The study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Anhui Medical University and written informed consent was obtained from each individual. This study included 90 type 2 diabetes patients and 62 age- and sex-matched healthy controls (HCs). Clinical characteristics of the patients and controls are shown in Supplementary Table S1 . The inclusion criterion was more than 5 years of diabetes. Exclusion criteria included stage 3 and 4 chronic kidney disease, history of malignancy, acute diseases, immunosuppressive therapy, and chronic or acute hepatitis B or AIDS. In addition, female and male subjects aged under 40 or over 80 years were excluded. The diagnostic criteria for type 2 diabetes are based on the World Health Organization (WHO) diagnostic criteria (http://www.who.int/diabetes/publications/diagnosis_diabetes2006/en/).

Peripheral blood mononuclear cells (PBMCs) were separated by density gradient centrifugation using human peripheral blood lymphocyte isolation fluid. (TBD Science, Tianjin, China). For flow cytometry staining of PBMCs, the following anti-human monoclonal antibodies were obtained from BioLegend:CD3 (HIT3a, PerCP-Cy5.5), CD3 (HIT3a, FITC), CD56 (HCD56, BV510), CD56 (HCD56, PE-Cy7), CD45 (2D1, APC-Cy7), Tim-3 (F38-2E2, PE), NKG2A (S19004C, APC), TIGIT (A15153G, APC), NKG2D (1D11, APC), CD16 (3G8, APC), NKP46 (9E2, FITC), CD27 (M-T271, FITC), CD69 (FN50, PE-Cy7), CD14 (63D3, PE), Galectin-9 (9M1-3, PE-Cy7), CD4 (OKT4, APC-Cy7), LAG-3 (11C3C65, BV421), TNF-α (W19063E, PE), IFN-γ (4S.B3, APC), and CD107a (H4A3, PE-Cy7). Isotype control immunoglobulin G was used as control. Anti-CD66 (B1.1, FITC), anti-CD19 (HIB19, PerCP-Cy5.5), anti-Ki67 (B56, Alexa-Fluor 647) and anti-HLA-DR (L203.rMAb, PE) was purchased from BD Bioscience. PBMCs were incubated with antibodies at 4°C for 30 min. Cell staining was evaluated using a BD FACSCanto flow cytometer, and data analysis was performed with FlowJo software v10 (TreeStar, USA). For intracellular cytokine assays, isolated PBMCs were stimulated with medium containing 10% fetal bovine serum (FBS, Biosharp, China) in the presence of 50 ng/ml phorbol 12-myristate 13-acetate, 1 µg/ml ionomycin, and 10 µg/ml monensin (all from Sigma-Aldrich, USA) for 4 h. Following immunolabeling of surface molecules, cells were fixed, permeabilized, and labeled with specific cytokine-targeting antibodies. After washing off excess antibodies, signal detection was carried out by flow cytometry. The gating strategy is shown in Figure 1 .

pLVX-EF1α-IRES-puro vector containing human Galectin-9 and CD66a gene was generated by Azenta. Lentiviral supernatants were generated by co-transfecting HEK293T cells with pLVX-EF1α-IRES-puro, psPAX2 (VT1444, YouBio, Changsha, Hunan, China) and pMD2G (VT1443, YouBio, Changsha, Hunan, China) plasmids using Lipofectamine 3000 reagent (Invitrogen, Carlsbad, CA, USA). Culture supernatants collected 48 h after transfection were added directly to K562 cells plated in 6-well culture plates, and Polybrene (Sigma Aldrich, St. Louis, MO, USA) was added at a final concentration of 5 μ g/mL for 24 h. Transduced cells were selected by adding 5 μ g/mL puromycin (Invitrogen, Carlsbad, CA, USA). K562 cells transduced with Galectin-9 and CD66a were named K562-Galectin-9 and K562-CD66a, respectively. Transfection efficiency was determined by flow cytometry.

Apoptosis detection in NK cells was performed using an APC Annexin V Apoptosis Detection Kit (BioLegend). Cells were washed twice with PBS and resuspended in 1× binding buffer at a concentration of 1×10⁶ cells/ml. Then, a 100 µL sample (1×10⁵ cells) was transferred to a 5 ml test tube with 5 µL APC Annexin V and 7-AAD. The cells were then gently vortexed and incubated for 15 min in the dark at room temperature (25°C). Finally, 400 µL of 1× binding buffer was added to each tube before analysis by flow cytometry.

The level of Galectin-9 in the human serum was measured using Human Galectin-9 ELISA kit (Proteintech, USA) according to the manufacturer’s protocol.

Statistical analysis was performed using GraphPad Prism 9.0 software. Quantitative data were expressed as mean ± standard deviation (SD). Comparisons were made between two groups using an unpaired t-test or the Mann-Whitney U test, depending on whether group values conformed to a normal distribution. Pearson’s correlation test was used to identify correlations between cell receptor expression and biochemical variables. P<0.05 was considered significant.

To investigate whether NK cells are affected in T2DM patients, we recruited 90 patients with T2DM and 62 age- and sex-matched healthy volunteers and analyzed their PBMCs. Gating strategies for total lymphocytes and NK cells are shown in Supplementary Figure S1A . We found that the frequency and absolute number of total NK cells were significantly lower in patients with T2DM compared to healthy volunteers ( Figure 1B ). We further analyzed the distribution of the two NK cell subpopulations, i.e., CD56^dim and CD56^bright NK cells (see Figure 1A for details). In line with the observed changes in total NK cell numbers, the frequency and absolute number of CD56^dim and CD56^bright NK cells were also significantly lower in T2DM patients ( Figures 1C, D ). No significant differences between groups were noted in the analysis of CD3⁺ T lymphocytes and CD14⁺ monocytes ( Supplementary Figure S1B ).

To determine whether T2DM affects receptor expression on NK cells, we analyzed the expression of the inhibitory receptors Tim-3, NKG2A, LAG-3, and TIGIT, as well as the expression of the activating receptors NKG2D, NKp46, and CD69 on peripheral blood NK cells. We found that the mean fluorescence intensity (MFI) and the frequency of Tim-3⁺ and LAG-3⁺ NK cells were increased, while the expression of NKG2A and TIGIT was not affected ( Figures 2A, C and Supplementary Figure S2A ). In line with data from total NK cells, overexpression of both Tim-3 and LAG-3 was detected in both CD56^dim and CD56^bright NK cell subsets. MFI and frequency of NKG2D⁺ NK cells were decreased in T2DM patients, while activating receptor NKp46 expression was not affected ( Figures 2B, D , and Supplementary Figure S2B ). Similar expression patterns for these two markers were also recorded in CD56^dim and CD56^bright NK cell subsets. Indicating NK cell activation, the expression of CD69 was upregulated in samples from T2DM patients compared to controls ( Figure 2D ). We also analyzed the expression of CD27, a marker of NK cell differentiation and maturation (26). Results showed that CD27 expression was significantly higher in CD56^bright compared to CD56^dim NK cells in both diabetic and non-diabetic subjects ( Figure 2D and Supplementary Figure S2B ).

Among the inhibitory receptors assessed, Tim-3 was the most abundantly expressed on NK cells ( Figure 2C ). In turn, both MFI and percentage of Tim-3⁺ NK cells were higher in T2DM patients relative to controls. Consistent with previous studies (24), there was no significant difference in Tim-3 expression on CD3⁺ T cells from T2DM patients compared to healthy volunteers ( Supplementary Figure S2C ). These results suggest that peripheral blood NK cells in T2DM patients are under chronic stimulation and proceed into an immunosuppressed state, denoted by increased Tim-3 and decreased NKG2D expression.

To assess potential functional changes in NK cells in T2DM, we examined by flow cytometry their cytokine production capacity upon PMA/ionomycin stimulation (see Supplementary Figure S3A for gating details). As shown in Figure 3A , both the percentage and absolute number of TNF-α⁺ NK cells were significantly decreased in T2DM patients ( Figure 3B ). Moreover, NK cells from T2DM patients also exhibited impaired degranulation, as evidenced by decreased CD107a expression compared to healthy volunteers ( Figure 3C ). The percentage of IFN-γ⁺ NK cells was also decreased in T2DM patients, though this decrease did not achieve statistical significance, and the lower absolute number of IFN-γ⁺ NK cells may be attributed to their lower NK cells ( Figure 3D ). These data suggest that NK cells have reduced cytokine synthesis and are less likely to mediate cytotoxicity in patients with T2DM compared to healthy individuals.

Glycosylated hemoglobin (HbA1c) is a standard indicator for glycemic control and prediction of diabetes complications (27). Patients with poorly controlled T2DM have higher HbA1c and elevated fasting blood glucose (FBG) levels. Therefore, the correlation between NK cell receptor expression and HbA1c and FBG levels was analyzed in the two cohorts. As shown in Figure 4A , Tim-3 expression on NK cells from T2DM patients was positively correlated with both HbA1c and FBG levels (r=0.29, P=0.04, and r=0.30, P=0.03, respectively). Likewise, and suggestive of NK cell activation, a positive correlation between CD69 expression and both HbA1c and FBG levels were also detected in T2DM samples (r=0.37, P=0.008 and r=0.38, P=0.007, respectively; Figure 4B ). In contrast, no correlation between the expression of other NK cell receptors (NKG2A, LAG-3, TIGIT, NKG2D, NKp46, and CD27) and either HbA1c or FBG levels was observed in T2DM patient samples ( Figures 4C–H ). To further determine whether high concentrations of glucose affect the expression of Tim-3 on NK cells in vitro, PBMCs isolated from healthy donors were cultured in medium containing 5.5 mM, 25 mM, and 50 mM of glucose and Tim-3 expression was analyzed by flow cytometry. We found that high glucose treatment did not change the expression of Tim-3 directly ( Supplementary Figures S4A, B ).

Despite its characteristic abundance, the regulatory function of Tim-3 on NK cells in the microenvironment of T2DM is unclear. Interestingly, we found that in NK cells from T2DM patients the expression of Tim-3 was negatively correlated with both the percentage and absolute number of circulating NK cells (r=–0.38, P=0.007 and r=–0.32, P=0.02, respectively; Figure 5A ). Consistent with these data, we observed that the frequencies of Annexin V⁺ NK cells were higher in T2DM patients than in healthy donors, while the frequencies of proliferative (Ki67⁺) cells among two cohorts were similar ( Figure 5B ). Further analysis showed that Tim-3⁺ NK cells from T2DM patients showed a greater propensity to undergo apoptosis compared with Tim-3^- NK cells ( Figure 5C ). Overall, these results indicate that in poorly controlled T2DM, the expression of Tim-3 on NK cells is higher, which may cause NK cell dysfunction and apoptosis.

To determine whether Tim-3 overexpression impairs NK cell function, we compared the expression of NKG2A, TNF-α, and CD107a in Tim-3⁺ and Tim-3^- NK cells from T2DM patients. We found that NKG2A expression was higher in Tim-3⁺, compared to Tim-3^-, NK cells ( Figures 6A, B ). In addition, Tim-3⁺ NK cells produced significantly less TNF-α in response to PMA and ionomycin stimulation ( Figures 6A, C ) and showed also decreased CD107a production, although this change did not achieve statistical significance ( Figures 6A, D ). Furthermore, we found that Tim-3⁺ CD4⁺ T cells from T2DM patients have a much weaker capacity to produce TNF-α and IFN-γ after PMA/ionomycin induction ( Supplementary Figure S5A ). Importantly, a negative correlation was established between the percentage of TNF-α⁺ NK cells and that of Tim-3⁺ NK cells in patients with T2DM (r=–0.44, P=0.01; Figure 6E ). These data indicate that Tim-3 overexpression is correlated with NK cell exhaustion in patients with long-term T2DM.

It has been reported that Tim-3 exerts inhibitory function in T cells by interacting with Galectin-9, while the reports on CD66a/Tim-3 interaction remain conflicting (16, 28). No significant differences in Galectin-9 and CD66a/c/d/e expression were detected among monocytes, CD4⁺ T cells, CD8⁺ T cells, NK cells, NKT cells and dendritic cells from T2DM and control study participants ( Supplementary Figures S6A–D ). Consistent with previous studies (24, 29), we observed elevated serum Galectin-9 levels in patients with T2DM ( Supplementary Figure S6E ), suggesting the soluble Galectin-9 may negatively regulate the immune response via Galectin-9/Tim-3 pathway.

Then, we evaluated the influence of Galectin-9/Tim-3 or CD66a/Tim-3 interaction on NK cell functions using NK92 cell line. Mild expression of Tim-3 on NK92 cells were confirmed using flow cytometry. K562 cells, the commonly used NK target cells, rarely express Galectin-9 or CD66a/c/d/e. We constructed K562-Galectin-9 and K562-CD66a cells by overexpressing the respective receptors and expression of Galectin-9 and CD66a was confirmed with flow cytometry ( Supplementary Figure S7A ). K562, K562-Galectin-9 and K562-CD66a cells were subjected to NK92-mediated 4 hour cytotoxicity assay or 24h coculture assay. Overexpression of Galectin-9 on K562 cells significantly inhibited NK mediated cytotoxicity in 4h cytotoxicity assay as well as IFN-γ production in 24h coculture assay, while overexpression of CD66a has no impact ( Supplementary Figures S7B, C ). Overall, our results suggested that Galectin-9/Tim-3 interaction significantly inhibited NK cell function while CD66a/Tim-3 interaction had rare impact on NK cell function.

We further assessed the influence of Galectin-9/Tim-3 blockade on T2DM NK cells function. We found that overexpression of Galectin-9 on K562 cells had no impact on T2DM NK cells mediated cytotoxicity and blocking Galectin-9/Tim-3 interaction with antibodies did not affect T2DM NK cells mediated cytotoxicity against K562-Galectin-9 ( Figure 7A ). In the IL-12 and IL-18 stimulating assay, it was observed that the production of both TNF-α and IFN-γ were upregulated upon Tim-3 and Galectin-9 blockade, while the production of CD107a did not achieve statistical significance ( Figures 7B, C ). Our results suggested that Tim-3 and Galectin-9 blockade recovered IFN-γ and TNF-α production of T2DM NK cells but did not affect T2DM NK cell-mediated cytotoxicity.