Hyperglycemia and Not Hyperinsulinemia Mediates Diabetes-Induced Memory CD8 T-Cell Dysfunction

With an increasingly obese population, type 2 diabetes (T2D) and its comorbidities are a progressively large health problem. Modern antidiabetes drugs, therefore, aim not only to lower blood glucose levels but also to diminish risks of key complications of T2D, such as cardiovascular disease and chronic kidney disease. An understudied comorbidity of T2D is reduced immune function. People with T2D are well known to have more frequent infections, such as urinary tract infection, surgical site infection, and skin infections. Moreover, patients with T2D also have a longer disease duration and typically have more severe disease upon infection (1–3). Reduced immune function was never prioritized as a target of therapeutic intervention, even though this complication may lead to life-threatening disease, such as gangrene of diabetic foot, pneumonia, and severe influenza infection (1,3). Notably, in infection with severe acute respiratory syndrome coronavirus 2, T2D gives a strongly increased risk of developing severe disease, which is particularly true for people with poor glycemic control (4,5). As such, it is of acute interest to uncover exactly how T2D weakens immune function and how this is affected by antidiabetes therapy.

Blood glucose levels have a direct impact on immune cells, as hyperglycemia was shown to negatively regulate immune functionality (1,6). Nevertheless, glycemia does not appear to be the only factor mediating immune dysfunction in T2D. Whereas immune function has never been a primary outcome of studies that measured the impact of antidiabetes drugs on study populations, several research initiatives have included this parameter. This revealed that there are great differences in the ability of antidiabetes drugs to lower the risk of infection. A nested case-control study in Taiwan of newly diagnosed patients with T2D started on oral antidiabetes drugs showed that use of metformin significantly reduced the risk of sepsis (odds ratio [OR] 0.80, 95% CI 0.77–0.83), as did thiazolidinedione (OR 0.83, 95% CI 0.73–0.94), whereas meglitinide increased this risk (OR 1.32, 95% CI 1.25–1.40) (7). These findings were confirmed in further studies (8). The impact of newer drugs, such as incretin mimetics or sodium–glucose cotransporter 2 inhibitors, on the risk of infection has not been extensively studied, even though no causal relationship between their use and more serious infections was found (9,10). Surprisingly, whereas insulin is one of the most potent glucose-lowering drugs, its use appears to increase risk of infection. In a Danish cohort study, use of insulin was associated with a significantly increased risk (hazard risk 1.63, 95% CI 1.54–1.72) of hospital-treated infection (10). A recent retrospective study revealed that insulin treatment of patients with T2D was associated with increased mortality (hazard ratio 5.38, 95% CI 2.75–10.54) following infection with severe acute respiratory syndrome coronavirus 2 (11). These findings imply that insulin may have a negative impact on immune function, especially in the context of T2D.

Endocrine hormones can have a profound impact on immune cell function, which can be both stimulating and inhibitory. Many immune cells express receptors for hormones such as leptin and adiponectin (12). Leptin promotes cytotoxicity of immune cells, such as CD8 T cells and natural killer (NK) cells, and enhances their ability to proliferate and produce cytokines, whereas adiponectin has the opposite effect (13,14). We and others recently showed that CD8 T cells express the receptor for insulin and that this hormone promotes their antiviral response (15,16). In addition to hyperglycemia, T2D is often associated with chronic hyperinsulinemia, especially early after diagnosis. However, how insulin and hyperglycemia impact the memory CD8 T-cell response is currently unknown.

We hypothesized that in context of T2D, memory CD8 T cells become refractory to insulin, which reduces their functionality following viral infection. Alternatively, hyperglycemia may directly impair memory CD8 T-cell function. We therefore investigated the importance of hyperglycemia and insulin receptor signaling on memory CD8 T cells in health and in the context of T2D. We find that an obese, hyperinsulinemic environment impairs the ability of memory CD8 T cells to control viral replication or tumor growth. Using CD8 T cell–specific deficiency of the insulin receptor, we could show that direct insulin signaling on these cells is redundant for their functionality, both in lean subjects and in the context of T2D. Instead, we demonstrate that hyperglycemia has a direct detrimental impact on cytokine production by memory CD8 T cells and in a reduced capacity to lower viral titers upon reinfection. Thus, memory CD8 T-cell dysfunction in diabetes is independent of direct insulin signaling to these cells but negatively impacted by hyperglycemia. Our findings indicate that blood glucose–lowering therapy, including insulin administration, will have a beneficial impact on the protection of people with T2D against viral infection.

Mice were strictly age and sex matched within experiments and handled in accordance with institutional and national guidelines. Male mice were used, unless stated otherwise. All mice were housed and bred under specific pathogen-free conditions at the animal facility of the Medical Faculty, University of Rijeka. Wild-type (WT) C57BL/6J (B6; strain 664), B6 Ly5.1 (2014), OT-1 (3831), and IR^Lox (6955) mice were obtained from the Jackson Laboratory. CD4cre mice were provided by D. Littman (New York, NY). All genetically modified animal models were generated on the C57BL/6J background or backcrossed at least 10 times with C57BL/6J mice. Male mice (8–12 weeks old) were fed ad libitum with a normal chow diet (NCD; ssniff) or with a high-fat diet (HFD) in which 50% of calories were derived from animal fat (Bregi). All animal experiments were approved by the National Committee for Welfare of Animals (no. UP/I-322-01/20-01/41).

CD8 T cells were purified by positive selection using magnetic beads (Miltenyi Biotec). Cells were cultured in RPMI 1640 medium (PAN-Biotech), supplemented with 10% FCS (PAN-Biotech) and 2-mercaptoethanol (Sigma-Aldrich). For in vitro memory differentiation, 30,000 cells per well in a U-bottom 96-well plate (Cellstar) were stimulated for 30 h with 1 ng/mL SIINFEKL (N4) peptide (Genscript) in the presence or absence of 0.5 mg/mL aCD28 (37.51; eBioscience) and in the presence or absence 1 unit/mL insulin (aspart; Novo Nordisk), washed, and cultured for an additional 5 days with 50 ng/mL interleukin (IL)-15 (PeproTech) and in the presence or absence of 1 unit /mL insulin (aspart, Novo Nordisk). For proliferation assays, cells were labeled with carboxyfluorescein succinimidyl ester (CFSE), and CFSE dilution was determined by flow cytometry after 72 h. For cytokine production, cells were restimulated with 10 ng/mL N4 peptide in the presence of brefeldin A. After 4 h, cytokines were measured by intracellular flow cytometry. For phosphorylation kinetics of S6, cells were rested for 2 h and stimulated with 1 unit/mL fast-working insulin (aspart, Novo Nordisk) for 0 and 15 min. Cells were fixed with 2% paraformaldehyde, permeabilized in 90% methanol, stained for pS6 (Cell Signaling), visualized with anti-rabbit phycoerythrin (eBioscience), and fold induction of S6 phosphorylation was determined by flow cytometry. For glucose uptake assay, cells were incubated with 2-NBDG (2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-seoxyglucose) (Cayman Europe 186689-07-6), a fluorescent glucose analog, for 10 min at 37°C. After the incubation, cells were washed, and 2-NBDG was quantified by flow cytometry. For hypo- and hyperglycemia experiments, cells were cultured in RPMI medium lacking glucose (PAN), supplemented with the indicated amounts of glucose.

Live CD8 T cells were sorted using the FACSAria III cell sorter (BD Biosciences) after culture. Cells were plated in XF medium consisting of nonbuffered RPMI 1640 (Sigma-Aldrich) supplemented with 2 mmol/L l-glutamine (Sigma-Aldrich) and 1 mmol/L sodium pyruvate (Sigma-Aldrich). Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured in response to 5 mmol/L d-(+)-glucose (Sigma-Aldrich), mitochondrial inhibitors, 1 μmol/L oligomycin, 1.5 μmol/L carbonyl cyanide-4 (trifluoromethoxy), phenylhydrazone (FCCP), and a mixture of 200 nmol/L rotenone/1 μmol/L antimycin A using a XFe96 Extracellular Flux Analyzer (Seahorse Bioscience).

For microarray analysis, 100 ng total RNA was amplified using the GeneChip 3′ IVT Plus Reagent kit (Thermo Fisher Scientific) generating biotinylated complementary RNA. The labeled samples were hybridized to GeneChip HT MG-430 PM arrays (Thermo Fisher Scientific). Washing, staining, and scanning were performed using the GeneTitan Wash, Stain Kit for 3′ IVT Array Plates and the GeneTitan Instrument (Thermo Fisher Scientific). Data were analyzed using the R2 Genomics analysis and visualization platform (https://r2.amc.nl). The microarray data have been deposited in the Gene Expression Omnibus database under code GSE102189.

For flow cytometry, single-cell suspensions of spleen and liver were prepared according to standard protocols. Flow cytometric analysis were performed by using anti-mouse mouse CD8b (53-6.7), CD62L (MEL-14), CD44 (IM7), CD127 (A7R34), KLRG1 (2F1), Fixable Viability Dye eFluor 780, CD45.2 (104), CD45.1 (A20), tumor necrosis factor (TNF) (MP6-XT22), interferon-γ (IFN-γ) (XMG1.2), IL-2 (JES6-5H4), and granzyme B (NGZB) from eBioscience, preceded by blocking of Fc receptors using 2.4G2 antibodies (in-house generated). To measure cytokine production, cells were stimulated with 10 ng/mL SIINFEKL (N4) peptide for 4 h in the presence of brefeldin A (10 mg/mL; eBioscience). MHC class I tetramers were provided by A. ten Brinke (Amsterdam, the Netherlands). For intracellular staining, permeabilization and fixation of cells was done with the Fix/Perm kit (BD Biosciences). All data were acquired using a FACSVerse (BD Biosciences) and analyzed using FlowJo software (Tree Star).

The bacterial artificial chromosome–derived murine cytomegalovirus (mCMV) strain MW97.01 (in-house produced) has previously been shown to be biologically equivalent to the mCMV Smith strain (VR-1399; ATCC) and is referred to as mCMV. mCMV-N4 was generated as described (17). Both mCMVs were propagated on mouse embryonic fibroblasts and purified by standard protocol. Animals were infected intravenously with 2 × 10⁵ plaque-forming units. Viral titers were determined on mouse embryonic fibroblasts by standard plaque assay. Lymphocytic choriomeningitis virus (LCMV) Armstrong strain (Armstrong E-350; ATCC) LCMV-N4 (18) were propagated on baby mouse kidney cells according to standard protocol. Animals were infected intraperitoneally with 10⁶ plaque-forming units. The B16-N4 cell line was provided by Tim Sparwasser (TWINCORE, Hannover, Germany). Depletion of NK cells was accomplished by a single intraperitoneal injection of 200 μg anti-NK1.1 clone PK136 (BioXcell). Cells were cultured in 10% DMEM (Lonza) supplemented with β-mercaptoethanol (PAN-Biotech) and under G418 selection (InvivoGen). Mice were rechallenged by 10⁵ B16-N4 cells intravenously. Prior to quantification of metastases, lungs were bleached in Fekete solution. To induce hyperglycemia, mice were injected with a mixture of alloxan (ALX) and streptozotocin (STZ) (both 100 μg/mouse; Sigma-Aldrich), 3 days and 1 day before infection and 2 weeks after infection. Hyperglycemia was confirmed using an automated blood glucometer.

Data are presented as mean ± SEM. Statistical significance was determined by two-tailed unpaired Student t test, Mann Whitney U test, or one-way ANOVA with Bonferroni correction using GraphPad Prism 5 software. A value of P > 0.05 was deemed not statistically significant (ns).

Microarray data have been deposited in the Gene Expression Omnibus database under code GSE102189. All other data generated or analyzed during this study are included in the published article. All unique resources generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

To investigate whether diabetes affects the antiviral CD8 T-cell response, we used the well-established diet-induced obesity model. In this model, mice receive food of which 60% of calories consists of animal fat (HFD). HFD-fed animals typically develop insulin resistance and glucose intolerance in 12 weeks (Fig. 1A) (19), and we therefore here refer to them as “diabetic.” To compensate for insulin resistance, animals develop hyperinsulinemia (Fig. 1B). After induction of diabetes, mice were transferred with 10⁴ OT-1 cells (CD45.1⁺), which are CD8 T cells that express a transgenic T-cell receptor specific for the SIINFEKL (N4) epitope of ovalbumin. This model allows for tracking of virus-specific cells. Next, animals were infected with mCMV expressing N4 (mCMV-N4) (17). At the acute phase of the antiviral T-cell response, on day 8 after infection, donor CD8 T cells were analyzed in spleen and liver. No differences were observed between NCD- and HFD-fed mice in both relative and absolute donor OT-1 cell numbers (Fig. 1C and D). We then investigated the frequency of memory precursor cells (MPECs; CD127⁺KLRG1⁻), short-lived effector cells (CD127⁻KLRG1⁺), and effector memory cells (CD44⁺CD62L⁻) within the donor cell population. None of these immune cell populations was affected by diabetes (Fig. 1E–G). Moreover, cytokine production was not different between OT-1 donor cells in NCD- and HFD-fed mice (Fig. 1H), indicating that functionality of these cells was not affected by the obese, hyperinsulinemic environment.

As OT-1 cells were only injected 1 day before infection, we argued that the exposure of these cells to the obese, hyperinsulinemic environment was possibly too brief to have a functional impact. However, the endogenous response to CMV was also not impaired in diabetic mice (Fig. 1I and Supplementary Fig. 1). In addition, the frequency of MPECs was not altered among recipient virus-specific CD8 T cells in spleen and liver (Fig. 1J). Finally, cytokine production by these cells was not impaired (Fig. 1K).

Thus, we conclude that the acute CD8 T-cell response is not affected in the diet-induced obesity model.

People with T2D are well documented to have recurrent infections (3). We therefore investigated the ability of CD8 T cells to form long-term protective memory in context of diabetes. OT-1 T cells were transferred to lean or obese mice and infected with mCMV-N4. Despite a slightly higher frequency of virus-specific donor cells in the blood of lean mice, the number of OT-1 cells at memory time points was not different from those observed in obese animals (Fig. 2A). As previously reported (20), the endogenous virus-specific T-cell response also did not differ between lean and obese animals (Fig. 2B). To investigate whether these cells were affected at a phenotypic level, we determined the number of cells with a CD127⁺KLRG1⁻ phenotype, which is associated with memory formation and of CD127⁻KLRG1⁺ cells, which represent terminal effectors (21). Again, we did not observe a difference in the expression of these immune cell subsets at any time point (Fig. 2C and D). Finally, we did not observe differences in the viral load between lean and obese mice upon primary infection (Supplementary Fig. 2A). Thus, memory CD8 T-cell formation is not affected by a diabetic environment.

To investigate the ability of memory cells to respond to virus reencounter, lean and obese mice that had received OT-1 cells and were primed with mCMV-N4 42 days prior, were reinfected with LCMV expressing N4. In this model, only N4-specific cells respond to the secondary infection, allowing for a detailed analysis of memory responses. Recall responses were analyzed after 7 days. As shown previously (20), the number of virus-specific CD8 T cells was not different between healthy and diabetic animals (Fig. 2E). In contrast, cells from obese animals produced significantly fewer cytokines upon in vitro restimulation (Fig. 2F). This was observed both after restimulation with viral peptides and with phorbol myristate acetate/ionomycin, indicating that the intrinsic capacity of these cells to produce cytokines is reduced, rather than upstream T-cell receptor signaling (Fig. 2F). This phenotype was observed both in male and female animals, excluding a major effect of sex hormones (Supplementary Fig. 2B). To determine the physiological relevance of reduced CD8 T-cell functionality, lean and obese animals were transferred with OT-1 cells and infected with LCMV-N4. After 30 days, mice were reinfected with mCMV-N4, and viral titers were determined in liver after 5 days. Indeed, obese mice had significantly higher viral titers than lean animals, which was the result of reduced OT-1 T-cell function (Fig. 2G).

To confirm these observations in a second model, lean and obese mice were transferred with OT-1 T cells and infected with mCMV-N4. After 30 days, mice were injected with B16 melanoma cells overexpressing the N4 peptide (18). Recall responses were analyzed 14 days later, and tumor metastases were quantified in the lung. As after viral reinfection, we did not observe a quantitative difference in the number of antigen-specific donor CD8 T cells upon tumor challenge (Fig. 2H). In contrast, the ability of CD8 T cells to produce cytokines was significantly impaired in rechallenged obese animals compared with lean controls (Fig. 2I). Accordingly, the number of metastases in the lung was significantly higher in obese animals compared with lean mice, which was dependent on impaired OT-1 T-cell function (Fig. 2J and Supplementary Fig. 2C). Chronic insulin receptor stimulation is known to reduce sensitivity to its ligand. We therefore investigated the ability of rechallenged memory CD8 T cells to respond to insulin. Indeed, OT-1 T cells from rechallenged obese mice showed significantly less pS6 induction than cells from lean animals upon insulin stimulation (Fig. 2K).

Whereas OT-1 cells are a powerful tool, they may not be representative of a physiological, polyclonal response. To confirm that the memory CD8 T-cell response is also altered in a nontransgenic model, lean and obese mice were infected with LCMV-N4. After 35 days, mice were reinfected with mCMV-N4, and 4 days later, the recall response against the N4 epitope was determined. As was observed for the OT-1 response, CD8 memory T cells of obese animals had a reduced capacity to produce cytokines upon in vitro restimulation (Fig. 2L).

In summary, obesity impairs the ability of activated memory CD8 T cells to produce cytokines, which is associated with a reduced ability to respond to virus reencounter or tumor challenge.

Our findings imply that an obese, hyperinsulinemic environment is detrimental for memory CD8 T-cell responses, which may depend on direct insulin signaling. As obesity impacts many endocrine signals, we first determined which hormone receptors are expressed on CD8 T cells. Total transcriptome analysis revealed that CD8 T cells express a number of receptors for endocrine signals, including the insulin receptor (Fig. 3A). We then applied a well-established model for in vitro generation of memory CD8 T cells (18,22) to investigate how activation impacts hormone receptor expression. We observed that the insulin receptor, as well as several other hormone receptors, was downregulated in these cells upon activation and again upregulated as memory differentiation progresses (Fig. 3B). Accordingly, CD8 T cells became refractory to insulin stimulation early after activation, but again were sensitive to this molecule upon memory differentiation (Fig. 3C).

These findings imply that insulin may affect memory CD8 T-cell function. As we and others have reported previously, we could show that insulin promotes acute cytokine production of effector cells after in vitro stimulation (Fig. 3D). To interrogate the impact of insulin on memory formation, OT-1 T cells were differentiated to memory cells in vitro in the presence or absence of short-acting insulin and analyzed by flow cytometry. Insulin regulates glucose uptake from the blood by inducing glucose transporter expression on insulin-sensitive organs. We therefore hypothesized that insulin stimulation impacts the ability of memory T cells to internalize glucose. We observed that insulin stimulation indeed increased expression of the glucose transporter GLUT-1 on these cells (Fig. 3E). Accordingly, insulin-treated memory cells were able to take up more glucose than cells cultured in the absence of this hormone, as determined by their ability to accumulate the fluorescent glucose analog 2-NBDG (Fig. 3F). However, insulin stimulation did not impact viability of these cells or their expression of key cell surface markers of memory formation (Fig. 3G and H). Importantly, insulin stimulation also did not affect their ability to produce cytokines or proliferate (Fig. 3I–K). To confirm these findings in vivo, OT-1 T cells were transferred to WT recipients, which were subsequently infected with mCMV-N4. In addition, animals were daily injected with long-acting (basal) insulin and analyzed 7 days after infection. We observed that donor cells expressed significantly higher levels of GLUT-1 in mice treated with insulin (Fig. 3L). However, their ability to generate memory precursors was not affected (Fig. 3M).

Thus, we conclude that insulin stimulation can increase GLUT-1 expression and glucose uptake in memory CD8 T cells. However, insulin-mediated glucose uptake does not further promote the functionality of memory CD8 T cells.

Under normal cell culture conditions, insulin is present in the media in bovine serum. Whereas additional insulin does not further increase memory CD8 T-cell function, this does not exclude a role for this molecule. To investigate this, we crossed CD4^Cre mice with animals with a floxed insulin receptor gene (IR^Flox) and an OT-1 allele. This allows deletion of the insulin receptor specifically on OT-1 transgenic CD8 T cells. First, we analyzed the impact of insulin receptor deficiency on memory formation in vitro. Purified CD45.1⁺ insulin receptor–deficient (CD4^CreInsR^FloxOT-1[IR^CKO]) and WT (CD45.1/2⁺) OT-1 CD8 T cells were mixed in an equal ratio, labeled with CFSE, and differentiated to memory cells in vitro.

Next, cells were analyzed by flow cytometry. Before the start of stimulation, IR^CKO CD8 T cells expressed equal amounts of GLUT-1 on their cell surface but had a reduced ability to upregulate expression of this molecule upon activation, which persisted into memory (Fig. 4A). Accordingly, IR^CKO cells had a lower rate of 2-NBDG uptake (Fig. 4B). In response to insulin, cells induce both glycolysis and oxidative phosphorylation. We therefore questioned whether deficiency of insulin receptor signaling impacts glucose utilization in effector and memory CD8 T cells. Purified WT and IR^CKO OT-1 cells were stimulated in vitro. Next, the OCR and ECAR of cells, which are indicative of oxidative phosphorylation and glycolysis, respectively, were analyzed on day 3 and day 8 after start of stimulation by extracellular flux analyzer. Measurement of OCR revealed no differences in basal respiration, maximal respiration, ATP production, or spare respiratory capacity on either day. Similarly, we found no differences in the glycolytic capacity of WT and IR^CKO OT-1 cells (Fig. 4C–F). In accordance with a lack of impact of insulin signaling on activated CD8 T-cell metabolism, we did not observe that WT cells have a competitive advantage over IR^CKO cells, as the ratio between these two populations did not change over time (Fig. 4G). Also, IR^CKO cells did not show differences in proliferation compared with WT cells (Fig. 4H). We then questioned whether insulin receptor deficiency impacts the phenotype and function of in vitro generated memory CD8 T cells. We found comparable levels of memory-associated markers CD127 and CD62L, as well as similar expression of CD44 on WT and IR^CKO cells memory CD8 T cells (Fig. 4I). Moreover, insulin receptor deficiency did not impair the ability of in vitro generated memory CD8 T cells to produce cytokines (Fig. 4J).

Standard culture medium of lymphocytes (RPMI 1640) contains relatively high levels of glucose (11.1 mmol/L). We argued that that the impact of insulin receptor signaling on memory formation was not observed because glucose concentrations are not limiting in standard culture media. We therefore mixed purified WT and IR^CKO OT-1 cells in a 1:1 ratio and differentiated them to memory cells in vitro under physiological (5 mmol/L) and low (1 mmol/L and 2.5 mmol/L) glucose concentrations in the presence or absence of exogenous insulin. However, we did not observe any differences in their ability to proliferate, produce cytokines, and acquire a memory phenotype (Supplementary Fig. 3A–C). In addition, we did not observe any changes in the ratio between the in vitro expanded WT and IR^CKO populations (Supplementary Fig. 3D), indicating that survival of these cells was also not affected under glucose-limiting conditions.

Thus, insulin receptor signaling is redundant for in vitro-generated CD8 memory T-cell formation, survival, proliferation, metabolism, and function.

Whereas in vitro models are valuable tools to elucidate the function of immune cells, they do not replicate the competitive environment that CD8 T cells encounter upon their activation in the body. Therefore, we investigated whether insulin receptor deficiency has an impact on CD8 T-cell memory formation and function in vivo. WT (CD45.1/2⁺) and IR^CKO (CD45.1⁺) OT-1 cells were mixed in equal ratios and subsequently transferred to WT (CD45.2⁺) recipients. Next, mice were infected with mCMV-N4, and responses were followed over time. We did not observe any difference in the number of WT and IR^CKO cells following primary infection at all time points measured (Fig. 5A). In addition, we did not observe a difference in the relative number of CD127⁺KLRG1⁻ memory cells or CD44⁺CD62L⁺ central memory cells (Fig. 5B and C). Finally, WT and IR^CKO cells showed an equal ability to produce cytokines upon in vitro restimulation (Fig. 5D).

To assess the recall capacity of memory cells, animals were reinfected with LCMV-N4 30 days after primary infection and analyzed 7 days later. We did not observe any difference between WT and IR^CKO cells in their ability to form secondary effector cells (Fig. 5A–C). Also, cytokine production was similar between the two donor cell populations (Fig. 5E). IR^CKO cells did have a reduced ability to take up fluorescent glucose, but this did not impact these cells on a functional level (Fig. 5F).

Thus, whereas insulin receptor signaling can promote glucose uptake in memory CD8 T cells, its role is redundant for normal CD8 memory T-cell formation and function.

Whereas insulin receptor signaling appears to be redundant for memory CD8 T-cell responses under physiological conditions, it is possible that it has a detrimental effect in the context of diabetes. To investigate this possibility, WT mice were fed the HFD for 12 weeks, after which they received WT and IR^CKO OT-1 cells mixed in a 1:1 ratio. Next, animals were infected with mCMV-N4, and donor cell expansion was followed over time. We did not observe any differences in the ratio between WT and IR^CKO cells at any time point after infection (Fig. 6A). We also did not observe any differences in the fraction of CD127⁺KLRG1⁻ memory cells or of CD44⁺CD62L⁺ central memory cells (Fig. 6B and C). Finally, we did not observe a difference in the capacity of WT and IR^CKO cells to produce IFN-γ, TNF, IL-2, or granzyme B (Fig. 6D). Mice were reinfected with LCMV-N4 39 days after the primary infection, and recall responses were analyzed 7 days later. No differences were observed in recall capacity or phenotype between WT and IR^CKO cells. Also, the ability to produce cytokines was comparable between donor cell populations (Fig. 6A–D).

To confirm our findings in a second model, WT mice were fed the HFD for 12 weeks, after which they received WT and IR^CKO OT-1 cells mixed in a 1:1 ratio. Next, animals were infected with LCMV-N4. After 28 days, mice were reinfected with mCMV-N4, and recall responses were analyzed 7 days later. We did not observe differences in the ability of donor cells to form secondary effector cells, nor did these cells display a different phenotype (Fig. 6E and F). In addition, the ability to produce cytokines upon in vitro restimulation was comparable between the two donor cell populations. Finally, WT donor cells had an increased ability to take up 2-NBDG, both within effector and MPECs, but this did not impact these cells functionally (Fig. 6G and H).

In summary, we conclude that insulin receptor signaling is redundant for memory CD8 T-cell formation and function. Moreover, a chronic hyperinsulinemic environment does not negatively impact the formation and function of CD8 T-cell memory.

Having shown that insulin does not impact memory CD8 T-cell biology, we hypothesized that hyperglycemia may directly impact cytokine production of these cells in the context of metabolic disease. To investigate this possibility, we generated memory CD8 T cells in vitro in the presence of low (<5 mmol/L), physiological (5 mmol/L), or high (>5 mmol/L) concentrations of glucose in the medium. As expected, when glucose levels were very low (i.e., not provided through media but only by FBS; here indicated as 0 mmol/L), cells failed to proliferate, and viability was low after memory formation (Supplementary Fig. 4A). As reported previously (23), increasing the glucose concentration enhanced proliferation and survival, but these processes were already optimal in cells cultured under subphysiological glucose concentrations (1 mmol/L). Notably, further increasing the glucose concentration to hyperglycemia (>10 mmol/L) did have a detrimental impact on survival of memory cells in vitro (Fig. 7A).

When the phenotype of in vitro-generated memory cells was analyzed, we observed only minor differences. Cells cultured under low glucose concentrations favored a more effector memory-like phenotype (CD44⁺CD62L⁻), whereas cells exposed to hyperglycemia demonstrated a more central memory-like phenotype (CD44⁻CD62L⁺) (Fig. 7B). However, it remains unclear whether this effect is a direct impact of glucose on memory differentiation or the result of selective survival. Upregulation of the memory cell marker CD127 was optimal in cells cultured under physiological glucose concentrations and was lower in cells cultured under both high and low glucose concentrations (Fig. 7B). Importantly, cytokine production was greatly suppressed by hyperglycemia. Whereas no major differences were observed between cells cultured under low and physiological glucose concentrations, production of IFN-γ, TNF, and IL-2 was strongly reduced in cells cultured at glucose concentrations of ≥10 mmol/L (Fig. 7C). To confirm that hyperglycemia negatively affects their antiviral capacity, memory OT-1 cells were generated in vitro under low glucose (1 mmol/L) or hyperglycemic conditions (30 mmol/L) and transferred to naïve recipients. Next, animals were infected with mCMV-N4, and viral titers were determined in livers after 4 days. Animals that received memory cells that had been generated under hyperglycemic conditions had higher viral titers compared with mice that received cells cultured under low glucose conditions (Fig. 7D).

To validate these findings in vivo, we induced hyperglycemia in mice by injecting them with a mixture of STZ/ALX, which eliminates pancreatic β-cells (Supplementary Fig. 4B). Next, we transferred OT-1 cells to these animals and infected them with mCMV-N4. We observed a minor reduction in the number of effector cells on day 7 after infection in hyperglycemic mice, but this did not affect the amount of memory cells or their phenotype (Fig. 7E and Supplementary Fig. 4C and D). Notably, upon in vitro restimulation, memory cells derived from hyperglycemic animals produced fewer cytokines, which was consistent with our in vitro observations (Fig. 7F). Finally, 43 days after primary infection, animals were reinfected with LCMV-N4, and recall responses were analyzed 5 days later. Again, we found that the hyperglycemic environment did not impact the number and phenotype of donor cells (Fig. 7E and Supplementary Fig. 7E) but resulted in reduced cytokine production, especially IFN-γ, by donor cells (Fig. 7G). Finally, to confirm the impaired antiviral capacity of memory cells formed under hyperglycemic conditions in vivo, STZ/ALX-treated mice or PBS-treated controls were transferred with OT-1 cells and infected with LCMV-N4. After 1 month, OT-1–carrying mice and naïve controls were infected with mCMV-N4, and viral titers were determined in livers after 4 days. Indeed, animals with memory CD8 T cells generated under hyperglycemic conditions showed significantly higher titers than PBS-treated controls (Fig. 7H).

In summary, our findings indicate that hyperglycemia and not hyperinsulinemia negatively impacts memory CD8 T-cell functionality in the context of metabolic disease and significantly impairs their ability to protect against viral reinfection.

People with T2D have greater susceptibility to infection (3,5). Insulin therapy was shown to further increase the risk of infection and infection-induced morbidity and mortality (7,11), but the underlying mechanism is still largely unclear. Here we show that T2D indeed impairs the ability of memory CD8 T cells to fight both infection and cancer. However, whereas insulin signaling did increase the ability of memory CD8 T cells to upregulate GLUT-1 and promote glucose uptake, it was redundant for their functionality. Moreover, in a model for metabolic disease, deficiency of the insulin receptor on memory CD8 T cells did not impact their ability to respond to infection. In contrast, hyperglycemia did impair cytokine production by memory CD8 T cells both in vitro and in vivo. Thus, hyperglycemia rather than direct insulin signaling is responsible for changes in memory CD8 T-cell functionality in context of diabetes.

Whereas the immune and endocrine systems were long thought to operate independently, their close interaction has become evident in recent years. Cytokines of the immune system are well known to contribute to the development of systemic insulin resistance in context of obesity (14,19). Indeed, we recently uncovered that both in humans and mice, the antiviral immune system actively induces insulin resistance in skeletal muscle to generate a state of hyperinsulinemia that promotes effector T-cell responses (15). Insulin was shown to promote proliferation and cytokine production by CD4 and CD8 T cells upon viral infection (15,16). In retrospect, this observation is not surprising considering that intracellular components of hormone receptors overlap with those of cytokine receptors and costimulatory molecules (13). For example, both the insulin receptor and CD28, an essential costimulatory molecule for CD8 activation, converge on PI3K signaling, and both receptors therefore induce upregulation of glucose transporters on the cell surface, leading to increased glucose uptake (13,23,24). Indeed, we found that insulin promoted GLUT-1 expression on memory precursor CD8 T cells and that insulin receptor deficiency impaired glucose uptake in memory cells. It is therefore surprising that insulin receptor deficiency did not impact memory CD8 T-cell functionality. However, during memory differentiation, CD8 T cells switch from glycolytic to oxidative metabolism (25). Memory cells therefore require less glucose to fulfill their energetic needs and use this capacity to rapidly gain effector function upon antigen reencounter (26). Thus, the increase of glucose uptake mediated by insulin is likely not needed for these cells to obtain full functionality. As a result, even if memory CD8 T cells become less responsive to insulin following chronic exposure in the context of T2D, they are not affected by this event.

Notably, we also observed that insulin receptor deficiency had a minimal impact on the effector CD8 T-cell response in vivo. This appears to be in contrast with previous reports from us and others that demonstrated that insulin deficiency or insulin receptor deficiency on T cells impairs acute CD8 T-cell responses (15,16). However, in these models, both CD4 and CD8 T cells were affected, whereas we specifically investigated the impact of insulin signaling on CD8 T cells. Thus, the impact of insulin appears to be more prominent in CD4 T cells in particular during the effector phase of the antiviral response.

Our findings show that memory CD8 T cells are negatively impacted by hyperglycemia, which has previously been shown for other immune cell subsets (27,28). How this is mediated is currently unknown but may be associated with aberrations in cellular metabolism. The aforementioned metabolic shift between naïve, effector, and memory CD8 T cells is directly linked to the functional properties of these cells. If effector cells fail to upregulate glycolysis, they are functionally greatly impaired (29). Similarly, if activated T cells cannot induce a metabolic switch back from glycolytic to oxidative metabolism, memory cell numbers and functionality are strongly reduced (25). Hyperglycemia was shown to impair mitochondrial respiration by promoting the formation of advanced glycation end products, providing a link between hyperglycemia and memory CD8 T-cell functionality (30). Our findings show that hyperglycemia has a clear detrimental effect on memory CD8 T-cell function and strongly impairs their ability to protect against viral reinfection. Memory cells showed a strong reduction in their ability to produce cytokines when subjected to hyperglycemia and were significantly less able to lower viral titers following infection in vivo. Notably, hyperglycemia did not impact the ability of CD8 T cells to proliferate, resulting in equal cell numbers at memory time points. In line with these findings, people with T2D do not appear to have a reduction in memory CD8 T-cell numbers, whereas they do have a worse prognosis in response to viral infection (3).

Insulin therapy was shown to impair the response of patients with T2D to infection, but our data indicate that this is not mediated through a negative impact on CD8 T cells. If not through CD8 T cells, then how does insulin therapy reduce the response of patients to infection? We found that insulin also does not impact the antiviral response of NK cells and macrophages (data not shown). However, various other immune cell populations, including CD4 T cells, express the insulin receptor (16), and a direct detrimental impact of insulin on the immune system in diabetes cannot be excluded at this point. An alternative explanation is that insulin only has a negative impact on disease outcome in comparison with other antidiabetes drugs, most notably metformin. Metformin directly promotes metabolic fitness of CD8 T cells by increasing mitochondrial mass, viability, and cytokine production (31). Moreover, metformin was shown to promote the antitumor function of CD8 T cells by inhibiting glycolytic metabolism and promoting memory formation in vivo (32). Similar beneficial effects were found for other antidiabetes drugs, such as α-glucosidase inhibitors (33). Indeed, in a recent study of patients with T2D in which insulin treatment was shown to be detrimental for coronavirus disease 2019 outcome, all other antidiabetes drugs tested appeared to improve survival compared with nontreated patients, even though a direct comparison between these groups was not made (11).

Insulin is one of the most potent treatment options for lowering of blood glucose levels and thereby highly efficient at reducing the risk of developing dangerous complications of T2D, such as chronic kidney disease and cardiovascular disease. However, in light of the current coronavirus disease 2019 pandemic, the impact of insulin on the immune system cannot be ignored, especially if it may have a detrimental effect. Whereas CD8 T cells are responsive to insulin, our findings show that this hormone is redundant for normal memory function. Moreover, we show that direct insulin signaling does not impair the memory CD8 T-cell response in the context of metabolic disease. Instead, we show that hyperglycemia has a direct detrimental effect on memory CD8 T-cell functionality. Any antidiabetes therapy that lowers blood glucose levels is therefore likely to have a positive impact on the functionality of these cells in people with T2D. Our findings provide better understanding of the way in which the immune and endocrine systems interact in context of diabetes and suggest that lowering of blood glucose levels is of key importance for preventing severe disease of patients with T2D.

Funding. This work was supported by a University of Rijeka Support grant 19-41-1551 and the Croatian Science Foundation (Hrvatska Zaklada za Znanost) (IP-2016-06-8027 and IP-CORONA-2020-04-2045 to F.M.W., IP-2020-02-7928 to T.T.W., and IP-2016-06-9306 to B.P). The study was supported by the Scientific Centre of Excellence for Virus Immunology and Vaccines and co-financed by the European Regional Development Fund (grant KK.01.1.1.01.0006). N.A.L. was supported by the Deutsche Forschungsgemeinschaft, Collaborative Research Center (CRC) 1292, TP11.

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

Author Contributions. I.K., M.K., and A.B-C. designed and performed experiments. I.K., M.K. and F.M.W. wrote the paper. B.P., T.T.W., and K.P.J.M.v.G. helped in experimental design and critically read the paper. N.A.L. provided reagents and helped in experimental design. F.M.W. supervised the project. F.M.W. 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.