ABHD11 is required for murine and human T-cell effector function
Given that reduced expression of ABHD11 within CD4+ T-cells is associated with remission in RA, we sought to better understand the role of ABHD11 in T-cell biology. Firstly, we established the degree of ABHD11 protein expression in human CD4+ T-cells, wherein ABHD11 was expressed at low levels in unstimulated T-cells, but became notably upregulated upon TCR-mediated activation (Fig. 1a). To determine the importance of ABHD11 function, human CD4+ T-cells were treated with ML-226, a highly-selective inhibitor that targets the active site serine of ABHD11, whilst concomitantly being activated by anti-CD3 and anti-CD28 for 24 h. Here, ML-226 significantly impaired cytokine production, with striking reductions in the release of IL-2, IL-10, IL-17, IFNγ and TNFα (Fig. 1b). We also analysed underlying mRNA levels, where comparable reductions were observed at the gene transcript level (Fig. 1c). Although there were modest reductions in CD25, CD44 and CD69 expression following ABHD11 inhibition, the proportion of cells that upregulate the expression of these markers following TCR stimulation remained similar (Fig. 1d; Supplementary Fig. 1a). Despite the loss of effector function, ABHD11 inhibition did not induce any notable reduction in T-cell size or proliferation (Fig. 1e, f). In line with this, global protein translation, as measured by puromycin incorporation, was not significantly altered by ABHD11 inhibition (Supplementary Fig. 1b). Importantly, cell viability remained intact in T-cells exposed to ML-226 (Supplementary Fig. 1c, d), confirming that the observed loss of effector function is attributable to ABHD11 inhibition rather than compromised cell viability.
We next examined proximal TCR signalling events in the presence or absence of ABHD11 inhibition. Interestingly, nascent phosphorylation events (≤5 min) were unperturbed following ABHD11 inhibition, however, all subsequent time points analysed (≥15 min) evidenced compromised TCR signalling, suggesting that T-cells are unable to sustain proximal TCR signalling events in the presence of ABHD11 inhibition (Fig. 1g). Given that our data suggest ABHD11 inhibition impairs global T-cell function, we next assessed its effect on regulatory T-cells (Treg). Here, polarisation of CD4+ naïve T-cells towards the Treg compartment was significantly reduced by ABHD11 inhibition (Supplementary Fig. 1e). Moreover, Tregs cultured in the presence of ML-226 for 24 h displayed significantly reduced FOXP3 expression (Supplementary Fig. 1f). In order to determine whether pharmacological inhibition of ABHD11 could be replicated using genetic ablation, we developed a series of Jurkat T-cell clones with ABHD11 gene expression knocked down using CRISPR/Cas9 (Fig. 1h). In agreement with pharmacological blockade, ABHD11 knockdown resulted in a significant reduction of IL-2 production upon stimulation with PMA/ionomycin (Fig. 1i).
To further investigate the impact of ABHD11 function on T-cell fitness, we activated murine T-cells in the presence and absence of WWL222, a potent and selective inhibitor targeting ABHD11 in mice. WWL222 is structurally distinct from ML-226, containing a different chemical scaffold, and, importantly, targets an alternate region of ABHD11. Specifically, murine T-cells were polarised in vitro towards distinct T-cell lineages in the presence and absence of WWL222 to assess its subset-specific effect on cytokine production. As with their human counterparts, cytokine production was impaired by ABHD11 inhibition, with a reduction in both the frequency of IFNγ-, IL-13- and IL-17-producing cells, as well as the quantity of each cytokine, under Th1-, Th2- and Th17-polarising conditions, respectively (Fig. 1j-l). These findings were largely recapitulated, with the exception of Th1-mediated IFNγ production, when murine T-cells were treated with ML-226, an inhibitor primarily associated with the human ortholog of ABHD11, but has some affinity for the equivalent murine protein (Supplementary Fig. 1g-i). It should be noted that WWL222 does not bind human ABHD11, therefore we did not perform the reciprocal experiments. Assessment of ABHD11 inhibition on murine Treg polarisation did not reveal any defects in FOXP3 expression when cultured with either WWL222 or ML-226 (Supplementary Fig. 1j). Furthermore, no compensatory increases in FOXP3 were observed under any of the three polarising conditions (Th1, Th2 or Th17) upon treatment with either ML-226 or WWL222 (Supplementary Fig. 1k-m). Together, these data indicate that ABHD11 is essential for optimal T-cell function.
Given that ABHD11 primarily maintains the function of α-KGDH within the TCA cycle, we next investigated the impact of ABHD11 inhibition on human T-cell metabolism. Firstly, we assessed whether ML-226 directly inhibits the catalytic activity of α-KGDH. Although modest, we recorded a significant reduction in α-KGDH activity (Fig. 2a), indicating that ML-226 can inhibit ABHD11 oxidative decarboxylation of α-KG in T-cells. Interestingly, inhibition of α-KGDH using an alternative small molecule inhibitor (CPI-613; currently undergoing clinical evaluation [NCT05325281]) was able to reproduce the reduction in cytokines observed upon ABHD11 inhibition (Supplementary Fig. 2a). Subsequently, we activated T-cells in the presence and absence of ML-226 and examined cellular metabolism using a mitochondrial stress assay. Here, there was a significant reduction in oxygen consumption rate (OCR) following ABHD11 inhibition, with consistent reductions in basal respiration, ATP-linked respiration, maximal respiratory capacity and spare respiratory capacity (Fig. 2b, c). As expected, these changes precede impaired ATP production from OXPHOS (Fig. 2d). Furthermore, we utilised targeted mass spectrometry analysis to understand potential alterations at the metabolite level. Here, ABHD11 inhibition revealed a compromised TCA cycle, in which succinate levels were significantly reduced following ABHD11 inhibition, in addition to a striking accumulation of acetyl-CoA (Fig. 2e). To determine whether the source of acetyl-CoA was citrate-derived, and thus dependent on ATP citrate lyase (ACLY), we cultured T-cells in the presence and absence of a specific ACLY inhibitor, BMS-303141. To this end, we did not observe a rescue in cytokine production upon culture with BMS-303141 (Supplementary Fig. 2b), suggesting that this conversion of mitochondrial citrate to cytosolic acetyl-CoA is dispensable. An alternative mechanism of buffering excess mitochondrial acetyl-CoA, independent of citrate, is conversion to acetyl-carnitine. Analysis of our mass spectrometry data confirmed heightened levels of acetyl-carnitine following ABHD11 inhibition (Supplementary Fig. 2c), which together suggests that the carnitine shuttle buffers excess mitochondrial acetyl-CoA upon ABHD11 inhibition. Although α-KG levels remain similar following ABHD11 inhibition (Fig. 2e), there is a marked increase in the observed α-KG / succinate ratio within these cells (Fig. 2f), further indicating impaired α-KGDH activity. However, intracellular 2-HG levels were unchanged (Supplementary Fig. 2d), suggesting that there is alternative mechanism underpinning the observed phenotype, rather than the epigenetic mechanism previously described following ABHD11 loss. Moreover, the amino acid pool is also noticeably diminished in these cells -- with the exception of glutamine, whereby increased levels are reported -- with observed reductions in glutamate, aspartate and asparagine (Fig. 2g). Aspartate plays an important role in nucleotide synthesis; therefore, it is unsurprising that there is a concomitant reduction in the production of several nucleotides following ABHD11 inhibition (Supplementary Fig. 2e). Interestingly, T-cells treated with ML-226 appeared to have greater mitochondrial mass (Fig. 2h), though this translated into no discernible effect on mitochondrial depolarisation (Supplementary Fig. 2f). Blockade through α-KGDH is associated with the production of mitochondrial reactive oxygen species (mitoROS), therefore we assessed both mitoROS and total ROS levels using mitoSOX and cellROX, respectively. Here, mitoROS levels were initially heightened 1 h post-ABHD11 inhibition, but this difference later dissipated at 24 h (Fig. 2i and Supplementary Fig. 2g). Total ROS levels were elevated at 1 and 24 h (Supplementary Fig. 2h), in agreement with the observed reduction in the glutathione pool present in these cells (Supplementary Fig. 2i). Taken together, there is substantial evidence that the TCA cycle is compromised following ABHD11 inhibition, altering the metabolic landscape within the mitochondria.
Given the impact on OXPHOS, we next determined whether ABHD11 inhibition perturbed lactic acid excretion, measured using the extracellular acidification rate (ECAR). We observed a modest reduction in ECAR, whereby glycolytic capacity was attenuated by ABHD11 inhibition (Fig. 2j, k). This translated into a reduction in glycolytic ATP production (Fig. 2l). However, these changes are not necessarily underpinned by depletion of the metabolite pool, as intracellular levels of glycolytic intermediates are not consistently changed following ABHD11 inhibition (Supplementary Fig. 2j). A substantial increase of intracellular lactate levels was observed upon ABHD11 inhibition (Fig. 2m). Importantly, lactate accumulates within these cells and is not exported at a higher rate (Fig. 2n), which agrees with the modest reduction in extracellular acidification rate with ABHD11 inhibition (Fig. 2j-l). Together, these data show that ABHD11 inhibition rewires mitochondrial metabolism, establishing a compromised TCA cycle that promotes the accumulation of acetyl-CoA and lactate.
To determine the source of the elevated lactate, we next employed liquid chromatography mass spectrometry to track the fate of cellular metabolites. Here, we performed stable isotope labelling using uniformly-labelled C-glucose to follow the incorporation of these carbons into downstream metabolites. An elevated percentage of C is incorporated into lactate (Supplementary Fig. 3a), which perhaps suggests a feedback mechanism from the compromised TCA cycle. We also observed C incorporation into acetyl-CoA. Strikingly, C incorporation into succinate is unchanged and there was relatively little incorporation of glucose carbon downstream of α-KG, perhaps due to timepoint restrictions. Therefore, we utilised uniformly-labelled C-glutamine (almost directly upstream of α-KGDH) to measure glutamine anaplerosis (Supplementary Fig. 3b). Here, reduced C incorporation into succinate was more pronounced, whilst incorporation into other TCA cycle intermediates did not appear to be significantly altered (Supplementary Fig. 3b). Whilst we did not observe a build-up of m + 5 α-KG, inhibition of α-KGDH is reflected by a significant reduction in m + 4 succinate and an increased ratio of α-KG/succinate (Supplementary Fig. 3c-e). These data would suggest that the accumulation of lactate and most likely acetyl-CoA is sustained by glucose following ABHD11 inhibition.
To further our understanding of the mechanisms underpinning rewired T-cell metabolism and suppressed effector function following ABHD11 inhibition, we sought to investigate changes in the T-cell transcriptome following ML-226 treatment. RNA-Seq analysis revealed 69 genes that were differentially-expressed upon ABHD11 inhibition, of which 38 were upregulated and 31 were downregulated (Fig. 3a). To realise the biological relevance of these changes, we performed pathway enrichment analysis to determine which pathways become up- and downregulated upon ABHD11 inhibition. Unsurprisingly, several pathways associated with T-cell function, such as cellular response to cytokine stimulus and inflammatory response, were downregulated following ABHD11 inhibition (Fig. 3b), supporting our observed effects on cytokine production (Fig. 1b, c). Conversely, the vast majority of the pathways upregulated in response to ABHD11 inhibition were associated with either sterol or fatty acid metabolism, with sterol biosynthetic process emerging as the most enriched pathway overall (Fig. 3b). Interestingly, RNA-Seq analysis of ABHD11 knockdown Jurkat T-cell clones highlighted the upregulation of similar metabolic pathways (sterol and cholesterol biosynthetic processes; Supplementary Fig. 4a). These data would suggest that the immunomodulatory phenotype observed are likely attributed to sterol biosynthetic processes, rather than alternative acetyl-CoA-mediated processes.
To determine what drives the observed changes in gene expression and, ultimately, the suppressed effector function observed following ABHD11 inhibition, we performed transcription factor analysis. Here, enriched transcription factors are predicted using the differentially-expressed genes identified -- the lower the "average rank" score, the more enriched that transcription factor and its activity. Once more, there was a clear association with lipid metabolism, whereby PPARG, SREBF1 and SREBF2 were all amongst the most enriched transcription factors (Fig. 3c). In fact, SREBF1 and SREBF2 are significantly upregulated at the transcript level following ABHD11 inhibition (Fig. 3a). In further support of this, sterol biosynthetic process is the pathway most significantly enriched following ABHD11 inhibition (Fig. 3d) -- a process tightly regulated by the signalling of sterol regulatory element binding proteins (SREBPs) encoded by SREBF1 and SREBF2. There is recent evidence demonstrating the existence of a lactate-SREBP2 signalling axis within human immune cells. Therefore, given the heightened intracellular lactate levels we observe following ABHD11 inhibition (Fig. 2m), we activated T-cells in the presence of either ML-226 or lactic acid and assessed whether there was a comparable effect on SREBF2 expression and cytokine production. Here, treatment with lactic acid phenocopied ABHD11 inhibition, with a similar increase in SREBF2 mRNA levels accompanied by a corresponding reduction in IL-2 and IFNγ production (Supplementary Fig. 4b, c). These data suggest that lactate drives SREBP activation and the subsequent upregulation of sterol biosynthesis pathways.
To establish the significance of augmented sterol biosynthesis following ABHD11 inhibition, we next carried out specialised mass spectrometry to measure the intracellular levels of various sterol species. Surprisingly, we observed a reduction in total sterol levels following ABHD11 inhibition (Fig. 3e). This is primarily attributed to a reduction in non-oxygenated sterols, wherein there is a trend towards decrease across the group, with several significantly reduced (Fig. 3f; Supplementary Fig. 4d). Conversely, we identified heightened oxysterol levels within these cells (Fig. 3g), which in turn increased the oxysterol/sterol ratio (Supplementary Fig. 4e). Of the oxysterols measured, 27-hydroxycholesterol (27-HC) and 24,25-epoxycholesterol (24,25-EC) emerged as the two most significantly elevated species, particularly 24,25-EC whose levels are approximately 5-10 times higher following ABHD11 inhibition (Fig. 3h). Interestingly, there does not appear to be a general increase across all oxysterol species, as 7α-hydroxycholesterol (7α-HC) and 24-hydroxycholesterol (24-HC) levels were reduced upon ABHD11 inhibition (Fig. 3h). Intriguingly, we report activation of a shunt pathway -- branching from the classical mevalonate pathway at oxidosqualene -- which synthesises 24,25-EC from acetyl-CoA (Supplementary Fig. 4f).
27-HC and 24,25-EC are potent activators of liver X receptor (LXR) signalling. To establish whether the increased production of 27-HC and 24,25-EC following ABHD11 inhibition leads to LXR activation, we compared treatment with ML-226 versus treatment with GW3965 -- a synthetic LXR agonist. Interestingly, recent work by Waddington et al. assessed the transcriptional effects of LXR activation on human CD4+ T-cell function using RNA-Seq. In comparison to our dataset, 19 of the 65 LXR-regulated transcripts identified were also differentially regulated by ABHD11 inhibition (Fig. 3i), which accounts for approximately 27.5% of all differentially-expressed transcripts (Supplementary Fig. 4g), indicating significant overlap between both nodes. As such, we hypothesised that LXR activation would phenocopy, at least partially, ABHD11 inhibition in human T-cells. To this end, we treated T-cells with ML-226 and GW3965 in parallel before assessing cytokine production. Here, we observed comparable reductions in IFNγ and IL-2 production between ABHD11 inhibition and LXR activation (Fig. 3j). This effect was also observed when treating T-cells with the intracellular oxysterol species that are increased following ABHD11 inhibition -- 27-HC and 24,25-EC (Fig. 3k). Interestingly, co-treatment with the LXR antagonist, GSK2033, ablated the mRNA levels of LXR targets SREBF1 and SCD following ABHD11 inhibition (Supplementary Fig. 4h), and subsequently rescued the impaired cytokine response (Fig. 3l). Together, these data suggest that LXR activation drives, at least in part, the anti-inflammatory phenotype observed in human T-cells following ABHD11 inhibition.
Given that acetyl-CoA both accumulates within the cell (Fig. 2e) and fuels many of the pathways that become upregulated following ABHD11 inhibition (Fig. 3b), we next explored some of the multi-faceted roles of acetyl-CoA. Firstly, we observed no difference in any of the histone acetylation marks measured or total protein acetylation following ABHD11 inhibition (Supplementary Fig. 5a, b). Intriguingly, lipidomic analysis demonstrated markedly increased levels of triacylglycerols (TAGs) in response to ABHD11 inhibition (Supplementary Fig. 5c-e), accompanied by a trend towards increase in diacylglycerol (DAG) levels (Supplementary Fig. 5c, d). These findings are in agreement with a previous study by Waddington et al., wherein pharmacological activation of LXR within human T-cells was shown to significantly heighten the levels of several TAG species. Overall, these data indicate that surplus acetyl-CoA within ABHD11-inhibted T-cells is used to fuel the lipid-associated processes that are upregulated by LXR activation, generating increased levels of TAGs.
Our findings thus far have outlined that ABHD11 inhibition suppresses T-cell effector function, underpinned by a compromised TCA cycle that promotes 24,25-EC synthesis and consequently activates LXR signalling (Fig. 4). This highlights the exciting potential of manipulating ABHD11 function for therapeutic benefit in T-cell-mediated autoimmune disease, where reversing the hyper-activation and -function of pathogenic T-cells would be valuable. We explored this possibility initially in two autoimmune patient cohorts, isolating CD4+ T-cells from rheumatoid arthritis (RA) and type 1 diabetes (T1D) patients and activating them in the presence and absence of ML-226 ex vivo (Fig. 5a). Here, we again observed a significant reduction in the production of a diverse range of cytokines by RA and T1D T-cells following ABHD11 inhibition (Fig. 5b, c), alongside a modest, but significant reduction in T-cell activation (Fig. 5d, e). Again, there was a minimal reduction in cell size in both patient cohorts (Supplementary Fig. 6a, b), whilst impaired effector function was independent of any changes in viability (Supplementary Fig. 6c, d).
To consolidate these findings, we assessed the efficacy of ABHD11 inhibition on synovial fluid mononuclear cells (SFMCs) isolated from the site of inflammation in a cohort of RA patients (Fig. 5f). To this end, we observed targeted reductions in IL-17 and IFNγ production following ABHD11 inhibition, whilst IL-2, IL-10 and TNFα production remained unchanged (Fig. 5g). In line with our earlier activation data (Fig. 1d), CD4+ T-cell activation remained intact following ABHD11 inhibition (Fig. 5h), with a modest, but significant increase in cell size (Supplementary Fig. 6e). Crucially, these changes did not result from compromised viability (Supplementary Fig. 6f). Together, these findings demonstrate that ABHD11 inhibition retains its suppressive effect on T-cell function in patients with RA and T1D, including those present at the site of inflammation.
To support our findings in humans, we next investigated whether ABHD11 regulates T-cell fate and function in a murine model of accelerated T1D. Here, we initially assessed the effect of ABHD11 inhibition on antigen-specific CD4+ T-cells, whereby diabetogenic H2-Ag-restricted BDC2.5 CD4+ T-cells were stimulated in vitro in the presence of their cognate antigen (a hybrid insulin peptide [HIP]) and treated with the murine ABHD11 inhibitor, WWL222. There was a minimal reduction in proliferation observed following ABHD11 inhibition (Fig. 6a). Despite a modest, but significant reduction in CD25 and CD69 expression on proliferating cells, the proportion of cells expressing these markers was unchanged (Fig. 6b), again indicating that their activation is intact. Importantly, proinflammatory cytokine production is impaired by ABHD11 inhibition, with marked reductions in IFNγ, IL-2, IL-17 and TNFα (Fig. 6c). Interestingly, this suppression appeared to be limited to proinflammatory cytokines, as we observed a striking increase in IL-10 production following ABHD11 inhibition (Fig. 6c), which might indicate that ABHD11 inhibition induces an anti-inflammatory phenotype in antigen-specific T-cells, rather than generally inhibiting effector function.
We furthered these investigations using an in vivo murine model of accelerated T1D. To this end, immunocompromised Rag-deficient female mice were injected with 5 × 10 peptide-activated BDC2.5 CD4+ T-cells and WWL222 (or vehicle control), with the drug dose repeated daily for the course of the experiment (Fig. 7a). It is well-documented that the Rag1 adoptive transfer model presents a rapid and high-penetrance disease course, thereby small delays in onset can reflect meaningful biological activity. Encouragingly, ABHD11 inhibition by low-dose WWL222 treatment significantly delayed the development of T1D (Fig. 7b), attributable to their glycaemic stability versus the vehicle control group (Supplementary Fig. 7a). At the end of the observation period, immune cells were harvested from the spleen of diabetic mice to elucidate the changes that underpin delayed onset of disease. Here, CD4+ T-cells from WWL222-treated mice displayed impaired activation (Fig. 7c), and produced significantly less IFNγ, TNFα and IL-2 (Fig. 7d; Supplementary Fig. 7b, c) versus those harvested from untreated mice. Additionally, TNFα was reduced in myeloid cell populations, meaning ABHD11 inhibition may skew the cytokine profile through multiple cell types (Supplementary Fig. 7d-f). To strengthen and validate these findings, we utilised an additional murine model in which immunocompromised Rag-deficient female mice were injected with 12 × 10 splenocytes from newly diabetic mice and treated as previously described with WWL222 (Fig. 7e). Excitingly, ABHD11 inhibition again significantly delayed the onset of T1D (Fig. 7f) by maintaining blood glucose levels (Supplementary Fig. 7g). Assessment of splenocytes harvested post-observation period revealed that this alleviation of T1D was underpinned by a reduction in the frequency of IFNγ-producing CD4+ T-cells (Fig. 7g), with no change in TNFα production (Supplementary Fig. 7h). These data show that targeting ABHD11 can improve outcomes in settings of T-cell-mediated inflammation, highlighting the significant potential of developing drugs that inhibit this metabolic node as a novel treatment strategy.