Regulatory T (Treg) cells are a distinct T-cell lineage characterized by sustained Foxp3 expression and potent suppressor function, but the upstream dominant factors that preserve Treg lineage-specific features are mostly unknown. Here, we show that Lkb1 maintains Treg cell lineage identity by stabilizing Foxp3 expression and enforcing suppressor function. Upon T-cell receptor (TCR) stimulation Lkb1 protein expression is upregulated in Treg cells but not in conventional T cells. Mice with Treg cell-specific deletion of Lkb1 develop a fatal early-onset autoimmune disease, with no Foxp3 expression in most Treg cells. Lkb1 stabilizes Foxp3 expression by preventing STAT4-mediated methylation of the conserved noncoding sequence 2 (CNS2) in the Foxp3 locus. Independent of maintaining Foxp3 expression, Lkb1 programs the expression of a wide spectrum of immunosuppressive genes, through mechanisms involving the augmentation of TGF-β signalling. These findings identify a critical function of Lkb1 in maintaining Treg cell lineage identity.
Regulatory T (Treg) cells preserve immune homeostasis by suppressing autoreactive immune responses1,2. The Treg cell lineage can be defined by two basic characteristics, stable expression of the transcription factor Foxp3 (forkhead box P3) and potent suppressive capacity3,4. Treg cells are stable and usually retain lineage characteristics in vivo5,6. However, under some conditions, Treg cells can lose or alter their lineage identity, resulting in immune disturbance and development of diseases3,4,6,7. Thus, delineating the molecular mechanisms that maintain the Treg lineage identity is important for understanding and treating Treg cell-related immune diseases.
Foxp3 is the most specific marker for distinguishing Treg cells from other T cells and an important regulator required for programming Treg suppressive function1,2. Lineage tracing in mice has shown that thymus-derived mature Treg cells have relatively stable Foxp3 expression in both homeostasis and inflammatory conditions5. Other reports indicate that epigenetic demethylation of the conserved noncoding sequence 2 (CNS2, also known as Treg cell-specific demethylation region) in the Foxp3 locus can ensure stable Foxp3 expression in Treg cells8,9,10. However, the upstream signalling checkpoints that activate chromatin, and the demethylation status, of Foxp3 locus are not clear.
Treg cells suppress immune responses through diverse mechanisms, such as the modulation of antigen presentation function (via CTLA4 (cytotoxic T-lymphocyte associated protein 4)), the production of inhibitory cytokines (for example, interleukin (IL)-10 and IL-35) and metabolites (for example, reactive oxygen species and adenosine), the deprivation of the T-cell growth factor IL-2 (via CD25), and the direct killing of target cells (via granzyme B and perforin)2,11,12. Although Foxp3 has an important function in programming Treg cells by controlling the expression of a large number of immunosuppressive genes, Foxp3 alone is not sufficient to confer this function1,2. In addition, T-cell receptor (TCR) signalling is also important to promote Treg cell function13,14,15. Nevertheless, little is known about other upstream ‘master regulators’ that broadly control the expression of these Treg cell-associated immunosuppressive genes.
Liver kinase b1 (Lkb1) is a tumour suppressor, and is mutated in Peutz–Jeghers cancer syndrome, cervical carcinoma and many sporadic non-small-lung carcinomas16,17,18. Under energy-stressed conditions, Lkb1 is an important upstream kinase that phosphorylates AMP-activated protein kinase (AMPK) and AMPK-related kinases that coordinate cell growth with metabolism16,17,18. Lkb1 has been shown to restrain the activation and proinflammatory function of conventional T cells18,19.
In this study, we find that Lkb1 protein is specifically increased in Treg cells upon TCR stimulation. To understand the function and mechanism of Lkb1 in Treg cells, we generate a mouse line with Lkb1 specifically deleted in Treg cells. These mice develop a fatal early-onset autoimmune disease with defective maintenance of stable Foxp3 expression and suppressive capacity in Treg cells. Mechanistically, Lkb1 restrains STAT4 (signal transducer and activator of transcription 4) activation partially through suppressing nuclear factor-κB (NF-κB) signalling, and thus prevents STAT4-mediated methylation of CNS2 in the Foxp3 locus, resulting in stable Foxp3 expression. Meanwhile, Lkb1 promotes the expression of a large number of immunosuppressive genes partially through augmenting transforming growth factor-β (TGF-β) signalling. Our study identifies Lkb1 as a critical determinant of Treg cell linage identity.
Lkb1 protein is increased in Treg cells upon TCR stimulation
TCR stimulation is essential for Treg cells to exert their optimal function13,14,15. Lkb1 protein expression in Treg cells were slightly lower compared with conventional T cells without stimulation (Fig. 1a and Supplementary Figs 1a and 8a,b). However, upon TCR stimulation, Lkb1 protein expression was markedly upregulated in Treg but not conventional T cells (Fig. 1a and Supplementary Fig. 1a), implying that Lkb1 might be particularly important for Treg cells to execute their immunoregulatory effect.
Deletion of Lkb1 in Treg cells leads to a fatal autoimmunity
To investigate the role of Lkb1 in Treg cells, we generated a mouse line with Lkb1 conditionally depleted in Treg cells by crossing Foxp3YFP-Cre (Foxp3Cre, which ensure Treg cell-specific deletion of a target gene) mice20 with Lkb1f/f mice21. Lkb1 protein was depleted in CD4+YFP+ Treg cells from Foxp3CreLkb1f/f mice (Supplementary Figs 1b and 8c,d). Strikingly, loss of Lkb1 in Treg cells caused early moribund at ∼35 days of age (Fig. 1b). Foxp3CreLkb1f/f mice exhibited smaller size, decreased mobility, hunched posture, collapsed ears and tail skin lesions (Fig. 1c), and barely increased in body weight with age (Fig. 1d). Foxp3CreLkb1f/f mice displayed splenomegaly and lymphadenopathy (Fig. 1e), and they had increased cell numbers in secondary lymphoid organs (Fig. 1f). Histopathological analysis revealed massive infiltration of lymphocytes into multiple organs such as skin, lung, liver and stomach (Fig. 1g), suggesting T-cell autoimmunity. In support of this, Foxp3CreLkb1f/f mice had substantial expansion of CD4+ conventional T cells (Supplementary Fig. 1c), associated with drastically increased percentages of cells that displayed a CD44highCD62Llow effector/memory phenotype (Fig. 2a and Supplementary Fig. 1d). These mice also had higher percentages of CD4+ and CD8+ conventional T cells that expressed the proliferation marker Ki67 and acute activation markers CD25 and CD69 (Fig. 2b and Supplementary Fig. 1e), and produced the inflammatory cytokines interferon-γ, IL-4 and IL-17A (Fig. 2c and Supplementary Fig. 1f). These phenotypes were similar in severity to those observed in mice deficient of Foxp3 (ref. 22) or depleted of Treg cells23, indicating a severe defect of immune suppression mediated by Treg cells.
Lkb1 maintains Foxp3 expression and CNS2 demethylation
Treg cells normally stay at a relatively constant abundance within the CD4+ T-cell population, and expand concomitantly with effector/memory T cells to preserve immune homeostasis during inflammation24. Although not altered at early ages, the percentages of peripheral Treg cells among CD4+ T-cell populations continuously dropped to ∼3% by 4 weeks of age when the mice became moribund (Fig. 3a). Despite greatly increased numbers of effector/memory T cells in Foxp3CreLkb1f/f mice (Supplementary Fig. 1d), the absolute numbers of Treg cells were only comparable with those in wild-type mice (Supplementary Fig. 1g), suggesting that Lkb1-deficient Treg cells fail to accumulate concomitantly with effector/memory T cells in Foxp3CreLkb1f/f mice at later ages that might contribute to the autoimmunity in Foxp3CreLkb1f/f mice.
The survival and proliferation of Lkb1-deficient Treg cells were barely altered compared with wild-type control cells (Supplementary Fig. 2a,b). Thus, other reasons might account for the defective accumulation of Lkb1-deficient Treg cells. Treg cells are considered a stable lineage marked by sustained Foxp3 expression5, but under certain conditions they might lose Foxp3 expression and develop into cells resembling effector T cells (designated ex-Treg cells)25. To investigate whether Lkb1-deficient Treg cells were prone to lose Foxp3 expression, we introduced the Rosa26YFP allele (a loxp-site-flanked STOP cassette followed by the YFP-encoding sequence was inserted into the Rosa26 locus, and the expression of YFP from the Rosa26 locus was dependent on the expression of Cre recombinase) into wild-type and conditional knockout mice26. CD4+Rosa26-YFPhigh cells (100% Lkb1-deficient Treg cells) sorted from Foxp3CreLkb1f/fRosa26YFP mice contain much higher percentages of Foxp3− cells (which represented ex-Treg cells that ever experienced Foxp3 expression but ceased to express it later) than that from Foxp3CreRosa26YFP mice (Fig. 3b). Therefore, a large proportion of Lkb1-deficient Treg cells lost Foxp3 expression and became ex-Treg cells that might result in the decrease of Treg cell abundance in vivo.
Foxp3 is an X-chromosome-encoded transcription factor. Due to random X inactivation, female mice heterozygous for Foxp3Cre (Foxp3Cre/+) should contain ∼1:1 ratio of Foxp3Cre-positive Treg cells to Foxp3Cre-negative Treg cells. Hence, the Foxp3Cre-heterozygous female Foxp3Cre/+Lkb1f/f mice were devoid of autoimmune diseases due to the presence of Treg cells lacking Foxp3Cre expression but retaining a wild-type Lkb1 allele8. Proinflammatory cytokines may impair Treg cell stability3,6. To determine whether the instability of Treg cells was due to either the inflammatory environment in Foxp3CreLkb1f/fRosa26YFP mice or the intrinsic defect of the cells, we utilized Foxp3Cre-heterozygous female Foxp3Cre/+Lkb1f/fRosa26YFP mice that were devoid of autoimmune diseases. Despite the presence of wild-type Treg cells devoid of the expression of Foxp3Cre in Foxp3Cre/+Lkb1f/fRosa26YFP mice, the CD4+Rosa26-YFPhigh cells from these mice were cells that had ever experienced the expression of Foxp3Cre that resulted in the deletion of Lkb1 and expression of Rosa26-YFP. Similarly, we observed substantially higher proportions of Foxp3− cells among sorted CD4+Rosa26-YFPhigh cells from Foxp3Cre/+Lkb1f/fRosa26YFP mice than that from Foxp3Cre/+Rosa26YFP mice (Supplementary Fig. 2c), ruling out the possibility that overproduction of proinflammatory cytokines in Foxp3CreLkb1f/f mice were the causative reasons for Treg cell instability.
Foxp3−Rosa26-YFP+ ex-Treg cells might also arise from recently activated T cells that transiently expressed Foxp3 but did not develop into a full Treg program (which were also considered as poorly committed Treg cells)27. To determine whether the full committed Treg cells lose stability, we doubly sorted CD4+CD25+YFP+ cells from CD45.1+CD45.2+ wild-type Foxp3Cre and CD45.1−CD45.2+Foxp3CreLkb1f/f mice, mixed them together with congenitally marked CD45.1+ CD45.2− naive T (Tn) cells and transferred them into Rag1−/− mice. Strikingly, we observed that nearly all the Lkb1-deficient Treg cells lost Foxp3 expression 3 weeks after transfer (Fig. 3c), despite the fact that most wild-type Treg cells retained Foxp3 expression. This was not due to the contamination of non-Treg cells because the starting population contained >99.5% Foxp3+ Treg cells after double sorting (Supplementary Fig. 2d). These results confirm that Foxp3 expression is lost in a large proportion of fully committed mature Lkb1-deficient Treg cells, rather than in a subset of poorly committed Foxp3+ cells.
Next, we investigated the molecular mechanisms by which Lkb1 promoted Treg cell stability. Because Foxp3 protein expression level was barely altered in Lkb1-deficient Treg cells (Supplementary Fig. 2e), Lkb1 seemed to not control Treg stability through affecting Foxp3 protein production or degradation. The epigenetic status across the Foxp3 locus has been shown to be critical for sustained Foxp3 expression9. Two recent reports demonstrated that deletion of CNS2 region in the Foxp3 locus led to the loss of Treg cell stability8,10, suggesting that CNS2 is pivotal for maintaining stable Foxp3 expression in Treg cells. CNS2 is completely demethylated in Treg cells, but not in conventional T cells9,28,29,30. Treg cell lineage development is partially dependent on the demethylation of CNS2 (refs 31, 32), and CNS2 has enhancer activity that is markedly reduced after methylation31. Therefore, we examined the methylation of CpG motifs across the Foxp3 locus by bisulfite sequencing in CD4+YFP+ Treg cells from Foxp3CreLkb1f/f mice and Foxp3Cre mice, and found a significant increase (nearly 25%) in methylation of CpG at CNS2 in Lkb1-deficient Treg cells ex vivo (Fig. 3d). Treg cells that already lost Foxp3 expression were excluded in this analysis. Although Foxp3 promoter and other potentially regulatory regions are associated with Treg cell lineage identity29,33,34, the methylation of CpG at sites including −1.5 kb region, promoter, exon 1, exon 8 and exon 11 across the Foxp3 locus in Lkb1-deficient Treg cells was comparable to the wild-type Treg cells (Fig. 3d). Therefore, we hypothesize that methylation of CpG at CNS2 may contribute to the instability of Lkb1-deficient Treg cells. It seems that the methylation of CNS2 is not a secondary effect of Foxp3 reduction, because Foxp3 protein level was barely altered in Lkb1-deficient Treg cells that already had substantial CNS2 methylation. Given the critical role of CNS2 in conferring Treg cell stability established by previous studies, our results together suggested that CNS2 methylation might lead to the instability of Lkb1-deficient Treg cells.
Lkb1 functions in Treg cells independent of AMPK
AMPK is the well-known Lkb1 downstream target critical for coordinating metabolism13. Interestingly, we found no obvious decrease of AMPK activity in Lkb1-deficient Treg cells (Supplementary Figs 3a and 8e–h). Therefore, other kinases such as calmodulin-dependent protein kinase kinase-β (CamKKβ) and transforming growth factor-β-activated kinase-1 (TAK1)15 may promote AMPK activation in Treg cells. These data argue that Lkb1 is not important for AMPK activation in Treg cells, and Lkb1 substrates other than AMPK are responsible for the observed phenotype of Lkb1-deficient Treg cells. Consistent with this, Treg cells that lacked the expression of AMPKα1/2 (from Foxp3CreAMPKα1f/fAMPKα2f/f mice, Supplementary Fig. 3b) did not exhibit any adverse phenotype and in vivo functional impairment (Supplementary Fig. 3c,d).
Lkb1 prevents STAT4 activation and STAT4 binding to CNS2
STAT3 and STAT6 are respectively the downstream mediators of the inflammatory cytokines IL-6 and IL-4 that bind to Foxp3 gene to mediate CpG methylation and Treg cell instability8, whereas the Treg cell growth factor IL-2 signals via STAT5 to counteract the destabilizing effect of these inflammatory cytokines8. Therefore, we examined whether the STAT family protein activations were altered in Lkb1-deficient Treg cells. Although STAT3 and STAT5 phosphorylations were barely affected, we detected a drastic increase of STAT4 phosphorylation and a slight increase of STAT6 phosphorylation in Lkb1-deficient Treg cells with or without the stimulation of relevant cytokines (Fig. 4a and Supplementary Figs 4a,b and 8i–k). The in vivo maintenance of Treg cells depends on their contact with dendritic cells (DCs)35,36, and hence we developed an in vitro DC-Treg cell co-culture system that could mimic the physiological situation to test which signalling would cause the instability of Treg cells. Lkb1-deficient Treg cells slightly lost stability when co-cultured with DCs without exogenous cytokines (Fig. 4b). Treatment with IL-2 alone, or IL-2 in combination with IL-4 or IL-6, did not lead to a significant loss of Foxp3 protein expression in Lkb1-deficient Treg cells after 4 days of culture (Fig. 4c). Strikingly, addition of IL-12, a potent STAT4 activator, caused the loss of Foxp3 expression in more than 50% Lkb1-deficient but not wild-type Treg cells (Fig. 4c), suggesting that STAT4 hyperactivation is the causative reason for the instability of Lkb1-deficient Treg cells. In addition, the apoptosis and proliferation were barely changed in Lkb1-deficient Treg cells compared with wild-type Treg cells (Supplementary Fig. 4c,d), suggesting that the loss of Foxp3 expressing Lkb1-deficient Treg cells in response to IL-2 + IL-12 is not due to their altered apoptosis or proliferation. Intriguingly, both divided and nondivided Lkb1-deficient Treg cells lost their stability (Fig. 4d). To determine whether the instability of Lkb1-deficient Treg cells was due to an intrinsic signalling alteration or in vivo selection of the Treg subpopulations with reduced stability, we used ERT2CreLkb1f/fRosa26YFP mice in which Lkb1 could be induced to be acutely deleted in Treg cells (Supplementary Figs 4e and 8l,m), thus avoiding the long-term in vivo selection. A similar IL-12-induced loss of stability was observed in Treg cells after induced Lkb1 deletion with 4-hydroxytamoxifen in vitro (Fig. 4e), implying an intrinsic role for STAT4 hyperactivation in destabilizing Lkb1-deficient Treg cells.
We then tested whether STAT4 could bind to and induce methylation of the Foxp3 locus in Lkb1-deficient Treg cells. Chromatin immunoprecipitation (ChIP) experiments showed that Lkb1-deficient Treg cells had significantly more STAT4 binding to the CNS2 than did the wild-type cells in response to IL-2 and IL-12 (Fig. 5a and Supplementary Datas 1 and 2). Additionally, compared with IL-2 treatment alone, IL-2 and IL-12 treatment decreased STAT5 binding to the CNS2 in Lkb1-deficient Treg cells (Fig. 5a), suggesting that STAT4 might outcompete STAT5 in DNA binding in the Foxp3 locus. Since DNA methyltransferase (Dnmt) 1 and 3a were critical for maintaining and inducing DNA methylation respectively37, we tested whether STAT4 was capable of recruiting Dnmt1 or 3a to the Foxp3 locus. Lkb1-deficient Treg cells had significantly more Dnmt1 binding to the CNS2 than did the wild-type cells in response to IL-2 and IL-12 (Fig. 5b), consistent with the results of STAT4 (Fig. 5a). Notably, STAT4 associated with Dnmt1 but not Dnmt3a in Treg cells (Fig. 5c and Supplementary Fig. 8n,o). Furthermore, we examined the methylation of CNS2 in Foxp3Cre and Foxp3CreLkb1f/f Treg cells treated with IL-2, IL-2+IL-12 or IL-2+IL-12+5-aza-deoxycytidine (5-aza-dC, a DNA methyltransferase inhibitor) to investigate the relationship between STAT4 and CNS2 methylation. In Foxp3CreLkb1f/f Treg cells, the methylation of CNS2 was higher after the treatment of IL-2+IL-12 compared with the treatment of IL-2 that could be reversed by 5-aza-dC (Fig. 5d). Indeed, the addition of 5-aza-dC rescued the loss of Foxp3 expression in Lkb1-deficient Treg cells (Fig. 5e). Since 5-aza-dC changes the global DNA methylation status, leading to the global changes in gene expression, we cannot exclude its effect on the expression of other genes. Nevertheless, these results together suggest that Lkb1 maintains stable Foxp3 expression in Treg cells under inflammatory conditions by preventing STAT4-mediated recruitment of Dnmt1 to and subsequent methylation of the Foxp3 CNS2.
Lkb1 restrains STAT4 and NF-κB signalling
The Stat4 and Il12rb2 mRNA and IL-12Rβ2 protein expression levels were elevated in Lkb1-deficient Treg cells that could explain the increased STAT4 phosphorylation in these cells in response to IL-12 (Supplementary Fig. 5a,b). To determine which signalling promotes Stat4 and Il12rb2 expression in Lkb1-deficient Treg cells, we examined the activation of NF-κB and AKT, the major signalling mediators driving T helper type cell development38,39. We found that NF-κB p65 but not AKT was hyperactivated in Lkb1-deficient Treg cells (Fig. 6a and Supplementary Figs 5c,d and 8p–r), and the chemical inhibitors to NF-κB reduced Stat4 and Il12rb2 mRNA expression (Fig. 6b), as well as STAT4 phosphorylation (Fig. 6c). In addition, conserved NF-κB binding sites were present at the Stat4 and Il12rb2 loci (Supplementary Data 1), and ChIP experiments demonstrated more binding of NF-κB p65 to Stat4 and Il12rb2 in Lkb1-deficient Treg cells than in wild-type cells (Fig. 6d and Supplementary Datas 1 and 2), suggesting a direct role of NF-κB activation in promoting Stat4 and Il12rb2 transcription. Of note, chemical inhibitors of NF-κB could partially rescue the instability of Lkb1-deficient Treg cells (Fig. 6e). These results collectively indicate an involvement of NF-κB activation in driving STAT4 activation and instability of Lkb1-deficient Treg cells.
IκB (inhibitor of NF-κB) kinase (IKK) phosphorylates IκB, leading to their degradation and activation of transcription factor NF-κB. IKK is composed of three subunits, the similar protein kinases IKKα and IKKβ, and a regulatory subunit IKKγ40. The phosphorylations serines (Ser) in IKKα (Ser 176/180)/β (Ser 177/181) and IκBα (Ser 32/36) were all increased in Lkb1-deficient Treg cells (Fig. 6f,g), suggesting that Lkb1 might dampen NF-κB signalling by inhibiting IKKα/β activation.
Lkb1 promotes the expression of immunosuppressive genes
We further explored whether there are defects in Lkb1-deficient Treg cells that did not lose Foxp3 expression yet. The T-cell activation phenotype was already obvious (Fig. 7a,b) in 9–11-day-old Foxp3CreLkb1f/f mice that had normal percentages of Treg cells among CD4+ T cells (Fig. 7c), suggesting that the suppressor function of Lkb1-deficient Treg cells might be impaired. Indeed, in vitro Treg suppression assay showed that with different Treg to Tn ratios, Lkb1-deficient Treg cells could not efficiently suppress the proliferation of responder T cells (Fig. 7d), indicating an impaired suppressive capacity of Lkb1-deficient Treg cells.
To investigate the molecular mechanisms by which Lkb1 controls Treg cell function, we compared the transcriptional profiling of CD4+YFP+ Treg cells sorted from Foxp3Cre/+ mice with that from Foxp3Cre/+Lkb1f/f mice lacking autoimmune diseases, thus avoiding the secondary impact of the inflammatory environment on Treg cell gene expression (Supplementary Data 3). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that the most altered pathways were enriched for genes regulating immune function (Supplementary Fig. 6a). Intriguingly, Lkb1-deficient Treg cells have reduced expression of a wide variety of genes critically involved in Treg cell suppressor function including those encoding secreted molecules associated with immune suppression (Fgl2, Il10 and Il18)11,41, factors in wound-healing processes (Tfpi)11, regulators of reactive oxygen generation (Cybb)11, enzymes catalysing the generation of adenosine that directly inhibit proliferation of effector T cells (Entpd1 and Nt5e)11,42, chemokine receptors critical for Treg cell migration (Ccr6)43 and cell surface molecules facilitating Treg cell suppressor functions in either tumour environment (Nrp1)44 or organ-specific inflammatory conditions (Itgae and Il1rl1)45,46 (Fig. 7e and Supplementary Fig. 6b). On the other hand, the loss of Lkb1 promoted the expression of genes encoding proinflammatory mediators (Tnfsf8)11 (Fig. 7e and Supplementary Fig. 6b). It was further confirmed that Lkb1-deficient Treg cells from Foxp3Cre/+Lkb1f/f mice had decreased protein levels of IL-10, Nt5e, Ccr6, Nrp1, Itgae and Il1rl1 (Fig. 7f). Most gene expression alterations were also recapitulated in Treg cells from Foxp3CreLkb1f/f mice (Supplementary Fig. 6b). These results clearly showed that Lkb1 promoted the immunosuppressive programs in Treg cells.
TGF-β signalling in Lkb1-implemented Treg cell function
TGF-β signalling plays crucial roles in Treg cell generation and function47. Transcriptional profiling analysis showed that Lkb1-deficient Treg cells express lower levels of Tgfbr1 and Tgfbr2 mRNA that was confirmed by real-time PCR and western blot (Fig. 8a and Supplementary Figs 7a and 8s,t). Hence, we proposed that this pathway might be defective in Lkb1-deficient Treg cells. Indeed, Smad2 (Ser 456/457)/Smad3 (Ser 423/425) phosphorylation was decreased (Fig. 8b), and Smad-DNA binding assay indicated that Smad transcriptional activity was impaired in Lkb1-deficient Treg cells (Fig. 8c). To determine whether impaired TGF-β signalling contributed to the impairment of Lkb1-deficient Treg cells, we conducted loss-of-function experiments by generating a mouse line with TGF-βR2 specifically deleted in Treg cells (Foxp3CreTgfbr2f/f, Supplementary Figs 7b and 8u,v) that displayed decreased Smad activation (Supplementary Fig. 7c). Although Foxp3CreTgfbr2f/f mice did not show overt autoimmune disease, increased percentages of CD44highCD62Llow effector/memory T cells (Fig. 8d) were observed despite a slightly increased Treg cell frequency among CD4+ T cells (Fig. 8e), indicating the functional impairment of TGF-βR2-deficient Treg cells. Consistently, TGF-βR2-deficient Treg cells displayed impaired suppressive capacity in vitro (Fig. 8f). To determine the global effects of defective TGF-β signalling on gene transcription, we conducted transcriptional profiling and found that TGF-βR2 deficiency led to the decreased expression of many genes important for Treg cell suppressor function (Supplementary Data 4). Remarkably, TGF-βR2 deficiency recapitulated part of the suppressor gene expression alterations in Lkb1-deficient Treg cells, including Nt5e, Ccr6, Nrp1, Itgae and Il1rl1 (Fig. 8g and Supplementary Data 4), that was also confirmed at protein levels (Fig. 8h,i). However, the Lkb1-regulated Fgl2, Il10, Tfpi, Entpd1 and Tnfsf8 transcripts were not altered in TGF-βR2-deficient Treg cells, suggesting that Lkb1 also utilize TGF-β-independent mechanisms to promote Treg cell function. To further determine whether lower expression of TGF-βR2 was the causative reason for impaired suppression function of Lkb1-deficient Treg cells, we generated Treg cells (from Foxp3Cre and Foxp3CreLkb1f/f mice) transduced with retrovirus carrying TGF-βR2 complementary DNA (cDNA). TGF-βR2 was successfully expressed in RFP+YFP+ Treg cells (Supplementary Fig. 7c). Overexpression of TGF-βR2 could partially rescue the expression of suppressor genes and suppression function of Lkb1-deficient Treg cells (Fig. 8j and Supplementary Fig. 7d,e). Together, these results indicate that Lkb1 implements Treg cell function partially through promoting TGF-β signalling. TGF-βR2-deficient Treg cells did not lose stability when co-transferred with Tn cells into Rag1−/− mice (Fig. 8k), suggesting that the decreased expression of TGF-β receptors does not account for the instability of Lkb1-deficient Treg cells.
Here, we demonstrate that mice with Treg cell-specific deletion of Lkb1 develop a fatal, early-onset, autoimmune disease with moribund at ∼35 days of age. Such a fulminant disease is comparable to that observed in mice devoid of Treg cells or deficient in Foxp3. Similar devastating diseases have also been reported to be present in mice with selective deficiency of Foxo1 (ref. 48) or Raptor, the rapamycin sensitive partner of mTORC1 (mammalian target of rapamycin complex 11)26 in Treg cells, due to impaired Treg cell function. Intriguingly, Lkb1-deficient Treg cells lost their linage-specific features and exhibited severe impairment in the maintenance of Foxp3 expression and suppressor function that may underlie the extremely severe autoimmunity in Foxp3CreLkb1f/f mice. Although Lkb1 expression is not restricted to Treg cells, Treg cells substantially upregulate Lkb1 protein expression upon TCR stimulation. Given that TCR signalling is essential for Treg cell homeostasis and function, TCR-induced upregulation of Lkb1 protein may represent a critical feed-forward loop to stabilize Treg cell lineage identity.
It is generally recognized that Foxp3 expression is highly stable in Treg cells. Whether a small proportion of Treg cells can lose Foxp3 expression in certain conditions is a controversial topic6. Some studies showed that Treg cells could lose Foxp3 expression and become ‘ex-Treg’ cells under certain conditions25, while others argued that the so-called ‘ex-Treg’ cells just might be the conventional T cells that transiently express Foxp3, namely poorly committed Treg cells27. Here, we showed that deficiency of Lkb1 caused very severe defects in Treg cell stability. More than 70% of Treg cells lost stability in Foxp3CreLkb1f/fRosa26YFP mice. In addition, the majority of the Lkb1-deficientTreg cells lost Foxp3 expression after being transferred into Rag1−/−mice, indicating that Lkb1-deficient mature Treg cells continuously lost Foxp3 expression, rather than that a subset of Lkb1-deficient T cells only transiently expressed Foxp3 (poorly committed Treg cells). Such a severe defect was rarely reported in literatures, placing Lkb1 as a particularly important maintainer of Foxp3 stability. The demethylated status of Foxp3 CNS2 is critically involved in the maintenance of stable Foxp3 expression in Treg cells8,9,10,28,29,30,31,32. Substantial methylation of CNS2 but not promoter was observed in Lkb1-deficient Treg cells. Treatment of IL-12 further enhanced the methylation of CNS2 and caused the loss of Foxp3 expression in Lkb1-deficient but not wild-type Treg cells. Moreover, the DNA methyltransferase inhibitor 5-aza-dC reduced the methylation of CNS2 and rescued the instability of Lkb1-deficient Treg cells. These results suggest that Lkb1 stabilizes Foxp3 expression by building up the demethylated status of CNS2. It has been recently reported that deletion of CNS2 caused cell-division-dependent loss of Foxp3 expression8,10. In contrast, both divided and nondivided Lkb1-deficient Treg cells lost Foxp3 expression, suggesting that Lkb1 might also promote Foxp3 expression by CNS2-independent mechanisms, which need future exploration.
The role of STAT4 in Treg cells is not well defined. Here we showed that deficiency of Lkb1 in Treg cells resulted in markedly more IL-12-induced STAT4 activation and STAT4 binding to CNS2. STAT4 associated with Dnmt1 and DNA methyltransferase chemical inhibitor rescued the instability of Lkb1-deficient Treg cells in response to IL-12. Thus, Lkb1 suppresses IL-12/STAT4 activation to stabilize Foxp3 expression in Treg cells. We further showed that NF-κB signalling is activated in Lkb1-deficient Treg cells, that increased Il12rb2 and Stat4 mRNA expression and STAT4 activation. However, inhibition of NF-κB signalling can only partially rescue the instability of Lkb1-deficient Treg cells, implicating that NF-κB-independent signalling mechanisms might also contribute to the unstable Foxp3 expression in Lkb1-deficient Treg cells.
Independent of maintaining Foxp3 expression, Lkb1 also promotes Treg cell-suppressive capacity. Loss of Lkb1 in Treg cells results in the moderate downregulation of a number of immunosuppressive genes (Fgl2, Tfpi, Cybb, Entpd1, Nt5e, Nrp1, Itgae, Il18, Il10, Ccr6, Il1rl1), suggesting that Lkb1 implements Treg cell function through broadly promoting Treg suppressor gene expression program. Lkb1 has been previously shown to positively or negatively regulate TGF-β/Smad signalling in a cell type-dependent manner49,50, but the precise mechanism is not fully elucidated. Here, we found the effect of Lkb1 on promoting Treg suppressor function was partially dependent on its augmentation of TGF-β signalling. TGF-β pathway is important for Treg cell development in the thymus and periphery47,51,52, but whether and how it affects Treg cell function remain poorly understood. Our results demonstrate that TGF-βR2-deficient Treg cells have impaired suppressive capacity associated with decreased expression of a number of genes associated with immune suppression (Nt5e, Ccr6, Nrp1, Itgae and Il1rl1), partially recapturing the suppressor gene expression alterations in Lkb1-deficient Treg cells. Furthermore, overexpression of TGF-βR2 partially rescued the expression of certain suppressor genes and the suppressor function of Lkb1-deficient Treg cells, supporting that impaired TGF-β signalling contributed to the impaired suppressor function of Lkb1-deficient Treg cells. However, numerous Lkb1-regulated genes, including Fgl2, Tfpi, Cybb, Entpd1 and Il18, were not changed in TGF-βR2-deficient Treg cells, suggesting that TGF-β-independent mechanisms were also applied by Lkb1 to promote Treg cell suppressor function.
AMPK-induced fatty acid oxidation is a critical metabolic feature of Treg cells53. To our surprise, AMPK activation is not dramatically altered in Lkb1-deficient Treg cells, and Treg cell-specific deletion of both AMPK1α and AMPK2α does not cause Treg cell abonrmality and immune disturbance, suggesting that Lkb1 does not control Treg cells by promoting AMPK activity in steady-state conditions.
Lkb1 has been previously shown to promote thymocytes development, maintain survival and proliferation of peripheral T cells and restrain peripheral conventional T-cell activation and proinflammatory cytokine production15,16,54 (we got similar results in certain experiments after inducing Lkb1 deletion in conventional T cells from ERT2CreLkb1f/f mice, Supplementary Fig. 7f,g). Here, we find that Lkb1 maintains the lineage identity of Treg cells through mechanisms involving the regulation of NF-κB/STAT4 and TGF-β signallings (Fig. 8). NF-κB/STAT4 and TGF-β signallings are separately controlled by Lkb1 (Supplementary Fig. 7h–j), suggesting a function of Lkb1 in setting up the optimal activation thresholds of multiple intracellular signallings in Treg cells. Furthermore, the upregulation of Lkb1 protein by TCR signalling in Treg but not in conventional T cells indicates a specific regulation of Lkb1 expression required for preserving the Treg cell lineage. Future studies will determine whether and how the environmental cues change Lkb1 in Treg cells to affect their stability/function under physiological and pathological conditions, and whether Lkb1 can be targeted to treat Treg-cell-related immune diseases.
All animals were maintained in specific pathogen-free barrier facilities and were used in accordance with protocols approved by the institutional animal care and user committee at the Institute of Hematology, Chinese Academy of Medical Sciences. C57BL/6, CD45.1+, Lkb1f/f, AMPKα1f/f, AMPKα2f/f, Tgfbr2f/f, ERT2Cre, Foxp3YFP-Cre (Foxp3Cre), Rosa26YFP and Rag1−/− mice were purchased from Jackson Laboratories. All mice had been backcrossed with C57BL/6 mice for at least 7 generations. Lkb1f/f, AMPKα1f/f, AMPKα2f/f, Tgfbr2f/f mice were crossed with Foxp3Cre mice to generate Foxp3CreLkb1f/f, Foxp3CreAMPKα1f/fAMPKα2f/f and Foxp3CreTgfbr2f/f mice, respectively. Foxp3Cre/+Lkb1f/f mice were crossed with Rosa26YFP mice to generate Foxp3CreLkb1f/fRosa26YFP mice. Lkb1f/f mice were crossed with ERT2Cre and Rosa26YFP mice to generate ERT2CreLkb1f/fRosa26YFP mice. Sample size for various animal experiments was chosen based on prior data generated in the laboratory, and no mice were excluded from experiments. The details of mice age and sex are provided in the figure legends.
Cell purification and flow cytometry
Single-cell suspensions were prepared from spleen and peripheral lymph nodes for staining or cell purification. CD4+ T cells and CD11c+ DCs were purified with Dynabeads Untouched Mouse CD4 Cells Kits (Invitrogen, 11415D) or CD11c MicroBeads (Miltenyi Biotec, 130-108-338), respectively. Indicated T-cell populations were sorted from purified CD4+ T cells with FACSAria III (BD Biosciences), and the sorted populations were >98% pure unless otherwise specified. Flow cytometry of cell surface molecules was performed as previously described55, and the antibody (Supplementary Data 5) was diluted as the instructions of the manufacturers suggested. Intracellular staining of Foxp3, cytokines (spleen cells were stimulated with phorbol myristate acetate (50 ng ml−1) and ionomycin (500 ng ml−1) for 4 h before analysis of cytokine expression in indicated populations) and other proteins were performed with Foxp3 staining kits (eBioscience, 00-5523). Intracellular staining of phosphorylated proteins was performed on cells fixed with methanol and permeabilized with Triton X-100. The antibodies were obtained from eBioscience, Biolegend, BD Biosciences, Cell Signalling Technology, R&D Systems and Invitrogen, and listed in Supplementary Data 5. Flow cytometry data were acquired on LSR II, LSRFortessa (BD Biosciences) or FACSCanton II (BD Biosciences) and analysed with Flowjo software (Tree Star).
Sorted CD4+YFP+ Treg cells from Foxp3CreLkb1f/f and Foxp3Cre mice were labelled with carboxyfluorescein succinimidyl ester (CFSE) Cell Proliferation Kits (Invitrogen, C34554). CFSE-labelled Treg cells were co-cultured with CD11c+ cells purified from CD45.1+ mice, supplemented with indicated combinations of cytokines of recombinant murine IL-2, IL-12, IL-6 and IL-4 (all 100 ng ml−1, PeproTech), NF-κB inhibitors sodium 4-aminosalicylate (4-ASA) or pyrrolidinedithiocarbamic acid (PDTC) or DNA methyltransferase inhibitor 5-aza-dC)in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (Gibco) for 4 days. For inducible Lkb1 deletion, Treg cells were respectively sorted from ERT2CreLkb1f/fRosa26YFP and ERT2CreRosa26YFP mice and cultured for 5 days as described above, with the addition of 4-hydroxytamoxifen (1 μM, Sigma) in the culture. In other experiments, sorted Treg cells were incubated in plates precoated with anti-CD3 (2 μg ml−1; 145-2C11; eBioscience) and anti-CD28 (2 μg ml−1; 45.21; eBioscience) in the presence of IL-2 (100 ng ml−1) for the indicated days. CFSE profiles and Foxp3 expressions were examined by flow cytometry after culturing.
In vivo Treg cell maintenance
CD4+CD25+YFP+ Treg cells were doubly sorted (>99.5% pure) from 3-week-old CD45.1− CD45.2+ Foxp3CreLkb1f/f and CD45.1+CD45.2+Foxp3Cre mice, and 2 × 105 of each type of cells were mixed with 4 × 105 CD4+CD25−CD44lowCD62Lhigh Tn cells sorted fromCD45.1+ CD45.2− mice and transferred into the Rag1−/− mice by intraperitoneal injection. The recipient mice were analysed 3 weeks after transfer.
In vitro Treg cell suppression
CD4+CD25−CD44lowCD62Lhigh Tn cells sorted from CD45.1+ mice were labelled with CFSE and used as responder cells (Tresp). Tresp cells (5 × 104) were cultured for 72 h with DCs (1 × 105) and soluble anti-CD3 (2 μg ml−1) in the presence or absence of the indicated numbers of CD4+YFP+ Treg cells sorted from Foxp3CreLkb1f/f or Foxp3Cre mice.
Foxp3 gene methylation assay
Genomic DNA from sorted cells was bisulfite converted with EZ DNA Methylation-Direct kits according to the manufacturer’s protocol (Epigentek, P-1026-050). Methylation-specific PCR primers (listed in Supplementary Data 2) were used for amplification of the promoter and intron 1 of Foxp3 (corresponding to Foxp3 conserved noncoding sequence 2). PCR products were subcloned into pGEM-T Easy vectors (Promega) and sequenced.
CD4+YFP+ Treg cells and CD4+YFP− conventional T cells were sorted from Foxp3Cre and Foxp3CreLkb1f/f mice and treated as indicated, and then cells were lysed and western blot was carried out as previously described55 with antibodies to Lkb1 (D60C5, Cell Signalling Technology), phosphorylated AMPK (40H9, Cell Signalling Technology), IKKβ (2C8, Cell Signalling Technology), phosphorylated IKKα/β (16A6, Cell Signalling Technology), phosphorylated IκBα (14D4, Cell Signalling Technology), phospho-NF-κB p65 (93H1, Cell Signalling Technology), NF-κB p65 (D14E12, Cell Signalling Technology), STAT4 (C46B10, Cell Signalling Technology), AMPKα (D63G4, Cell Signalling Technology), phospho-Acetyl-CoA Carboxylase (D7D11, Cell Signalling Technology), phospho-STAT4 (Abcam), TGF-βR2 (Abcam) and GAPDH (D16H11, Cell Signalling Technology).
Microarray and quantitative real-time PCR
CD4+YFP+ Treg cells were sorted from spleen and lymph nodes of mice for RNA extraction with Trizol reagent (Invitrogen). Total RNA was reverse transcribed, amplified, labelled and hybridized to Mouse Genome 2.0 arrays (Affymetrix). Microarray data sets were analysed with Agilent Genespring GS 11 software. RNA of different samples was obtained in the same manner as in microarray analysis, and real-time PCR was performed with SYBR Green PCR Master Mix (ABI) as previously described55. The sequences of the primer pairs used are listed in Supplementary Data 2.
DNA binding ELISA
TransAM Flexi kits (Active Motif, 40098) were used to test the activity of Smad transcription factors according to the manufacturer’s protocol. Briefly, nuclear extract from stimulated Treg cells sorted from Foxp3Cre and Foxp3CreLkb1f/f mice were mixed with biotinylated oligos (listed in Supplementary Data 2) recognized by transcription factors of interest, and incubated in a streptavidin-coated plate. After washing, primary antibodies specific for the bound transcription factor and horseradish peroxidase-conjugated secondary antibodies were subsequently added, and transcription factor binding activities were quantified using a microplate reader (Synergy H4, BioTek).
Sorted wild-type CD4+YFP+ Treg cells were expanded in plates coated with anti-CD3 (2 μg ml−1) and anti-CD28 (2 μg ml−1) in the presence of IL-2 (100 ng ml−1) for 5 days and collected. Cell lysis was prepared using Co-immunoprecipitation kits according to the manufacturer’s protocol (Sigma, FLAGITP1), and incubated with antibodies against Dnmt1 (60B1220.1, Abcam), Dnmt3a (Abcam) and protein G agarose (Cell Signalling Technology). Immunoprecipitated proteins were detected as described above with antibodies against STAT4 (C46B10, Cell Signalling Technology) and STAT5 (Cell Signalling Technology).
ChIP assays were performed using the ChIP kits (Active Motif, 53040) according to the manufacturer’s protocol. Precipitated DNA and input DNA were assessed by real-time PCR using primers listed in Supplementary Data 2.
TGF-βR2 cDNA was subcloned into the retroviral vector pMYs-IRES-RFP. Retroviruses were produced by transfection of the Plat-E cells with polyethylenimin. Treg cells (YFP+) were cultured in plates coated with anti-CD3 and anti-CD28 transduced in virus-containing media supplemented with polybrene by centrifuging for 1 h at 500 RCF. After 24 h, cells were expanded and then RFP+ YFP+ cells were sorted before further analysis. The Plat-E cell line was provided by the Cell Resource Center, Institute of Hematology and Hospital of Blood Diseases, Chinese Academy of Medical Sciences/Peking Union Medical College. The Cell Resource Center confirmed the species origin with PCR and checked free of mycoplasma contamination by PCR and culture.
Skins, lungs, livers and stomachs were removed from 3–4-week-old mice. Samples were formalin fixed, paraffin embedded and stained with haematoxylin and eosin before tissue histology. Photomicrographs were taken at × 10 or × 20 magnifications.
An unpaired two-tailed Student’s t-test (for two group comparisons) or a two-way analysis of variance (for more than two group comparisons) were performed using Prism (GraphPad) to calculate statistical significance of the difference in mean values and P values. A P value of <0.05 was considered statistically significant. *P<0.05; **P<0.01. No specific randomization or blinding protocols were used.
The data that support the findings of this study are available from the corresponding author on reasonable request. The microarray data have been deposited in the Gene Expression Omnibus under accession GSE97840.
How to cite this article: Wu, D. et al. Lkb1 maintains Treg cell lineage identity. Nat. Commun. 8, 15876 doi: 10.1038/ncomms15876 (2017).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank Professor Malcolm White FRSE, University of St Andrews, Professor Dangshen Li, Chinese Academy of Sciences and Professor Hui Xiao, Institute Pasteur of Shanghai, Chinese Academy of Sciences, for critical comments and corrections of the manuscript. This work was supported by the National Basic Research Program of China (2015CB964400 to X.F. and Y.Y., and 2013CB966904 to X.F.), the National Natural Science Foundation of China (81322007, 81273217 and 81670107 to X.F., 81370104 to G.X., 81421002 to T.C. and 81401295 to Y.L.), the Recruitment Program of Global Youth Experts (to X.F.), the CAMS Innovation Fund for Medical Sciences (CIFMS, 2016-I2M-1-003), Postdoctoral Science Foundation Program of CAMS & PUMC to D.W. (2015), the Tianjin Research Program of Application Foundation and Advanced Technology (15JCQNJC45200 to Y.L.) and the PUMC Youth Fund and the Fundamental Research Funds for the Central Universities (3332015126 to Y.L.).
About this article
Cellular and Molecular Immunology (2017)