 mmol/mol]) or more than two episodes of ketoacidosis in the preceding 12 months; 3) history of severe familial hypercholesterolemia; 4) previous myocardial infarction, stroke, or preexisting kidney disease; 5) pregnancy; 6) any existing medication other than insulin or hormonal contraceptive agents; 7) autoimmune diseases including uncontrolled celiac disease or Addison disease; 8) any congenital condition resulting in insulin-requiring diabetes; 9) diagnosed eating disorders; and 10) acutely unwell at recruitment visit.

This research was approved by the Human Research Ethics Committees of Mater Misericordiae Limited (HREC_15_MHS_35T1DM) and the University of Queensland (UQ 2015–000–958). Written informed consent was obtained from all participants and, if under 18 years of age, from their legal guardian in addition to the participant assent, prior to inclusion in the study. All investigation was conducted according to the principles of the Declaration of Helsinki.

Data obtained at recruitment included chronological age, sex, height, weight, medications, year of diabetes diagnosis, and resting seated mean systolic blood pressure. Historical HbA_1c, nonfasted plasma glucose, and uACR were collated from patient histories back to diabetes diagnosis if available. The mean observation periods for HbA_1c and plasma glucose were: low risk, 7.6 ± 2.7 years and high risk, 7.0 ± 3.4 years. For uACR, the mean observation period for the low-risk group was 5.0 ± 1.7 years and for the high-risk group 4.8 ± 2.5 years. HbA_1c, fructosamine albumin, serum creatinine, uACR, and total cholesterol were assessed in a centralized Mater Pathology laboratory. Urine albumin (range 2–4,400 mg/L) and creatinine were measured by a Cobas 6000 Analyzer (Roche/Hitachi, Indianapolis, IN) by turbidimetry or crep-2 (range 5–1,290 mg/dL), respectively. Quantikine ELISA kits (R&D Systems, Minneapolis, MN) were used for plasma cystatin C, KIM-1, tumor necrosis factor receptor-2, and fibroblast growth factor-21 according to the manufacturer’s instructions.

Estimated glomerular filtration rate (eGFR; in milliliters per minute per 1.73 m²) was calculated from plasma creatinine using the Chronic Kidney Disease Epidemiology Collaboration eGFR (eGFR_CKD-EPI) equation, eGFR = 133 × min (S_Cr/0.8, 1)^−0.49 × max (S_Cr/0.8, 1)^−1.328 × 0.996^Age × 0.932 (if female), and by the modified Schwartz equation, eGFR_SCHWARTZ = 42 × height (centimeters)/S_Cr (micromoles per liter), for all participants irrespective of age given their previous discordance in young adults (30). Hyperfiltration was defined if the average of eGFR_CKD-EPI and eGFR_SCHWARTZ was >95th percentile for a healthy age-matched population (31). In all individuals, the second morning urine void was collected and averaged with the two previous consecutive clinic visits to calculate mean uACR values for DKD risk. Albuminuria was defined according to guidelines from the Australasian Proteinuria Consensus Working Group and the International Society of Pediatric and Adolescent Diabetes as uACR >2.5 mg/mmol in males and >3.5 mg/mmol in females (28,32).

Primary human renal proximal tubule epithelial cells (PCS400010; ATCC, Manassas, VA) at passages 3–5 in triplicate were exposed to 4% patient plasma added into RPMI/F12 media as previously described (33), with a final d-glucose concentration of 5 mmol/L and FCS concentration of 0.5% in the presence and absence of anti–KIM-1 antibody pretreatment (5 μg/mL; R&D Systems). Sodium–glucose cotransporter 2 (SGLT2) (5 μg/mL; ab37296, Abcam, Cambridge, MA) and collagen IV (10 μg/mL; AB769, Merck Millipore, Darmstadt, Germany) were quantified using flow cytometry or IN Carta Image Analysis Software (General Electric, Boston, MA), respectively.

Peripheral blood mononuclear cells (PBMC) were incubated with anti-CD3/CD28 MACSiBeads (4:1) and then stained with antibodies against KIM-1 (219211; R&D Systems), CD3 (OKT3; BioLegend, San Diego, CA), CD4 (RPA-T4; BD Horizon; BD Biosciences, Piscataway, NJ), CD8 (SK1; BioLegend), CD25 (2A3; BD Horizon), CD127 (HIL-7R-M21; BD Horizon), CD45RA (H100; BD Horizon), HLA-DR (L243; BioLegend), and FOXP3 (POCH101, RPA-T4, and BC96; eBioscience, Waltham, MA; fixed panel only). Dead cells were excluded using a Fixable Viability Stain (BD Horizon). Cells were characterized using BD LSR Fortessa X-20 (SORP; BD Biosciences) and analyzed using FlowJo (>99% regulatory T [T_reg] cell and conventional T [T_conv] cell purity). T_reg cells were visualized using an Amnis ImageStreamX (Luminex Corporation, Austin, TX).

Tissue sections were obtained from the healthy portion of a tumor nephrectomy (N = 4) or from patients with minimal change disease (N = 4), primary focal segmental glomerulosclerosis (FSGS; N = 4), or DKD (N = 4), following informed patient consent and approval by the Royal Brisbane and Women’s Hospital Human Research Ethics Committee (2002/011 and 2006/072). DKD was confirmed by a renal histopathologist by the presence of classical lesions (i.e., Kimmelstiel-Wilson nodules and mesangial matrix expansion). Pathological features of other renal diseases were not identified in renal tissue from donors with DKD. Frozen 7-μm human tissue sections were fixed with 25% ethanol/75% acetone at room temperature for 5 min, followed by a protein block with 10% donkey serum for 30 min. Sections were subsequently probed with primary antibodies against KIM-1 (goat polyclonal IgG; R&D Systems), aquaporin-1 (monoclonal mouse IgG1; Santa Cruz Biotechnology, Dallas, TX), and CD3 (rabbit polyclonal IgG; Agilent Technologies, Santa Clara, CA) at room temperature for 2 h. Fluorescent detection was obtained by secondary incubation with Alexa Fluor Plus 488 anti-goat IgG, Alexa Fluor Plus 555 anti-rabbit IgG, and Alexa Fluor Plus 647 anti-mouse IgG (all from Life Technologies, Grand Island, NY) at room temperature for 30 min. Nuclei were stained with DAPI (Sigma-Aldrich, St. Louis, MO). Slides were coverslipped in fluorescence mounting medium (Agilent Technologies), visualized using a Zeiss 780 NLO confocal microscope (Carl Zeiss, Hamburg, Germany), and quantified using ImageJ (v2.0.0; National Institutes of Health, Bethesda, MD) and QuPath (v0.2.0; University of Edinburgh, Edinburgh, U.K.).

Transient intermittent hyperglycemia in the absence of diabetes was induced by four intraperitoneal d-glucose bolus doses (2 g/kg) at 2-h intervals with sham control mice administered 0.9% saline as described previously (34). This regimen was repeated weekly from adolescence at week 6 of age for 10 weeks in male apolipoprotein E–deficient (ApoE^−/−; C57BL/6J background) mice prone to DKD and atherosclerosis. Immunofluorescence was performed using rabbit anti–Kim-1 antibody (10 μg/mL; ab47635; Abcam) at 4°C overnight and then secondary antibody donkey anti-rabbit IgG Highly Cross-Absorbed (0.5 μg/mL; AF568-A10042; Thermo Fisher Scientific) for 1 h at room temperature on fixed mouse tissue sections post–citrate buffer, pH 6.0, antigen retrieval. Sections were then washed, counterstained with DAPI (2 μg/mL), and visualized using an Olympus FV3000 confocal microscope. At the study end, circulating CD4⁺ and CD8⁺ T lymphocyte surface Kim-1 expression was assessed by a flow cytometry panel consisting of Ghost Dye 510 Viability dye (13–0870-T500; Tonbo Biosciences), Kim-1 (CD365; RMTI-4; BioLegend), CD3 (17A2; BioLegend), CD4 (RM4–5; BD Biosciences), and CD8b (Ly-3; YTS156.7.7; BioLegend). uACR was calculated following measurement using albumin ELISA (R&D Systems) and creatinine enzyme assay (Crystal Chem). Serum cystatin C was assayed by ELISA (R&D Systems). Renal cortical gene expression of Kim-1 (Havcr1), collagen IV (Col4a1), and SGLT2 (Slc5a2) was measured using quantitative PCR.

Diabetes was induced in male ApoE^−/− (C57BL/6J background) mice at 5 weeks of age using multiple low-dose intraperitoneal (i.p.) injections of the β-cell toxin streptozotocin (55 mg/kg) (35). All mice (N = 10) with plasma glucose concentrations of >15 mmol/L 1 week later at 6 weeks of age were included in the study. Mice were followed concurrently for 10 weeks from week 6 to 16 of life, with those given chronic glucose injections as described above until week 16 of life.

Data are shown as mean ± SD or median (interquartile range) unless otherwise stated. Normality testing (Shapiro-Wilk) was performed on all data prior to analysis. Parametric data were analyzed by one-way ANOVA with Tukey post hoc for multiple comparisons or Kruskal-Wallis with Dunn post hoc testing for low- and high-risk groups compared using Student t test or Mann-Whitney U. Univariate modeling with Holm correction was used to determine interdependence of variables for inclusion in eGFR linear models and to generate Pearson coefficients for mouse uACR/T-cell KIM-1 associations. General linear models in R were used to examine the relationships between uACR and eGFR surrogates or plasma KIM-1 with correction for covariates age, diabetes duration, sex, and HbA_1c. All statistical analyses were performed using GraphPad Prism (v8.2.1; GraphPad Software, San Diego, CA). P values ≤0.05 were considered statistically significant.

The data set from Heidkamp’s group analyzed during the current study is available in the Stemformatics repository (https://www.stemformatics.org). The data sets and resources generated and analyzed during this study are available from the corresponding author upon reasonable request.

The uACR was examined in a cohort of 100 youth with type 1 diabetes, with a mean age of 20.0 ± 2.8 years of age, a ratio of male to female patients of 54:46, HbA_1c of 66.1 ± 12.3 mmol/mol (8.2 ± 0.7%), BMI of 24.5 (5.3) kg/m², and a diabetes duration of 10.7 ± 5.2 years (Table 1). Those with previously diagnosed kidney disease were excluded from this cohort. The study population was plotted in order of increasing uACR (Fig. 1A) and divided by uACR tertiles (Table 1) aligning with low (uACR ≤0.66 mg/mmol; n = 33), medium (uACR 0.67–1.16; n = 33), or high risk for DKD (uACR ≥1.17; n = 34). The risk calculations were performed on uACRs collected over a mean of 3.6 ± 1.0 years for the low-risk group and 3.9 ± 0.9 years for the high-risk group with ∼90% of participants having three uACR measurements taken during that time. Despite those with previously diagnosed kidney disease being excluded from our study, 11% (∼13% of male and 8.7% of female patients) of young people with diabetes in our cohort had early clinical evidence of kidney disease when their past three uACR results were combined (microalbuminuria in Table 1). This small subgroup with microalbuminuria also had higher systolic blood pressure than other high-risk subjects without albuminuria, but not increased lipids (Table 1).

Kidney function using eGFR showed that many individuals with diabetes in our human cohort had significantly increased glomerular filtration (hyperfiltration). This was most pronounced in the high-risk group in which 52.9% of youth had hyperfiltration compared with 30.3% of youth in both the low- and medium-risk groups (Table 1 and Supplementary Fig. 1A–C). The relationship between DKD risk factors log uACR and eGFR corrected for age, sex, diabetes duration, and HbA_1c,was examined using both adult eGFR_CKD-EPI (Fig. 1B) and pediatric eGFR_SCHWARTZ (Fig. 1C) equations and the eGFR surrogate serum cystatin C (Supplementary Fig. 1C). eGFR consistently showed a strong positive relationship with log uACR in individuals at low and medium risk for DKD with adjustment for age, diabetes duration, sex and HbA_1c, but this was lost in high-risk individuals, evident using either the adult (eGFR_CKD-EPI; low, r = 0.53, P_adj = 0.002; high, r = −0.09, P_adj = 0.63) (Fig. 1B) or pediatric eGFR (eGFR_SCHWARTZ; low, r = 0.58, P_adj = 0.0007; high, r = −0.28, P_adj = 0.12) (Fig. 1C) formulas. In high-risk individuals, lower eGFR was seen in the context of greater log uACR, reminiscent of changes seen in early CKD (Fig. 1B and C and Supplementary Table 1). This was also seen when the relationship between eGFR by the surrogate cystatin C and log uACR was examined (Supplementary Fig. 1D).

Diabetes duration did not differ between risk groups (Table 1). uACR was elevated at or soon after diabetes diagnosis in those individuals at high compared with low risk for DKD (Fig. 1D) (area under the curve [AUC], low vs. high risk, P < 0.01) (Fig. 1G). Historically, both the long-term marker of glucose control HbA_1c and point of care random blood glucose concentrations did not differ among risk groups for DKD (Fig. 1E and F and as AUC in Supplementary Fig. 1E and F). However, at recruitment (∼10 years after diabetes diagnosis), poorer glucose control as demonstrated by higher HbA_1c (Fig. 1H) and fructosamine albumin (Fig. 1I) and greater glucose variability using 1,5 anhydroglucitol (36) (Fig. 1J) were evident in high- compared with low-risk subjects for DKD. If 1,5 anhydroglucitol is <2, as seen in the high-risk group (Fig. 1J), this translates to mean average daily glucose excursions >16.1 mmol/L (290 mg/dL).

Youth at high risk of DKD had elevated plasma KIM-1 concentrations (Fig. 2A). Other biomarkers of early kidney injury did not differ among DKD risk groups, including plasma tumor necrosis factor receptor-2, fibroblast growth factor-21, and neutrophil gelatinase-associated lipocalin (Supplementary Fig. 2A–C). Plasma KIM-1 concentrations also increased across risk groups (Fig. 2A) (P_trend = 0.0382) and were positively associated with log uACR (Fig. 2B) in both medium- (r = 0.11; P_adj = 0.002) and high-risk individuals (r = 0.28; P_adj = 0.0003).

When healthy human kidney cells (PTECs) were exposed to 4% plasma from high-risk youth, they showed damage akin to DKD, including greater production of the fibrosis marker collagen IV (Fig. 2C and E-H) and cell surface expression of SGLT2 (Fig. 2D and E–H and Supplementary Fig. 2D and E), but viability was unaffected (Supplementary Fig. 2F and G). These DKD-like changes were not seen when PTECs were exposed to plasma from low-risk individuals pairwise matched for age, sex, diabetes duration, BMI, random blood glucose, and HbA_1c (Supplementary Fig. 2I). Further, preincubation with anti–KIM-1–neutralizing antibodies alleviated the increases seen in collagen IV (Fig. 2C and E–H) and SGLT2 (Fig. 2D and E–H and Supplementary Fig. 2D and E) following exposure to high-risk plasma, suggesting that plasma KIM-1 was responsible. Urinary KIM-1 was not different between DKD risk groups (Fig. 2J).

Glucose variability in the absence of diabetes was compared with type 1 diabetes (Fig. 3A) to examine if this was sufficient to elevate DKD risk measured as uACR and Kim-1. Adolescent ApoE^−/− mice (6 weeks of age) who were susceptible to kidney and CVD received four i.p. injections of glucose (2 g/kg) or isovolumetric saline, delivered 2 h apart (Fig. 3A and B), achieving plasma glucose variations peaking between ∼15 and 20 mmol/L (Fig. 3B). No changes in body or kidney weight, random blood glucose (Supplementary Fig. 3), or HbA_1c concentrations (Fig. 3C) were seen in saline- or glucose-injected groups, but as expected, HbA_1c was elevated in the diabetes group by week 16 of life (Fig. 3C). uACR, which defined DKD risk in our human cohort, was also elevated by glucose variations in the absence of diabetes by week 16 (Fig. 3D). As expected, diabetes induced hyperfiltration, seen by decreases in the GFR surrogate plasma cystatin C (Supplementary Fig. 3) and structural injury seen as increased kidney collagen IV and Kim-1 expression compared with saline control mice (Supplementary Fig. 3).

Kidney Kim-1 protein expression in the group given chronic glucose injections was not increased (Supplementary Fig. 3) but was increased on the cell surface of circulating T cells (Fig. 3E and F and gating strategy in Supplementary Fig. 4). Kim-1 was particularly increased in CD8⁺ T cells (Fig. 3F and H), and this was strongly associated with uACR in the glucose-injected group (Fig. 3G). There was no relationship between CD4⁺ T-cell Kim-1 expression and uACR in any group (Supplementary Fig. 3).

Using the data set from Heidkamp et al. (37) acquired from healthy human donors, we showed that peripheral blood cell expression of the KIM-1 gene HAVCR1 is highest in CD8⁺ T cells but is also present in CD4⁺ T cells and various dendritic cell subsets (Supplementary Fig. 5).

In our young cohort with diabetes, we found no differences in KIM-1 expression on circulating mononuclear cells (PBMC) that were CD3⁻ (non-T cells) between individuals with different risk for DKD (gating strategy in Supplementary Figs. 6 and 7A–C). However, when just T cells were studied, greater cell surface KIM-1 expression in individuals at high risk for DKD compared with low risk was evident (CD3⁺) (Supplementary Fig. 7B). T-cell KIM-1 was greatest on CD8⁺ T cells when compared with CD4⁺ T cells in youth with type 1 diabetes (Fig. 4A and B). With further characterization of T_conv subsets based on CD45RA and HLA-DR cell surface expression, youth at high risk for DKD had greater cell surface KIM-1 expression on naive T_conv cell subsets (CD45RA⁺HLA-DR⁻) in both CD8⁺ (Fig. 4C) and CD4⁺ (Supplementary Fig. 7G) compartments. KIM-1 expression was also increased in CD8⁺ T_conv double-positive (CD45RA⁺HLA-DR⁺) cells, but not any other CD8⁺ (Supplementary Figure 7D–F) or CD4⁺ T_conv subsets (Supplementary Figure 7G–I).

KIM-1 was more highly expressed by T_reg compared with T_conv cells, predominantly in the CD8⁺ T_reg cell population (Fig. 4D). KIM-1 was further increased in both CD8⁺ and CD4⁺ T_reg cells in youth at greater risk for DKD, in whom increased frequencies of KIM-1⁺CD8⁺ T_reg cells were present in high-risk youth as compared with those at low risk for DKD (Fig. 4D and E). T_reg cells (either CD4⁺ or CD8⁺CD25⁺CD127^lo/−) each had >90% FOXP3 expression (Supplementary Fig. 8A–D). As for T_conv cells, KIM-1 was predominantly expressed by naive T_reg cells in both the CD8⁺ (Fig. 4F) and CD4⁺ (Supplementary Fig. 8H) compartments and was significantly elevated in youth with high DKD risk. Coexpression of cell surface KIM-1 and intracellular FOXP3 by naive CD8⁺ T_reg cells was confirmed by single-cell imaging (Fig. 4G). CD8⁺ double-positive (CD45RA⁺HLA-DR⁺) T_reg cells also had higher KIM-1 expression in youth at high risk for DKD as compared with low risk (Supplementary Fig. 8G). Other circulating T_reg subsets did not differ in KIM-1 expression according to DKD risk (Supplementary Fig. 8E–J). For all T-cell subsets examined, there were no differences in frequencies between high- and low-risk subjects (Supplementary Fig. 9A–X).

Kidney tissue taken from individuals with impairment in kidney function (Supplementary Table 1) (i.e., those with either primary FSGS or DKD) had increased proximal tubule KIM-1 when compared with healthy kidney or kidney tissue taken from subjects with minimal change nephropathy (Fig. 5A–D). Kidney tissue from individuals with FSGS or DKD had greater T-cell infiltration (Fig. 5E), consistent with previous reports (38,39). However, the proportion of T cells expressing KIM-1 was significantly greater in those tissue donors with established DKD when compared with all other groups (Fig. 5F), including those with FSGS who had a similar degree of functional impairment (Supplementary Table 1).

Historically, strict glucose control shows a legacy effect to decrease the incidence of chronic diabetes complications (40,41), but this has been challenged by more recent large-scale clinical trials (42,43). Further, youth with diabetes often cannot achieve the strict clinical targets required to reduce their risk for DKD and CVD (44) due to factors including puberty, insulin resistance, and physical growth. This has shifted the focus to blood glucose fluctuations, which are now clearly evident with continuous blood glucose monitoring and show a strong relationship with risk for DKD (45) and mortality (11). Certainly, in the current study, youth at high risk for DKD had increased glucose variability. In both youth at higher risk for DKD and in susceptible mice with glucose fluctuations or diabetes, glucose variability was associated with elevations in uACR and increased expression of KIM-1 on circulating T cells. Further, T-cell KIM-1 was a likely source of measured increases in plasma KIM-1, an early biomarker of progressive DKD in previous studies in high-risk youth (18,27,46). In the future, it would be interesting to track variability in blood glucose in real time using continuous blood glucose monitoring and examine acute and chronic effects on KIM-1 expression in T cells. This would allow better ascertainment of causality in a clinical context.

Glucose variability in the absence of diabetes can activate innate immune cells, resulting in atherosclerosis (34,47). In type 2 diabetes, dynamic modulation of T-cell subsets during oral glucose tolerance testing, irrespective of baseline glucose tolerance, has also been shown (48). In the current study, glucose variations stimulated the expression of the immunomodulatory molecule KIM-1 on circulating T cells in youth at high risk for DKD. While DKD is not considered as a primarily immune-mediated form of CKD (49), extensive evidence does support involvement of many components of the immune system in DKD progression (23). Certainly, CD4⁺ and CD8⁺ T cells infiltrate the kidney parenchyma in diabetes (24), and modest associations between urinary albumin excretion and fewer T_reg cells, commonly beneficial in disease, have been demonstrated in type 2 diabetes (25). Dysfunctional CD8⁺ cytotoxic T cells (part of the T_conv subset) are definitely seen in youth with type 1 diabetes (50,51), but also in obese individuals with glucose intolerance at risk for kidney and CVD (52). In our study, there were fewer circulating KIM-1⁺CD4⁺ T cells, but a greater proportion of CD8⁺ T cells in youth at high risk for DKD, the majority of which were naive T_reg cells. Given the heterogeneity of T_reg cells, a more detailed study of their functionality in populations at risk for DKD is warranted. Interestingly, in a preclinical model of diabetes, short-term interruption of T_reg cell function exacerbates kidney damage (53). This is consistent with other contexts, in which interruption of CD25 either via loss-of-function mutations in humans (54) or via specific knockout in humanized mice (55) resulted in loss of self-tolerance/autoimmunity and enhanced proliferation of CD8⁺ cells, as seen in the current study in high DKD risk youth. These findings might suggest that glucose variability in high-risk youth is the result of persistent autoimmunity and that CD8⁺ T_reg cells adaptively increase in an attempt to combat this. The greater proportion of KIM-1⁺ naive CD8⁺ T_reg cells to both activated or memory CD8⁺ T_reg cells could also allude to a functional inability of this pathway to regulate and suppress autoimmunity or chronic inflammation, supporting this postulate. Although beyond the scope of the current study, the reasons for this compensatory or alternatively dysfunctional CD8⁺ T_reg cell response in those more susceptible to DKD should be examined in the future, since previous studies show that particular MHC haplotypes that associate with autoimmunity also confer greater risk for developing vascular complications in diabetes (56).

Circulating concentrations of the immunomodulatory molecule KIM-1 were also elevated in the plasma of subjects at high risk of DKD, which increased collagen IV and SGLT2 expression when exposed to healthy kidney cells (renal PTECs), and this was prevented with KIM-1 blockade. KIM-1 is hypothesized to act as a double-edged sword in the injured kidney (57), where it acutely facilitates repair, removing cell debris from tubule lumens, and elicits anti-inflammatory antigen presentation to control immune responses (58). By contrast in CKD, persistent kidney KIM-1 expression stimulates inflammation and fibrosis, eventually spilling over into the urine (21). In this study, supported by our clinical and murine findings, we postulate that increases in circulating KIM-1 and T-cell KIM-1 expression precede changes in kidney or urinary KIM-1, which are known to occur later in DKD (59). In agreement, two previous studies in individuals with type 1 diabetes also demonstrated that circulating soluble KIM-1 was elevated prior to urinary KIM-1 concentrations (18,27,46). However, the source of these circulating increases in KIM-1 in diabetes has remained elusive. As supported by data in the current study, we hypothesize that KIM-1 is released into the circulation by T cells in response to glucose fluctuations. Certainly, neutralization of soluble KIM-1 in plasma from high-risk youth alleviated subsequent injury upon exposure to human primary PTECs. This does not, though, prove that the source of plasma KIM-1 was from T cells, although this is highly likely given that only T cells and injured kidney cells highly express this molecule. It is also possible that soluble KIM-1 from the bloodstream and produced locally by infiltrating T cells, as seen in our kidney biopsies, each compete with kidney cell KIM-1 for ligands (e.g., phosphatidylserine and TIM-4), thereby inhibiting local anti-inflammatory responses. Indeed, a recent study indicated that surprisingly, healthy kidneys had a high proportion of T cells in which physiological function remains relatively unknown (60). It would be interesting in the future to examine what specific T-cell subsets are present in healthy kidneys and if they express KIM-1 under specific conditions. If local or infiltrating kidney T cells chronically express and release soluble KIM-1 in response to glucose variations, this might interfere with the ability of PTECs to phagocytose other damaged kidney cells, blocking a known protective response by sequestering self-antigens from presentation to the immune system. PTECs can also directly present antigens to the immune system as seen in other auto-immune kidney diseases such as crescentic glomerulonephritis, in which KIM-1 is known to play a role (61). This might explain the paradox seen between the seemingly protective KIM-1⁺CD8⁺ and CD4⁺ regulatory profile and the increases in uACR, indicating changes in kidney function in our high-risk youth.

There were of course, limitations to our research. The first is that we do not have prospective data to indicate whether the youth categorized as high risk go on to develop overt DKD, since these young adults were between 20 and 29 years of age at recruitment. However, there is strong evidence from previous prospective studies that >85% of those youth with type 1 diabetes in the upper tertile of uACR do develop DKD. This lack of prospective follow-up is offset somewhat by our historical uACR data, in which persistent elevation in uACRs was seen from close to diabetes diagnosis until recruitment (with a mean observation period of ∼5 years), in youth at high risk for DKD. We also did not measure autoantibodies at the time of recruitment to assess autoimmunity, given that our youth had type 1 diabetes for 10 years on average, but maturity-onset diabetes of the young had been previously excluded. In addition, we did not examine the functionality of the KIM-1⁺ T_conv and T_reg cells, which were increased in youth at risk for DKD, to see if this aligned with either autoimmunity or uACR. The kidney biopsies studied were also from older individuals with established DKD and represent a later stage of disease than present in our youth with type 1 diabetes. Therefore, future studies should examine T-cell infiltration and KIM-1 expression in early DKD in kidney biopsies such as those previously studied by Caramori et al. (62). However, many pathological pathways to DKD appear common to type 1 and type 2 diabetes (49,63). Reverse causality must also be considered for the association between increased KIM-1 in T cells and high risk for DKD in patients, but this is offset somewhat by the murine and PTEC studies. Despite these limitations, we have presented evidence of a novel pathway, by which kidney injury seen as increased uACR can be initiated via recruitment and expansion of KIM-1⁺ T cells as the result of glucose variations in early diabetes.

The data presented in this study suggest that glycemic variations align with risk for DKD in diabetes and increased CD8⁺ T-cell production of KIM-1. This highlights other pathways for consideration when designing therapies for the prevention of DKD and highlights the clinical importance of considering glycemic variability rather than purely relying on HbA_1c.

Acknowledgments. The authors thank the outstanding Mater Young Adult Health Centre diabetes clinic nursing and administration staff as well as the flow cytometry and microscopy core facility staff at the Translational Research Institute (Brisbane, Queensland, Australia).

Funding. This work was supported by the National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases (DIACOMP 25034-61), Kidney Health Australia, National Health and Medical Research Council of Australia (GNT1102935 and GNT1160428), and the Mater Foundation (Equity Trustees and the Laurie Gertrude McCallum Estate and George Weaber Foundation Trusts). A.K.F. was supported by an Australian Government Research Training Program Scholarship. The Translational Research Institute is supported by a grant from the Australian Government.

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

Author Contributions. J.M.F. designed the experiments, analyzed and interpreted data, prepared the manuscript, and gained funding. D.A.M., A.J.K., A.K.F., K.T.K.G., A.G., M.A.S., R.W., K.J.R., N.F., and S.S.B. completed experiments including the ELISAs, flow cytometry, and PTEC experiments, analyzed and interpreted data, and assisted in manuscript preparation. D.A.M., A.J.M., M.C.F., and P.C. completed the mouse experiments. T.B., N.D’S., J.N., A.M., S.T., H.L.B., and T.O’M.-S. assisted type 1 diabetes clinical cohort design and collation. A.R., N.I., T.J., J.C., H.H., M.H., K.D., D.W.J., A.C., and H.L.B. assisted in study conceptualization and design and gaining financial support. All authors edited and approved the final manuscript. J.M.F. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented at the Australasian Diabetes Congress 2020, 11–13 November 2020.