Research ArticleSystemic Lupus Erythematosus

Normalization of CD4+ T cell metabolism reverses lupus

See allHide authors and affiliations

Science Translational Medicine  11 Feb 2015:
Vol. 7, Issue 274, pp. 274ra18
DOI: 10.1126/scitranslmed.aaa0835


Systemic lupus erythematosus (SLE) is an autoimmune disease in which autoreactive CD4+ T cells play an essential role. CD4+ T cells rely on glycolysis for inflammatory effector functions, but recent studies have shown that mitochondrial metabolism supports their chronic activation. How these processes contribute to lupus is unclear. We show that both glycolysis and mitochondrial oxidative metabolism are elevated in CD4+ T cells from lupus-prone B6.Sle1.Sle2.Sle3 (TC) mice as compared to non-autoimmune controls. In vitro, both the mitochondrial metabolism inhibitor metformin and the glucose metabolism inhibitor 2-deoxy-d-glucose (2DG) reduced interferon-γ (IFN-γ) production, although at different stages of activation. Metformin also restored the defective interleukin-2 (IL-2) production by TC CD4+ T cells. In vivo, treatment of TC mice and other lupus models with a combination of metformin and 2DG normalized T cell metabolism and reversed disease biomarkers. Further, CD4+ T cells from SLE patients also exhibited enhanced glycolysis and mitochondrial metabolism that correlated with their activation status, and their excessive IFN-γ production was significantly reduced by metformin in vitro. These results suggest that normalization of T cell metabolism through the dual inhibition of glycolysis and mitochondrial metabolism is a promising therapeutic venue for SLE.


CD4+ T cells are active mediators of systemic lupus erythematosus (SLE) pathogenesis. Autoreactive nucleosome-specific T cells provide help to anti-DNA–specific B cells in mice (1) and SLE patients (2). Lupus-prone mice carry large numbers of activated T cells (3), and signaling defects have been found in SLE CD4+ T cells (4). In addition, SLE patients and lupus-prone mice show an increase in T helper 1 (TH1) and TH17 subsets (5), as well as a dysregulated expression of interferon-γ (IFN-γ) and interleukin-2 (IL-2) (68).

The activation, proliferation, and differentiation of CD4+ T cells are tightly regulated by cellular metabolism. Quiescent T cells have a low energy demand and can interchangeably use glucose, lipid, and amino acids to fuel oxidative metabolism (9, 10). Activation through the T cell receptor increases both glycolysis and mitochondrial metabolism (1114). The up-regulation of glycolysis results from a transcriptional reprogramming regulated by c-MYC, HIF1α (hypoxia-inducible factor 1α), and ERRα (estrogen-related receptor α) (12, 15, 16). The importance of glucose metabolism in autoimmunity was demonstrated by the overexpression of the glucose transporter Glut1 in mice that led to CD4+ T cell hyperactivation, hypergammaglobulinemia, and immune complex–mediated nephritis (17). Cellular metabolism also regulates effector T cell differentiation and memory cell formation. TH1, TH2, and TH17 cells are highly glycolytic, whereas Foxp3+ regulatory T (Treg) cells and long-lived memory T cells have high lipid oxidation rates (18, 19).

Several studies have suggested that SLE CD4+ T cells present defective mitochondrial functions. Splenocytes from (NZB × NZW)F1 (NZB/W) mice showed an enhanced pyruvate oxidation (20). Moreover, mitochondrial membrane hyperpolarization was associated with an elevated mammalian target of rapamycin (mTOR) activity in CD4+ T cells from SLE patients (21). Accordingly, SLE patients treated with rapamycin showed reduced disease manifestations (22). Furthermore, CD4+ T cells use oxidative metabolism for chronic activation in response to alloantigens (23), and adenosine triphosphate (ATP) synthase inhibitors eliminated pathogenic T cells (23, 24). Given these conflicting results between the preponderance of either glycolysis or mitochondrial metabolism to support the function of activated effector T cells, the metabolism of SLE CD4+ T cells remains poorly understood.

The B6.Sle1.Sle2.Sle3 lupus-prone mouse model (a triple congenic strain hereafter called TC) contains three NZM2410-derived lupus susceptibility loci—Sle1, Sle2, and Sle3—on a non-autoimmune C57BL/6 (B6) background (25). TC mice spontaneously develop symptoms similar to SLE patients, including the production of anti–double-stranded DNA (dsDNA) immunoglobulin G (IgG) and a high penetrance of immune complex–mediated fatal glomerulonephritis (GN). The Sle1c2 susceptibility locus corresponds to the reduced expression of the Esrrg gene, which contributes to CD4+ T cell activation and increased IFN-γ secretion (26). Esrrg controls cellular metabolism by up-regulating mitochondrial oxidative phosphorylation (OXPHOS) (27, 28). This finding led us to hypothesize that cellular metabolism contributes to lupus pathogenesis through T cell activation and that metabolism modulators could be used to reduce or revert disease. We tested this hypothesis in TC mice, which express Sle1c2, and CD4+ T cells from SLE patients. Here, we show that both glycolysis and mitochondrial oxygen consumption are elevated in CD4+ T cells from TC mice. This enhanced metabolism was observed in naïve T (Tn) cells and was amplified with age and activation. In vitro, treatment with mitochondrial electron transport chain complex I inhibitor metformin (Met) or glucose metabolism inhibitor 2-deoxy-d-glucose (2dG) normalized IFN-γ production by TC CD4+ T cells. In vivo, treatment of TC mice and other lupus models with the combination of Met and 2DG normalized T cell metabolism and reversed disease phenotypes. We also demonstrated that CD4+ T cells from SLE patients exhibited both a higher glycolysis and mitochondrial metabolism as compared to healthy controls (HCs), which correlated with T cell activation and subset distribution. Moreover, their excessive IFN-γ production was significantly reduced by Met in vitro. These results suggest that targeting T cell metabolism represents a promising therapeutic strategy for SLE.


CD4+ T cells from TC mice exhibited elevated T cell metabolism

TC CD4+ T cells present with several immune abnormalities that are typical of lupus pathogenesis (25, 29), including T cell hyperactivation (fig. S1A and Fig. 5B), accumulation of CD44+CD62L effector memory (Tem) and CD44+CD62L+ central memory T (Tcm) cells (fig. S1B and Fig. 5C), as well as increased IFN-γ production (fig. S1C and Fig. 3H). To test whether these CD4+ T cell phenotypes were associated with alterations in cellular metabolism, we measured their extracellular acidification rate (ECAR), which is primarily attributed to glycolysis, and the oxygen consumption rate (OCR), which corresponds to OXPHOS. CD4+ T cells from 2-month-old predisease TC mice showed enhanced ECAR and OCR compared to age-matched B6 counterparts. This difference in CD4+ T cell metabolism became more pronounced in 9-month-old TC mice, which have developed clinical disease (Fig. 1, A to C). CD4+ T cells from 9-month-old TC mice also showed a higher spare respiratory capacity (SRC) (Fig. 1D), an indication of cellular energy reserve that is essential for memory T cell formation and function (30). The enhanced glycolysis of TC CD4+ T cells was confirmed by increased extracellular lactate production (Fig. 1E). Despite increased glycolysis and OXPHOS, TC CD4+ T cells presented intracellular ATP levels comparable to B6, both ex vivo and after activation (Fig. 1F). This result suggests that the increased metabolism leads to ATP consumption by TC CD4+ T cells to support elevated effector functions. Overall, CD4+ T cells from TC mice present with an enhanced cellular metabolism that precedes disease manifestation and increases as T cells become more activated and disease progresses.

Fig. 1. CD4+ T cells from TC mice show an enhanced metabolism.

(A to D) ECAR (A), OCR (B and C), and SRC (D) measured in total CD4+ T cells from 2- and 9-month-old B6 and TC mice. (B) Representative OCR in 9-month-old B6 and TC CD4+ T cells. FCCP, carbonyl cyanide p-trifluoromethoxyphenylhydrazone. (E) Extracellular lactate production from 3-month-old B6 and TC CD4+ T cells. (F) ATP production by B6 and TC CD4+ T cells stimulated with PMA/ionomycin (IO) or anti-CD3/CD28. RQ, relative quantity. (G and H) ECAR (G) and OCR (H) in Tn and Tem from 9-month-old B6 and TC mice. (I and J) ECAR (I) and OCR (J) in B6 and TC Tn after 24-hour stimulation with anti-CD3/CD28. n = 3 to 6.

Naïve and activated CD4+ T cell subsets have different metabolic profiles (31). The expansion of Tem cells in TC mice (fig. S1B) could be the source of the elevated metabolism observed in total CD4+ T cells. Tem cells showed a significantly higher ECAR and, to a lesser extent, OCR than did Tn cells in both B6 and TC mice (Fig. 1, G and H). This result implies that the higher percentage of Tem cells in TC mice contributes to the higher metabolism of total CD4+ T cells. In addition, TC Tn cells also showed a higher ECAR and OCR than did B6 Tn cells. To directly characterize the metabolism of TC CD4+ T cells after activation, we compared the metabolic profiles of B6 and TC-sorted Tn cells after stimulation with anti-CD3 and anti-CD28 for 24 hours. In vitro stimulated TC Tn cells exhibited significantly higher ECAR and OCR as compared to B6 (Fig. 1, I and J). The activity of mTOR complex 1 (mTORC1), a sensor for cellular energy (32), was elevated in TC CD4+ T cells, as shown by an increased phosphorylation of ribosomal protein S6 and 4E-BP1 (eukaryotic translation initiation factor 4E–binding protein 1) and the increased expression of CD98 and CD71 (Fig. 2A and fig. S2). The mTORC1 inhibitor rapamycin reduced glycolysis and mitochondrial oxygen consumption in anti-CD3/CD28–activated B6 CD4+ T cells (Fig. 2, B and C), linking mTORC1 activity to both glycolysis and mitochondrial metabolism. These results indicate that naïve TC CD4+ T cells exhibit both an increased glycolysis and mitochondrial oxygen consumption ex vivo and in response to in vitro activation.

Fig. 2. CD4+ T cells from TC mice show an increased mTORC1 activity.

(A) S6 and 4E-BP1 phosphorylation and expression of CD98 and CD71 in total CD4+ T cells as well as Tn, Tem, and Tcm subsets from 2-month-old mice. n = 3 to 4. (B and C) ECAR (B) and OCR (C) in B6 CD4+ T cells stimulated with anti-CD3/CD28 with or without rapamycin (Rapa) (100 nM) for 24 hours. Representative graphs of two independent assays each performed with n = 7 technical replicates.

TC CD4+ T cells showed an elevated expression of glycolytic genes Hif1a, Hk2, and Slc16a3 (fig. S3A). On the other hand, the level of Pdk1, which inhibits pyruvate oxidation, was significantly lower in TC than in B6 CD4+ T cells (fig. S3B). The expression of Cpt1a, a key fatty acid oxidation regulator (30), was higher in TC CD4+ T cells than in B6 (fig. S3C). Consistently, TC CD4+ T cells showed an enhanced uptake of fatty acids (fig. S3D). Finally, TC CD4+ T cells showed a higher expression of Gls2 and Odc (fig. S3E), two genes involved in amino acid metabolism (12). In summary, CD4+ T cells from lupus-prone TC mice present an enhanced metabolism fueled through both glycolysis and mitochondrial metabolism, with evidence that glucose, fatty acids, and glutamine may all contribute to the latter.

TC CD4+ T cell dysfunction was normalized by metabolic modulators in vitro

We tested whether targeting glycolysis and mitochondrial metabolism could normalize TC CD4+ T cell functions in vitro. As expected, 2DG dose-dependently inhibited CD4+ T cell ECAR (Fig. 3A) but also decreased OCR (Fig. 3B), presumably by decreasing glucose oxidation. Consistent with its inhibition of mitochondrial complex I (33), Met dose-dependently decreased OCR (Fig. 3C) and increased ECAR at a high concentration (Fig. 3D), likely as an alternative pathway to generate ATP. We examined whether IFN-γ production could be modulated by Met or 2DG in vitro. TC CD4+ T cells produced more IFN-γ than did B6 T cells after a short stimulation with PMA (phorbol 12-myristate 13-acetate)/ionomycin, which was decreased by treatment with Met (Fig. 3E). The ATPase inhibitor oligomycin and the combination of complex I/III inhibitors rotenone/antimycin A also inhibited IFN-γ production, indicating that mitochondrial metabolism is required for IFN-γ production during early CD4+ T cell activation. 2DG, however, had no effect, indicating that IFN-γ production does not rely on glycolysis at that stage. Under TH1 polarizing condition, TC CD4+ T cells also generated more IFN-γ than did B6 T cells (Fig. 3F). Met added at the beginning of the culture dose-dependently decreased IFN-γ production (Fig. 3F), which could be observed as early as day 1 and throughout the 3-day culture (Fig. 3G). When added at day 2 of polarization, Met had no effect, whereas 2DG markedly reduced IFN-γ production in both strains (Fig. 3H). These results show that, in both normal and lupus mice, IFN-γ production by CD4+ T cells depends on mitochondrial metabolism in the activation phase when it can be reduced by Met, whereas its production during the proliferation phase depends on glycolysis and can be reduced by 2DG.

Fig. 3. Metabolic modulators normalized TC CD4+ T cell effector functions in vitro.

(A to D) ECAR (A and C) and OCR (B and D) in B6 CD4+ T cells stimulated with anti-CD3/CD8 for 24 hours in the presence of 2DG or Met. Representative graphs of three assays each performed with technical replicates. (E) IFN-γ production in CD4+ T cells stimulated with PMA/ionomycin/GolgiPlug for 6 hours (Ctrl), in the presence of Met (2 mM), 2DG (5 mM), antimycin A/rotenone (A&R) (both 0.5 μM), or oligomycin (Olig) (1 μM). (F to H) IFN-γ production in CD4+ T cells cultured under TH1 condition for 3 days with Met [0 to 5 mM in (F) or 1 mM in (G)] or 2DG (1 mM) added from day 0 (d0) (F and G) or day 2 (d2) (H). (I and J) IL-2 production in CD4+ T cells stimulated with anti-CD3/CD28 for 3 days with 0 to 1 mM (I) or 1 mM (J). (G) Statistical significance of a two-way analysis of variance (ANOVA) between B6 and TC. n = 3 to 8.

IL-2 production is defective in lupus T cells (7). Anti-CD3/CD28–stimulated TC CD4+ T cells produced less IL-2 than did B6 T cells, and Met treatment increased IL-2 production in both strains in a dose-dependent manner (Fig. 3, I and J, and fig. S4). In summary, TC CD4+ T cells display excessive IFN-γ as well as a defective IL-2 secretion. Both Met and 2DG inhibited their capacity to produce IFN-γ, but at different stages of activation, and Met promoted IL-2 production. This resulted in an overall normalization of TC CD4+ T cell functions in vitro.

A treatment combining Met and 2DG reversed disease in TC mice

To target glucose metabolism and mitochondrial metabolism simultaneously, we treated 7-month-old TC mice with a combination of 2DG and Met (Met + 2DG) in drinking water for 3 months. At that age, TC mice are at the early stage of clinical disease that includes splenomegaly, anti-dsDNA IgG production, and accumulation of activated T (fig. S1) and B cells, as well as splenic plasma cells (25). The Met + 2DG treatment significantly decreased ECAR (Fig. 4A) and OCR (Fig. 4B) in TC CD4+ T cells to levels similar to that of B6 T cells, indicating that the treatment effectively targeted CD4+ T cell metabolism in vivo. B6 CD4+ T cells were, however, not affected, suggesting that the treatment is selective to the more metabolically active TC cells. A further characterization showed that the in vivo treatment selectively down-regulated metabolism in TC effector T cells (fig. S5). The Met + 2DG treatment significantly reduced splenomegaly in TC mice but had little effect in B6 mice (Fig. 4C). The treatment significantly decreased the production of anti-dsDNA IgG and anti-nuclear autoantibodies (ANAs), whereas it increased in untreated TC mice (Fig. 4, D and E). The inhibitory effect of the treatment on autoantibody production was confirmed with antigens arrays, in which four of five treated TC mice showed a profile similar to B6 mice (Fig. 4F). The Met + 2DG treatment also reduced the level of C3 and IgG2a immune complex deposition in TC kidneys (Fig. 4G). Furthermore, GN was less severe in treated TC mice, assessed either by severity rank or by score distribution (Fig. 4, H and I).

Fig. 4. Treatment with Met + 2DG for 3 months reversed disease in 7-month-old B6 and TC mice.

(A and B) ECAR (A) and OCR (B) in CD4+ T cells. (C) Spleen weight (representative spleens on the right of 14 for TC and 5 for B6). (D) Serum anti-dsDNA IgG in TC mice (two-way ANOVA). (E) Initial (I) and terminal (T) serum ANA from TC mice. Representative images of 14 sera per group and ANA intensity quantification, with each linked symbol representing a mouse. Untreated B6 mice are shown as control (paired t tests). (F) Autoantibody microarray analysis of terminal sera (IgG). ssDNA, single-stranded DNA; H2A, histone 2A; H2B, histone 2B. (G) Immune complex deposition in TC glomeruli. Representative images of 14 kidneys per group with C3 and IgG2a deposits (left) and C3 intensity (three to six glomeruli per mouse). (H) Renal pathology assessed by severity rank (median and interquartile range, two-tailed Mann-Whitney test) and GN score distribution (χ2 test). (I) Representative glomeruli (periodic acid–Schiff stain) from untreated mice (left) showing large subendothelial deposits (arrows) and from treated mice (right) showing open capillaries and reduced hypercellularity (stars) (scale bars, 25 μm). n = 14 TC and 5 B6 mice per group.

Met + 2DG decreased the percentage of total splenic CD4+ T cells in TC mice (Fig. 5A), as well as activation measured as the percentages of CD69+ expression (Fig. 5B) and Tem (Fig. 5, C and D). Globally, Met + 2DG decreased the expression of markers associated with activation and effector functions in TC CD4+ T cells and increased their expression of the inhibitory molecule CTLA-4 (cytotoxic T lymphocyte–associated protein 4) (fig. S6). For most markers, the expression level on CD4+ T cells from treated TC mice was similar to that of age-matched B6 mice. Both follicular helper (Tfh) and follicular regulatory (Tfr) T cell subsets were reduced in TC spleens after Met + 2DG treatment (Fig. 5, E to G), which reflects the overall decrease in FO T cells. The expression levels of PD-1, CXCR5, and BCL6 were drastically reduced by the treatment (fig. S6).Within the FO T cells, the treatment expanded Tfr at the expense of Tfh population (P = 0.04). The high frequency of germinal center (GC) B cells in TC mice (34) was also reduced after treatment (Fig. 5H). Finally, Met + 2DG restored the impaired production of IL-2 by TC CD4+ T cells (Fig. 5I), and reduced mTORC1 activity (Fig. 5, J to M). In conclusion, Met + 2DG treatment resulted in a profound and global reversal of T cell activation in TC mice, whereas it had minimal effects on B6 T cells.

Fig. 5. A 3-month Met + 2DG treatment normalized CD4+ T cell phenotypes in aged TC mice.

(A to C) Percentage of total splenic CD4+ T cells (A), CD69+ (B), and Tem (C) CD4+ T cells in B6 and TC-treated and control (Ctrl) mice. (D) Representative CD4+-gated FACS (fluorescence-activated cell sorting) plots showing the CD62L+CD44 Tn and CD62CD44+ Tem subsets. (E to G) Frequency of Tfh (E) and Tfr (F) CD4+ T cells and representative CD4+-gated FACS plots showing the PD-1hiCXCR5hiBCL6+Foxp3Tfh and PD-1hiCXCR5hiBCL6+Foxp3+ Tfr subsets (G). (H) Frequency of GC CD19+ B cells. (I) IL-2 production in CD4+ T cells stimulated with anti-CD3/CD28 for 24 hours. (J to M) Effect of treatment on phosphorylation of S6 (pS6) (J) and 4E-BP1 (p4E-BP) (K), as well as expression of CD98 (L) and CD71 (M) in total, Tn, and Tem T cells. n = 4 to 14. MFI, mean fluorescence intensity.

Met + 2DG–treated TC and B6 mice maintained their body weight and blood glucose (fig. S7). The treatment did not result in a general humoral suppression because total serum IgM and IgG were unchanged in TC mice (fig. S8A). The reason for the increased total IgG in treated B6 mice is not clear at this point. Furthermore, Met + 2DG treatment did not impair the humoral response in response to a foreign antigen (fig. S8B). Met + 2DG–treated B6 mice immunized with nitrophenyl –keyhole limpet hemocyanin (NP-KLH) produced the same amount of anti-NP IgM and IgG1 with no difference in affinity or kinetics. The number of B cells, GC B cells, and plasma cells (fig. S8C), Tem CD4+ T cells, Tfh cells, and Tfr cells, as well as their expression of effector surface markers were similar between treated and untreated mice (fig. S8, D and E). A 1-month treatment also decreased the metabolism of TC CD4+ T cells (fig. S9, A and B), and there was a trend for the reduction of spleen size (fig. S9C). The treatment reduced the production of anti-dsDNA IgG (fig. S9D) and ANA (fig. S9, E and F), the percentage of CD4+ T cells (fig. S9G), CD4+ T cell activation (fig. S9, H and I), as well as the percentage of GC B cells (fig. S9J). The short-term treatment also decreased glomerular IgG2a immune complexes (fig. S7, K and L) and improved renal pathology (fig. S9, M and N). Thus, 1 month of Met + 2DG treatment reversed disease biomarkers in TC mice.

Met + 2DG treatment reversed disease phenotypes in other mouse models of SLE

We tested how Met + 2DG treatment affects disease phenotypes in two additional models of SLE. A 1-month Met + 2DG treatment of 7-month-old NZB/W mice reduced their CD4+ T cell metabolism, although the difference was significant only for OCR (Fig. 6, A and B). It also decreased the production of anti-dsDNA IgG (Fig. 6C) and tended to reduce serum ANA levels, despite an increase in control mice (Fig. 6D). Contrary to TC mice, Met + 2DG significantly decreased total IgG (Fig. 6F), and there was a trend in the same direction for IgM (Fig. 6E). The total CD4+ T cell percentage was unaffected by treatment (Fig. 6G), but CD4+ T cell activation (Fig. 6H), the percentage of Tem (Fig. 6I), Tfh, and Tfr CD4+ T cells (Fig. 6, J and K) as well as GC B cells (Fig. 6L), and mTROC1 activity (Fig. 6, M to P) were all decreased in treated NZB/W mice. Renal pathology was, however, unchanged (Fig. 6, Q and R). Thus, a 1-month Met + 2DG treatment reversed the immunophenotypes of NZB/W mice.

Fig. 6. Met + 2DG treatment for 1 month reversed immunophenotypes in NZB/W mice.

(A and B) ECAR (A) and OCR (B) in splenic CD4+ T cells. (C) Change between the terminal and the initial serum anti-dsDNA IgG for each mouse. (D) Serum ANA intensity, with each linked symbol representing the initial and final values for each mouse. (E and F) Change between the terminal and initial total serum IgM (E) and IgG (F). (G to L) Frequency of total CD4+ T cells (G); CD69+ (H), Tem (I), Tfh (J), and Tfr (K) CD4+ T cells; and GC B cells (L). (M to P) Effect of treatment of mTORC1 targets: phosphorylation of S6 (M) and 4E-BP1 (N) and expression of CD98 (O) and CD71 (P) in total CD4+ T cells. (Q and R) Renal pathology assessed by severity rank (median and interquartile range) (Q) and distribution of GN mesangial (Mn 2–3 and 4) scores (R). n = 4 to 5.

The impact of Met + 2DG treatment was also assessed in the chronic graft-versus-host disease (cGVHD)–induced model of lupus (35). Treatment did not affect the induced splenomegaly (fig. S10A), but it reduced the production of anti-dsDNA IgG (fig. S10B), as well as the percentage of Tcm, but not Tem, CD4+ T cells (fig. S10, C and D). Therefore, Met + 2DG prevented the induction of T cell–dependent production of autoantibodies. Met monotherapy reduced splenomegaly (fig. S10A) and the size of the memory T cell subset (fig. S10, C and D) in this model, but it had no effect on the induction of anti-dsDNA IgG (fig. S10B). This suggested that the combination of the two metabolic inhibitors is required for the effective suppression of autoimmunity.

Met and 2DG showed a synergistic effect in vivo

We evaluated the relative contributions of 2DG and Met to disease reversal in 7-month-old TC mice treated for 1 month at the same dose used for the combination treatment. 2DG decreased ECAR but had no effect on OCR in TC CD4+ T cells (fig. S11, A and B). We did not observe any effect on spleen weight (fig. S11C), autoantibody production (fig. S11, D to F), CD4+ T frequency (fig. S11G), CD4+ T and B cell activation, or effector subset distribution in these treated mice (fig. S11, H to L). Although 2DG lowered immune complex deposition in the kidneys (fig. S11, M and N), renal pathology was not ameliorated (fig. S11, O and P). Met monotherapy did not alter either ECAR or OCR in TC CD4+ T cells (fig. S12, A and B) or any of the biomarkers that were affected by the combination treatment (fig. S12, C to N). Therefore, 1-month treatment of neither 2DG nor Met alone had an effect on the established autoimmune pathology of TC mice. Although it is possible that monotherapy for a longer time could have stronger effect on the disease, it is clear that Met + 2DG treatment is superior to Met or 2DG alone, which strongly suggests a synergistic effect between 2DG and Met in reversing lupus phenotypes.

CD4+ T cells from SLE patients exhibited elevated cellular metabolism

SLE patients present with a reduced percentage of CD4+CD45RA+CCR7+ Tn cells and a corresponding increase in CD4+CD45RACCR7+ Tcm cells, leading to an altered ratio of Tn/Tcm subsets as compared to HCs (fig. S13, A to D and F). No significant difference was found in the percentage of circulating CD4+CD127−/loCD25+ Treg cells (fig. S11, E and G). We compared ECAR, OCR, and SRC in CD4+ T cells purified from peripheral blood of SLE patients and HCs, with and without anti-CD3 and anti-CD28 activation (Fig. 7, A and B). Given that the SLE patients were treated with at least one immunosuppressive drug (table S1), we tested whether three of the most commonly used drugs could affect CD4+ T cell metabolism. HC CD4+ T cells were treated for 24 hours with either dexamethasone (Dex), hydroxychloroquine (HCQ), or mycophenolate mofetil (MMF) at a concentration that did not affect cell viability. Dex inhibited ECAR but not OCR, whereas HCQ and especially MMF reduced both ECAR and OCR (fig. S13, H and I). Despite the decreased metabolism induced by these treatments, activated CD4+ T cells showed a higher ECAR in SLE than in HC samples (Fig. 7C), and both resting and activated SLE CD4+ T cells showed a higher OCR than did HC T cells (Fig. 7D). Notably, anti-CD3/CD28 activation induced a rapid increase in both ECAR and OCR (Fig. 7, A and B), suggesting that both pathways are important for the early activation of human CD4+ T cells. Moreover, both resting and activated SLE CD4+ T cells have a significantly higher SRC (Fig. 7, B and E). These metabolic parameters were correlated with the distribution of circulating CD4+ T cell subsets. The percentage of Tn was inversely correlated with both activated ECAR (Fig. 7F) and basal OCR (Fig. 7G) levels, whereas that of Treg was positively correlated with activated ECAR (Fig. 7H). These results strongly suggest that, as shown in the mouse, cellular metabolism regulates human CD4+ T cell activation and effector functions. Further, our results show a strong association between a dysregulated homeostasis in SLE CD4+ T cells and an elevated cellular metabolism. Consistent with our findings in mice (Fig. 3), SLE CD4+ T cells produced more IFN-γ than did HCs after TH1 induction (Fig. 7I). Met decreased IFN-γ production from SLE and HC CD4+ T cells to comparable levels between the two cohorts (Fig. 7I). Overall, these results showed that Met treatment normalized the in vitro IFN-γ production by SLE CD4+ T cells.

Fig. 7. CD4+ T cells from SLE patients have an enhanced metabolism.

(A and B) Representative ECAR (A) and OCR (B) graphs of human CD4+ T cells during a mitochondrial stress test. Anti-CD3/CD28 or isotope controls, oligomycin, FCCP, and antimycin A/rotenone (Rote + Anti) were added to the cells as indicated. (C to E) ECAR (C), OCR (D), and SRC (E) in HC and SLE CD4+ T cells, with and without anti-CD3/CD28 activation. n = 19 HCs and 20 SLE patients. (F to H) Correlations between Tn percentages and activated ECAR (F) or basal OCR (G) and between Treg percentages and activated ECAR (H). The significance and correlation coefficient of Pearson tests are shown. (I) IFN-γ production in TH1-polarized CD4+ T cells with or without Met. Two-tailed paired t tests compared the effect of treatment within each cohort. n = 6.


CD4+ T cells up-regulate glycolysis after activation, a phenomenon very similar to the Warburg effect described in cancer cells (36). Glycolysis is critical for T cell effector functions, and increased glycolysis leads to autoimmunity (13, 17). More recently, the field has gained a better appreciation for mitochondrial oxidative metabolism in mediating T cell activation and proliferation (12, 14, 23). Here, we showed that CD4+ T cells from the TC lupus model as well as from SLE patients exhibit elevated glycolysis and mitochondrial oxidative metabolism as compared to non-autoimmune controls, both ex vivo and after in vitro activation. We also showed that the enhanced oxidative metabolism is attributed to a combination of higher glucose, fatty acid, and glutamine oxidation. Moreover, the enhanced metabolism in CD4+ T cells is associated with an increased mTORC1 activity. These findings led us to hypothesize that lupus could be treated with metabolic modulators that target both glycolysis and mitochondrial metabolism. We found that treating TC mice with a combination of Met and 2DG normalized T cell metabolism and reversed disease phenotypes, including T cell activation, autoantibody production, and renal disease. The combination treatment was also effective on the NZB/W spontaneous lupus model and the cGVHD-induced model. This result suggests that dysregulated T cell metabolism may be underlying SLE pathogenesis and that T cell metabolism can be a therapeutic target.

Metabolic modulators have been evaluated for a number of diseases. Rapamycin reduced disease activity in MRL/lpr mice (37) and in SLE patients (22). Treatment with N-acetylcysteine, a precursor of glutathione that blocks mTOR, improved disease outcomes in NZB/W mice (38) and SLE patients (39). A clinically approved inhibitor of glycosphingolipid biosynthesis, N-butyldeoxynojirimycin, normalized the functions of CD4+ T cells from SLE patients (40). Met is commonly used to treat type 2 diabetes, and 2DG has been tested to treat cancer (41) and Alzheimer’s disease (42). Either 2DG or Met alone inhibited the development of experimental autoimmune encephalomyelitis, a mouse model of multiple sclerosis (13, 43). 3PO, a small-molecule inhibitor of PFKFB3, an enzyme that controls a rate-limiting step of glycolysis, prevented the development of T cell–mediated delayed hypersensitivity and imiquimod-induced psoriasis in the mouse (44). To the contrary, we showed in this study that a dual inhibition of glycolysis and mitochondrial metabolism synergizes for the reversal of lupus in the TC model, which highlighted the complexity of this disease. A synergistic effect of Met and 2DG has also been shown in cancer models (45). Because recent studies have revealed the importance of glycolysis and mitochondrial metabolism in controlling T cell effector functions, dual inhibition of glycolysis and mitochondrial metabolism may have broader applications to other immune-mediated diseases besides lupus.

Our study provides possible mechanisms by which Met and 2DG reversed immunophenotypes in vivo. Met regulates cellular metabolism through complex mechanisms (46), including the inhibition of the mitochondrial respiratory chain complex I (33). Here, the Met + 2DG treatment decreased T cell mitochondrial oxygen consumption as well as mTORC1 signaling in both the TC and NZB/W models. Therefore, inhibiting mitochondrial metabolism and antagonizing mTOR signaling are the likely mechanisms by which Met normalized TC CD4+ T cell metabolism in vivo. We also observed a decrease in glycolysis after the Met + 2DG treatment, which is likely the mechanism by which 2DG normalized TC CD4+ T cell metabolism in vivo.

In summary, we have identified defective CD4+ T cell metabolism as a therapeutic target in both murine and human SLE. Besides CD4+ T cells, Met + 2DG may also target directly other immune cell types in vivo. Glucose metabolism is essential for B cell functions (47). Both glycolysis and mitochondrial metabolism are important for the activation and maturation of dendritic cells (48, 49), which can indirectly affect T cells. It will be important to find out whether lupus mice have abnormal metabolism in other immune cell compartments and whether the metabolism of those cells is altered after Met + 2DG treatment in vivo. Although Met + 2DG was effective at reversing disease phenotypes, whereas the monotherapies at the same doses were not, it is possible that different doses or a longer duration would be effective for either Met or 2DG alone. We have also not yet investigated whether the treatment needs to be maintained to prevent flares or whether it resets the immune system for an extended period of time. Another limitation of the present study is that 2DG blocks the entire glucose metabolic pathway and does not discriminate between the contribution of glycolysis and glucose oxidation. Finally, a comprehensive evaluation of the effect of Met, with and without glucose inhibitors, on the functions of human CD4+ T cells should be performed.


Study design

The objective of the study was to define the metabolism of mouse and human SLE CD4+ T cells and to explore potential metabolism-based therapeutic strategies. The metabolism of mouse CD4+ T cells was first evaluated ex vivo and after in vitro stimulation in the presence of metabolic inhibitors. Six- to 7-month-old anti-dsDNA IgG-positive lupus mice and age-matched B6 controls were randomized to treatment and control groups. Treatment was performed for 1 or 3 months with Met (3 mg/ml) or 2DG (5 mg/ml), or a combination of the two, dissolved in drinking water. Age-matched control mice received plain drinking water. Peripheral blood was collected biweekly to analyze autoantibody production; body weight and blood sugar levels were monitored weekly and biweekly, respectively. At the end of the treatment, spleens were collected for flow cytometry and metabolic analysis of CD4+ T cells, and kidneys were evaluated for renal pathology in a blinded fashion. All experiments were conducted according to protocols approved by the University of Florida (UF) Institutional Animal Care and Use Committee. For human subjects, at least 20 ml of peripheral blood was obtained after signed informed consent in accordance with UF Institutional Review Board. Female SLE patients fulfilled at least four of the revised SLE criteria of the American College of Rheumatology. Healthy female volunteers with no family history of autoimmune disease served as age- and ethnicity-matched HCs. All patients were treated with at least one medication, none of them with a biologic treatment. The demographics and treatment regimens of the patients and HCs are summarized in table S1.


TC mice have been described previously (18). C57BL/6J (B6), B6(C)-H2-Ab1bm12/KhEgJ, and (NZB × NZW)F1 mice were purchased from The Jackson Laboratory. Only female mice were used in this study. cGVHD was induced as previously described (29).

Metabolic measurements

Splenocyte suspensions were enriched for CD4+ T cells by negative selection with magnetic beads (Miltenyi). Naïve (Tn, CD4+CD44CD62L+) and effector memory (Tem, CD4+CD44+CD62L) subsets were sorted with a FACSAria cell sorter (BD Biosciences). Sorted mouse Tn were activated with plate-bound anti-CD3e (145-2C11, 2 μg/ml) and soluble anti-CD28 (37.51, 1 μg/ml) for 24 hours. ECAR and OCR were measured using either an XFe24 or an XFe96 Extracellular Flux Analyzers under mitochondrial stress test conditions (Seahorse). Assay buffer was made of nonbuffered RPMI medium (Sigma) supplemented with 2.5 μM dextrose, 2 mM glutamine, and 1 μM sodium pyruvate. Baseline ECAR and OCR values were averaged between technical replicates for these first three successive time intervals. Extracellular lactate production was measured using L-Lactate Assay Kit (Abcam). Intracellular ATP was measured using the ATP Determination Kit (Life Technologies). Fatty acid uptake by CD4+ T cells was measured by flow cytometry with the Bodipy dye (Life Technologies).

Gene expression analysis

SYBR Green (Applied Biosystems)–based qPCR (quantitative polymerase chain reaction) was performed on complementary DNA from bead-enriched CD4+ T cells with primer sequences shown in table S2. mRNA levels were expressed as relative quantities to Ppia (cyclophilin A).

Flow cytometry

The immunophenotypes of human CD4+ T cells were determined with antibody panels emulated by the Human Immunophenotyping Consortium (table S4). The other antibodies used in this study are listed in table S3. Cytokine production was analyzed in cells treated with the leukocyte activation cocktail (BD Biosciences) and the Fixation/Permeabilization kit (eBioscience).

Cell culture

Mouse splenic CD4+ T cells were stimulated by anti-CD3e and anti-CD28 for 3 days and IL-12 (10 ng/ml) for TH1 differentiation. IL-2 was measured in CD4+ T cells stimulated by anti-CD3e and anti-CD28. Human CD4+ T cells were enriched from peripheral blood using the RosetteSep Enrichment Cocktail (StemCell Technologies). For TH1 polarization, cells were stimulated in complete RPMI with beads coated with anti-CD3 and anti-CD28 (Dynabeads Human T-Activator, Life Technologies) for 6 days with IL-12 (10 ng/ml), IL-2 (20 U), and anti–IL-4 (1 μg/ml). In some assays, Dex, HCQ, or MMF (1 μM, Sigma) was added, and ECAR and OCR were measured 24 hours later.

Antibody measurement

Serum anti-dsDNA, anti-chromatin IgG, and ANA were measured as previously described (25). Autoantibody microarray analysis was performed at the Microarray Core Facility in University of Texas Southwestern Medical Center.

Renal pathology

GN was scored, and the glomerular deposition of C3 and IgG2a immune complexes was evaluated on frozen kidney sections as previously described (50).

Statistical analysis

Statistical analyses were performed using the GraphPad Prism 6.0 software. Unless indicated, data were normally distributed, graphs show means and SEM for each group, and results were compared with two-tailed t tests with a level of significance of P < 0.05. Bonferroni corrections were applied for multiple comparisons. Each in vitro experiment was performed at least twice with reproducible results.


Fig. S1. TC mice showed enhanced CD4+ T cell activation.

Fig. S2. Representative FACS plots of ex vivo CD4+-gated T cells from B6 (black) and TC (red) mice for pS6, p4E-PB1, CD98, and CD71.

Fig. S3. CD4+ T cells from TC mice showed an altered expression of metabolic genes.

Fig. S4. IL-2 gating in CD4+ T cells.

Fig. S5. The Met + 2DG treatment normalized the metabolism of effector CD4+ T cells.

Fig. S6. A 3-month Met + 2DG in vivo treatment resulted in a global normalization of TC CD4+ T cell effector phenotypes.

Fig. S7. The Met + 2DG treatment did not have adverse effects on body weight and blood sugar.

Fig. S8. Met + 2DG treatment did not affect normal humoral response.

Fig. S9. One-month Met + 2DG treatment reduced disease severity in TC mice.

Fig. S10. Met + 2DG treatment prevented autoantibody production in the cGVHD model.

Fig. S11. 2DG treatment failed to reverse immunophenotypes in TC mice.

Fig. S12. Met treatment failed to reverse immunophenotypes in TC mice.

Fig. S13. CD4+ T cell immunophenotypes in SLE patients.

Table S1. Human subject demographics and disease parameters.

Table S2. Primers for qPCR assays.

Table S3. Antibodies used in this study.

Table S4. Source data (Excel file).


  1. Acknowledgments: We thank C. Mathews for helpful suggestions and N. Kanda and L. Zeumer for expert technical help. Funding: Supported by NIH grants R01 AI045050 and ALR-TIL 0000075018 (to L.M.), P01 AI42288 (to T.M.B.), and a JDRF grant (CDA 2-2012-280) (to T.M.B.). Author contributions: Y.Y., T.M.B., and L.M. conceived and designed the experiments; Y.Y., S.-C.C., Z.X., D.J.P., and H.S. performed the experiments; Y.Y., S.-C.C., D.J.P., B.P.C., T.M.B., and L.M. analyzed data; E.S.S. contributed materials/samples; and Y.Y. and L.M. wrote the paper. Competing interests: Y.Y. and L.M. have filed a patent application entitled “Treatment of lupus using metabolic modulators” (U.S. Provisional Application No.: 62/107,160). The other authors declare no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article