A unique inflammaging profile generated by T cells from people with obesity is metformin resistant

Abstract

The alarmingly high prevalence of obesity in older adults coupled with the negative health effects of chronic inflammation in both obesity and aging highlight the importance of studies investigating the impacts of obesity on age-related inflammation. Since shifts in peripheral T-cell metabolism and function drive systemic inflammation in both obesity and aging, we hypothesize that obesity impacts the Th17-dominated inflammaging profile we identified in lean subjects and thus modifies the anti-inflammatory effects of geroprotective drugs like metformin. New cytokine profiling data showed that CD4⁺ T cells from older people with obesity generate a profile that specifically excludes Th17 cytokines. Metformin failed to change the age-associated T-cell profile in obesity, despite lowering both mitochondrial respiration and reactive oxygen species (ROS) production. Metformin did not improve macroautophagy in T cells from older people with obesity, in sharp contrast to the ability of metformin to promote autophagy in T cells from older lean subjects. These data indicate that body mass index modifies the mechanisms supporting inflammaging in T cells from older subjects, and that metformin-mediated restoration of redox balance is insufficient to stem obesity-associated inflammaging. We conclude that obesity fundamentally changes the mechanisms that promote inflammaging, and thus obesity becomes a critical consideration for clinical trials of geroprotective agents such as metformin.

Supplementary information

The online version contains supplementary material available at 10.1007/s11357-024-01441-4.

Introduction

The systemic inflammation that develops during aging is a main driver of age-related comorbidities and mortality risk. Obesity similarly fuels systemic inflammation and exacerbates risk of age-related diseases like neurodegeneration [1, 2], in part by increasing biological and immunological age [3, 4]. Analysis of individual characteristics has highlighted some similarities between obesity- and age-associated inflammation (e.g., metaflammation and inflammaging, respectively) [5–7], but in-depth analyses to test the idea that obesity is a state of accelerated aging remain forthcoming.

CD4⁺ T cells play a central role in systemic inflammaging [7] and in obesity-mediated comorbidities such as type 2 diabetes (T2D) [8–10]. Among the CD4⁺ T effector cell subsets (Teff), Th17s and Th1s disproportionately contribute to obesity- and T2D-associated metaflammation [6], while Th17s, with little Th1 contribution, dominate inflammaging in lean/healthy adults [7]. Such changes in T-cell function with age or obesity could explain the increased predisposition to infections and the poor vaccine response in both the elderly [11–13] and middle-aged/younger people with obesity [14, 15]. The impact of obesity on T cell–mediated inflammaging has not been reported.

Obesity- and age-related changes in CD4⁺ T-cell cytokine production associate with metabolic reprogramming. While the field originally proposed that pro-inflammatory Teff cells rely on glucose metabolism in the cytosol (glycolysis) to produce energy, and that anti-inflammatory regulatory T cells rely on pyruvate and fatty acid oxidation in the mitochondria (oxidative phosphorylation, OXPHOS), recent evidence demonstrates that this dogma is overly simplistic. Circulating CD4⁺ T cells from older people who are lean/healthy generate Th17 cytokine profiles through mitochondrial metabolism rather than glycolysis, in part based on low expression of lactate dehydrogenase [7]. Another example showed that in vitro polarized Th17 cells from people with obesity upregulate the fatty acid metabolism enzyme ACC1 to support IL-17 production [16], consistent with our previous demonstration that changes in fatty acid flux support Th17 cytokine production in T2D [17]. Roles for other Teff subsets, including Th2s and their close family member Th9s, are underexplored in both aging and obesity.

Metformin is an intensely studied geroprotective agent [18] that ameliorated the age-related Th17 profile generated by CD4⁺ T cells from lean/healthy adults ex vivo [7]. Metformin intervention also lowered non-specific indicators of inflammation in middle-aged subjects in the Diabetes Prevention Program [19], raising the possibility that metformin may be a useful tool for promoting healthspan in an ongoing clinical trial that is using a generally overweight/obese population perhaps more representative of the US population [20]. To establish likely outcomes of metformin geroprotective trials [21] from an inflammaging viewpoint in an overweight/obese population, we first identified a cytokine profile generated by CD4⁺ T cells from older people with obesity (but not T2D) that fundamentally differs from the Th17 profile generated by CD4⁺ T cells from lean, older people. In contrast to metformin-mediated improvements in autophagy and amelioration of the Th17 inflammaging profile in CD4⁺ T cells from lean/healthy older adults [7], metformin failed to alter autophagy and the broad inflammaging profile in T cells from older people with obesity. However, metformin lowered OXPHOS and hydrogen peroxide production by CD4⁺ T cells from older people with obesity, mirroring the impact of metformin on reactive oxygen species (ROS) produced by T cells from lean/healthy older adults. Our data highlight metformin’s ability to mitigate ROS in line with previous work [22], but raise the new possibility that obesity fundamentally modifies mechanisms that drive chronic inflammaging to uncouple ROS and mitochondrial function from inflammaging.

Material and methods

Human subjects

The University of Kentucky Institutional Review Board approved this study in accordance with the Declaration of Helsinki. All procedures followed institutional guidelines. Subjects were recruited from the local community through the University of Kentucky Center for Clinical and Translational Sciences, who obtained written informed consent. All participants with glycemic control data did not have diabetes (% HbA1c < 6.5) as categorized by the American Diabetes Association, and were lean (BMI 18–23 kg/m²) or with overweight/obesity (herein “obesity” BMI 26–47 kg/m², Tables 1 and 2). These cohorts were segregated by age into younger (23–38 yo) and older (≥ 58 yo) adults. Exclusion criteria include inflammatory or auto-immune diseases (IBD, RA, type 1 diabetes, etc.), NSAIDs < 72 h before blood draw, heart failure, CKD (eGFR < 45), liver disease, glycemic control drugs, and cancer < 5 years past. All exclusions are based on published evidence of impacts on immune cell function and/or inflammation. Participants with stable coronary disease taking statins, 81 mg aspirin/day, and ACE inhibitors were included.

Table 1.

Description of research participants with obesity

	Younger	Older
Total N	16	21
Age, years [mean, range]*	29.9 (23–38)	65.38 (58–73)
BMI, kg/m2 [mean, range]	30.40 (27–36.5)	33.36 (26–46.7)
Females [N (%)]	10 (62.5)	12 (58.8)
Males [N (%)]	6 (37.5)	9 (41.2)

^*p < 0.0001 Student’s t-test young vs. older

Table 2.

Description of research lean participants

	Younger	Older
Total N	3	4
Age, years [mean, range]*	26 (25–28)	67 (64.3–74)
BMI, kg/m2 [mean, range]	23.3 (23.03–23.5)	22.34 (17.6–24.1)
Females [N (%)]	3 (100)	2 (50)
Males [N (%)]	0 (0)	2 (50)

^*p < 0.0001 Student’s t-test young vs. older

CD4⁺ T-cell isolation and culture

CD4⁺ T cells were freshly purified from human PBMCs by negative selection using a human CD4⁺ T-cell isolation kit and magnetic separation following the manufacturer’s protocol (Miltenyi Biotec) as we previously published [8]. Cells were slow frozen then archived in vials of 5 million cells/0.5 mL of 10% DMSO 90% fetal bovine serum (FBS) in liquid nitrogen. Cells were thawed in batches (ensuring a minimum of n = 3/per group/batch/experiment) at 37 °C, and washed 1 × in 5 mL of RPMI1640 supplemented with 2 mM l-glutamine, 25 mM HEPES, 1% 100 × penicillin/streptomycin, 10% FBS (all from Gibco), and 1% pyruvate (Corning; complete RPMI-1640 herein). Pelleted cells were resuspended in complete RPMI-1640 at 10⁶ cells/mL and stimulated with human CD3/CD28 antibody-coated beads (1 bead/cell; Gibco) mixed in by gently pipetting. Flow cytometry of stimulated cells determined cell viability based on staining with Zombie Aqua dye, and cell purity was determined based on the surface marker CD4 (see “Flow cytometry” section below). Metformin (100 μM final concentration) or an equal volume of diluent (water) was added at the same time as beads to cultures as indicated. Cells and supernatants were harvested at 40 h, then supernatants were stored at − 80 °C in aliquots for analysis after ≤ 2 freeze/thaw cycles.

For TEMPOL and metformin titration analyses, archived PBMCs were cultured in complete RPMI media supplemented with IL-2 (10 ng/mL) for 8–10 h prior to CD4⁺ T-cell isolation, using a magnetic bead based kit per above, followed by confirmation of purity per above. TEMPOL at a final concentration of 10 μM was added 3 h after the start of CD3/CD28 activation to ensure >90% viability.

Cytokine quantification

We quantified cytokines using Milliplex human Th17 25-plex kits (Millipore) in 384-well plates in duplicate, as previously published [6]. Wells in which bead counts were less than 50% of the added beads were excluded, as were samples with a CV > 25%. Plates were washed between incubations using a BioTek 406 Touch plate washer (BioTek) and read using the Luminex FlexMap 3D system (Luminex xPONENT 4.2) and Bio-plex Manager (Bio-Rad) software. Cytokines above the standard curve were measured after a 1:10 dilution. All cytokines represented showed a tenfold concentration above the lowest standard. Eighteen of the 25 cytokines measured by this kit were produced by > 50% of the subjects’ T cells and include IL−2, −4, −5, −6, −9,−10, −12, −13, −17A/F, −21, −22, −23, CCL20, GMCSF, IFNγ, and TNFα/β. The remaining seven cytokines (IL-1β, −15, −17E, −27, −28A, −31, −33) were almost exclusively below the level of detection and thus were omitted from the analysis.

Partial least squares discriminant analysis

We used partial least squares discriminant analysis (PLS-DA) [23, 24] to overcome the problem of limited N’s and small absolute differences in cytokines by integrating the information across multiple cytokines to highlight differences with small/medium effect sizes that, when combined, may have a large effect. PLS-DA thus complements traditional analyses to better capture the biology by assessing overall “inflammation,” a sum of numerous pro- and anti-inflammatory cytokines that work in unique combinations for each condition/person. For PLS-DA, analysts were blinded to subject cohort identity, and cytokine data were log-transformed to mitigate skewing. Due to the biological properties of antibodies, amounts of different cytokines cannot be compared directly, and cytokines with a small range of values should have the option to be as or more important than cytokines with a large range of values. We thus scaled each cytokine to zero mean and unit variance before applying PLS-DA. After data preprocessing based on exploratory analysis, we applied PLS-DA using the R function splsda() in the R package “mixOmics” [25] in R4.2.2. We conducted PLS-DA based on all the cytokines passing quality control and then, for some studies, cytokines with variable importance in projection (VIP) scores > 1.0. VIP is a multidimensional reduction tool that reflects the relative importance of a cytokine in the PLS-DA model, and quantifies the impact of each cytokine on overall inflammation to enable more impactful cytokine selection and model interpretation [26]. A VIP score greater than 1 generally indicates that the variable (in this case, cytokine) is important for the model's predictive capability and should be prioritized in model interpretation. We used the first two PLS-DA components for each comparison as indicated to display the classification results based on the cytokines with VIP scores > 1.0. The top 10 ranked cytokines are shown for most analyses. N’s exceeded the number of subjects required to build predictive models; accuracies by both leave-one-out and K-fold cross-validation were > 70%. For the PLS-DA models in this study, we adopted fivefold cross-validation to prevent overfitting and to enable robust model selection and evaluation. Cytokine interactions are implicit to the models and the concept of statistical power is not applicable. No data were missing.

Flow cytometry

Cells were stimulated with CD3/CD28 beads for 40 h (per above) before harvest. For intracellular staining, cells were treated with Brefeldin A (Invitrogen eBioscience) during the last 5 h of stimulation. Stimulated cells were pelleted, and beads were removed with a magnetic separator. All cells were resuspended in Zombie Aqua (V405 Biolegend) viability stain at 1:250 in 1 × dPBS and incubated at 4 °C for 30 min. Cells were washed, pelleted, and resuspended with antibodies that recognize CD4 (FITC; Biolegend) then incubated at 4 °C for 30 min. Before intracellular cytokine staining, cells were treated with fixation buffer at room temperature for 30 min, washed in 1 × permeabilization buffer (Biolegend), and pelleted. Cells were resuspended in anti-IFNγ (PECF594; BD Pharmingen), IL-17A (1:50 BV421; BD Pharmingen), or IL-17F (1:50 AF647; BD Pharmingen) prepared as in the indicated dilution in 1 × permeabilization buffer and incubated at 4 °C for 30 min. Stained cells were pelleted and resuspended in 200 μL 1% paraformaldehyde (ChemCruz) in 1 × dPBS and analyzed on a BD Symphony A3. UltraComp eBeads (Invitrogen) were used for compensation. Data was analyzed by FlowJo.

Immunoblotting

Immunoblotting quantified protein expression as we published [7, 27]. Briefly, 30 µL of 1 × cell lysis buffer (Cell Signaling Technology, Danvers, MA) was added to 1 × 10⁶ cells and incubated on ice for 20 min. Cells were centrifuged at 13,000 rpm for 20 min and supernatant was collected. A bicinchoninic acid assay (Thermo Fisher Scientific, 23,225) assessed protein concentration. Fifteen micrograms of protein was loaded onto polyacrylamide gels, and electrophoresis was performed at 100 V for 1 h. Transfer of protein to polyvinylidene difluoride (PVDF) membrane was performed at 35 V for 5 h. The membrane was blocked for 30 min at room temperature (RT) in a blocking buffer containing 2% bovine serum albumin in TBST followed by overnight incubation at 4^○C in primary antibody. The membrane was washed 2 × with 1 × TBST and incubated with secondary antibody for 2 h at RT, then washed and imaged. All primary antibodies were used at a dilution of 1:500 except β-actin which was used at 1:10,000. All secondary antibodies were used at a concentration of 1:5000 except the secondary antibody for β-actin which was used at 1:10,000. We quantified protein expression on western blots using Image Studio Lite software (Licor, Lincoln, NE).

Immunofluorescence

Cells (200,000) were collected after 40 h of stimulation ± respective treatments and plated on coverslips in 12-well plates coated with poly-d-lysine. Plates were centrifuged (1200 rpm, 10 min) to adhere cells to the coverslip, then washed two times with 1 × PBS and fixed in 4% paraformaldehyde (30 min at RT). The coverslips were washed twice with PBS and 0.1% triton X-100 (PBST) and blocked for 30 min in 5% BSA/PBST. Antibodies for detection of LC3 and LAMP1 (Cell Signaling Technology, Danvers, MA) were added at 1:50 dilution with incubation overnight at 4 °C. The coverslips were washed 2 × with PBST and incubated with fluorophore-tagged secondary antibodies (anti-mouse Alexa 488 or anti-rabbit Alexa 680; Rockland Immunochemicals, Limerick, PA; 1/500) for 2 h at RT. The coverslips were washed twice with PBST and mounted on glass slides using Fluromount G (Southern Biotech, Birmingham, AL). Cell imaging under a 63 × oil immersion lens was performed in a Zeiss LSM 800 confocal microscope. Approximately 5–7 fields/slide were imaged on N = 3–4 subjects per treatment, and data were analyzed using FIJI/ImageJ [28, 29]. Protein expression was reported as mean fluorescence intensity (MFI) and colocalization between two proteins was reported as Pearson’s colocalization coefficient (PCC). PCC is a well-established method that quantifies the degree of overlap between fluorescence in the two channels, and is unaffected by changes to the offset and independent of gain. PCC has a range of + 1 (perfect correlation) to − 1 (perfect but negative correlation). PCC is not sensitive to differences in signal intensity between the components of an image caused by different labeling with fluorochromes, photobleaching or different settings of amplifiers [27, 30].

ROS detection assays

Stimulated cells (200,000; 40 h) were plated in black-walled, clear-bottom plates in non-red phenol RPMI media and incubated for 20 min with chromomethyl 2′,7′-dichlorofluorescin diacetate (CM-H₂DCFDA, 10 mM; Sigma-Aldrich, D6883) to detect total peroxides, MitoSox red (MitoSox, 5 mM; Thermo Fisher Scientific, M36008) to detect production of mitochondrial superoxide, and dihydroethidium fluorescence (DHE, 30 mM; Thermo Fisher Scientific, D11347) to detect total superoxide. After incubation, cells were washed and the fluorescence signal was measured in a plate reader (Varioskan; Thermo Fisher). To measure hydrogen peroxide production over time, Amplex red (Invitrogen) was used following the manufacturer’s instructions.

Glutamate quantification

Glutamate measures were performed using the Glutamine/glutamate-Glo Assay (Promega) following the manufacturer’s protocol. Briefly, CD4⁺ T cells (250 k) were stimulated for 40 h per above and washed twice with 200 mL of cold dPBS. After washing, 30 mL of dPBS and 15 mL of the inactivation solution (0.3 N HCl) were added to the cells and mixed on a plate shaker for 5 min followed by the addition of 15 mL of 450 mM Tris Solution I, pH 8.0. Then, 35 mL of the enzyme glutaminase in glutaminase buffer was added to each well. Plates were shaken for 60 s and incubated for 40 min at room temperature. Glutamate detection reagent was added and mixed by shaking the plate for 60 s, then incubated for 60 min. Luminescence was recorded in a Varioskan plate reader (Thermo Fisher). Standards were prepared by following the manufacturer’s instructions.

Seahorse

Anti CD3/CD28-stimulated cells plated at 10⁶ cells/mL were harvested at 40 h and cell viability was assessed by Trypan Blue, then 250,000 live cells/well were plated in quadruplicate wells of Seahorse XF plates previously coated with poly-d-lysine (Agilent, Santa Clara, CA) in extracellular flux assay media (non-buffered DMEM containing 11 mM of glucose, 2 mM l-glutamine, and 1 mM sodium pyruvate). OCR and ECAR were measured using the mitochondrial stress test for basal OCR followed by sequential addition of 3.5 µM oligomycin (Calbiochem, San Diego, CA) to block ATP synthase (Complex V) at ~ 15′, 1 µM fluoro-carbonyl cyanide phenylhydrazone (FCCP) (Enzo Life Sciences, Farmingdale, NY) to drive maximal respiration at ~ 35′, and 14 µM antimycin A (Enzo Life Sciences; as previously titrated [31]) to block Complex III at ~ 55′ in the XFe96 Extracellular Flux Analyzer (Agilent, Santa Clara, CA) as previously described [31].

Statistical analysis

We used GraphPad Prism for two-way ANOVA with Bonferroni’s multiple comparisons and a 95% confidence interval to assess cohort- or treatment-specific differences in cytokine competency, OCR, and ECAR. Subject characteristics were analyzed by Student’s t-test. P < 0.05 indicated significant differences.

Results

Peripheral CD4⁺ T-cell inflammaging in people with obesity is defined by a mixed Teff profile

To investigate the impact of obesity on inflammaging in human CD4⁺ T cells, we stimulated CD4⁺ T cells freshly isolated from peripheral blood mononuclear cells (PBMCs) of non-diabetes younger (23–38 yo; Ob/Y) or older (58–73 yo; Ob/O) adults with obesity (BMI 27–47 kg/m²) (Table 1), and quantified supernatant cytokines. PLS-DA calculated the relative importance of the multiple cytokines produced in higher amounts by cells from Ob/O compared to Ob/Y adults (Fig. S1) to define obesity-associated inflammaging. PLS-DA showed that overall “inflammation” calculated using a combination of reliably quantifiable cytokines discriminated T cells from the two age-constrained cohorts with an accuracy of 74% (Fig. 1A). Cytokines with variable importance projection (VIP) values > 1.0, and thus the most important in defining CD4⁺ T-cell inflammaging, including those generally produced by Th1 (IL-12, IFNg) and Th2/9 (IL-6, −4, −9, −13) cells, distinguished profiles generated by Ob/O and Ob/Y T cells (Fig. 1B). The top-ranked cytokine, IL-21, is made by multiple CD4⁺ T-cell subsets [32, 33]. The signature Th17 cytokines IL-17A and IL-17F that dominate lean inflammaging [7] were unimportant for defining obesity-associated inflammaging (Fig. 1B), despite our ability to reproduce higher production of these cytokines from cells of a small independent group of lean/healthy sexagenarians (Fig. S2A, Table 2). Flow cytometry confirmed freshly isolated CD4⁺ T-cell viability and purity (Fig. S2B), and demonstrated age- and obesity-independent frequency and mean fluorescence intensity (MFI) of Th17 cytokine-producing cells (Fig. 1C). We conclude that aging impacts inflammatory profiles produced by CD4⁺ T cells from people with obesity by promoting a mixed Teff profile that excludes Th17 cytokines.

Fig. 1.

A mixed Teff profile defines CD4⁺ T-cell inflammaging from people with obesity. (A) PLS-DA shows compendium measures of “inflammation” generated by combining the 18 detectable cytokines (Fig. S1) from Ob/Y (green) and Ob/O (black) CD4⁺ T cells stimulated with CD3/CD28 beads for 40 h. Cross-validation (>70% for informative models) was calculated by two approaches described in the “Materials and methods” section. (B) Bars show variable importance projection (VIP) scores to rank cytokines from the most (leftmost) to the least (rightmost) important in discriminating compendium cytokine profiles between Ob/Y and Ob/O cells shown in panel (A). A VIP > 1 (horizontal line) identifies the more important variables (cytokines) for distinguishing groups in PLS-DA models. (C, D) Percentages (left) and mean fluorescence intensity (MFI; right) of (C) CD4⁺IL-17A⁺ or (D) CD4⁺IL-17F⁺ cells determined by flow cytometry; L/Y (circle), Ob/Y (square), Ob/O (triangle). (A, B) N = 14 for Ob/Y; N = 13 for Ob/O. (C, D) N = 3 per group. Lack of differences in (C) and (D) were calculated using a one-way ANOVA with Bonferroni post hoc correction

High OXPHOS fuels the Th17 profile characteristic of lean/healthy inflammaging [7]. To determine whether differences in CD4⁺ T-cell metabolism similarly fuel (non-Th17) inflammaging in obesity, we measured the oxygen consumption rate (OCR), a direct measure of OXPHOS, by extracellular flux (Seahorse) during a Mito Stress test. CD4⁺ T cells from Ob/O have higher maximal OCR than cells from Ob/Y subjects (time points 40–50 min), and both Ob/Y and Ob/O had higher OCR than cells from L/Y subjects at basal (time points 0–15 min) and maximal (time points 40–50 min) respiration (Fig. 2A). CD4⁺ T-cell extracellular acidification rate (ECAR), an indirect measure of glycolysis, was higher with obesity (Ob/Y and Ob/O > L/Y) but lower with age (Ob/O < Ob/Y), consistent with lactate amounts (Fig. 2B, C). The OCR/ECAR ratio, which indicates relative utilization of OXPHOS versus non-mitochondrial glycolytic energy production, was higher in Ob/O compared to Ob/Y cells, indicating that Ob/O cells are more reliant on mitochondrial respiration (Fig. 2D). Higher maximal OCR in Ob/O cells corresponded with higher mitochondrial superoxide (Fig. 2E), while other ROS subspecies remained unchanged (Fig. S2C). Glutamine metabolism was similar in Ob/Y and Ob/O T cells, as evidenced by cellular glutamate, the immediate downstream metabolite of glutamine (Fig. 2F). These data support the interpretation that enhanced pyruvate oxidation, rather than glutaminolysis, mediates higher OXPHOS in Ob/O compared to Ob/Y cells, and thus may preferentially support age-mediated changes in the inflammatory profiles of CD4⁺ T cells in obesity.

Fig. 2.

CD4⁺ T cells from older people with obesity have higher mitochondrial OXPHOS than cells from younger people. CD4.⁺ T cells from lean younger (L/Y, circle) and younger (Ob/Y, square) or older (Ob/O, triangle) people with obesity were CD3/CD28 stimulated for 40 h and 250,000 live cells were assayed for (A) oxygen consumption rate (OCR) and (B) extracellular acidification rate (ECAR) in a Mito Stress test using a Seahorse XFe96 analyzer. Results are shown as avg. ± SEM of quadruplicate cultures of cells from multiple subjects. (A, B) N = 4–8; differences were calculated by two-way ANOVA with Bonferroni post hoc correction, ****p < 0.001 L/Y vs. Ob/Y and Ob/O; ++ p = 0.008 Ob/Y vs. Ob/O. (C) Lactate production in cells from indicated subjects. (D) OCR/ECAR ratio was calculated based on basal OCR and ECAR from panels (A) and (B). (E) Mitochondrial superoxide and (F) glutamate in cells from the indicated subjects; N = 5–7. Differences in panels (C), (D), (E), and (F) were calculated using unpaired Mann–Whitney U tests. *p < 0.05, **p < 0.007

Metformin modestly shifts the inflammatory profile produced by CD4⁺ T cells from older adults with obesity

Metformin has been tested as an anti-inflammaging geroprotective, but the impact of excess weight without metabolic decline on this activity has been largely underappreciated. To test the influence of metformin on the newly defined inflammaging profile in obesity, we stimulated CD4⁺ T cells in the presence or absence of metformin and analyzed cytokine secretion. One-by-one cytokine analysis showed little impact of metformin on cytokine production across a range of physiologically achieved concentrations (Figs. S3 and 4). The PLS-DA model using data from all detectable cytokines confirmed a modest impact of metformin on cytokine profiles from Ob/O cells (accuracy of 67%; cross-validation < 50% Fig. 3A, left). Follow-up analysis on IL-9, the cytokine ranked as most important in the modest metformin response (Fig. 3A, right), using a confusion matrix showed that IL-9 predicted metformin treatment in eight out of nine samples, but that vehicle outcomes were unable to distinguish whether the sample was treated with metformin or vehicle (Fig. 3B). The model generated from cytokine production by Ob/Y cells also showed no impact of metformin on cytokine profiles (Fig. 3C; 52% accuracy and < 50% cross-validation accuracy approximates random chance). We conclude that obesity blunts the broad and robust response of inflammaging to metformin.

Fig. 3.

Metformin modestly impacts obesity-associated inflammaging profile. CD4⁺ T cells from younger (Ob/Y) and older (Ob/O) people with obesity were CD3/CD28 stimulated in the presence of vehicle (black) or metformin (100 μM; yellow) for 40 h. PLS-DA (left panel) shows a compendium of “inflammation” generated by combining the 18 detectable cytokines (Fig. S3) made by (A) Ob/O (N = 9) cells. Bar graph at right ranks cytokines important for the modest metformin-mediated differences in overall inflammation based on component 1 of the model (left). (B) Confusion matrix to determine how often observed amounts of IL-9 predict the addition of metformin or vehicle during T-cell stimulation. (C) PLS-DA showing a compendium of “inflammation” generated by combining the 18 detectable cytokines (Fig. S3) made by Ob/Y (N = 12) cells

To begin testing mechanisms that limit the responses of T cells from Ob/O and Ob/Y subjects to metformin, we compared Seahorse-generated OCR and ECAR measures from Ob/Y and Ob/O T cells stimulated in the presence and absence of metformin. Metformin significantly lowered OCR in Ob/O but not Ob/Y cells (Fig. 4A), and also lowered ECAR in Ob/Y cells (Fig. 4B). Metformin did not change the OCR/ECAR ratio in Ob/Y or Ob/O cells (Fig. 4C). Preliminary analyses with Mitotracker green did not conclusively show that changes in mitochondrial mass could account for changes in mitochondrial function (not shown). We conclude that metformin lowers mitochondrial metabolism only in Ob/O cells while insignificantly changing the ratio between mitochondrial and non-mitochondrial metabolism.

Fig. 4.

Metformin reduces OCR and hydrogen peroxide production by CD4⁺ T cells from older people with obesity. CD4⁺ T cells from younger (Ob/Y, square) and older (Ob/O, triangle) people with obesity were CD3/CD28 stimulated in the absence (black) or presence of metformin (100μM, yellow) for 40 hr. (A) oxygen consumption rate (OCR) or (B) extracellular acidification rate (ECAR) were measured in a mito stress test assayed in a Seahorse analyzer. (C) OCR/ECAR ratios were calculated based on basal OCR and ECAR from panels A and B. (D) hydrogen peroxide (H₂O₂) production after 0–120min of stimulation in Ob/O cells. N=3–6. Differences were calculated by two-way-ANOVA with Bonferroni post hoc in panels A, B and D *p<0.01, **p<0.003, ***p<0.0002, and ****p<0.0001 and a paired Wilcoxon test in panel C (no difference)

Because (1) metformin improved redox balance in CD4⁺ T cells from lean/healthy older adults [7], (2) mitochondria are a major source of reactive oxygen species (ROS), (3) mitochondrial superoxide is higher in Ob/O compared to Ob/Y T cells (Fig. 2E), and (4) metformin lowered mitochondrial OXPHOS in CD4⁺ T cells from Ob/O subjects (Fig. 4A), we tested whether metformin mitigates ROS species. Metformin did not impact total peroxides, total superoxide, or mitochondrial superoxide 20 min post-stimulation (Fig. S5A–C) although a trend toward lower total peroxide appeared (Fig. S5A, bottom) that led us to investigate the peroxide species hydrogen peroxide. Time-course detection demonstrated that metformin significantly lowered hydrogen peroxide production (Fig. 4D). The effect of metformin on superoxide dismutase 1 and 2 (SOD 1/2) was highly variable on western blots (Fig. S5D, E), thus uninterpretable. The nitric oxide species (NOS) scavenger TEMPOL had no effect on cytokine production by Ob/O cells (Fig. S5F). We conclude that metformin lowers an important source of ROS in CD4⁺ T cells from Ob/O subjects, echoing the ability of metformin to lower total peroxides from lean/older adults’ cells [7]. This effect, in contrast to the effect in L/Y cells, appeared to be independent of NOS.

Metformin-mediated improvement of macroautophagy is limited in CD4⁺ T cells from older adults with obesity

Metformin also improved autophagy in CD4⁺ T cells from lean/healthy older adults [7], and thus we quantitated indicators of macroautophagy by confocal analysis to determine the impact of obesity on metformin in T cells. The autophagosome formation protein LC3B and the lysosomal marker LAMP1 were less abundant in Ob/O compared to Ob/Y T cells (Fig. 5A, B). However, more co-localization of LC3B and LAMP1 in Ob/O compared to Ob/Y cells indicated that the frequency of those (less abundant) autophagy indicators in autophagolysosomes was higher in Ob/O cells (Fig. 5C). Metformin had no impact on LC3B/LAMP1 expression/co-localization (Fig. 5B), mTORC1/2 activation (Fig. S6A–C), or the autophagy activator glutamate (Fig. S5D). Confocal controls included nuclear (DAPI) staining, primary/secondary antibody exclusion, and everolimus as a positive control for autophagolysosome formation (Fig. S7A, B). We conclude that metformin lowers OXPHOS and hydrogen peroxide production, but does not change autophagy, mTORC activation, or glutamate to fuel the modest shifts in cytokine production by CD4⁺ T cell-generated inflammation in obesity-associated aging.

Fig. 5.

Metformin does not affect macroautophagy in CD4⁺ T cells from people with obesity. CD4⁺ T cells from younger (Ob/Y) and older (Ob/O) people with obesity were CD3/CD28 stimulated in the presence or absence of metformin (100 μM) for 40 h. Three to five fields/slide were imaged by using 63 × oil immersion in a Zeiss confocal microscope. (A) Representative images for the macroautophagy markers LC3B (green) and LAMP1 (red). Yellow shows merge in cells from subjects indicated at left. (B) Average fluorescence/field for LC3B and LAMP1. (C) Pearson’s co-localization coefficient for LC3B/LAMP1. N = 3–4. Differences were calculated by two-way ANOVA with Bonferroni post hoc testing. *p < 0.05, ***p < 0.001, and ****p < 0.0001 when comparing Ob/Y versus Ob/O. Comparisons between vehicle and metformin p value or no significance (ns) displayed under x-axis

Discussion

One of the limitations in many human inflammaging studies is lack of appreciation of BMI as a critical variable. Given our results, this oversight could bias clinical studies for decades, similar to what had occurred with a historical lack of appreciation for sex differences. We show herein that CD4⁺ T-cell inflammaging in obesity is dominated by a mixed Teff profile that excludes the Th17 cytokines that defined inflammaging in lean/healthy adults [7]. In line with our results showing IL-9 involvement in obesity-associated inflammaging, Hu et al. [34] showed in T-cell differentiation studies that IL-9 producing Th9s were more frequent in older compared to younger adults of unspecified BMI, while Th17 frequency (rather than function as assessed herein) was not changed by age. Concordance of these previous findings with our new and/or published work [7] awaits follow-ups with BMI measures and functional cellular analyses. Although we did not have the decades of medical records required to rigorously confirm that aging of our subjects occurred in the presence of obesity (e.g. “obesity-associated aging”), CDC data showing ~ 40% of US adults have obesity across all age groups [35], and that obesity for individuals tends to peak in middle-age, suggest that aging of our older subjects occurred in the context of at least some excess weight. Longer-term electronic medical records and characterization of metabolic status by state-of-the-art methods in future studies will further our understanding of obesity as an important modifier of inflammaging.

Similar to our findings in cells from lean/healthy older subjects [7], the age-associated shift in the inflammatory profile of CD4⁺ T cells from people with obesity associates with enhanced mitochondrial metabolism. Indications of age-mediated metabolic reprogramming have been also reported in human CD4⁺ T cells from the Baltimore Longitudinal Study on Aging (BLSA), where proteomic analysis of OXPHOS-related proteins in cells from older men are overrepresented compared to cells from younger men (BMIs unspecified). Indicators of defective autophagy in older men’s cells were also noted [36]. Despite higher OXPHOS-related proteomes with age, mitochondrial function was lower in the same cells from BLSA participants, indicating that proteomics alone can be deceiving. Additionally, non-mitochondrial respiration was higher in cells from older adults in the BLSA study, and thus opposite our demonstration that lactate is lower in cells from lean/older compared to lean/younger adults [7], and in Ob/O compared to Ob/Y cells. Older age of BLSA subjects (70-93 yrs) may explain this seeming discrepancy. In another study lacking BMI/metabolic data, memory CD4⁺ T cells from older subjects (> 65 yo) showed higher mitochondrial respiration that, in accordance with our results, associated with higher mitochondrial superoxide [37].

The electron transport chain (ETC) complexes, especially CI and CIII, are main ROS generators, and thus we speculate that metformin-mediated reduction of OCR in Seahorse (e.g., fewer electrons fed into the ETC) is the underlying mechanism for lower hydrogen peroxide production in Ob/O T cells. Metformin similarly lowered total peroxide production by CD4⁺ T cells from lean/healthy older adults [7]. Regardless of obesity-associated differences between metformin mechanisms of action in CD4⁺ T cells from older adults, work on lean and obese cohorts converges on the likely role of ROS species in inflammaging. ROS are known to be crucial for early stages of T-cell cytokine production [38]. The mechanisms underlying this critical need for early ROS generation in T-cell activation is likely multifactorial and includes activating numerous signaling pathways and transcription factors like NF-kB. ROS also acts epigenetically by impacting DNA methylation, although connections between any of these mechanisms and ROS in obesity remain future directions. Metformin mechanisms of action are similarly multifactorial, although we showed that the stereotypical ability of metformin to improve autophagy is absent in T cells from older adults with obesity. The analysis of different ROS species in obesity shown herein adds nuance to our original “two-hit” hypothesis of inflammaging ([7]; ROS + autophagy defects = inflammaging), by focusing on metformin-mitigated hydrogen peroxide production, and highlighting outcomes that may be obesity specific: control of IL-9. Ours is the first evidence of IL-9 roles in obesity-associated inflammaging. Th9s, which have been ignored for decades due to their overlapping characteristics and roles with Th2s, are considered the primary producers of IL-9 [39]. Th2s have been identified as intermediates for the differentiation of naïve CD4⁺ T cells into Th9 [40, 41] and classified as “anti-inflammatory” in obesity-induced insulin resistance [8], leading us to speculate about a relationship between metformin-mediated reductions in maximal respiration and hydrogen peroxide production, and reduction of IL-9 production in Ob/O T cells.

Clinical trials testing the geroprotective actions of metformin, interpreted in light of our new work, must stratify for metabolic differences for rigorous interpretation, as is being done for the ANTHEM clinical trial [19] and analysis of inflammation in a related study, iANTHEM (R01AG079525). Metabolic stratification of subjects for inflammation-related outcomes (which differ from our combinatorial analysis herein) will likely be possible in the TAME study, a multi-site metformin intervention with a planned enrollment of 3000 and an eye toward geroprotection [21, 42, 43]. These planned studies can include analysis of metformin effects on other cellular sources of inflammaging, such as the expansion of myeloid cells, specific B-cell subsets, and the recently identified granzyme K⁺/CD8⁺ T cells [44] noting that relative contributions of different immune cell subsets to overall inflammaging remain to be detailed.

At present, links between clinical outcomes of metformin intervention and molecular mechanisms that explain these outcomes remain largely disconnected and are largely extrapolated from superphysiological metformin doses in animals or cell culture models. In our studies, metformin did not improve autophagy in Ob/O or Ob/Y T cells. Instead, age positively associated with higher mass of autophagically active organelles (LC3B/LAMP-1 co-localization), despite less LC3B/LAMP1 signal in Ob/O compared to Ob/Y cells. Together, these data show that obesity modifies age-related declines in autophagy of T cells from lean/healthy subjects and renders these cells refractory to metformin-mediated improvements. Whether the greater percentage of co-localized LC3B and lysosomes in T cells from Ob/O subjects compensate for the overall lower signal of each autophagolysosome component will require detailed follow-ups using indicators of autophagic flux that go beyond our focus on metformin mechanisms of action as a candidate geroprotective.

Limitations of the study include low N’s of the lean comparator groups that we intentionally limited to an N to validate, rather than completely replicate, our extensive study of T cells from lean/younger and lean/older subjects in previous work [7]. The lean/young group was all female, which introduces the possibility of gender bias. BMI is an imperfect measure of obesity, although the vast majority of our subjects had central obesity. Recruiting males for obesity studies is a challenge in the field, and relative to many such studies, our recruitment of males is reasonable. Future work will be needed to explore the possibility that higher peroxide requires a partner to drive obesity-associated inflammaging via a variation of the two-hit inflammaging model we posited for lean/healthy T cells [7]. Analysis by race remains an important future direction, as would more detailed metabolic characterization of subjects, and similar analysis across the age spectrum by including middle-aged subjects.

Supplementary information

Below is the link to the electronic supplementary material.

Author contribution

Conceptualization—SSCC and BSN. Methodology—SSCC, GHK, SS, DZ, and LPB. Investigation—SSCC, GHK, SS, and LPB. Adquisition and formal analysis—SSCC, SS, GHK, EZ, HM, AJ, FG, and LPB. Supervision—BSN. Writing—BSN and SSCC. Editing—BSN and SSCC. Funding—LPB, DZ, PAK, and BSN. Resources—BSN, DZ, and PAK. BSN is the guarantor of this work and, as such, had full access to all data in the study and takes responsibility for the integrity of this data and the accuracy of the data analysis.

Funding

This work was supported by R56AG06985, 1R01AG079525, and P20GM148326 (BSN); The Barnstable Brown Diabetes and Obesity Research Center (BSN and XDZ); 1U01DK135111, UL1TR001998, and P30 DK020579 (XDZ); R15AG068957 (LPB); T32DK007778 and TL1UL1TR001998 (GHK). The project described was supported by the NIH National Center for Advancing Translational Sciences through grant number UL1TR001998 (PAK). This research was supported by the Flow Cytometry and Immune Monitoring shared resource of the University of Kentucky Markey Cancer Center (P30CA177558). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. No permissions are needed. All work is original.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

No permissions are needed. All work is original.

Conflict of interest

The authors declare no competing interests.