Cryopreserved PBMC were acquired as part of an ancillary study approved by TrialNet from the Pathway to Prevention study of 1st and 2nd degree relatives of individuals with type 1 diabetes. Aside from the familial association, study eligibility is limited to those 1st and 2nd degree relatives who are not on medication to control hyperglycemia, not currently using immunosuppressive or immunomodulatory agents, have no known severe active disease, and do not have diabetes or a history of diabetes. From this pool of subjects, we received 79 blinded samples comprising 3 subgroups of subjects. Twenty-five samples were from autoantibody negative (AA–) 1st and 2nd degree relatives of diagnosed type 1 diabetics or seroconverted, at-risk individuals. Having no autoantibodies, yet sharing some environmental and genetic risk, these samples constitute our control group. Importantly, these AA– subjects were not related to the seroconverted, at-risk individuals analyzed in our study. The remaining 54 subjects, labeled “seroconverted,” had previously seroconverted for one or more autoantibodies and are at increased risk of developing T1D (15, 16). Autoantibody testing was performed by TrialNet. Individuals are screened for GAD65, IA-2, and IAA. If any of these measurements were positive, they were also then screened for ZnT8 and ICA. Positivity thresholds were as follows: IAA, >0.010; GAD65, >0.032; ICA512, >0.049; GAD65H, >20; IA-2H, >5; ZnT8, >0.020; ICA, ≥10. Once an individual has tested positive for an autoantibody twice within one year, they are considered “autoantibody positive” or seroconverted to the production of autoantibodies targeting islet antigens. Among the seroconverted subset, 26 samples were from individuals who have not yet progressed to clinical disease, and for our analysis these are labeled “non-progressor.” The remaining 28 samples were from individuals who ultimately progressed to clinical disease. This subgroup is labeled as “progressor.” Our request for samples from TrialNet did not specify particular HLA-risk alleles, therefore our sample set constituted a mix of HLA types. Although we did request samples from subjects between 8 and 11 years old, upon unblinding the samples we discovered our subjects spanned a much broader age range. Patient data is presented in Table 1.

Surface panel design and analysis was performed using a BD LSR II equipped with 5 lasers: UV (355 nm) for Brilliant Ultraviolet (BUV) 395 and 737 and LiveDead UV Blue; Violet (405 nm) for Brilliant Violet (BV) 421, 510, 605, 650, 711, and 785; Blue (488 nm) for FITC and PerCP-Cy5.5; Yellow-Green (561 nm) for PE, PE-CF594, and PE-Cy7; and Red(633 nm) for APC and Alexa Fluor(AF) 700. The following antibodies and reagents were used: from Biolegend CCR7(PE, clone G043H7), CCR4(PE-Cy7, clone L291H4), CD14(BV711, clone M5E2), CD19(BV711, clone HIB19), CD27(BV785, clone O323), CD57(FITC, clone HCD57), CD45RA(PerCP-Cy5.5, clone HI100), Vα7.2(BV605, clone 3C10), CD45(AF700, clone HI30), Streptavidin-APC; from BD Biosciences CXCR5(BV421, clone RF8B2), CD3(BUV737, clone UCHT1), CD4(BUV395, clone RPA-T4), CD8(BV510, clone RPA-T8), CD28(PE-CF594, clone CD28.2), CD161(BV650, clone DX12); from eBioscience CD127(Biotin, clone eBioRDR5); from life technologies LiveDead UV Blue. Cryopreserved PBMC were thawed and rested overnight in X-VIVO 15 (Lonza) supplemented with human AB serum (MP Biomedicals) at 2% v/v at 37° C with 5% CO₂. Samples were then counted and distributed to 96 well plates for labeling following methods described previously (17) with two modifications: Fc receptors were blocked using human IgG (Sigma Aldrich) and antibody cocktails were prepared using Brilliant Stain Buffer (BD Biosciences).

For FOXP3 analysis, lymphocytes from healthy human donors were acquired from the UNMC elutriation core facility, cryopreserved, thawed, rested and surface labeled as described above. We then utilized the True-Nuclear Transcription Factor Buffer set (BioLegend) according to manufacturer's recommendations. In panel validation, we compared PE-conjugated FOXP3 clones 206D (BioLegend), PCH101 (eBioscience), and 259D/C7 (BD Biosciences) and clone 206D gave the greatest signal-to-noise ratio (data not shown). The final panel included the aforementioned FOXP3, CD25 (VioBright FITC, clone 4E3, Miltenyi), and the conjugated-clones and reagents listed above for LiveDead, CD3, CD4, CXCR5, CCR4, and CD127 with streptavidin-APC.

Data analysis was conducted in a blinded fashion using FlowJo (v10.2, TreeStar). All samples were pregated to exclude doublets, debris, dead cells, and non T cells (lineage negative) and then multiple subsets were identified and gated according to known T cell differentiation pathways. A representative gating procedure for all subsets explored is shown in Supplemental Figure 1. Following analysis, raw data files and tabular results were uploaded to TrialNet, upon which we received sample identification, clinical data, and autoantibody levels in order to perform statistical analysis.

We received blinded vials of plasma from the majority of the subjects in our PBMC analysis: 25/25 AA–, 23/26 non-progressors, and 27/28 progressors. Prior to performing assays, the vials of plasma were thawed, divided into several aliquots, refrozen, and stored at −80°C in order to run several assays and limit sample degradation from multiple freeze/thaw cycles. The following anti-viral immunoglobulin ELISAs from IBL America were performed according to manufacturer's directions: CMV IgM (IB19206), CMV IgG (EG101), EBV VCA IgM (IB79231), EBV VCA IgG (IB79230), enterovirus IgM(IB05053), and enterovirus IgG(IB05052). All sample dilutions contained Heteroblock (Omega Biologicals Inc.) to eliminate potential interference resulting from the presence of heterophilic antibodies. Optical densities were measured using a Synergy H1 Hybrid Multi-Mode Reader (BioTek Instruments, Inc.). Sample positivity for viral immunoglobulins was determined following calculations defined per assay. If sample positivity was equivocal, the sample was retested. If the sample was also equivocal on the follow-up test, it was labeled as “gray” for purposes of stratification and treated as negative. Upon completion of our plasma analysis, tabular data was uploaded to TrialNet. We then received sample identification data for pairing with our PBMC results and subsequent statistical analysis.

For mean comparisons, natural log transformation was applied to the data prior to analysis. For comparison of two groups, we used Student's T-test and for comparisons of three groups, we used analysis of variance (ANOVA). If the results of the ANOVA were significant, then pairwise comparisons were made, adjusting for multiple comparisons with Tukey's method. For correlations, data were not transformed and we used Spearman's Rank-Order correlation. In all cases, figures and descriptive data depict non-transformed data to facilitate interpretation. Figures were made using Prism (v6.03 GraphPad Software Inc.), and data was analyzed using Prism and SAS software versions 9.4 (SAS Institute Inc., Cary, NC).

MAIT cells are regularly found in human blood, the liver, and the gastrointestinal mucosa (18), and respond to microbially-derived vitamin B metabolites presented through MR1 (19). MAIT cells are capable of directly lysing bacterially-sensitized and infected target cells (20, 21) and can produce the proinflammatory cytokines IFN-γ, TNF-α, and IL-17 (18). Phenotypically, in the absence of MR1-loaded multimers, MAIT cells are unambiguously identifiable in blood by the high CD161 expression, presence of Vα7.2 T cell receptor, and T cell lineage marker CD3 (18, 22). The circulating MAIT cell lineage in healthy subjects is also regularly CD28+, CD127+, CD45RA^low, and CCR7^low (18, 23), but exceptions to this phenotype do exist in certain disease states [e.g., chronic HIV (24)], as well as in fetal tissue and cord blood (25, 26). MAIT cells are generally CD8+ (~85%) with the remaining being chiefly DN (CD4–, CD8–), although CD4+ and DP (CD4+, CD8+) groups are frequently observable in scant amounts in healthy donors. The functional differences among these MAIT cell subsets have not been fully clarified, although it has recently been reported that the CD8+ subset possesses somewhat greater effector potential than the DN subset (27).

We identified total MAIT cells as CD3+, Vα7.2+, CD161+, CD45RA^low, CCR7^low, CD28+, CD127+ events and also divided them by CD4 and CD8 expression (Figure 1A). With this approach, we observed a sharply reduced frequency of MAIT cells among seroconverted individuals (Figures 1B,D). This reduction extended into all MAIT cell subsets as defined by CD4 and CD8 expression, and appeared most acute within the major compartments identified as CD8+ and DN (Supplemental Figures 2A,B). When evaluating the seroconverted subjects according to disease progression, we found the non-progressors harbored the fewest MAIT cells, while the reduction among the progressors was not significant (Figures 1C,D). These data indicate that a MAIT cell reduction is associated with seroconversion and this effect is slightly ameliorated among those who progress to disease.

Conventional αβ CD8 T cells are the chief cytotoxic arm of adaptive immunity. Upon licensing and activation, CD8 T cells rapidly proliferate in order to destroy microbially-infected or transformed cells (28). Following the eradication of targets, effector cells die off while memory populations remain in tissue and circulation. This memory provides a quick recall response to repeat offenders and is considered a hallmark of T cell immunity. Outside of this role as defenders, CD8 T cells have been incriminated in the pathogenesis of several autoimmune diseases (29).

In our analysis of CD8 T cells, we first divided CD8 T cells according to presence or absence of costimulatory protein CD28. Among the CD28+ CD8 T cells, we removed MAIT cells and analyzed the remaining CD8 T cells for expression levels of CCR4, CXCR5, CD45RA, CD27, CD127, CCR7, and CD57. Similar to what was observed among MAIT cells, seroconverted subjects possessed reduced proportions of CD45RA+, CCR7^dim/−, CD57–, CD27+, CD127^high, CXCR5–, CCR4– CD8 T cells compared to AA– subjects (Figures 2A,B,D and Supplemental Figure 3). This combination of markers can best be described as a CD45RA+ memory subset following current differentiation paradigms (30, 31). Upon dividing the seroconverted into non-progressor and progressor subsets, we found that the non-progressors were significantly reduced in proportion, while the progressors did not approach significance (Figures 2C,D and Supplemental Figure 3).

The absence of CD28 on human T cells is indicative of chronic stimulation (32). We divided CD28- CD8 T cells by expression of CD57, CD127, and CD27 to characterize this antigen-experienced compartment. Our analysis revealed an elevated frequency of CD127–, CD27–, CD57+, CD28– CD8 T cells among seroconverted subjects (Figures 3A,B,D). This combined phenotype indicates terminal differentiation, cytotoxic potential via perforin production, and short-lived status (33–36). Thus, we abbreviated their phenotype as SLEC for short-lived effector-like cells. Intriguingly, this expansion of SLEC was most prominent among progressors (Figures 3C,D), suggesting an acute pathogen response may be associated with disease progression.

C-C chemokine receptor 4 (CCR4) expression among CD4 T cells suggests previous T cell receptor engagement (37, 38) as well as chemotactic responsiveness to thymus and activation-regulated chemokine (TARC), macrophage-derived chemokine (MDC), and chemokine like factor 1 (CKLF1) (39–41). Rather than solely indicating “Th2” status, CCR4– expressing CD4 T cells can be enriched for IFN-γ, IL-22, IL-17, and/or IL-4 production, as well as possess regulatory function (42–46).

Our analysis of CD4 T cells from TrialNet donors revealed reductions in frequency of three CCR4– expressing subsets among seroconverted subjects. Memory (CD127^bright, CD27+, CCR7–, CCR4+, CXCR5–; Figures 4A–D), Treg-like (CD127^dim, CD27+, CCR7–, CCR4+, CXCR5–; Figures 5A–D), and T follicular helper-like (CCR4+, CXCR5+, CD161–; Supplemental Figure 4) CD4 T cell subsets were all reduced among seroconverted subjects. For each compartment, the reduction in frequency was most profound among non-progressors and approached “normal” levels among progressors.

As our T cell panel did not include CD25 nor FOXP3, we were unable to cross-confirm regulatory phenotype among the Treg-like, CCR4-expressing CD4 T cells that we observed contracted among TrialNet samples. However, we analyzed CD4 T cells from 3 healthy human donors and found that the CCR4+, CXCR5– compartment possesses a substantial FOXP3+, CD25+ subset, which is also CD127^dim/−, a combined phenotype indicating regulatory potential (Figure 6). This is in agreement with previous studies indicating CCR4+ and CD127^dim/low CD4 T cells contain Tregs (47–49). Thus, it's highly likely that the observed Treg-like CCR4+ CD4 T cells we found to be reduced among seroconverted subjects contains T cells with regulatory potential.

We observed altered frequencies of MAIT, CCR7^dim/− CD45RA+ CD8 memory, SLEC, and CCR4+ CD4 T cells, including memory and Treg-like subsets. As mentioned previously, it's widely presumed that T1D is driven by autoreactive CD8 T cells which have escaped tolerance. Since Tregs can effectively suppress T cell proliferation in vitro (50), it's plausible that such subsets may impact T cell expansions and contractions in vivo. Therefore, we examined some of the CD4 and CD8 T cell subsets described above for relationships with our CD127^dim Treg-like subset.

First, we compared frequencies of T cell subsets using Spearman's Rank-order correlation. We observed no significant correlations among CD127^dim Treg-like CD4 T cells with either MAIT cells, SLEC CD8 T cells or CCR7^dim, CD45RA+ CD8 T cells (Figures 7A,C,D). Alternatively, among seroconverted subjects CD127^bright memory CD4 T cells were positively and significantly correlated with the frequency of CD127^dim Treg-like CD4 T cells, while there was no significant relationship among AA– subjects (Figure 7B). Further division of the seroconverted group into non-progressor and progressor demonstrated similar strong and significant positive correlations among both groups (data not shown). In total, these correlation data suggest that the frequency dynamics of the CD127^dim Treg-like cells do not negatively regulate frequencies of the noted T cell compartments in the periphery. However, they do demonstrate comparable and unique dynamics for CD127^bright memory CD4 T cells and CD127^dim CD4 T cells among only the seroconverted subjects. This could be due to similar effects driving the reductions observed in these two subsets.

In addition to the correlations comparing frequency of T cell subsets, we also examined relationships between autoantibody levels and frequency of T cells. Such a correlation could reveal if any of these subsets are associated with islet immunity. Interestingly, we observed a positive and significant correlation with islet cell antibody (ICA)values and the CCR7^dim CD45RA+, CD8 T cell subset (Figure 7E). Furthermore, we found that the SLEC subset was positively and significantly correlated with harmonized insulinoma antigen 2 (IA-2H) levels (Figure 7F). We observed no significant correlations among other T cell subsets with autoantibody levels (data not shown).

Along with these correlations, we examined ratios of frequency of effectors to frequency of regulators. Here, MAIT cells were hypothesized to function as either effector or regulator, as a recent report has suggested MAIT cells may have a regulatory function (51). We observed that the ratios of SLEC to CD127^dim or MAIT cells were significantly elevated among seroconverted, as well as when divided among non-progressors and progressors (Figures 8A,B,G,H). Alternatively, the ratio of MAIT cells to CD127^dim was relatively similar when comparing AA- and seroconverted subjects, including progressors and non-progressors (Figures 8C,D). The ratio of CCR7^dim CD45RA+ CD8 T cells to CD127^dim CD4 T cells was also similar among AA– and seroconverted subjects, and this continued when seroconverted subjects were divided into non-progressors and progressors (Figures 8E,F).

To determine if viral exposure was influencing the T cell changes associated with seroconversion, we tested plasma for abundance of anti-viral immunoglobulins using ELISA. We then performed correlations comparing T cell frequency with relative abundance of viral immunoglobulins as gauged by optical density values. Neither MAIT cells nor CD127^dim Treg-like cells were correlated with abundance of CMV, EBV VCA and enterovirus IgG (Figures 9A,D) nor IgM (data not shown). This was not the case for conventional αβ CD8 T cells. We observed a weak but significant association between EBV VCA IgG and CCR7^dim CD45RA+ CD8 T cells (Figure 9B). However, there was no significant association among this subset with CMV IgG or enterovirus IgG (Figure 9B), nor any viral IgM tested (data not shown). Finally, we observed a strong and significant association among SLEC and CMV IgG, but not with EBV VCA IgG, enterovirus IgG, nor any viral IgM tested (Figure 9C and data not shown).

Since the SLEC subset appeared to be strongly influenced by CMV, we stratified our subjects according to CMV IgG positivity and then retested the entirety of the CD28– CD8 T cell compartment by ANOVA. We observed that the SLEC expansion is greatest among CMV+ progressors (Figure 10A). Furthermore, we observed that the CMV+ progressors possessed elevated frequencies of additional chronically-stimulated CD28– subsets (Figures 10B–E). While these lacked the expression of CD57, they were heterogeneous for CD127 and CD27 expression. In total, these data reveal a relationship among type 1 diabetes progression, terminally-differentiated effector-like CD8 T cells, and existing cytomegalovirus infection among at-risk, seroconverted subjects.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.