Clinical

SUMMARY Augmented T cell function leading to host damage in autoimmunity is supported by metabolic dysregulation, making targeting immunometabolism an attractive therapeutic avenue. Canagliﬂozin, a type 2 diabetes drug, is a sodium glucose co-transporter 2 (SGLT2) inhibitor with known off-target effects on glutamate dehydrogenase and complex I. However, the effects of SGLT2 inhibitors on human T cell function have not been extensively explored. Here, we show that canagliﬂozin-treated T cells are compromised in their ability to activate, proliferate, and initiate effector functions. Canagliﬂozin inhibits T cell receptor signaling, impacting on ERK and mTORC1 activity, concomitantly associated with reduced c-Myc. Compromised c-Myc levels were encapsulated by a failure to engage translational machinery resulting in impaired metabolic protein and solute carrier production among others. Importantly, canagliﬂozin-treated T cells derived from patients with autoimmune disorders impaired their effector function. Taken together, our work highlights a potential therapeutic avenue for repurposing canagliﬂozin as an intervention for T cell-mediated autoimmunity.


INTRODUCTION
Upon activation, human T cells adopt an anabolic metabolism to fuel the biosynthetic and energetic requirements of effector function. 1,2 This metabolic rewiring supports blastogenesis and the production of various immune-based mediators, such as cytokines, to drive the adaptive immune response. 3 One of the hallmarks of autoimmune disorders is aberrant T cell activation, which is supported by elevated metabolic demands. [4][5][6] For example, in rheumatoid arthritis (RA), patient-derived pathogenic T cells display dysregulated cellular metabolism under-pinned by an enhanced requirement for energy and biosynthetic intermediates. 4,7,8 Furthermore, T cell metabolic rewiring, such as elevated glucose metabolism, lipid synthesis and glutaminolysis, and mitochondrial dysfunction, is a major driver of systemic lupus erythematosus (SLE) pathogenesis. 5,9,10 Taken together, these studies highlight that targeting the general immunometabolic profile of T cells could be therapeutically beneficial.
Appraising and repurposing type 2 diabetes (T2D) medications that target cellular metabolism as treatments for autoimmunity has significant potential. [11][12][13][14][15] For example, the biguanide and complex I inhibitor, metformin has been repurposed in RA, 16,17 SLE, 11,18 multiple sclerosis, 19,20 and myasthenia gravis. 21 In a recent clinical trial, metformin reduced the occurrence of severe flare-ups in patients with SLE. 11 Furthermore, the inhibitor rosiglitazone from the thiazolidinedione class suppressed pathogenic Th17-driven atopic dermatitis in obese mice. 22 These studies highlight the potential for targeting altered metabolic signatures characteristic of pathogenic T cells, using T2D drugs for therapeutic benefit in autoimmunity.
Sodium glucose co-transporter 2 (SGLT2) inhibitors are a novel class of T2D drugs approved for the treatment of T2D since 2012. 23 SGLT2 inhibitors, such as canagliflozin (cana) and dapagliflozin (dapa), target the SGLT2 transporter and prevent glucose reabsorption in the kidneys. Here, increased glucose excretion in the urine and a reduction in blood glucose improves glycemic control. 24,25 Interestingly, cana, as opposed to dapa, has been shown to have off-target effects, inhibiting mitochondrial glutamate dehydrogenase (GDH) and complex I. 26,27 Previous work has highlighted the beneficial effects of cana in renal and cardiovascular function in patients with chronic kidney disease and heart failure, respectively. 28,29 However, the effects of the gliflozins, specifically cana, on human T cell function is limited. As levels of SGLT2 are barely detectable in human T cells, 30 the aforementioned off-target effects of cana on mitochondria pose an interesting avenue to investigate in T cells.
Here, we found that cana but not dapa is a potent inhibitor of human T cell effector function. We reveal that in addition to the known off-target effects on GDH and complex I, cana inhibited T cell receptor (TCR) signaling resulting in compromised ERK and mTORC1 activity. These events were associated with a concomitant reduction in c-Myc levels, curbing metabolic reprogramming in activated T cells. This blunted metabolic response was associated with mitochondrial dysfunction, ultimately leading to impaired T cell effector function in multiple human T cell subsets. Importantly, cana was able to suppress T cell effector function in two autoimmune patient cohorts-SLE and RA-highlighting the potential of repurposing cana for improved treatment of T cell-mediated autoimmunity.

Canagliflozin modulates activated T cell function
We sought to better understand the effects of SGLT2 inhibitors on human CD4+ T cells-the key drivers in autoimmunity. 31 Here, we examined the effects of two different SGLT2 inhibitors; cana, which displays off-target effects on GDH and complex I, and dapa, which has no known off-target effects. 32 We purified human CD4+ naive T cells (Tnv) and activated them with anti-CD3 and anti-CD28 for 24 h in the presence or absence of physiologically relevant doses of either cana or dapa and assessed downstream effector function. 33,34 Cana significantly reduced IL-2 production in a dose-dependent manner, whereas dapa had a modest but non-significant effect ( Figure 1A). This observation was not due to compromised cell viability ( Figure 1B). Cana further constrained T cell activation by reducing expression of activation markers CD25, CD44, and CD69 ( Figure 1C).
To confirm our findings in a more physiological setting, we utilized the recently developed human-plasma-like media (HPLM). The formulation of HPLM more accurately reflects the nutrient levels of human plasma and has profound effects on T cell func-tion in comparison with traditional, non-physiological media, such as RPMI. 35 Once again, using HPLM, cana significantly impaired IL-2 production ( Figure 1D) and activation marker expression ( Figure 1E). Importantly, cana did not alter the phenotype of unstimulated Tnv based on measurement of activation markers ( Figure S1A). These results demonstrate that cana treatment impairs T cell activation. Closer inspection of publicly available data (ImmGen [human] and ImmPRes [mouse]) indicate that SLC5A2 or ''SGLT2'' is minimally expressed in human T cells (normalized expression value of SLC5A2 = 8, whereas SLC2A1 (GLUT1) = 145, SLC2A3 (GLUT3) = 277) 30 and undetectable in mouse T cells (www.immpres.co.uk). Taken together, this suggests that it is the ''off-target'' effects of cana rather than the known SGLT2 inhibitory effects that mediate inhibition of T cell activation programs.

Canagliflozin impairs T cell blastogenesis and proliferation
To better understand the impact of cana treatment on T cell function and phenotype, we next measured gene transcription in HPLM cultured, activated T cells. Using a NanoString human autoimmune panel, we revealed 3 downregulated and 39 upregulated genes upon cana treatment (Figure 2A). Our previous findings were confirmed at the gene transcript level with downregulation of IL2 ( Figure 2B). In contrast to our flow cytometry data, levels of the CD25 transcript, IL2RA, were significantly elevated upon cana treatment ( Figure 2B) indicating potential issues at the translational level. We also observed significantly increased expression of SELL (CD62L; Figures 2B and S1B) in cana-treated cells compared with controls, which was also confirmed at the protein level ( Figure S1C). These data also revealed that cana was associated with the downregulation of CSF2 and CCL20 genes-commonly associated with a Th17 signature 36 ( Figure 2B). No significant transcript alterations were observed of T cells treated with dapa ( Figure S1D).
Given the striking effects of cana on IL-2 production, we next assessed blastogenesis. Here, cana significantly reduced T cell blastogenesis ( Figure 2C). This observation was in line with repressed downstream signaling targets involved in biogenesis, mainly mammalian target of rapamycin complex 1 (mTORC1). Two markers of mTORC1 activity, S6 ribosomal protein (S6) and 4E-BP1 phosphorylation, were both decreased at 4 and 24 h post-activation without changes in AMPK activity (a negative regulator of mTORC1 37 ) and its downstream target acetyl-CoA carboxylase ( Figures 2D and S1E).
To determine whether the effects of cana on T cells were long lived, we followed the activation of T cells out to 72 h of culture. Again, cana significantly reduced T cell proliferation ( Figure 2E), IL-2 production, and STAT5 phosphorylation (Figures S1F and S1G) revealing that cana inhibits both short-and long-term T cell activation in physiological HPLM. Given the significant defects in IL-2 production in both HPLM and RPMI, we sought to identify whether the addition of exogenous IL-2 was able to rescue the observed phenotype. Supplementation with IL-2 did not rescue the cana-dependent effect on blastogenesis, indicating a more global effect on T cell suppression ( Figure 2F).
Next, we assessed whether cana could suppress previously activated T cells and employed two strategies to determine this. First, we activated T cells for 48 h and subsequently introduced cana for the final 24 h of culture. Here, cana retained an ability to reduce blastogenesis in activated T cells (Figures S1H and S1I). Second, we activated purified total or ''pan'' CD4+ T cells for 24 h, treating them for a further 72 h (supplemented with IL-2) with increasing doses of cana. Here, cana was able to suppress total CD4+ T cell counts and a reducing trend in IFNg production, but at higher concentrations than those when starting from a naive T cell population (Figures S1J and S1K). Taken together, our data demonstrate that cana suppresses T cell activation in a physiological environment.

Canagliflozin suppresses c-Myc signaling
To better understand the underlying mechanisms responsible for the impact of cana treatment on activated T cells, we assessed immunometabolic changes at the gene transcript level. Using a NanoString human metabolism panel we revealed that cana treatment resulted in 24 downregulated and 14 upregulated genes ( Figure 3A). Pathway analysis of downregulated genes identified functions linked to glycolysis, c-Myc, and cell cycle, which supported our earlier findings ( Figures 3B and 3C). At the transcript level, c-Myc expression was unaffected (Figure 3D), whereas it was reduced at the protein level upon cana treatment at both an early (4 h) and late time point (24 h), indicating a translational rather than transcriptional perturbation ( Figure 3E).
As c-Myc is a master transcription factor, integral to metabolic reprogramming of T cells upon activation, 38,39 we explored our findings further by carrying out label-free liquid chromatography-mass-spectrometry-based whole-cell proteomics to investigate the impact of cana on the global activated T cell proteome. In line with cana reducing cell size ( Figure 2C), cana-treated cells had a significantly lower protein mass in comparison with those treated with the vehicle control ( Figure 3F). A total of 5,655 proteins were detected and differential expression analysis revealed that at the copy-number level, 4,421 proteins were significantly reduced whereas one protein was upregulated-the cell-cycle inhibitor, CDKN1B ( Figures 3G and S2A). Importantly, we were not able to detect SLC5A2 or SGLT2 in our proteomic dataset. We next took into consideration the concentration of the detected proteins, irrespective of cell size. Here, 481 proteins were downregulated in response to cana treatment, while 203 proteins were upregulated (log 2 foldchange of ±0.585, p value % 0.05; Figures 3H and S2A). Of these downregulated proteins, we confirmed that c-Myc was significantly reduced at both the copy-number and concentration level ( Figure 3I).
To better understand the biological relevance of these protein changes, ingenuity pathway analysis (IPA) was employed to consider the contribution of individual proteins toward various canonical cellular pathways and their enrichment following cana treatment. Here, ''cell-cycle control of chromosomal replication'' emerged as the most significantly downregulated pathway, supporting the observed effects on T cell proliferation ( Figure 3J). Given the known effects of cana on mitochondrial proteins (GDH and complex I), it was not surprising that ''oxidative phosphorylation (OXPHOS)'' and ''mitochondrial dysfunction'' were the predominant upregulated pathways in cana-treated T cells ( Figure 3K). Cana treatment resulted in the compensatory enrichment of several ETC complex-associated proteins, with concomitant increases in mitochondrial mass and membrane potential as determined by flow cytometry (Figures S2B-S2D). Concomitant with mitochondrial dysfunc-tion, other upregulated pathways included proteins associated with ''sirtuin signaling'' and ''glutathione redox reactions'' (Figure 3K), which was consistent with elevated early mitochondrial ROS production ( Figure S2E) and heightened levels of proteins associated with the response to oxidative stress (Figures S2F-S2J) in cana-treated T cells.
Subsequent analysis allowed us to identify potential upstream regulators (e.g., transcription factors, microRNAs, kinases, small molecule inhibitors) associated with changes in the activated T cell proteome following cana treatment. Here, the activation Z score predicts the activation state of each predicted upstream regulator. In line with our NanoString and immunoblot data, reduced expression of 52 distinct targets predicted that c-Myc might be an upstream regulator which is repressed following cana treatment ( Figure 3L). Interestingly, the drug torin1 and protein RICTOR-important upstream regulators of the mTOR axis-were identified as potential upstream regulators that were suppressed based on proteins upregulated in cana-treated T cells ( Figure 3M).
Given the agreement between transcriptional and proteomic analysis on c-Myc-associated pathways, we further interrogated the proteomics dataset and observed several c-Myc-associated metabolic targets that were significantly inhibited with cana treatment including SLC2A1 (GLUT1), hexokinase 2 (HK2),

Figure 2. Canagliflozin inhibits T cell proliferation
T cells were activated with anti-CD3 (2 mg/mL) and anti-CD28 (20 mg/mL) and treated with or without canagliflozin (10 mM) for 24 h. (A) NanoString differential expression analysis of autoimmunity-associated genes in canagliflozin-treated T cells versus vehicle control. Blue and red data points represent downregulated and upregulated genes, respectively. Genes with an adjusted p value < 0.05 and log 2 fold-change <À1 and >1 were considered differentially expressed.   Figure 3N) and solute transporters SLC7A5, SLC38A1, SLC38A2, and SLC1A5 ( Figures 3O and  S2K). In fitting with a reduction in c-Myc activity, we observed significant decreases in translational machinery in concert with no difference in the concentration of the translational inhibitor, PDCD4 ( Figures S2L-S2R). Collectively, these data demonstrate that cana treatment blunts c-Myc signaling and results in concomitant mitochondrial dysfunction, leading to impaired T cell metabolism and function.

Canagliflozin impairs TCA cycle metabolism via GDH inhibition
Given that cana blunts c-Myc induction, we next sought to directly measure cellular metabolism in response to cana treatment. We activated naive CD4+ T cells for 24 h in the presence or absence of cana and monitored metabolic alterations using a mitochondrial stress assay. Interestingly, cana perturbed the extracellular acidification rate (ECAR; Figures 4A, S3A, and S3B) and oxygen consumption rates (OCRs; Figures 4B and S3C-S3F). Specifically, cana reduced adenosine triphosphate (ATP) production from glycolysis ( Figure 4C) and affected the maximal respiration rates leading to a significant drop in ATP production from OXPHOS ( Figure 4D). Moreover, cana significantly reduced the bioenergetic scope of activated T cells by impacting on their ability to generate ATP at the maximal rate ( Figures S3G and S3H). No metabolic perturbation was observed in activated T cells treated with dapa ( Figures S3G and S3H). Furthermore, activated T cells and effector T cells (Teff) cells in the presence of cana released significantly lower levels of lactate in comparison with the vehicle or dapa-treated cells ( Figures S3I  and S3J). Taken together, these results indicate that it is the offtarget effects of cana that perturb T cell metabolism rather than the ''on-target'' effects of SGLT2 inhibition.
Given the clear metabolic effects of cana on activated T cells, we next employed liquid chromatography-mass spectrometry (LC-MS/MS) to track cellular metabolites. The total abundance of individual tricarboxylic acid (TCA) cycle intermediates was reduced in the presence of cana ( Figure 4E), while levels of the amino acids, glutamate, and aspartate were unchanged (Figure 4F). To determine whether cana inhibits GDH in activated T cells, we performed stable isotope labeling using a uniformly labeled 13 C 5 -glutamine (Gln) probe to measure the incorporation of Gln-derived carbon through GDH and into the TCA cycle (Figure 4G). Total intracellular levels of 13 C-Gln and 13 C-aKG were reduced in cana-treated T cells, while there was no impact on 13 C-glutamate abundance, indicating impaired Gln uptake and inhibition at GDH ( Figure 4H). When we analyzed the total proportion of 13 C incorporation into metabolite pools irrespective of mass isotopolog distribution, levels of 13 C into aKG and citrate were significantly reduced with cana ( Figure 4I). Furthermore, levels of incorporation of 13 C into the m + 5 mass isotopolog of aKG were significantly reduced with cana treatment ( Figure 4J). This trend was apparent in other TCA cycle metabolites, including amino acids, glutamate, aspartate, and proline ( Figures S3K-S3M). In an attempt to rescue the effect of canamediated GDH inhibition, we cultured cells with membranepermeable aKG; however, this did not restore IL-2 production in cana-treated T cells ( Figure 4K).
As metabolic perturbation is associated with plasticity in human T cells, 1 we next wanted to determine the impact of the impaired glutamine utilization on 13 C 6 -glucose metabolism (Figure S3N). Cana treatment impaired lactate production from glucose, as illustrated by a reduction in the lactate m + 3 isotopolog ( Figure S3O), consistent with the reduced ECAR we observed ( Figure 4A). Analysis of the total 13 C incorporation into metabolite pools revealed significant increases in malate, glutamate and aspartate ( Figure S3P). Consistent with this, cana promoted the entry of glucose-derived carbon into the TCA cycle ( Figure S3Q) and amino acid (glutamate and aspartate) metabolite pools ( Figure S3R). These data suggest that T cells metabolically adapt to cana-induced GDH inhibition by becoming more reliant on glucose-derived TCA cycle intermediates.
To determine whether the effects observed can be explained by inhibition of complex I, another known off-target effect of cana, we activated T cells in the presence or absence of cana, alongside other known complex I inhibitors-rotenone and high-dose metformin (to bypass any transporter-specific uptake). Cana was the superior T cell inhibitor when assessing IFNg release, CD69 expression, and blastogenesis (Figures S3S-S3U) in comparison with rotenone or metformin. Given this, we next considered whether the combined effect of complex I and GDH inhibition might underpin the impaired effector function of T cells. Therefore, we activated T cells in the presence or absence of cana, piericidin A (an alternative complex I inhibitor), R162 (GDH inhibitor) or in combination. Cana was again, the most effective T cell inhibitor when assessing IFNg release (

Canagliflozin impairs T cell receptor signaling
We have demonstrated that cana significantly impacts on key intracellular metabolic nodes-mTOR and c-Myc. As both signaling molecules are sensitive to downstream signals from the TCR, 40 we sought to investigate the role of cana on T cell signaling. Here, we cultured T cells in the presence or absence of cana and assessed the early signaling events downstream of the TCR in ZAP70, LAT, and PLCg. Cana impaired phosphorylation of all TCR targets assessed ( Figure 5A) and subsequently blunted ERK phosphorylation in line with a MAPK inhibitor protein level. The Z score predicts the likely activation states of upstream regulators; a positive Z score suggests activation, while a negative Z score suggest inhibition.   Figure 5B). In addition, cana-treated cells showed delayed CD69 expression (downstream ERK target; Figure 5C). Given the impact of cana on TCR signaling, we sought to determine whether cana would affect human CD8+ T cells in the same manner. Here, cana significantly suppressed blastogenesis, activation marker expression and IL-2, IFNg and granzyme B production in CD8+ T cells (Figures S4A-S4D).
Next, we tested whether the impaired TCR signaling caused by cana was responsible for the observed defects in cytokine production. In an attempt to rescue cytokine production, we stimulated T cells with the diacylglycerol (DAG) mimetic-PMA and ionomycin-in order to bypass the TCR, and thus any downstream effects of cana on the TCR. T cells treated with PMA/ionomycin in the presence of cana did not have any defect in cytokine production in comparison with the vehicle control ( Figure 5D).
Given the dysregulation of critical signaling nodes (ERK and mTORC1) by cana, we hypothesized that cana treatment would phenocopy inhibition of these nodes. Here, we treated T cells with cana, rapamycin (mTORC1 inhibitor) and a MAPK inhibitor (ERK inhibition) and measured early (4 h) c-Myc expression and early (4 h) and late (24 h) puromycin incorporation (a measure of translation). Here, cana phenocopied both mTORC1 and MAPK inhibition by reducing c-Myc expression ( Figure 5E) and puromycin incorporation at both 4 and 24 h (Figures 5F and  5G). Functionally, cana phenocopied both mTORC1 and ERK inhibition by significantly reducing blastogenesis ( Figure 5H) and IFNg production ( Figure 5I). Taken together, these data suggest that, mechanistically, cana impairs TCR signaling leading to suppressed signaling to metabolic nodes. Additionally, the data indicate that c-Myc protein expression is downstream of ERK and mTORC1 signaling in TCR activated human T cells.
Canagliflozin inhibits the diverse cytokine profile of T effector cells Thus far, we have shown that cana inhibits the function of activated T cells that display a restricted cytokine profile. To determine if cana suppressed the diverse cytokine response seen in antigen-experienced Teff, Teff were isolated from human blood and activated in HPLM in the presence or absence of cana. Cana potently inhibited the secretion of key effector cytokines (IFNg, IL-2, IL-4, IL-10, IL-17 and IL-21) at 24 h ( Figure 6A), followed by a sustained suppression of most of these cytokines at 72 h ( Figure 6B). Consistent with a critical role for cana in shaping activated Teff responses, we also saw a decrease in the master transcriptional regulators Tbet, GATA3, and RORgt ( Figure 6C) concurrent with a striking reduction in blastogenesis that did not compromise viability (Figures 6D and S5A). Similar to our previous observations, activated Teff cultured with cana had reduced expression of activation markers CD25, CD44, and CD69 ( Figure 6E).
One important effector CD4+ T cell population are Tregs. To determine the impact of cana on Treg biology we employed two approaches. Firstly, we sorted naive T cells and cultured them under Treg polarizing conditions for 6 days in the continuous presence of cana and determined FoxP3 expression. Here, treatment with cana did not significantly impact on FoxP3 expression ( Figure S5B). Secondly, we polarized naive T cells toward the Treg compartment for 6 days and then restimulated differentiated Tregs for up to 72 h in the presence or absence of cana. Cana treatment did not significantly impact on FoxP3 expression at either time point ( Figure S5C). We then assessed supernatant IL-10 levels of polarized Tregs cultured with cana for 24 h. In line with cana being a global T cell inhibitor, the presence of cana significantly impaired IL-10 production from Tregs ( Figure S5D).
As cana significantly impacted on activated and effector T cell metabolism and function, we next questioned whether cana treatment possibly promoted T cell exhaustion by measuring expression of the co-inhibitory receptors LAG3, PD-1 and Tim-3 ( Figure 6F). A negligible reduction in the expression of LAG3 and Tim-3 was seen, and a modest decrease in PD-1 expression. Here, our data indicate that cana is a global T cell inhibitor and its effects are mediated independently of T cell exhaustion.

Canagliflozin inhibits T cell function in systemic lupus erythematosus and rheumatoid arthritis
Thus far, we have demonstrated that cana is a global human T cell inhibitor, mechanistically disrupting both Myc-and mTORC1-driven programs leading to suppressed T cell metabolism and function. Therefore, cana may have significant potential as a repurposed drug for the treatment of T cell-mediated diseases. To this end, we isolated CD4+ T cells from two autoimmune patient cohorts (SLE and RA) and activated them for 24 h in the presence or absence of cana ( Figure 7A). Strikingly, cana impaired cytokine production (IL-2 and IFNg, Figures  7B and 7C, IL-17 and TNF; Figure S6A) and CD25, CD44, and CD69 expression ( Figures 7D and 7E) but did not result in a decrease in T cell size ( Figure S6B).
To account for the functional differences observed, we assessed whether the metabolic perturbations first observed in T cells from healthy donors upon cana treatment were recapitulated in T cells from patients with autoimmunity ( Figures 5A-5D). Here, using extracellular flux analysis cana suppressed ECAR and both basal and maximal glycolytic rate ( Figures S6C and  S6E), in addition to decreasing ATP levels derived from glycolysis at both the basal and maximal rates ( Figure S6F). In contrast, OXPHOS levels were impaired only at the maximal level, with basal levels unaffected ( Figures S6G and S6G-S6I). We observed a significant decrease in the spare respiratory capacity of autoimmune T cells exposed to cana ( Figure S6J   to a decrease in ATP levels derived from the maximal rate of OXPHOS ( Figure S6K). Finally, to determine whether cana has an effect at the site of inflammation, we isolated and treated synovial fluid mononuclear cells (SFMNCs) from RA patients undergoing arthroscopy for 24 h ( Figure 7F). Cana reduced TNF levels in 3 of the 4 patient samples and had a more profound effect decreasing IFNg and IL-17 levels, with the latter reaching statistical significance (Figures 7G and 7I). Cana had a largely inhibitory effect on reducing activation markers CD25, CD44 and CD69 and cell size in both CD4+ and CD8+ T cells analyzed with one patient outlier in the CD4+ T cell subsets analyzed (Figures S7A-S7D). Interestingly, this patient had a lower CD4+ T cell compartment (17.5% of CD3 gated) compared with the other patients (average: 47.4%).
Importantly, these data reveal that cana retains its ability to potently suppress T cell responses in patients with autoimmunity, underlining cana as a critical modulator of T cell metabolism and effector function with significant potential for the treatment of T cell-mediated autoimmunity.

DISCUSSION
In this manuscript, we investigate repurposing two types of SGLT2 inhibitors as potential therapeutic options for T cell-mediated autoimmunity. Importantly, we demonstrate that using physiologically relevant doses, cana, but not dapa, significantly impairs human CD4+ T cell-mediated inflammation. Our data indicate that this is not due to inhibition of SLC5A2 or SGLT2, which is not expressed by T cells, rather by cana-specific effects impairing T cell signaling, metabolic perturbation, and downstream T cell-activation-induced remodeling.
We reveal that cana impacts on T cell activation by specifically blunting TCR signaling. This impaired TCR response led to compromised metabolic reprogramming and effector function; in part via inhibition of key metabolic nodes downstream of the TCR (mTORC1 and Myc) and in part via off-target effects of cana on GDH and complex I. Inhibition of GDH by cana led to impaired glutamine anaplerosis within the TCA cycle, demonstrated by decreased incorporation of glutamine-derived carbon into TCA cycle metabolites. T cells were, however, able to compensate by enhancing incorporation of glucose-derived carbon, notably within aspartate and glutamate pools. This is in line with the study by Wu et al., which showed that impaired aspartate pools led to the biogenesis of TNF in RA. 41 Second, antagonism of mitochondrial complex I by cana led to increased early mitochondrial ROS production concomitant with mitochondrial dysfunction.
Mechanistically, cana impaired early TCR signaling events ultimately compromising cytokine production. Interestingly, this observation could be rescued when stimulating T cells with PMA/ionomycin, thus bypassing the TCR. Reduced TCR transduction, mediated by cana, led to decreased ERK and mTOR, ultimately blunting c-Myc expression and activity. The observed decrease in mTORC1 signaling and reduced c-Myc expression at the protein level correlates with the compromised metabolic profile of T cells. Specifically, downstream metabolic targets of c-Myc that were suppressed in the presence of cana were of significant interest because of their association with glucose uptake (SLC2A1), glycolysis (HK2), one carbon metabolism (DHFRalso a specific target of the autoimmune drug, methotrexate) and fatty acid synthesis (ACLY and FASN). These data are in line with previous studies demonstrating the catastrophic effect on the T cell proteome of impaired c-Myc signaling. 39 It is well known that aberrant T cell metabolism contributes to dysregulated function and breakdown of self-tolerance in autoimmunity. 42 Given the relative ineffectiveness alongside the crippling side-effects of certain autoimmune drugs, investigation into novel therapeutics is warranted. Multiple pre-clinical studies have demonstrated therapeutic benefit from targeting T cell metabolism in autoimmunity such as tetramization of the glycolytic enzyme pyruvate kinase, 43 inhibition of OXPHOS using oligomycin, 44 use of the glycolytic inhibitor 2-deoxy-D-glucose, 45 and glutaminase suppression. 46 However, while promising, the clinical translation of these studies remains a challenge due to the toxic nature of impairing metabolism at the whole-body level.
Interestingly, targeting immunometabolic processes using repurposed T2D drugs has previously demonstrated promise in multiple inflammatory disorders. 18,47 One of the most notable drug candidates is the complex I inhibitor metformin. In numerous studies, metformin has exhibited protective effects in a range of autoimmune disorders such as multiple sclerosis, SLE, and RA. 15,18,47 Aside from metformin, other T2D candidates have been investigated such as the PPAR-g agonist, pioglitazone, again with promising results in reducing autoimmuneassociated symptoms. 12,47 The novel class of recently approved T2D medication, SGLT2 inhibitors, have extensive roles beyond improved glycemic control. Additional benefits include the ability of cana to elicit protective clinical outcomes in cardiovascular disease (CVD) and chronic kidney disease. 28,29 Furthermore, cana has been demonstrated to promote PD-L1 degradation by endocytic recycling in SGLT2 positive non-small cell lung cells. 48 Given these encouraging precedents, we sought to investigate the effects of SGLT2 inhibitors on human T cell function.
Critically, our results show that cana retained its ability to suppress T cell-mediated inflammation in two human ex vivo autoimmune cohorts-SLE and RA. Here, cana treatment led to the suppression of effector function in peripheral blood T cells (potentially preventing the repopulation of inflammatory T cells) and ex vivo synovial fluid mononuclear cells-further advocating the repositioning of cana to treat autoimmunity. A further benefit of prescribing patients with autoimmunity cana is the protective effects against CVD-a comorbidity causing significant morbidity and mortality in multiple autoimmune disorders. 49 Together, the impaired effector function of T cells accompanied by systemic protective effects against CVD highlights the exciting potential in repurposing cana for autoimmunity. To conclude, a wealth of prior studies have reported that targeting T cell metabolism in autoimmunity can lead to therapeutic benefit. In our manuscript, we have demonstrated this by repurposing a clinically available T2D medication that modulates metabolism. Collectively, our manuscript demonstrates the inhibitory effects of cana on human CD4+ T cell function and provides a foundation for the clinical development of cana for the treatment of T cell-mediated autoimmune disease in humans.

Limitations of the study
Our study provides evidence in the repurposing of cana as a therapeutic intervention for T cell-mediated autoimmunity. RA is a complex and heterogeneous disease and further work is required to determine whether certain patient subgroups would preferentially benefit from cana therapy. It was beyond the scope of this study to investigate the impact of cana on other non-immune cell populations within the joint such as synovial fibroblasts and future work should determine whether inhibition of T cell inflammation by cana can reduce the inflammatory environment of the joint. Whether biologically relevant concentrations of cana would reach the inflamed joint is also a pertinent question and one that warrants further investigation. As cana is often prescribed in combination with metformin, it will therefore be poignant to determine whether there is a synergistic effect between the two drugs, potentially enhancing immunosuppressive capabilities in autoimmunity. Taken together, given that cana is an FDA-approved medication, a future clinical trial is warranted to determine whether cana has protective, clinical effects in patients with autoimmunity.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

INCLUSION AND DIVERSITY
We support inclusive, diverse, and equitable conduct of research. Demographic and clinical characteristics data of the study cohorts are presented in Tables S1 and S2.
Cell isolation and culture PBMCs were isolated by density gradient centrifugation using LymphoprepÔ (StemCell Technologies). CD4+ and CD8+ T-cells were purified from human PBMCs by negative selection using automated magnetic cell separation (Miltenyi). Cells were cultured in human plasma like medium (HPLM; Gibco), unless otherwise stated, and activated in the presence of plate-bound anti-CD3 (2 mg/ml; Biolegend) and soluble anti-CD28 (20 mg/ml; Biolegend) or left unstimulated. Cultures were supplemented by 10% FBS following 3 h incubation to avoid impaired T cell activation. T-cells were treated with canagliflozin (Cambridge Bioscience) or dapagliflozin (Combi-Blocks) at a concentration of 10 mM, unless otherwise stated.

METHOD DETAILS
Alternative culture conditions For certain experiments, IL-2 (10 ng/ml; Miltenyi) was added after 24 h. To bypass initial T cell receptor-dependent signalling, T-cells were activated using PMA (10 ng/ml; Merck) and ionomycin (500 ng/ml; Merck) for 4 h. For partial rescue, T-cells were activated in the presence and absence of canagliflozin, supplemented with dimethyl a-ketoglutarate (0.3 mM; Merck). For complex I inhibition experiments, T-cells were activated in the presence or absence of rotenone (1 mM; Merck) or metformin hydrochloride (10 mM; MedChemExpress). High-dose metformin hydrochloride was used to bypass any transporter-specific uptake. For combined inhibition of complex I and glutamate dehydrogenase, T-cells were activated in the presence and absence of piericidin A (500 nM; Enzo Life Sciences) and R162 (10 mM; Merck).

Human CD4+ regulatory T-cell differentiation
For regulatory T-cell (Treg) differentiation, naïve CD4+ T-cells were activated using Immunocult T cell activator (12 ml/ml; StemCell Technologies) and cultured with TGF-b (5 ng/mL; PeproTech) for 6 d in the presence and absence of canagliflozin (10 mM). Half of the medium was replaced at d3. At d6, cells were stained for FoxP3 (PE, mIgG1k, 206D, 320108, Biolegend) and supernatants harvested. Additionally, Tregs induced in the absence of canagliflozin were restimulated (12 ml/ml; StemCell Technologies) at d6 in the presence and absence of canagliflozin before staining for FoxP3 at 24 h, 48 h and 72 h.

Enzyme-linked immunosorbent assay
Human IL-2 (DY202), IL-4 (DY204), IL-10 (DY217B), IL-17 (DY317), IL-21 (DY8879), IFNg (DY285B), granzyme B (DY2906) and TNFa (DY210) were measured in cell-free culture supernatants according to the manufacturer's instructions (R&D Systems). ELISA plates were coated with the capture antibody and incubated overnight at 4 C. Wells were incubated at room temperature (RT) with gentle agitation with the following: appropriately-diluted cell-free supernatants and protein standards for 2 h, kit-specific detection antibody for 2 h, and streptavidin-horse radish peroxidase for 20 min. The plate was washed with 0.05% Tween-20 in PBS between each step. The plate was then incubated at RT with a 1:1 mixture of hydrogen peroxide and tetramethylbenzoic acid (BD Biosciences). Absorbance was measured at 450 nm after the addition of sulfuric acid (Merck) to each well and values were corrected to the blank.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical details of experiments can be found in the figure legends. All data are expressed as mean ± SEM as indicated in the figure legends. Statistical tests were selected based on appropriate assumptions with respect to data distribution and variance characteristics. For normally distributed data, statistical significance was determined using unpaired T test, one-way analysis of variance (ANOVA) followed by Dunnett's multiple comparisons test, or two-way ANOVA followed by Sidá k's multiple comparisons test. For not normally distributed data, statistical significance was determined using Mann-Whitney test, or Kruskal-Wallis test followed by Dunn's multiple comparisons test. For data normalised to the vehicle control group, statistical significance was determined using one-sample T test. Any data involving a large number of multiple comparisons were adjusted for the FDR using the methods outlined. All statistical analyses were performed using GraphPad Prism 9.0. Significant differences are indicated as follows: * p % 0.05, ** p % 0.01, *** p % 0.001, **** p < 0.0001.