Associations of Innate and Adaptive Immune Cell Subsets With Incident Type 2 Diabetes Risk: The MESA Study

Abstract

Objective

Cell-mediated immunity is implicated in glucose homeostasis and insulin resistance. Whether the levels of innate and adaptive immune cells in peripheral blood are risk factors for incident type 2 diabetes (T2D) remains unknown. We hypothesized that the proportions of naive, memory, CD28⁻, Th17, and T regulatory CD4⁺ cells would be associated with incident T2D. In secondary analyses, we evaluated the relationships of 28 additional immune cell phenotypes with T2D.

Design

Immune cell phenotypes (n = 33) were measured by flow cytometry using cryopreserved cells collected from 1113 participants of the Multi-Ethnic Study of Atherosclerosis (MESA) at the baseline examination (2000–2002). Cox proportional hazards models were used to evaluate associations of immune cell phenotypes with incident T2D over a median follow-up of 9.1 years, adjusted for age, sex, race/ethnicity, educational status, and body mass index.

Results

Incident T2D was observed for 120 participants. None of the cell phenotypes included in the primary hypotheses were significantly associated with T2D (all P > 0.05). Among the secondary immune cells studied, a higher proportion of CD19⁺CD27⁺ B cells was associated with a reduced risk of T2D (hazard ratio: 0.72 (95% confidence interval: 0.56, 0.93), per 1-standard deviation (16%) increase). This association was no longer significant after correction for the multiple cell phenotypes tested (P > 0.0015).

Conclusions

Our results suggest that the frequencies of several subsets of monocytes, innate lymphocytes, and CD4⁺ and CD8⁺ T cells in circulating blood are not related to the future onset of T2D. Higher levels of CD19⁺CD27⁺ B cells may be associated with decreased T2D risk.

Chronic inflammation is an established risk factor for insulin resistance (IR) and type 2 diabetes (T2D) (1). Higher concentrations of inflammation biomarkers (C-reactive protein [CRP], plasminogen activator inhibitor-1 [PAI-1], fetuin-A, serum amyloid A); cytokines (interleukin-1β [IL-1β], interleukin-6 [IL-6], interleukin-18 [IL-18]); and soluble adhesion molecules (intercellular adhesion molecule-1 [ICAM-1], E-selectin) are associated with incident T2D risk in several prospective epidemiologic studies (2–8). Inflammation also contributes to comorbidities associated with T2D including atherosclerosis, nephropathy, and retinopathy (1).

Elevated inflammation biomarker levels reflect activation of innate immunity, which occurs in response to metabolic challenges, such as obesity, associated with T2D risk (9). Inflammatory cytokines secreted by macrophages and adipocytes, including tumor necrosis factor-α (TNF-α), IL-6, and IL-1β, can inhibit insulin signaling by posttranslational modification of proteins in the insulin signaling pathway, such as insulin receptor substrate 1 (IRS-1) (10–12).

Innate and adaptive immune cell responses have increasingly been implicated in adipose tissue inflammation, IR, and T2D (13). Monocytes, T cells (including CD4⁺ T helper (Th) and CD8⁺ cytotoxic T (Tc) cells), and B cells are recruited into the visceral adipose tissue (VAT) of mice fed a high-fat diet. This recruitment shifts the cellular pool in the VAT towards higher levels of pro-inflammatory M1 macrophages, CD4⁺ T helper type 1 (Th1) cells, CD8⁺ T cells, and B-2 B cells, as well as reduced levels of anti-inflammatory Th2, T regulatory (Treg), and B-1 B cells (13, 14). CD4⁺ and CD8⁺ T cells and CD20⁺ B cells also infiltrate human adipose tissue and are associated with IR (15, 16). Distinct innate lymphocyte, T cell, and B cell subpopulations have been shown in mouse models to regulate or compromise glucose homeostasis, suggesting changes in immune cell distributions may be important (13, 16–19).

In clinical studies, subjects with T2D were reported to have higher circulating levels of Th1 and Th17 cells and lower Tregs than those without disease (20–22). Subjects with T2D were also reported to have altered monocyte (increased IL-6, IL-8, TNF-α, and IL-1β) and CD19⁺ B cell (increased IL-8 and decreased IL-10) cytokine production as well as decreased expression of natural killer (NK) cell activation receptors (23–25). In participants of the community-based Multi-Ethnic Study of Atherosclerosis (MESA), we demonstrated that high proportions of CD4⁺ memory cells and low proportions of CD4⁺ naive cells in circulating blood are associated with prevalent T2D (26). This finding suggests either a role of chronic adaptive immune activation in the etiology of T2D or activation of these cells as a consequence of the disease (27, 28).

To date, evidence implicating monocytes, innate lymphocytes, T cells, and B cells in the pathogenesis of T2D in humans are limited to case-control or cross-sectional studies and their importance as risk factors for disease onset is unknown. To address this limitation, we measured 33 innate and adaptive immune cell phenotypes from cryopreserved cell samples collected at the MESA baseline examination (Exam 1 in 2000–2002). We evaluated distributions of the immune cell subsets by flow cytometry and investigated their longitudinal associations with incident T2D over a median follow-up time of 9.1 years. Based on our prior cross-sectional results in MESA (26), and findings from others (20, 21, 29–32), we hypothesized that higher proportions of CD4⁺ memory, CD4⁺CD28⁻, and Th17 cells, and lower proportions of CD4⁺ naive and Treg cells, would be associated with incident T2D risk. In secondary analyses, we evaluated associations of monocyte subsets, innate lymphocytes, CD4⁺ and CD8⁺ T cell, and CD19⁺ B cell subsets with incident T2D.

Materials and Methods

Study cohort

MESA is a prospective epidemiological cohort study of subclinical cardiovascular disease (CVD) (33). During the baseline examination (Exam 1) in 2000–2002, 6814 men and women from 6 US communities (Baltimore, MD, Chicago, IL, Los Angeles, CA, St. Paul, MN, New York, NY, and Forsyth County, NC) were enrolled. Participants were aged 45 to 84 years and were from 4 race/ethnic groups (European-American, African-American, Hispanic-American, and Chinese-American). Those with a self-reported history of clinical CVD or undergoing active treatment for cancer, pregnancy, or amputation were excluded from enrollment in the study. Follow-up examinations occurred in 2002–2004 (Exam 2), 2004–2005 (Exam 3), 2005–2007 (Exam 4), 2010–2012 (Exam 5), and 2016–2018 (Exam 6). At each examination, participants went to the MESA Field Center clinic where they answered standardized questionnaires, underwent assessment for CVD risk factors, and provided a fasting blood sample (33).

The current study, MESA-MI, is derived from an ancillary case-cohort study nested within the main MESA study. MESA-MI included 1195 participants who had cryopreserved cells available from the baseline examination (Exam 1 in 2000–2002). The case-cohort sample included all cases of incident myocardial infarction (MI) and incident angina (n = 484) and a cohort random sample (n = 711). All participants provided written informed consent for participation in the study and all procedures were conducted under institutionally approved protocols for human subjects research.

Cellular phenotyping

Peripheral blood mononuclear cells (PBMCs) were isolated during the MESA baseline examination (Exam 1), cryopreserved, and stored at −135°C. Cells were thawed in a 37°C water bath for 15 minutes, immediately treated with benzonase (250 U/mL in RPMI medium) for 20 minutes, and diluted 10-fold by sequential additions of RPMI supplemented with fetal bovine serum, L-glutamine and penicillin-stretpomycin (fsRPMI) (all cell culture reagents from Thermo Fisher Scientific, Waltham, MA) (34). Samples were centrifuged for 5 minutes at 200g, and the cell pellet was treated with 250 U/mL benzonase in fsRPMI for 10 minutes. Samples were washed and the final cell pellet was resuspended in 1 mL of fsRPMI and filtered through a 70 μM filter.

For intracellular cytokine staining assays (ie, Th1, Th2, Th17, type 1 CD8⁺ cells [Tc1], Tc2, and Tc17), samples were activated with phorbol myristate acetate and ionomycin in the presence of Brefeldin A as described previously (34). Cells were resuspended in phosphate buffered saline (PBS) (pH 7.4) and stained with a live/dead stain (Thermo Fisher Scientific) for 15 minutes at room temperature. Cells were incubated for 15 minutes at room temperature in the dark with Per-CP-Vio-conjugated anti-CD4, Per-CP-Vio-anti-CD8, or an isotype control in staining buffer (PBS, 1% bovine serum albumin (BSA), 10 μg/mL Brefeldin A). Cell pellets were washed with staining buffer and fixed with 2% paraformaldehyde (Alfa Aesar, Tewksbury, MA) for 10 minutes. Cells were permeabilized with 0.1% saponin and incubated with cytokine-specific antibodies. Cells surface labeled with CD4 and CD8 were treated for 15 minutes with PEVio-anti-interferon gamma (IFN-γ), PE-anti-IL-4, and APCVio-anti-IL17, or with a mixture of isotype controls.

Cellular phenotypes measured by surface labeling assays (eg, NK cells, γδ T cells, monocyte, B cell, and activated and senescent T cell subsets) were performed as described above for CD4 and CD8. Cells were fixed with paraformaldehyde and stored at 4°C with protection from light until evaluated. All antibodies were from Miltenyi Biotec (San Diego, CA). The markers used to characterize each of the immune cell phenotypes are presented in Table 1.

Table 1.

Characterizations of Immune Cell Phenotypes in MESA-MI

Phenotype	Cellular Markers	Cells Expressed as
Monocytes
Classical monocytes	CD14++CD16-	% of CD14+ monocytes
Intermediate monocytes	CD14+CD16+	% of CD14+ monocytes
Non-Classical monocytes	CD14+CD16++	% of CD14+ monocytes
Innate Lymphocytes
Natural killer cells	CD3-CD16+CD56+	% of lymphocytes
γδ T cells	CD3+γδTCR+	% of CD3+ cells
CD4 + T helper (Th) Cells
Pan CD4+ T cells	CD4+	% of lymphocytes
Th1	CD4+IFN-γ +	% of CD4+ cells
Th2	CD4+IL-4+	% of CD4+ cells
Th17	CD4+IL-17A+	% of CD4+ cells
T regulatory (Treg) cells	CD4+CD25+CD127−	% of CD4+ cells
Naive CD4+	CD4+CD45RA+	% of CD4+ cells
Memory CD4+	CD4+CD45RO+	% of CD4+ cells
Activated or Treg CD4+	CD4+CD25+	% of CD4+ cells
Activated / Mature CD4+	CD4+CD38+	% of CD4+ cells
Differentiated / Senescent	CD4+CD28−	% of CD4+ cells
Differentiated / Senescent	CD4+CD57+	% of CD4+ cells
Differentiated / Senescent	CD4+CD28−CD57+	% of CD4+ cells
TEMRA	CD4+CD28−CD57+CD45RA+	% of CD4+ cells
CD8 + Cytotoxic T cells (Tc)
Pan CD8+ T cells	CD8+	% of lymphocytes
Tc1	CD8+IFN-γ +	% of CD8+ cells
Tc2	CD8+IL-4+	% of CD8+ cells
Tc17	CD8+IL-17A+	% of CD8+ cells
Naive CD8+	CD8+CD45RA+	% of CD8+ cells
Memory CD8+	CD8+CD45RO+	% of CD8+ cells
Activated / Mature CD8+	CD8+CD38+	% of CD8+ cells
Differentiated / Senescent	CD8+CD28−	% of CD8+ cells
Differentiated / Senescent	CD8+CD57+	% of CD8+ cells
Differentiated / Senescent	CD8+CD28−CD57+	% of CD8+ cells
TEMRA	CD8+CD28−CD57+CD45RA+	% of CD8+ cells
B cells
Transitional B cells	CD19+CD5+	% of CD19+ B cells
Transitional B cells	CD19+CD5+CD27-	% of CD19+ B cells
Memory B cells	CD19+CD27+	% of CD19+ B cells
Memory B cells	CD19+CD5−CD27+	% of CD19+ B cells

Cell phenotypes were measured by flow cytometry using an MQ10 (Miltenyi Biotec) and analyzed with MACS Quantify software (Miltenyi Biotec). The cytometer was calibrated daily using calibration beads. Singlecolor compensation controls, prepared contemporaneously with the samples, were utilized for each assay to set compensation. Isotype controls were used to set negative gates. Cells were expressed as a percentage of their respective parent populations as described in Table 1. The cell distributions stratified by incident T2D case status are shown in Table 2.

Table 2.

Baseline Characteristics of the MESA-MI Study Population Stratified by Incident Type 2 Diabetes

Variable	No Diabetes (n = 812)	Incident Type 2 Diabetes (n = 120)	Value
Age, years (mean, SD)	63.4 (10.4)	62.7 (9.6)	0.51
Sex (n, %)			0.84
Female	387 (48)	56 (47)
Male	425 (52)	64 (53)
Race (n, %)			0.03
White	371 (46)	39 (32)
African-American	202 (25)	35 (29)
Hispanic-American	149 (18)	32 (27)
Chinese-American	90 (11)	14 (12)
Education status (n, %)			0.01
Less than college degree	495 (61)	88 (73)
College degree or above	315 (39)	32 (27)
Current smoker (n, %)	105 (13)	16 (13)	0.91
BMI, kg/m2 (mean, SD)	27.6 (5.0)	30.2 (4.6)	<0.0001
Waist circumference, cm (mean, SD)	97.2 (14.2)	103.4 (13.4)	<0.0001
Systolic blood pressure, mm Hg (mean, SD)	128 (21)	130 (18)	0.19
Diastolic blood pressure, mm Hg (mean, SD)	72 (11)	73 (9)	0.85
Hypertension (n, %)	385 (47)	60 (50)	0.60
Hypertension medication use (n, %)	277 (34)	45 (37)	0.47
LDL cholesterol, mg/dL (mean, SD)	119 (30)	118 (31)	0.77
HDL cholesterol, mg/dL (mean, SD)	51 (15)	46 (12)	0.002
Statin use (n, %)	118 (15)	22 (18)	0.27
Immune cell phenotype, % (median, 25th, 75th)
Classical Monocytes	76.2 (69.3, 81.9)	75.6 (68.4, 82.8)	0.99
Intermediate Monocytes	16.9 (12.8, 22.1)	16.9 (13.5, 22.0)	0.79
Non-Classical Monocytes	5.1 (2.7, 9.3)	5.1 (3.0, 8.9)	0.83
Natural Killer cells	2.5 (1.1, 6.7)	2.6 (1.5, 8.0)	0.09
γδ T cells	4.9 (2.7, 8.9)	4.5 (2.8, 8.2)	0.56
Pan CD4+ T cells	50.3 (42.5, 57.8)	49.1 (42.2, 56.6)	0.41
Th1	13.4 (9.0, 19.4)	14.6 (10.8, 19.8)	0.10
Th2	2.4 (1.6, 3.7)	2.4 (1.6, 3.6)	0.80
Th17	1.7 (1.3, 2.5)	1.7 (1.2, 2.9)	0.60
T regulatory cells	4.7 (3.5, 6.0)	4.7 (3.4, 6.4)	0.73
Naive CD4+	25.5 (18.1, 34.4)	25.5 (16.8, 32.8)	0.46
Memory CD4+	50.6 (41.2, 59.3)	52.3 (44.2, 61.9)	0.10
CD4+CD25+	30.9 (23.5, 39.6)	32.2 (24.7, 42.0)	0.20
CD4+CD38+	25.4 (18.1, 34.9)	22.4 (16.3, 34.7)	0.28
CD4+CD28−	11.8 (6.6, 18.8)	12.1 (7.4, 18.7)	0.66
CD4+CD57+	21.2 (12.5, 30.6)	21.5 (12.6, 31.3)	0.73
CD4+CD28−CD57+	7.5 (3.8, 14.0)	7.6 (4.6, 13.0)	0.72
CD4+ TEMRA	4.2 (2.0, 7.8)	4.3 (2.4, 8.1)	0.59
Pan CD8+ T cells	22.3 (17.1, 29.3)	21.4 (16.4, 27.8)	0.27
Tc1	39.9 (28.0, 55.1)	41.8 (33.9, 53.3)	0.20
Tc2	5.8 (3.8, 8.9)	5.3 (4.0, 8.2)	0.53
Tc17	3.9 (1.8, 6.9)	4.4 (1.9, 7.4)	0.44
Naive CD8+	52.3 (42.6, 62.3)	51.6 (40.9, 60.6)	0.46
Memory CD8+	20.1 (14.2, 27.8)	20.1 (13.3, 26.7)	0.56
CD8+CD38+	21.3 (14.6, 30.9)	22.3 (13.0, 29.3)	0.66
CD8+CD28−	56.7 (45.3, 68.0)	55.8 (43.0, 66.2)	0.49
CD8+CD57+	61.0 (48.1, 70.8)	61.9 (50.4, 72.6)	0.43
CD8+CD28−CD57+	44.8 (32.6, 57.2)	44.3 (33.5, 53.4)	0.79
CD8+ TEMRA	31.8 (21.5, 42.5)	32.6 (21.3, 39.7)	0.65
CD19+CD5+ B cells	55.1 (38.4, 72.3)	50.7 (37.6, 69.5)	0.57
CD19+CD5+CD27-	27.3 (16.5, 42.4)	27.2 (17.0, 43.2)	0.74
CD19+CD27+	42.7 (30.5, 54.6)	39.8 (30.3, 52.2)	0.23
CD19+CD5−CD27+	16.0 (8.5, 26.2)	16.3 (10.7, 23.1)	0.86

Differences in demographic and clinical characteristics were compared using t-tests and analysis of variance for continuous variables and χ ² statistics for categorical variables. Differences in cell distributions evaluated by Wilcoxon rank sum tests.

Laboratory measurements and definitions

Fasting blood glucose and lipid measurements were performed by a central laboratory at the University of Minnesota Medical Center. Serum glucose was measured by the Vitros analyzer (Johnson & Johnson Clinical Diagnostics, Inc., Rochester, NY). Lipid measurements were performed on EDTA plasma samples as described (35). Cytomegalovirus (CMV) was measured from samples collected during Exam 1 as described (36). Medication use was determined by a standardized medication inventory as a component of the MESA examination.

T2D was defined as a fasting blood glucose ≥ 126 mg/dL or use of insulin or an oral hypoglycemic medication (37). Body mass index (BMI) was calculated by dividing weight in kilograms by height in meters squared (kg/m²). Waist circumference measurements were in centimeters. Smoking was defined as never, former (no cigarettes within the past 30 days), or current. Hypertension was defined as self-reported use of antihypertensive medications, a systolic blood pressure ≥ 140 mm Hg, or diastolic blood pressure ≥ 90 mm Hg. Education status was defined by less than college degree or college degree or above.

Statistical analyses

To replicate our previous results in MESA (26), we performed a complete-case cross-sectional analysis of T2D. Associations between immune cell subsets and prevalent T2D were estimated using logistic regression models, adjusted for age, sex, race/ethnicity, and education. A separate model adjusted for these variables plus BMI. Models used case-cohort weights (38) to correct for the oversampling of MI and angina cases in the MESA-MI study.

Associations between immune cell phenotypes, evaluated per 1-standard deviation (SD) higher proportions, and incident T2D were estimated by Cox proportional hazards models. Participants with T2D at baseline (Exam 1) (n = 181) were excluded from incident event analyses. Cox models used case-cohort weights to correct for the sampling design and were adjusted as above. Confidence intervals used robust (sandwich) standard errors. We used multiple imputation (100 imputations) with chained equations to impute missing cell phenotype data from among the 1113 participants with at least one cell phenotype measured. Missing data were due to insufficient numbers of viable cells or technical errors, which occurred at random, and were not considered to be related to participant characteristics. Immune cell phenotypes, covariates, censoring indicators, and time-to-event data were included in the multiple imputation model. Cell subpopulations were modeled singly to avoid collinearity. We used a Bonferroni correction to control for multiple hypothesis testing. The statistical significance threshold was defined as P < 0.01 for our 5 primary hypotheses and P < 0.0015 in secondary analyses.

In sensitivity analyses, we used parametric Weibull models that incorporated interval-censoring to account for T2D status being known only at study visits. We also evaluated nonlinear response relationships of the immune cell traits specified in our a priori hypotheses with T2D using cubic spline functions (39). In separate analyses, waist circumference replaced BMI as a covariate and additional adjustment for LDL, HDL, systolic blood pressure, hypertension, and CMV serology were included. CMV serology was included due to its known relationships with CD4⁺ memory and naive cells in MESA (40). Statistical analyses were performed using SAS 9.4 and R 3.2.3.

Results

In the cross-sectional replication analysis of 1113 participants with cell phenotypes measured from samples cryopreserved at the MESA baseline examination (Exam 1), there were 181 cases of prevalent T2D. Higher proportions (1-SD) of memory cells (13.4%; defined as CD4⁺CD45RO⁺) (odds ratio [OR]: 1.57; 95% confidence interval (CI): 1.22, 2.02) and lower proportions of naive cells (12.0%; defined as CD4⁺CD45RA⁺) (OR: 0.70; 95% CI: 0.53, 0.93) were associated with prevalent T2D (adjusted for demographics and BMI). Since the surface marker CD45RA⁺ may also include populations of differentiated CD45RA⁺ re-expressing effector memory (TEMRA) cells, we analyzed CD4⁺ TEMRA cells (defined as CD4⁺CD28⁻CD57⁺CD45RA⁺). These were not significantly associated with prevalent T2D (OR: 0.76; 95% CI: 0.54, 1.06 per 5.3% higher). There were also no relationships between Th1 (defined as CD4⁺IFN-γ ⁺) (OR: 0.86; 95% CI: 0.62, 1.19 per 9.0% higher) or Th2 cells (defined as CD4⁺IL-4⁺) (OR: 1.03; 95% CI: 0.77, 1.39 per 1.7% higher) and prevalent T2D, consistent with our previous results (26).

In a longitudinal analysis of 932 participants without T2D at the baseline examination (Exam 1), there were 120 incident T2D cases over a median 9.1 years of follow-up (interquartile range: 4.8–9.5 years). Those who developed T2D tended to be of non-European race/ethnicity, have higher BMI and waist circumference, and lower HDL-cholesterol and education status (Table 2).

Associations of the 5 cell populations specified in a priori hypotheses with incident T2D (CD4⁺ naive, CD4⁺ memory, senescent CD4⁺ cells [defined as CD4⁺CD28⁻], Th17 [defined as CD4⁺IL-17A⁺], and Tregs [defined as CD4⁺CD25⁺CD127⁻] are shown in Table 3. There were no statistically significant associations among any of these cell populations with T2D risk (all P > 0.05) in models adjusted for age, sex, race/ethnicity, and education status (demographic model). Results were similar with additional adjustment for BMI (adiposity model).

Table 3.

Hazards Ratios of Type 2 Diabetes per SD Higher Proportion of Immune Cells Specified As Primary Hypotheses

CD4+ T Cell Subset	Hazard Ratio (95% CI) Model 1	Hazard Ratio (95% CI) Model 2
Naive (CD4+CD45RA+) (12.0%)	1.06 (0.83, 1.35)	1.03 (0.81, 1.31)
Memory (CD4+CD45RO+) (13.4%)	0.99 (0.78, 1.27)	0.99 (0.78, 1.26)
Senescent (CD4+CD28−) (10.0%)	0.91 (0.70, 1.18)	0.90 (0.69, 1.18)
Th17 (CD4+IL-17A+) (1.4%)	0.99 (0.72, 1.37)	0.96 (0.69, 1.33)
T Regulatory (CD4+CD25+CD127−) (2.2%)	1.06 (0.82, 1.36)	1.05 (0.81, 1.36)

CD4⁺ T cell subsets were expressed as a percentage of CD4⁺ lymphocytes and analyzed per 1-SD higher values (shown in parentheses) using Cox proportional hazards models with sampling weights. Confidence intervals (CI) used robust standard error estimates.

Model 1: Age, sex, race/ethnicity, and education status; Model 2: Model 1 + BMI.

In secondary analyses, the associations of 28 additional innate and adaptive immune cell phenotypes with incident T2D were estimated (per 1-SD higher value). Phenotyping assays included monocyte subsets, innate lymphocytes (ie, natural killer, γδ T), activated, differentiated, and TEMRA CD4⁺ and CD8⁺ T cells, and CD19⁺ B cell subsets (as presented in Table 1). None of the associations of these cell populations with incident T2D were statistically significant (defined by a Bonferroni-adjusted P value < 0.0015) in either the demographic- or adiposity-adjusted models (Fig. 1). Trends towards an association (P < 0.05) were observed with incident T2D for 2 subsets of CD19⁺ B cells expressing the memory marker CD27. Each of the 2 CD19⁺CD27⁺ subsets showed trends for an association with lower T2D risk in adiposity-adjusted models (hazard ratio [HR] per 16.3% higher proportion of CD19⁺CD27⁺ cells: 0.72 [95% CI: 0.56, 0.93], P = 0.01; HR per 12.7% higher CD19⁺CD27⁺CD5⁻ cells: 0.80 [95% CI: 0.63, 0.99], P = 0.04). A third phenotype with suggestive trends for an associations with incident T2D was CD4⁺CD25⁺ activated or Treg cells (HR per 11.4% higher: 1.42; 95% CI: 1.07, 1.88; P = 0.01) (Fig. 1).

Figure 1.

Hazards ratios of type 2 diabetes per SD higher proportion of immune cells included as secondary analyses. Cell subsets were analyzed per 1-SD higher values using Cox models with sampling weights and robust standard error estimates. Models were adjusted for age, sex, race/ethnicity, education status, and BMI.

The results were similar in sensitivity analyses using parametric Weibull models in place of Cox proportional hazards models; evaluating nonlinear response relationships; adjusting for waist circumference in place of BMI; and including additional adjustment for LDL, HDL, systolic blood pressure, hypertension, and CMV serology.

Discussion

In our previous cross-sectional case-cohort study in MESA (during Exam 4), higher memory and lower naive CD4⁺ cells (defined by the surface markers CD45RO⁺ and CD45RA⁺, respectively and measured using unfrozen 12 to 24-hour-old blood samples), were associated with prevalent T2D (26). In this study, prior findings with prevalent T2D were replicated among a different participant sample at a different MESA examination year using cryopreserved cells. In our longitudinal analyses, however, we did not identify statistically significant associations of CD4⁺ memory, naive, senescent, Th17, or Treg cells with the onset of T2D, despite these populations being implicated in T2D risk in prior human studies. Lower CD19⁺CD27⁺ B cells, and higher CD4⁺CD25⁺ T cells, had suggestive trends for associations with increased T2D risk in secondary analyses, but these findings were not statistically significant after correcting for multiple cell phenotypes tested.

Although several mouse models have demonstrated a role for T and B cells in glucose homoeostasis and IR (13), limited human studies exist. Our cross-sectional findings in MESA are consistent with recent results from the Malmö SUMMIT cohort that demonstrates higher percentages of CD4⁺ central and effector memory subpopulations (defined as CD4⁺CD45RO⁺CD62L⁺ and CD4⁺CD45RO⁺CD62L⁻, respectively) and lower percentages of CD4⁺ naive cells (defined as CD4⁺CD45RO⁻CD62L⁺) among those with T2D than without (31, 41). Higher proportions of CD4⁺ cells expressing the markers CD45RA⁻CD45RO⁺ (characterized as antigen-experienced), and lower proportions of CD4⁺ cells expressing CD45RA⁺CD45RO⁻ (characterized as total naive cells) were also associated with T2D in the Malmö SUMMIT study, consistent with results in MESA. Similarly, CD4⁺ memory cells (defined by CD4⁺CD45RO⁺) were associated with features of the metabolic syndrome in a cross-sectional study of Japanese men (42).

In case-control studies, higher Th1, Th17, and senescent (defined by CD4⁺CD28⁻) cells, and lower Tregs (defined by CD4⁺CD25⁺Foxp3⁺ or CD4⁺CD25^hiCD127⁻) were observed in the peripheral blood of subjects with T2D (20, 21, 29). Prospective associations between lymphocyte subsets and incident T2D have not been reported, to our knowledge. Our results suggest variation in the proportions of peripheral blood CD4⁺ naive, memory, senescent, Th1, Th17, and Treg cells are not associated with incident T2D.

Our null results are generally consistent with clinical studies failing to show efficacy of immunosuppressive therapies on reducing incident T2D. In the CANTOS trial, which included 4960 patients with prediabetes, IL-1β inhibition by canakinumab over a 5-year period did not reduce the development of new-onset diabetes (43). Blockade of TNF-α has not demonstrated a benefit on insulin sensitivity (44), despite the known ability of TNF-α to inhibit insulin receptor signaling (10). Although some experimental evidence suggests metformin, the first-line medication for type 2 diabetes, has anti-inflammatory effects on immune cells (45), results from human studies are less clear. For example, in the LANCET Trial, metformin treatment did not reduce circulating levels of CRP, IL-6, or soluble TNF receptor 2 (sTNFr2) in patients with recent-onset T2D (46). Our results are also consistent with some murine studies demonstrating that T cell activation—or suppression—was not sufficient to modulate glucose homeostasis or IR (47, 48).

In contrast to preceding disease onset, our findings suggest that prevalent T2D, or its concomitant features (eg, altered glucose homeostasis, pharmacologic treatment, formation of advanced glycation end products), may alter the proportions of memory and naive CD4⁺ T cells. These alterations may reflect, or promote, the state of chronic inflammation associated with T2D. However, inflammatory T cells in human adipose tissue correlate with IR (15) and we cannot rule out a role of adipose tissue T cells towards T2D risk. Cell proportions in peripheral blood may not reflect the functional contributions that promote disease pathogenesis, such as cytokine secretion and intercellular signaling.

Although findings were only suggestive, and not statistically significant, after correcting for multiple hypothesis testing (using a conservative Bonferroni approach of P < 0.0015), the trends of higher proportions of CD19⁺CD27⁺ B cells with decreased T2D risk are consistent with emerging literature describing roles of distinct B cell subsets in attenuating IR. B-1a, B-1b, and B regulatory cells protected against murine IR through the secretion of natural IgM antibodies or IL-10 (16, 49, 50). In bariatric patients, IgM antibodies to specific oxidized epitopes correlated with lower IR (16). In contrast, B-2 B cells have been implicated in promoting glucose intolerance and IR in the mouse through the production of IgG antibodies (51). In obese men, a distinct profile of IgG autoantibodies was associated with IR (51). B cells measured from the blood of patients with T2D were suggested to promote inflammation through increased secretion of IL-8 and decreased secretion of IL-10 (24). The trend of higher CD19⁺CD27⁺ B cells with lower T2D risk observed here is consistent with a role of IL-10 and/or natural IgM antibody-producing B cells in attenuating inflammation and glucose intolerance.

High proportions of CD4⁺ T cells expressing CD25⁺ also showed suggestive trends for an association with increased T2D risk, although this did not exceed the Bonferroni significance threshold. CD25 (the IL-2 receptor alpha chain [IL-2ra]) is expressed on the surface of activated CD4⁺ and Treg cells and the structural IL2RA gene is a known type 1 diabetes risk locus (52). Tregs are immunomodulatory and require evaluation of additional phenotypic markers in humans (eg, FoxP3⁺, CD127⁻) (53). Since Tregs (defined by CD4⁺CD25⁺CD127⁻) were not associated with incident T2D in our analyses, the trend for an association between higher CD4⁺CD25⁺ cells and increased diabetes risk may reflect activated T helper cell populations. This interpretation is consistent with results from the German Diabetes Study showing higher CD4⁺CD25⁺ cells in those with T2D compared with controls (54), and with correlations of activated CD4⁺CD25⁺ cells with higher C peptide and lower insulin sensitivity in T2D subjects (55). This cell population, however, may be heterogeneous and these results should be interpreted with caution.

The strengths of this study include the longitudinal cohort design, composed of participants free of clinical CVD at the time cells were collected, and the broad panel of innate and adaptive immune cell phenotypes included in the analysis. Limitations include the case-cohort study design being selected for cases of incident MI and not incident T2D. The number of incident T2D cases may have been too small to detect moderate or weak associations. Our study also does not rule out the importance of immune cell function (eg, cytokine production) towards T2D risk. Differences in sensitivity of immune cell subsets to freezing and thawing is not well established, although possible. As such, the use of cryopreserved, as opposed to freshly-isolated, cells may be a limitation. However, the current replication of previously reported associations of memory and naive CD4⁺ cells (phenotyped in the prior study using unfrozen whole blood) with prevalent T2D (26) among a different MESA participant sample at a different MESA examination supports the validity of using cryopreserved cells.

In summary, our findings suggest that the frequencies of CD14⁺ monocyte, innate lymphocyte (NK and γδ T cells), and CD4⁺ and CD8⁺ T cell subsets in peripheral blood are not significant risk factors for incident T2D. In contrast, alterations in the proportions of memory and naive CD4⁺ T cells may reflect an outcome of having T2D or its physiologic alterations due to IR, obesity, or pharmacologic treatments. Higher levels of CD19⁺CD27⁺ B cells may reflect lower T2D risk, but these findings require confirmation in other prospective cohorts.

Glossary

Abbreviations

BMI

body mass index

CI

confidence interval

CMV

cytomegalovirus

CRP

C-reactive protein

CVD

cardiovascular disease

fsRPMI

Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with fetal bovine serum, L-glutamine, and penicillin-streptomycin

HDL

high-density lipoprotein

HR

hazard ratio

ICAM-1

intercellular adhesion molecule-1

IFN-γ

interferon-γ

IL-1β

interleukin-1β

IL-10

interleukin 10

IL-17

interleukin 17

IL-18

interleukin 18

IL-6

interleukin 6

IL-8

interleukin 8

IR

insulin resistance

LDL

low-density lipoprotein

MESA

Multi-Ethnic Study of Atherosclerosis

MI

myocardial infarction

NK

natural killer

OR

odds ratio

PAI-1

plasminogen activator inhibitor-1

SD

standard deviation

sTNFr2

soluble tumor necrosis factor receptor 2

T2D

type 2 diabetes

Tc

T cytotoxic

TEMRA

CD45RA+ re-expressing effector memory T cells

Th

T helper

TNF-α

tumor necrosis factor-α

Treg

T regulatory

VAT

visceral adipose tissue

Additional Information

Disclosure Summary: The authors have nothing to disclose.

Data Availability: As a National Heart, Lung, and Blood Institute-funded epidemiological cohort study, MESA follows the NHLBI’s policies for data and resource sharing. Researchers interested in working with MESA investigators, data, and resources are encouraged to contact the study’s Coordinating Center at chsccweb@u.washington.edu. Information for new investigators, including the study’s data distribution policy, instructions for completing the DMDA, and publications and presentations policies can be found here: https://www.mesa-nhlbi.org/Publications.aspx.