Research ArticleAUTOIMMUNITY

LAG3 limits regulatory T cell proliferation and function in autoimmune diabetes

See allHide authors and affiliations

Science Immunology  31 Mar 2017:
Vol. 2, Issue 9, eaah4569
DOI: 10.1126/sciimmunol.aah4569

Regulating the regulators

Inhibitory receptors on T cells, including lymphocyte activation gene 3 (LAG3), serve as brakes that limit immune-mediated damage to the host. LAG3 is expressed by exhausted conventional T cells in the tumor microenvironment and has emerged as a key target for tumor immunotherapy. The role of LAG3 in regulatory T cells (Tregs) has remained unclear. Using a mouse model of autoimmune diabetes, Zhang et al. report that Treg-specific deletion of LAG3 led to enhanced Treg proliferation and reduced the incidence of type 1 diabetes. Their studies highlight the cell-type dependence and context specificity of the role of LAG3 and call for a more holistic assessment of the functions of inhibitory receptors that emerge as targets for tumor immunotherapies.


Inhibitory receptors (IRs) are pivotal in controlling T cell homeostasis because of their intrinsic regulation of conventional effector T (Tconv) cell proliferation, viability, and function. However, the role of IRs on regulatory T cells (Tregs) remains obscure because they could be required for suppressive activity and/or limit Treg function. We evaluated the role of lymphocyte activation gene 3 (LAG3; CD223) on Tregs by generating mice in which LAG3 is absent on the cell surface of Tregs in a murine model of type 1 diabetes. Unexpectedly, mice that lacked LAG3 expression on Tregs exhibited reduced autoimmune diabetes, consistent with enhanced Treg proliferation and function. Whereas the transcriptional landscape of peripheral wild-type (WT) and Lag3-deficient Tregs was largely comparable, substantial differences between intra-islet Tregs were evident and involved a subset of genes and pathways that promote Treg maintenance and function. Consistent with these observations, Lag3-deficient Tregs outcompeted WT Tregs in the islets but not in the periphery in cotransfer experiments because of enhanced interleukin-2–signal transducer and activator of transcription 5 signaling and increased Eos expression. Our study suggests that LAG3 intrinsically limits Treg proliferation and function at inflammatory sites, promotes autoimmunity in a chronic autoimmune-prone environment, and may contribute to Treg insufficiency in autoimmune disease.


Inhibitory receptors (IRs) are pivotal in controlling and shaping the host immune response. Insufficient coinhibition can lead to the breakdown of self-tolerance, whereas chronic use of inhibitory pathways constitutes a major barrier to effective antitumor immunity (13). Tumor-infiltrating lymphocytes up-regulate multiple IRs and exhibit an exhausted T cell phenotype as a consequence of chronic antigen stimulation (3). Recent clinical trials have highlighted the therapeutic benefit of IR blockade, called checkpoint inhibition (e.g., anti-CTLA4, anti-PD1, and anti–PD-L1) (2, 3). However, not all patients benefit from this, leading to speculation that compensatory mechanisms may be elevated in nonresponders and that additional IRs may need to be blocked. Lymphocyte activation gene 3 (LAG3) is the most recent IR to be targeted in the clinic.

LAG3 is expressed on activated T cells, intrinsically limiting conventional T (Tconv) cell proliferation, expansion, and viability (47). LAG3 is required for maintaining self-tolerance in an autoimmune-prone environment (1, 8). Mice on the nonobese diabetic (NOD) background deficient in Lag3 exhibit accelerated autoimmune diabetes with 100% penetrance (8). In contrast, LAG3 contributes to tumor-mediated immunosuppression and promotes tumoral immune escape (9, 10). Although LAG3 blockade or deficiency alone only has a minimal effect on tumor growth in mouse models, combinatorial PD1 blockade or deletion leads to tumor clearance (9).

LAG3, as well as other IRs, are also highly expressed on regulatory T cells (Tregs) (1114), a critical suppressive subpopulation of T cells that prevents autoimmunity but limits antitumor immunity (15, 16). This complicates the analysis of studies using Lag3−/− mice or LAG3 blocking antibodies and limits the dissection of IR function on different T cell populations. Previous studies have suggested that LAG3 was used as a mechanism for Treg suppression (1113, 15). However, it is also possible that LAG3 may intrinsically limit Treg function in a manner commensurate to its role on other T cell populations. This ambiguity complicates defining the role of LAG3, and perhaps other IRs, on Tregs, assessing whether differential IR expression on Tregs affects the progression of autoimmune disease and interpreting the mechanistic basis of LAG3 blockade in mouse models and patients with cancer.

Because the activity of Tregs is likely to be near maximal in tumors (17), the full impact of LAG3 deficiency on Treg function and viability may not be evident in tumor models. Thus, we chose to assess Treg function in a model in which their activity delays disease onset but is ultimately insufficient (18). Several studies have shown that Tregs ultimately lose stability and function in an autoimmune-prone environment (1922). To assess the function of LAG3 on Tregs, we generated mice that specifically lack LAG3 on Foxp3+ Tregs and assessed their proliferation and function in a mouse model of autoimmune diabetes. These mice exhibited reduced autoimmune diabetes due to enhanced Treg proliferation and function in islets. The expression of LAG3 on Tregs may limit Treg proliferation and function by down-regulating Eos and the interleukin-2 (IL-2)–signal transducer and activator of transcription 5 (STAT5) pathway by reducing CD25 expression.


Intra-islet Tregs constitutively express LAG3

To assess the role of LAG3 on Tregs, we used the NOD mouse model of type 1 diabetes (23). We first assessed the expression of LAG3, as well as other IRs, on islet-infiltrating T cells using transcriptional analysis and flow cytometry. Multiple IRs were transcriptionally up-regulated in intra-islet Tregs compared with peripheral Tregs (Fig. 1A). LAG3 was substantially up-regulated on islet-infiltrating Tregs compared with peripheral Tregs in NOD mice, and the percentage of LAG3+ Tregs was also higher than the percentage of LAG3+ CD4+ and CD8+ Tconv cells (Fig. 1, B to D, and fig. S1), suggesting a role for LAG3 on Tregs.

Fig. 1 LAG3 is up-regulated on intra-islet Tregs.

(A) WT NOD Treg expression profiles shown in x-y dot plots. Values were averaged from independent triplicates (female, 8 weeks of age). Labels denote IR genes that are significantly up-regulated (red) or down-regulated (blue) in intra-islet Tregs compared with peripheral Tregs. FPKM, fragments per kilobase million. (B) Representative histograms of LAG3 expression on the cell surface of CD8+, CD4+Foxp3, and CD4+Foxp3+ T cells in female WT NODs at 8 weeks of age. (C) Percentage of LAG3+ Tregs (top) and mean fluorescence intensity (MFI) (fold change relative to isotype control) of LAG3 on Tregs (bottom) in WT NODs at 6, 8, and 10 weeks of age (n = 5 to 6, females). Data represent means ± SEM. Statistic significance was determined by comparison of islet and ndLN at each time point. (D) Proportion of CD4+Foxp3, CD8+, and CD4+Foxp3+ cells in intra-islet LAG3+ T cells. Nonparametric Mann-Whitney test was used in (C). *P < 0.05, **P < 0.01.

The absence of LAG3 on Tregs results in reduced autoimmune diabetes

To assess the importance of LAG3 expression on Tregs in controlling autoimmune diabetes, we generated Lag3L/L-YFP conditional knockout reporter mice (backcrossed to NOD/ShiLtJ for 12 generations; see Materials and Methods) that lack cell surface expression of LAG3 and thus cannot mediate signaling but continue to release soluble LAG3 (24, 25), specifically on Tregs, when crossed with the Foxp3Cre-GFP.NOD mice (Fig. 2A and fig. S2). Although there is no evidence that soluble LAG3 affects T cell function (24, 25), we took this approach to avoid this complication. This mutant mouse also incorporated an internal ribosomal entry site (IRES)–yellow fluorescent protein (YFP) cassette inserted into the 3′ untranslated region (UTR) as a reporter of Lag3 promoter activity (Fig. 2A). We initially crossed Lag3L/L-YFP.NOD mice with Cd4Cre.NOD mice to assess the phenotype after global loss of surface LAG3 on all T cells (Fig. 2B and fig. S2, B and C). Loss of LAG3 surface expression on all CD4+ and CD8+ T cells (Lag3L/L-YFPCd4Cre.NOD) resulted in a markedly accelerated onset of autoimmune diabetes with 100% penetrance by 12 weeks of age, which phenocopied our observations with Lag3−/−.NOD mice and suggested that the dominant function of LAG3 was restricted to T cell populations (Fig. 2B) (8).

Fig. 2 Loss of LAG3 on Tregs results in reduced autoimmune diabetes.

(A) Schematic of WT and Lag3L/L-YPF loci. (B) Diabetes onset and incidence monitored in Lag3L/L-YFPCd4Cre.NOD females and co-caged littermate controls. (C) Diabetes onset and incidence monitored in Lag3L/L-YFPFoxp3Cre-GFP.NOD females (top) and males (bottom) together with co-caged littermate controls. (D) Histological assessment of insulitis performed in female and male Lag3L/L-YFPFoxp3Cre-GFP.NOD mice together with co-caged controls at 6 and 10 weeks of age (n = 4 to 7). Horizontal black bars indicate the mean. Error bars indicate SEM. The log-rank test was applied to Kaplan-Meier survival function estimates to determine the statistical significance in (B) and (C), and nonparametric Mann-Whitney test was used in (D). ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Because LAG3 has been shown to be required for optimal Treg function (1113, 15), we reasoned that the accelerated autoimmune diabetes observed in the Lag3−/−.NOD and Lag3L/L-YFPCd4Cre.NOD mice might be partially due to the loss of LAG3 expression on Tregs. We then assessed the impact of the loss of LAG3 surface expression on Tregs by analyzing Lag3L/L-YFPFoxp3Cre-GFP.NOD mice (fig. S2). Unexpectedly, female Lag3L/L-YFPFoxp3Cre-GFP.NOD mice had a significantly delayed onset of autoimmune diabetes, decreased diabetes incidence (48% versus 84%) by 30 weeks of age, and male Lag3L/L-YFPFoxp3Cre-GFP.NOD mice were completely protected from autoimmune diabetes (Fig. 2C). Although we only assessed a small number of cohoused female Lag3+/L-YFPFoxp3Cre-GFP.NOD littermates, their diabetes incidence was 60%, suggesting that LAG3 exhibits haploinsufficiency. Although the absence of LAG3 on Tregs did not affect the degree of insulitis at 6 weeks of age, it led to a significant reduction at 10 weeks of age in female and male Lag3L/L-YFPFoxp3Cre-GFP.NOD mice (Fig. 2D), suggesting that the expression of LAG3 on Tregs had little impact on the initiation of islet infiltration but was critical to limit Treg-mediated self-tolerance and the onset of autoimmune diabetes. Together, these data suggest that LAG3 may limit Treg-mediated suppression of autoimmunity.

Consistent with the reduced insulitis and diabetes observed, there were reduced numbers of CD4+Foxp3 and CD8+ T cells in the islets of Lag3L/L-YFPFoxp3Cre-GFP.NOD mice (fig. S3A). Although the number of Tregs in the islets of Lag3L/L-YFPFoxp3Cre-GFP.NOD mice was decreased as a result of the reduced insulitis and inflammation, there was a trend, albeit not reaching significance, toward a lower ratio of Tconv cells to Tregs (fig. S3A). The proportion of chromogranin A–specific (BDC2.5mi+) CD4+ T cells and islet-specific glucose-6-phosphatase (IGRP)–specific (NRP-v7mi+) CD8+ T cells was not altered in the islets of Lag3L/L-YFPFoxp3Cre-GFP.NOD mice compared with controls, suggesting that Lag3-deficient Tregs may not selectively suppress specific subpopulations of diabetogenic T cells but rather globally affect all islet-infiltrating cells (fig. S3B). This observation may have been anticipated given that we have previously shown that only islet antigen–reactive T cells can enter the islets (26). The reduced number of Tconv cells in the islets of Lag3L/L-YFPFoxp3Cre-GFP.NOD mice was due to the decreased CD8+ T cell proliferation [assessed by Ki67 expression and 5-bromo-2′-deoxyuridine (BrdU) incorporation] and reduced expression of the antiapoptotic factor Bcl2 (27) in CD4+Foxp3 T cells in islets (fig. S4 and S5). Both CD4+Foxp3 and CD8+ T cells in the islets of Lag3L/L-YFPFoxp3Cre-GFP.NOD mice had a significantly reduced expression of tumor necrosis factor–α (TNFα) but not interferon-γ (IFN-γ) (fig. S6). A significant reduction in IL-2 production was also observed in CD4+Foxp3 T cells in the islets of Lag3L/L-YFPFoxp3Cre-GFP.NOD mice (fig. S6). There was a trend, albeit not reaching significance, toward an increased percentage of T helper 2 (TH2) and TH17 cells (fig. S6). Although the activation, terminal differentiation [as marked by killer cell lectin-like receptor subfamily G member 1 (KLRG1) expression (28)], and proliferation of Lag3-deficient and wild-type (WT) Tregs were comparable, Lag3-deficient Tregs expressed a higher level of the antiapoptotic factor Bcl2 in the islets (figs. S5 and S7A). Inducible T cell costimulator (ICOS, CD278), which has been shown to be critical for Treg homeostasis and functional stability in NOD mice (29), was also up-regulated on intra-islet Tregs in the absence of LAG3 (fig. S7A). The expression of multiple IRs [programmed cell death protein 1 (PD1), T cell immunoreceptor with Ig and ITIM domains (TIGIT), and T cell immunoglobulin and mucin-domain containing 3 (TIM3)] was slightly enhanced on Lag3-deficient Tregs (fig. S7A), inferring a cell-intrinsic process in Tregs to compensate for the loss of LAG3. However, the suppressive capacity of intra-islet and peripheral WT and Lag3-deficient Tregs in an in vitro microsuppression assay was comparable on a per-cell level (fig. S7B). Overall, these data suggest that Lag3-deficient Tregs appear to have an enhanced impact on islet antigen–reactive T cell proliferation, effector cytokine production, and probably viability (as suggested by Bcl2 expression) in vivo, perhaps at the population level as a result of their enhanced proliferative and survival capacity.

LAG3 affects the Treg transcriptome

To assess the impact of LAG3 deletion on the Treg cell transcriptome, we performed RNA sequencing (RNA-seq) of WT and Lag3-deficient Tregs from the islets and nondraining lymph nodes (ndLN). The whole-genome expression profiles, as well as the expression of previously defined Treg signature genes (30, 31), were affected by the islet microenvironment (Fig. 1A and fig. S8). A substantial number of genes and pathways were modulated by the loss of LAG3 expression of Tregs in the islets (Fig. 3A, fig. S9, and table S1). A group of genes were down-regulated in intra-islet WT but not in Lag3-deficient Tregs, compared with peripheral Tregs (Fig. 3A and fig. S10), suggesting that these genes might be required for optimal Treg function or survival and that LAG3 was limiting their expression in intra-islet Tregs.

Fig. 3 LAG3 alters the Treg transcriptome.

(A) Heat map of differentially expressed genes in Lag3-deficient versus WT ndLN and intra-islet Tregs. Fold change in gene expression was relative to WT Tregs from ndLN and shown as the means of independent triplicates (female, 8 weeks of age). The gene dendrogram was calculated based on standard Euclidean distance mean linkage clustering and rotated to sort values in islet Lag3L/L-YFPFoxp3Cre-GFP.NOD Tregs. The following gene sets were used for comparison: Treg signature from D’Alise et al. (30) and Fu et al. (31), IL-2–STAT5 targeted genes from mSigDB (HALLMARK_IL2_STAT5_SIGNALING), and si-Ikzf4 Tregs (Treg knockdown with Ikzf4 siRNA) and si-RL Tregs (Treg knockdown with renilla luciferase siRNA) from Pan et al. (32). (B) Scatterplot of the Eos targeted genes (32) in Lag3-deficient versus WT Treg expression profiles. (C) Differentially expressed IL-2–STAT5 targeted genes shown in x-y dot plots. Labels denote genes from mSigDB HALLMARK_IL2_STAT5_SIGNALING that are significantly up-regulated (red) or down-regulated (blue) in Lag3-deficient Tregs compared with WT Tregs. (D) Barcode plots depicting the enrichment of genes in the mSigDB WIERENGA_STAT5_TARGETS_UP and HALLMARK_IL2_STAT5_SIGNALING pathways. Statistics denotes the Wald statistic used to test the significance of coefficients in a negative binomial generalized linear model.

One of these genes was Ikzf4 (Eos), a corepressor of Foxp3 that prevents the expression of Tconv genes in Tregs (Fig. 3A and fig. S10) (3134). The expression profile of intra-islet WT Tregs resembled the previously published transcriptional signature in Ikzf4-knockdown Tregs, whereas the expression profile of intra-islet Lag3-deficient Tregs resembled the transcriptional signature in mock control Tregs (Fig. 3, A and B) (32). These data suggest that LAG3 might limit Eos expression and thus the function and maintenance of intra-islet Tregs. IL-2 has been shown to be essential for Treg cell maintenance, whereas defective IL-2 signaling in Tregs triggers autoimmune islet destruction (22, 35, 36). Genes modulated by IL-2–STAT5 signaling were substantially enhanced in the absence of LAG3 on Tregs (Fig. 3, C and D, and table S1). Overall, the transcriptome analyses suggest that LAG3 negatively regulates intra-islet Tregs by down-regulating key genes and pathways that are essential for Treg maintenance and function.

LAG3 intrinsically limits Treg proliferation

To directly assess whether Lag3-deficient Tregs had a proliferative advantage over WT Tregs and to determine whether the pathways identified by transcriptome analysis were intrinsically regulated by LAG3 in Tregs, we cotransferred an equal number of activated congenic marker–mismatched WT (Thy1.1+) and Lag3-deficient (Thy1.2+) Tregs into NOD (Thy1.1+Thy1.2+) hosts (Fig. 4A). Both WT and Lag3-deficient donor Tregs were sorted from mice that expressed the islet antigen–specific BDC2.5 T cell receptor (TCR), which facilitates islet entry (26, 37). Foxp3 expression was not altered in either Treg population after adoptive transfer (fig. S11, A and B), suggesting that LAG3 may not affect Treg stability. Lag3-deficient Tregs outcompeted WT Tregs in the islets (60% versus 40%, respectively) and in the pancreatic lymph node (PLN) (54% versus 46%, respectively), but not in the periphery (Fig. 4, B and C).

Fig. 4 LAG3 intrinsically limits Treg proliferation.

(A) Schematic of Treg cotransfer experiment. (B) Representative plots of Thy1.1+ (WT, blue) and Thy1.2+ (Lag3-deficient, red) Tregs (gated on CD4+Vβ4+) in NOD recipients (gray). (C) Proportion of WT and Lag3-deficient Tregs in the islets and lymph nodes assessed with the percentage of Thy1.1+ and Thy1.2+ in CD4+Vβ4+ cells after transfer (n = 17, three independent experiments; female mice). Data presented as means ± SEM. (D to I) Flow cytometry analysis of markers shown. Horizontal black bars indicate the mean [(D) and (E): n = 17, three independent experiments; (F) to (I): n = 12, two independent experiments; all female mice]. Nonparametric Mann-Whitney test was used in (C) to (I). **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Previous studies have shown that reduced CD25 and Bcl2 levels cause a decline in intra-islet Treg viability, whereas administration of low-dose IL-2 promotes Bcl2 expression and Treg survival (22, 38, 39). A higher percentage of intra-islet Lag3-deficient Tregs expressed Ki67 and Bcl2 compared with WT Tregs (Fig. 4, D and E, and fig. S11C). Although differences were also observed in the periphery in these cotransfer experiments, these were probably due to the activation of Tregs in vitro before adoptive transfer. Consistent with the transcriptomic analysis, Lag3-deficient Tregs exhibited higher CD25 expression and STAT5 phosphorylation compared with WT Tregs (Fig. 4, F and G, and fig. S11C). Furthermore, Eos (Ikzf4) expression was reduced in WT but not in Lag3-deficient intra-islet Tregs compared with peripheral Tregs, whereas another Ikaros family member, Helios (Ikzf2), was unaffected by LAG3 expression (Fig. 4, H and I, and fig. S11C).

Last, to determine whether LAG3 modulated Eos expression and whether direct modulation of Eos levels affected Treg proliferation, we first assessed Eos levels in WT and Lag3-deficient Tregs after stimulation in vitro. As anticipated from our transcriptomic analysis, Lag3-deficient Tregs expressed more Eos after stimulation than WT Tregs (Fig. 5). Likewise, activated Lag3-deficient Tregs exhibited increased proliferation, as measured by BrdU incorporation, over WT Tregs. Knockdown of Ikzf4 in Lag3-deficient Tregs reduced Eos expression and their proliferative capacity, whereas overexpression of human IKZF4 in WT Tregs enhanced their proliferation (Fig. 5 and fig. S12). Together, these data support a model in which LAG3 intrinsically limits Treg proliferation and viability by modulating pathways that are critical for Treg function and proliferation, particularly the IL-2–STAT5 and Eos pathways.

Fig. 5 LAG3 limits Treg proliferation through Eos pathway.

(A and B) Eos expression (top) and BrdU incorporation (bottom) assessed in activated Tregs after knockdown [(A) four independent experiments] or overexpression [(B) three independent experiments] of Ikzf4. WT, Foxp3Cre-GFP.NOD; KO, Lag3L/L-YFPFoxp3Cre-GFP.NOD; NTC, nontargeting control vector; EV, empty vector pMIA. Horizontal black bars indicate the mean. Fisher’s LSD test was applied to one-way ANOVA to determine the statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.


Our study supports a model in which the IR LAG3 intrinsically limits Treg proliferation and functionality by repressing pathways that promote the maintenance of Tregs at inflammatory sites. Lag3-deficient Tregs do not appear to have increased suppressive capacity on a per-cell basis. However, they do have an enhanced proliferative and survival advantage that potentiates their suppressive capacity at the population level and endows them with a critical advantage over time. As the disease progresses, subtle changes in Lag3-deficient Tregs allow them to accumulate over time, leading to a substantial impact on the development of chronic autoimmune diabetes. It is remarkable that this small, Treg-restricted genetic alteration renders male NOD mice resistant to diabetes and substantially limits autoimmune diabetes in female mice, whereas LAG3 deletion in all T cell subsets markedly accelerates the disease. Thus, in autoimmune environments where chronic inflammation dominates, LAG3 may be constitutively expressed on Tregs, thereby limiting their capacity to block the function of diabetogenic T cells and prevent autoimmune diabetes. This raises the possibility that constitutive IR expression on Tregs may underlie their insufficiency in autoimmune disease. It is also possible that increased or constitutive LAG3 expression on Tregs may also limit their suppressive capacity in inflammatory or infectious diseases where increased tissue damage or pathology is observed.

The onset and incidence of autoimmune diabetes is affected by sex hormones (40, 41). Our transcriptomic analysis suggested that some sex hormone–related pathways were affected by the presence or absence of LAG3 on Tregs (table S1). This may partially explain the differences observed between female and male NOD mice that had Lag3-deficient Tregs, suggesting that there may be additional noncanonical roles for LAG3 in modulating immune responses.

The impact of IRs on Treg function and maintenance has been controversial. It was reported that the absence of PD1 on Tregs led to the generation of ex-Foxp3 T cells (42). However, Foxp3 stability was maintained in the absence of LAG3 on Tregs, suggesting that distinct pathways are regulated by LAG3 in Tregs. Although our observations here do not preclude a role for LAG3 as a mechanism of Treg suppression (1113, 15), our data point to a dominant role for LAG3 in limiting Treg maintenance and proliferation. This may be mediated in part by the Foxp3 corepressor Eos, which is required for Treg maintenance (32, 33). There seemed to be a direct correlation between Eos expression and Treg proliferation because both were increased after stimulation of Lag3-deficient Tregs, and the overexpression or knockdown of Eos resulted in analog alterations in Treg proliferation (Fig. 5). Furthermore, enhanced IL-2–STAT5 signaling has been shown to promote Treg maintenance and survival (22, 35, 36, 38, 39). LAG3 appears to limit this pathway, thereby having a global impact on Treg function. Future studies may shed light on two further questions: (i) Do other IRs promote or limit Treg function, proliferation, and/or survival? (ii) Do these IRs affect these Treg parameters using comparable or distinct mechanisms?

These observations highlight the differential impact of LAG3 modulation on different T cell populations in vivo, where LAG3 modulation alleviates or exacerbates disease depending on whether Treg or Tconv cells are targeted. The impact of losing LAG3 on Tregs leads to enhanced immunosuppression and therefore may offset the effect of blocking the LAG3 pathway in Tconv cells. One wonders whether this might underlie the lack of efficacy observed on tumor growth with LAG3 blockade (9). These findings may also apply to other IRs (4, 14), whose intrinsic effect on Tregs might have been previously overlooked. Our findings may have clinical relevance in that patients who fail to respond to checkpoint blockade immunotherapy may do so because it has a greater impact on promoting Treg function than mitigating Tconv cell exhaustion. Thus, the differential impact of immunotherapy may be modulated by the Tconv/Treg cell ratio and/or IR expression on different T cell subsets. Given that anti-LAG3 has entered phase 1 clinical trials for multiple tumor types with the goal of enhancing the efficacy of PD1 blockade, we should consider the differential impact this might have on Treg function and survival versus Tconv cell exhaustion as well as the strategies that specifically target checkpoints on Tconv cells or Tregs to boost antitumor immunity or mitigate autoimmunity, respectively.


Mice and study design

NOD/ShiLtJ (stock #001976), Thy1.1 (stock #004483), and BDC2.5 (stock #004460) NOD mice were purchased from the Jackson Laboratory. Foxp3Cre-GFP.NOD mice were obtained from J. A. Bluestone (43). Lag3−/− C57BL/6 mice were obtained from Y. H. Chen with permission from C. Benoist and D. Mathis and bred onto a NOD background with 100% NOD, as determined by single-nucleotide polymorphism (SNP) microsatellite analysis (8, 44). Cd4Cre.NOD mice were obtained from A. Chervonsky.

All animal experiments were performed in the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC)–accredited, specific pathogen–free facilities in the Animal Resource Center of St. Jude Children’s Research Hospital (SJCRH) and the Division of Laboratory Animal Resources of the University of Pittsburgh School of Medicine (UPSOM). Animal protocols were approved by the Institutional Animal Care and Use Committees (IACUC) of SJCRH and UPSOM. Mice of different groups were cohoused and randomly assigned to analyses. Ten to 20 mice per group were used in diabetes incidence studies and followed up to 30 weeks of age. Three to five age-matched female mice per group were used in each analytical experiment, and two to four independent experiments were repeated. Three 8-week-old female mice per group were pooled and used in RNA-seq analyses, and three independent experiments were repeated. The genotypes were not blinded, except for the insulitis scoring. All data points were presented.

Generation of Lag3L/L-YFP mice

A 5.7-kb Xba I–Sal I fragment (5′ arm of homology) corresponding to exon 6 and the intronic region between exons 5 and 6 and a 4.1-kb Cla I–Eco RI fragment (3′ arm of homology) containing the polyA site (pA) were generated by polymerase chain reaction from C57BL/6J genomic DNA and cloned into pSP73. A fragment corresponding to exon 7 (containing the CP cleavage site and flanked by loxP sites) and exon 8 was inserted between the two homologous arms. An IRES-YFP fragment was inserted between the Lag3 stop codon and the pA. Just after the pA, an frt-flanked neomycin-positive selection cassette (Frt-Neo) was inserted. To increase the frequency of homologous recombination and reduce nonspecific integration, we cloned a diphtheria toxin cassette (DT-A) upstream of the 5′ homologous arm. The resulting plasmid was linearized with Ssp I and electroporated into E14 embryonic stem cells. After selection with G418, resistant clones were screened by Southern blot analysis, sequenced, and injected into blastocycts, and the resulting chimeras were bred to C57BL/6J for germline transmission. The mice were backcrossed 12 generations onto the NOD background and tested by microsatellite analysis. All 20 Idd loci were covered by 144 SNPs in the microsatellite test (45), and all tested SNPs were NOD.

Measurement of diabetes and insulitis

Diabetes and insulitis were assessed as previously described (8, 46). Briefly, diabetes incidence was monitored weekly by testing for the presence of glucose in the urine with Diastix (Bayer). Mice positive by Diastix were then bled and tested with a Breeze2 glucometer (Bayer). Mice were considered diabetic if the blood glucose level was ≥400 mg/dl.

Pancreata were embedded in a paraffin block and cut into 4-μm-thick sections at 150-μm step sections and stained with hematoxylin and eosin. Pancreata collected at SJCRH were processed at the Veterinary Pathology Core of SJCRH, and pancreata collected at UPSOM were repeated in the same way at the Histo-Scientific Research Laboratories (HSRL Inc.). An average of 60 to 80 islets per mouse were scored in a blinded manner. Two methods of insulitis measurement were used as previously described (46).

Islet isolation and lymphocyte preparation

Islets were isolated as previously described (26). Briefly, the pancreata were perfused with 3 ml of collagenase type 4 (Worthington) through the pancreatic duct and incubated in 3 ml of collagenase [600 U/ml in Hanks’ balanced salt solution (HBSS) with 10% fetal bovine serum (FBS)] for 30 min at a 37°C water bath. The pancreata were then distributed and washed twice with HBSS (Corning) with 10% FBS. The islets were picked under a dissecting microscope, distributed with 1 ml of cell dissociation buffer (Life Technologies), and incubated for 15 min at 37°C with vortexing every 5 min. After a final wash, the cells were resuspended, counted, and used.

Antibodies and flow cytometry

Single-cell suspensions were stained with antibodies against CD4 (clone GK1.5, BioLegend), CD8β (clone YTS156.7.7, BioLegend; clone H35-17.2, eBioscience), TCRβ (clone H57-597, BioLegend), Vβ4 (clone KT4, BD Biosciences), Thy1.1 (clone OX-7, BioLegend), Thy1.2 (clone 30-H12, BioLegend), CD45RB (clone C363-16A, BioLegend), CD44 (clone IM7, BioLegend), CD62L (clone MEL-14, BioLegend), CD25 (clone PC61, BioLegend), LAG3 (clone 4-10-C9, made in-house), Foxp3 (clone FJK-16s, eBioscience; clone 150D, BioLegend), Eos (clone ESB7C2, eBioscience), Helios (clone 22F6, BioLegend), Ki67 (clone B56, BD Biosciences), BrdU (clone Bu20a, BioLegend), Bcl2 (clone BCL/10C4, BioLegend), TNFα (clone MP6-XT22, BioLegend), IFN-γ (clone XMG1.2, BioLegend), IL-2 (clone JES6-5H4, BioLegend), IL-4 (clone 11B11, eBioscience), IL-17A (clone TC11-18H10.1, BioLegend), GATA3 (clone TWAJ, eBioscience), RORγt (clone B2D, eBioscience), PD1 (clone RMP1-30, BioLegend), TIM3 (clone RMT3-23, BioLegend), TIGIT (clone GIGD7, eBioscience), KLRG1 (clone 2F1, eBioscience), ICOS (clone C398.4A, BioLegend), and phospho-STAT5 (clone C71E5, Cell Signaling).

Surface staining was performed on ice for 15 min. For cytokine expression analysis, cells were activated with phorbol 12-myristate 13-acetate (PMA) (0.1 μg/ml; Sigma) and ionomycin (0.5 μg/ml; Sigma) in RPMI containing 10% FBS and monensin (eBioscience) for 5 hours. For intracellular staining of cytokines and transcription factors, cells were stained with surface markers, fixed in Fix/Perm buffer (eBioscience) for 0.5 to 2 hours, washed twice in permeabilization buffer (eBioscience), and stained with intracellular factors in permeabilization buffer for 30 min on ice. For phosphoprotein staining, cells were fixed with 1.6% paraformaldehyde (Alfa Aesar) for 15 min at 37°C, permeabilized with ice-cold methanol for 1 hour, and stained on ice for 1 hour. For BrdU incorporation analysis, mice were injected with 2 mg of BrdU (Sigma) in phosphate-buffered saline (PBS) intraperitoneally 8 hours ahead of death. After transcription factor staining, cells were incubated in Cytofix/Cytoperm buffer (BD Biosciences) for 10 min at room temperature, washed with Perm/Wash buffer (BD Biosciences), treated with deoxyribonuclease I (650 U/ml; Sigma) for 30 min at 37°C, and stained with anti-BrdU antibody in Perm/Wash buffer for 30 min at room temperature. The chromogranin A (BDC2.5 mimotope) tetramer for CD4+ T cells (AHHPIWARMDA/Ag7) and IGRP (NRV-v7 mimotope) tetramer for CD8+ T cells (KYNKANVFL/H-2Kd) were obtained from the National Institutes of Health Tetramer Core Facility, and cells were stained in RPMI containing 10% FBS for 40 min at room temperature. Cells were sorted using Aria II (BD Biosciences) or analyzed using Fortessa (BD Biosciences), and data analysis was performed on FlowJo version 9 (Tree Star).

Microsuppression assay

Splenic TCRβ+CD4+CD45RB+GFP cells were sorted as responder cells and labeled with CellTrace Violet (Life Technologies). T cell–depleted whole splenocytes were treated with mytomycin C (2 μg/ml; Sigma) for 30 min at 37°C, washed three times with PBS, and then used as antigen-presenting cells (APCs). Responder cells (4 × 103), APCs (8 × 103), and different concentrations of Tregs were activated with anti-CD3 (2 μg/ml; BioLegend) in a 96-well round-bottom plate with 100 μl of RPMI for 3 days. Suppression was calculated as previously described (47). Briefly, cells were acquired using BD Fortessa, and the division index (DI) of responder cells was analyzed using FlowJo based on the division of CellTrace Violet. Suppression was then calculated with the following formula: %Suppression = (1 − DITreg/DICtrl) × 100%, where DITreg stands for the DI of responder cells with Tregs and DICtrl stands for the DI of responder cells activated without Tregs.

Gene expression profiling by RNA-seq and bioinformatic analyses

Tregs (5 × 103) were sorted from three pooled mice of each group, and cDNAs were prepared using the SMARTer Ultra Low Input RNA Kit for Sequencing version 3 following the user manual (Clontech Laboratories). Sequencing libraries were prepared using the Nextera XT DNA Library Prep Kit (Illumina), normalized to 2 nM using tris-HCl (10 mM; pH 8.5) with 0.1% Tween 20, diluted and denatured to a final concentration of 1.8 nM using the Illumina Denature and Dilute Libraries for the NextSeq 500 protocol Revision D (Illumina). Cluster generation and 75–base pair paired-end dual-indexed sequencing was performed on the Illumina NextSeq 500 system.

The raw reads of RNA-seq were aligned to the mm10 genome using TopHat, and counts were computed relative to the RefSeq transcript annotation file provided in the cufflinks suite (48, 49). Genes whose mean count value (computed in log2 space) was below 32 (5 in log2 space) were removed from further processing, leaving 10,371 total genes. The counts were analyzed for differential expression using DESeq2 with a GC content and length-dependent offset computed by cqn R package (50, 51).

Pathway analysis

We performed gene set analysis using the Wilcoxon rank-sum test on the differential expression statistic (Wald statistic for the negative binomial coefficient) computed with the DESeq2 package. Significance was assessed with a parametric P value calculation followed by multiple hypothesis correction as well as sample permutation tests. Because there are three replicates of islet Treg samples of each genotype, there are 10 possible ways to divide those into two equal groups, and one of these corresponds to the correct grouping, leaving nine remaining permutations. Pathways that were significant at a false discovery rate of 0.2 but were not significant in any of the possible permutation tests were reported. Principal components analysis was performed using the prcomp R function on the log2-transformed normalized counts produced by the DESeq2 counts function with “normalized=T.”

Comparisons with Ikzf4-knockdown data set

We retrieved processed data from the Gene Expression Omnibus (accession no. GSE17166). Because this data set had no replicates, we used fold change between the Eos small interfering RNA (siRNA) and control siRNA as reference. Genes that had expression levels less than log2 (intensity) of 5 as well as genes that were affected more than twofold by the control siRNA were excluded from the analysis. The significance of the association between the two transcriptional signatures was assessed using a χ2 test on the contingency table summarizing the number of up- or down-regulated genes in si-Ikzf4 Tregs and intra-islet Tregs.

Treg expansion and adoptive transfer

Splenic TCRβ+CD4+GFP+CD45RBlow cells (Tregs) were sorted and activated with PMA (0.1 μg/ml; Sigma) and ionomycin (0.5 μg/ml; Sigma) with human IL-2 (hIL-2) (500 U/ml; Prometheus) for 2 days and then expanded for another 3 days with hIL-2. WT and Lag3-deficient Tregs were mixed at an equal ratio, and 2 × 106 total Tregs were cotransferred into 6- to 8-week-old WT NODs. Treg recipients were sacrificed and analyzed 4 days after the transfer.

Ikzf4 overexpression and knockdown in Tregs

Human IKZF4 open reading frame was amplified from IKZF4-pMIG construct [obtained from C. Benoist (31)] using primers (forward: 5′-CGC GGC TCT AGA TCT GCC AGC ATG CAT ACA CCA CCC GCA CTC C, reverse: 5′-CCT TCC ATC CCT CGA GCT AGC CCA CCT TAT GCT CCC CC), cut with Bgl II and Xho I restriction enzymes, and ligated into the pMI-Ametrine (pMIA) retroviral vector. Murine Ikzf4 targeting short hairpin RNA (shRNA) (3′-TCC AGA AAG AGG ATG CGG CAG T, 5′-CCT GCC GCA TCC TCT TTC TGG A, loop-TAG TGA AGC CAC AGA TGT A) and nontargeting control (3′-TAA CCT ATA AGA ACC ATT ACC A, 5′-CGG TAA TGG TTC TTA TAG GTT A, loop-TAG TGA AGC CAC AGA TGT A) retroviral constructs (transOMIC technologies) were cut with Bgl II and Mlu I restriction enzymes and inserted with the IRES-Ametrine cassette as a fluorescence reporter.

Sorted splenic Tregs were activated with αCD3/αCD28 Dynabeads (Invitrogen) and hIL-2 (500 U/ml) for 48 hours. Plat-E cells (obtained from H. Chi) were transiently transfected by retroviral vector along with the pCL-Eco helper plasmid (obtained from H. Chi). The viral supernatant was harvested 36 hours after transfection of Plat-E cells and then used for spin transduction of activated Tregs with polybrene (6 μg/ml; Sigma) at 2000 revolutions per minute for 1 hour. Transduced Tregs were sorted 48 hours after transduction, rested for 72 hours, and then restimulated with PMA (0.1 μg/ml) and ionomycin (0.5 μg/ml) with hIL-2 (500 U/ml) for another 48 hours. BrdU (10 μg/ml) was pulsed into Treg culture medium 2 hours before the staining.

Statistical analyses

Experiments were pooled for statistical analyses using Prism version 7 (GraphPad). The log-rank test was applied to Kaplan-Meier survival function estimates to determine the statistical significance of differences in diabetes incidence between experimental groups. Fisher’s least significant difference (LSD) test was applied to one-way analysis of variance (ANOVA) to determine the statistical significance in the Ikzf4 overexpression or knockdown experiments. Nonparametric Mann-Whitney test was used in all other instances.


Fig. S1. Up-regulation of LAG3 expression on islet-infiltrating lymphocytes.

Fig. S2. Generation of conditional Lag3-knockout mice.

Fig. S3. Reduced lymphocyte infiltration into islets in the absence of LAG3 on Tregs.

Fig. S4. Intrinsic and extrinsic impact of Treg-expressed LAG3 on T cell proliferation.

Fig. S5. Intrinsic and extrinsic impact of Treg-expressed LAG3 on Bcl2 expression.

Fig. S6. Effector cytokine production in the absence of LAG3 on Tregs.

Fig. S7. Phenotypic and functional analyses on Lag3-deficient Tregs.

Fig. S8. Treg signature genes are affected by islet microenvironment.

Fig. S9. Consistency between independent replicates of RNA-seq.

Fig. S10. A group of genes are down-regulated in intra-islet WT Tregs but are still maintained in Lag3-deficient Tregs.

Fig. S11. Lag3-deficient Tregs outcompeted WT Tregs in the same hosts.

Fig. S12. Eos expression and knockdown in Tregs.

Table S1. Pathways differentially regulated in intra-islet Lag3-deficient versus WT Tregs (Excel file).

Table S2. Raw data sets (Excel file).


Acknowledgments: We thank J. A. Bluestone (University of California, San Francisco) for Foxp3Cre-GFP.NOD mice, D. Mathis and C. Benoist for Lag3−/− mice and IKZF4-pMIG construct, A. Chervonsky for Cd4Cre.NOD mice, P. Murray for the embryonic stem cells, H. Chi for the Plat-E cells and pCL-Eco plasmid, and the NIH Tetramer Core Facility (contract HHSN272201300006C) for the tetramers. We also thank K. Forbes, A. Castellaw, and E. A. Brunazzi for maintenance, breeding, and genotyping of mouse colonies; T. Benos and W. Chen for additional advice regarding computational analysis of the RNA-seq data; D. Sawant for assistance with RNA-seq library preparation; A. Fergerson for assistance with next-generation sequencing; P. Brindle and L. Kasper for technical advice on targeting construct design and embryonic stem cell culture; and A. Visperas and A. H. Herrada for advice. We also thank R. Cross and G. Lennon (SJCRH) and H. Shen, D. Falkner, and A. Yates [University of Pittsburgh (UP)] for cell sorting; the staff of the Animal Resource Center (SJCRH) and the Division of Laboratory Animal Resources (UP) for animal husbandry; the Veterinary Pathology Core (SJCRH) for histological preparation; and the Genomics Core (UP) for library quantification. Funding: This work was supported by the NIH (R01 DK089125 and P01 AI108545 to D.A.A.V.), National Cancer Institute Comprehensive Cancer Center Support CORE grants (CA21765 and CA047904 to D.A.A.V.), the American Lebanese Syrian Associated Charities (to D.A.A.V.), and Pennsylvania Department of Health (SAP #4100068731 to M.C.). Author contributions: Q.Z. designed and performed most of the experiments, carried out statistical analyses, and wrote the manuscript. M.C. performed computational analysis of the RNA-seq data. A.L.S.-W. designed and made the Lag3L/L-YFP construct and generated the Lag3L/L-YFP mice. W.H. and J.K.K. sequenced RNA-cDNA libraries and performed initial data normalization. K.M.V. assisted in designing Ikzf4 knockdown and overexpression constructs. D.N. oversaw the statistical analyses. M.B. provided input in experiment design and data analysis. C.J.W. assisted in designing the mouse construct and experimental design and analysis. D.A.A.V. conceived the project, directed the research, and wrote the manuscript. All authors edited and approved the manuscript. Competing interests: D.A.A.V. and C.J.W. are inventors on issued patents [United States (8,551,481), Europe (1,897,548), Australia (200,427,526), and Hong Kong (1,114,339)] held by SJCRH and Johns Hopkins University that cover LAG3. An additional application is pending in Japan. All other authors declare that they have no competing interests. Data and materials availability: The RNA-seq data have been deposited in the Gene Expression Omnibus at the National Center for Biotechnology Information with accession no. GSE94581. The Lag3L/L-YFP.NOD mice will be freely distributed to investigators at academic institutions for noncommercial research when a material transfer agreement is signed. These mice will also be distributed to commercial entities upon completion of a licensing agreement with UP and SJCRH. Individual requests for shipment of mice to AAALAC-accredited institutions will be honored. The recipient investigators should provide written assurance and evidence that the animals will be used solely in accord with their local IACUC review, that the animals will not be further distributed by the recipient without consent from UP Office of Research, and that the animals will not be used for commercial purposes.
View Abstract

Stay Connected to Science Immunology

Navigate This Article