key: cord-0256176-inlehuk1
authors: Wilson, John J.; Wei, Jian; Daamen, Andrea R.; Sears, John D.; Bechtel, Elaine; Mayberry, Colleen L.; Stafford, Grace A.; Bechtold, Lesley; Grammer, Amrie C.; Lipsky, Peter E.; Roopenian, Derry C.; Chang, Chih-Hao
title: Augmented glucose dependency of autoreactive B cells provides a treatment target for lupus
date: 2022-02-03
journal: bioRxiv
DOI: 10.1101/2022.02.01.475510
sha: 9f0af15220f834769a377517a73119efb4565bf9
doc_id: 256176
cord_uid: inlehuk1

Heightened glycolysis is inherent to immune/inflammatory disorders, but little is known of its role in the pathogenesis of systemic lupus erythematosus (lupus). Here, we profile key autoimmune populations in acute and chronic lupus-prone models and their response to glycolytic inhibition. We demonstrate that glycolysis is specifically required for autoreactive germinal center B cells (GCB), but not for T follicular helper cells (Tfh) to survive. This augmented reliance on glucose oxidation to maintain ATP production in pathogenic GCB renders them highly susceptible to oxidative stress-induced apoptosis triggered by glycolysis blockade via 2-deoxyglucose (2DG). We show that 2DG can preferentially reduce GCB in lupus-prone mice, while sparing other autoreactive populations, including Tfh, but still significantly improving lifespan and kidney function. Furthermore, the subset of GCB expressing B-cell maturation antigen (BCMA) exhibits an exaggerated dependence on glycolysis to sustain their growth. Depletion of these cells with a proliferation-inducing ligand-based CAR T-cells leads to greatly prolonged lifespan of mice with severe autoimmune activation. These results reveal that glycolysis dependent GCB, especially those expressing BCMA, are key lupus mediators and highlight that they can be selectively targeted to improve disease outcomes for lupus patients.

The autoimmune disease systemic lupus erythematosus (SLE, lupus) is characterized by loss of self-tolerance, leading to a dysregulated expansion of hyperactivated T and B cells 1, 2 .

Underlying lupus is the induction of T-cell-dependent activation and clonal expansion of autoreactive B cells in both germinal centers (GCs) and extra-follicular foci, resulting in their differentiation into plasma cells that secrete pathogenic autoantibodies, which are the primary cause of lupus nephritis 3,4 . B-cell maturation antigen (BCMA, or TNFRSF17) is particularly upregulated in mature B cells as they differentiate toward plasma cells. Moreover, increasing expression of BCMA has also been observed on lupus B cells, and has been linked with increased activation via interactions with the cytokines TNFSF13 and 13B 5,6 . B cells and particularly activated GC-like B cells, therefore present an attractive target for treatment of lupus, but to date only belimumab, a monoclonal antibody to TNFSF13B has been approved as a targeted treatment for this autoimmune disorder 7 .

Autoreactive CD4 + T follicular helper cells (Tfh), are also critical drivers of lupus autoimmunity 8,9 by engaging follicular B cells in cognate and costimulatory interactions and secreting cytokines, especially interleukin (IL)-21, that promote germinal center B cells (GCB) proliferation and differentiation 8,10,11 . In turn, the expression of costimulatory molecules, such as ICOS-L and CD80/CD86 on activated B cells in GC further promotes Tfh activation, differentiation, and Bcl6 expression 12 . Because of the co-dependency of Tfh and GCB, both are critical regulators of spontaneous GC formation during lupus 2, 9 . A primary metabolic adaptation of activated lymphocytes is increased glucose metabolism 13 , likely mirrored in hyperactive autoimmune Tfh and GCB. This increased glucose metabolism not only provides a rapid source 4 of ATP, but also glycolytic intermediates required for enhanced proliferation and effector functions 14, 15 . With this accentuated glycolytic demand comes a potential avenue for therapeutic intervention by exploiting differential metabolic activity between these cell types for precise treatment of disease. However, the distinct metabolic requirements of discrete autoreactive T and B cells remain unclear.

Here, we demonstrate that activated autoimmune B cells closely resembling GCB (GL7 + B cells), but not autoreactive Tfh (CXCR5 + PD1 + CD4 + T cells), exhibit markedly enhanced glycolytic activity that makes them selectively susceptible to therapeutic intervention in spontaneous mouse models of lupus. Compared with Tfh, GCB show a higher glycolysis rate and glucose dependency. We found that 1 week of glycolytic inhibition (short-term) with 2deoxyglucose (2DG) preferentially reduced GCB through oxidative stress-induced apoptosis, while leaving Tfh numerically and functionally unaffected. This numerical reduction of GCB, significantly reduced mortality rates and protected mice from lupus nephritis. We also found that a subset of GCB expressing BCMA exhibit markedly elevated glycolytic rates and a correspondingly increased sensitivity to glucose inhibition. Reduction of these cells with TNFSF13 (APRIL)-based CAR-T cells significantly delayed lupus progression, suggesting the importance of these cells in the pathogenesis of lupus. Overall, this work serves as a pre-clinical proof-of-principle that metabolic modulation via 2DG preferentially targets and reduces autoreactive GCB, especially the BCMA-expressing subset. Moreover, this reduction is directly linked with reduced lupus severity, revealing an exploitable vulnerability to treat lupus, while minimizing the broad-spectrum immunosuppression.

Long-term glycolytic inhibition via 2DG can attenuate development of cellular disease phenotypes in chronic lupus-prone mice 16, 17 . We profiled autoreactive immune cell populations over the course of disease progression in the acute lupus-prone mouse model, BXSB.Cg-Cd8a tm1Mak Il15 tm1Imx Yaa (Yaa DKO) 18 , and asked whether long-term glycolytic inhibition could result in increased lifespan. Six-week-old pre-symptomatic Yaa DKO mice showed decreased frequencies of circulating CD4 + T cells when compared with healthy controls, increased frequencies of circulating Tfh (cTfh), effector T (Teff), CD25 + CD4 + T cells, and myeloid cells ( Figure 1A and Supplemental Figure 1A ). After 4-weeks (long-term) of 2DG treatment, many of the cellular aberrations observed in untreated Yaa DKO mice (decreased CD4 + T-cell frequencies, increased Teff and cTfh frequencies) were suppressed, whereas frequencies of CD25 + CD4 + T cells, total B cells, and myeloid cells were unaffected over the 7 weeks of treatment. Moreover, frequencies of B cells that were positive for GL7, a marker of activated B cells in the periphery 19,20 and of GCB in the splenic and lymphatic compartments 17,21 , were reduced to below those of control mice ( Figure 1A and Supplemental Figure 1A) . Flow cytometric analysis of spleen cells from untreated symptomatic Yaa DKO mice indicated expansion of Teff, Tfh, GCB, myeloid, plasmablasts/plasma, and CD25 + CD4 + T cells; 7 weeks of 2DG treatment reversed these abnormalities ( Figure 1B and Supplemental Figure 1B) . 2DGtreated mice showed complete survival with the first fatality in the treated group occurring ~10 weeks after 2DG withdrawal, by which time 95% of untreated mice had died ( Figure 1C) .

Notably, the significant reduction of peripheral GL7 + B cells was mirrored in splenic GCB, suggesting that long-term glycolytic inhibition had a comparable effect on GL7 + B cells in both 6 compartments. Overall, these results indicate that glycolytic inhibition via 2DG is effective in ameliorating autoimmune activation and increasing survival in the model of severe lupus.

Since the gene-expression patterns of T and B cells from lupus patients differ with those from healthy individuals 22,23 , we investigated whether 2DG treatment alters the dysregulated gene expression patterns in Yaa DKO mice. Initially, we carried out Gene Set Enrichment Analysis Figure 2) . Notably, although the activated T cell gene signature was also influenced, it was decreased to a much lesser degree. These data indicate that 2DG-mediated glycolytic inhibition preferentially influences metabolic regulation of activated B cell populations.

Gene expression analysis suggested that activated B lineage cells appear to be more sensitive to long-term glycolytic inhibition than T cells, consistent with the greater depletion of GCB (to below normal levels) than Tfh ( Figure 1B) . Given this disparity, we hypothesized that the metabolic requirements of autoreactive GCB and Tfh might differ and, specifically, that GCB might be more dependent on glycolysis. To test this, splenic GCB and Tfh from symptomatic Yaa DKO mice were sorted to assess their bioenergetic profiles. Autoreactive GCB were found to have higher extra-cellular acidification rates (ECAR), an indicator of glycolysis, and oxygen consumption rates (OCR), an indicator of mitochondrial OXPHOS, compared to Tfh ( Figure   3A ), and other effector T and non-GCB cells ( Figure 3B ). This high bioenergetic profile was also found in GCB isolated from symptomatic NZBWF1 late-onset lupus-prone mice (Supplemental Figure 3A) . To determine whether this elevated metabolic demand was diseaserelated, basal ECAR and OCR in GCB and Tfh from healthy, pre-symptomatic and symptomatic Yaa DKO mice were assessed. We found that symptomatic mice displayed significantly higher ECAR in both GCB and Tfh, but that only GCB showed elevated OCR compared to presymptomatic or healthy mice ( Figure 3C ). Autoreactive GCB also showed greater expression of the major glucose transporter Glut1 (Figure 3D) , and significantly higher 2-NBDG uptake than 8 did Tfh (Figure 3E ). In addition, GCB, cultured ex vivo, reduced the concentration of glucose in the medium to a significantly greater degree than Tfh ( Figure 3F ). Together, these data demonstrate that GCB exhibit a higher glucose usage and glycolysis rate than do Tfh. We next investigated whether enhanced glycolysis supports OXPHOS, by treating autoreactive GCB and Tfh with 2DG in the presence of glucose. Although 2DG reduced the basal ECAR of both GCB and Tfh to similar levels (Supplemental Figure 3B) , 2DG-sensitive maximal respiratory capacity (MRC) was significantly decreased (27%) in GCB, but not Tfh, non-GCB, nor in Teff (Figure 3 , G and H, and Supplemental Figure 3C ), after FCCP-mediated mitochondrial decoupling, an indicator of OXPHOS reliance on a nutrient substrate fueling mitochondria 27 . A similar 2DG sensitivity was also found in GCB (30%) from NZBWF1 mice (Supplemental Figure 3D) . These results suggest that 2DG-mediated glycolytic inhibition selectively impacts OXPHOS in GCB. To confirm whether fueling the mitochondrial matrix in GCB depends on glycolysis, we exposed cells to UK5099, which blocks the import of glycolysis-derived pyruvate into mitochondria. Treatment with UK5099 significantly impaired the MRC of GCB (15%) but had no effect on Tfh (Supplemental Figure 3E) . We further evaluated whether mitochondrial fueling of GCB also depends on other catabolic pathways.

Treatment of GCB and Tfh with BPTES, a glutaminase inhibitor 28 ; etomoxir, which suppresses mitochondrial fatty acid oxidation (FAO) or thioridazine, a selective inhibitor of peroxisomal FAO 29 ; resulted in little to no reduction in basal OCR in either cell type (Supplemental Figure   3 , F-H). Treatment with BPTES resulted in a small but significant reduction of MRC (7.7%) in GCB but not in Tfh (Supplemental Figure 3I) ; however, this reduction was much less than that found with glycolytic inhibition (Figure 3H) . Conversely, no change in MRC was observed in either GCB or Tfh treated with etomoxir or thioridazine (Supplemental Figure 3, J and K).

Together, these results are consistent with the mechanism of GCB reliance on glucose as the primary anaplerotic precursor to support mitochondrial function, whereas autoreactive Tfh exhibit greater metabolic flexibility.

To test whether glycolytic inhibition preferentially affects GCB survival, mouse splenocytes were activated in vitro. Numbers of differentiated GCB-like cells were significantly decreased by 2DG exposure, whereas Teff-like cells were unaffected ( Figure 3I) . A similar trend in 2DG sensitivity was found in activated human PBMCs ( Figure 3J ). Glycolytic blockade with the hexokinase inhibitor 3-bromopyruvate also selectively decreased numbers of murine GCB-like cells (Supplemental Figure 3L) , further confirming the efficacy of glycolytic inhibition. We next assessed the effect of 2DG treatment on the survival of splenic GCB and Tfh isolated from symptomatic Yaa DKO mice. One-day 2DG exposure significantly reduced GCB numbers, which decreased further after two days, whereas expansion of Tfh was observed ( Figure 3K) .

These data show that 2DG treatment preferentially affects survival of highly-glycolytic GCB, while sparing the less bioenergetic Tfh.

While early long-term 2DG treatment was effective at preventing disease in Yaa DKO mice, the effects of prolonged glycolytic inhibition might be too oppressive for clinical application. Given their increased glycolytic dependency, we hypothesized that short-term 2DG treatment might selectively inhibit GCB in vivo. First, we assessed circulating activated B cells and Tfh in symptomatic Yaa DKO mice as treatment progressed and found that GL7 + B cell numbers dropped significantly by day 7 of therapy with no cTfh reduction ( Figure 4A) . No other circulating populations showed significantly reduced numbers compared to untreated controls with this 1-week (short-term) treatment course ( Figure 4B and Supplemental Figure 4A ). Next, a similar reduction in GCB was observed in lymph nodes (Supplemental Figure 4B ) and spleens ( Figure 4B and Supplemental Figure 4C ) of the mice, with no significant numerical reductions in other populations. Numbers of splenic follicular regulatory T (Tfr), Treg, and transitional, follicular, and marginal zone B cells, were also unaffected following short-term treatment (Supplemental Figure 4 , D-F). 2DG-induced reduction of cell numbers was likewise observed in B220 + CD95 hi CD38 low gated GCB 29 (Supplemental Figure 4G) . Moreover, symptomatic NZBWF1 mice treated with short-term 2DG displayed reductions in both circulating GL7 + B cells and splenic GCB numbers, similar to those in Yaa DKO mice ( Figure   4C ). These data suggest that increased sensitivity to glycolytic inhibition is a uniform feature of autoreactive GCB, rendering them comparatively more vulnerable than other autoimmune cell types.

To test whether this GCB reduction can reverse autoimmune pathology in mice with severe clinical disease, we treated symptomatic Yaa DKO mice with 2DG for 1 week and monitored their survival. The mean lifespan of untreated mice was 15.4 weeks, in contrast to 24.1 weeks for 2DG-treated mice ( Figure 4D ). Of note, mice showed no significant decrease in frequencies of autoreactive Tfh and ICOS + T cells after 2DG removal (Supplemental Figure 4H) , suggesting that therapeutic efficacy was unlikely to be conveyed by a post-treatment reduction of these populations. NZBWF1 mice experience progressive proteinuria and death from lupus nephritis 11 caused by cumulative kidney damage owing to immune complex deposition 30,31 . Lifespan extension ( Figure 4E) , paired with persistently reduced proteinuria (Supplemental Figure 4I) , was also observed in NZBWF1 mice treated with 1-week 2DG. These data suggest that diverse murine lupus models can be positively impacted by this short-term 2DG treatment. To confirm the involvement of 2DG-responsive GCB in disease pathogenesis, an adoptive transfer approach was undertaken. As noted above, GCB numbers were reduced in Yaa DKO mice with 1-week of 2DG ( Figure 4B ). Following treatment, the mice underwent adoptive transfer of flow-sorted autologous GCB or non-GCB cells from symptomatic Yaa DKO mice. The mean lifespans of GCB-recipient mice were significantly shorter (19.0 weeks) than those receiving non-GCB cells (23.8 weeks) or no cells (24.1 weeks) ( Figure 4F ). These data demonstrate that, short-term 2DG treatment was remarkably effective in reversing disease trajectory with a concomitant decrease in lupus nephritis and mortality. We then examined the effect of short-term 2DG treatment on splenic morphology. Histologic examination of spleens showed that treatment of Yaa DKO mice with 2DG significantly reduced the size but not the number of GCs ( Figure 4G) . The effects were different in NZBWF1 mice with no significant effect on splenic GC morphology ( Figure   4H ), consistent with the lower efficacy of 2DG in reducing GCB numbers in NZBWF1 mice (Figure 4 , B and C) and implying a need to optimize 2DG treatment in these mice. Taken together, these results demonstrate that autoreactive GCB, shown to be extremely 2DG-sensitive, are highly effective at accelerating lupus.

Inhibition of glycolysis can cause cell death through apoptosis 32 . To determine the mechanism by which autoreactive GCB are reduced, we treated symptomatic Yaa DKO mice with 2DG to assess cellular apoptosis. By day 4 of treatment, there was a significant increase in relative frequencies of pre-apoptotic (Annexin + /PI -) GL7 + B cells but not cTfh in 2DG-treated mice ( Figure 5A and Supplemental Figure 5A ). With 2DG treatment, there was a trend toward increased apoptotic GL7 + B cells on day 7, but no such trend for cTfh ( Figure 5A ). Splenic GCB isolated from 7-day 2DG-treated mice showed increased pre-apoptosis ( Figure 5B ). In Tfh, frequencies of pre-apoptosis were unchanged, whereas apoptosis was significantly decreased ( Figure 5B ). The reduction in GCB numbers could not be explained by a loss of GL7 protein as 2DG did not affect GL7 expression in GCB from Yaa DKO mice in vitro (Supplemental Figure   5B ). Moreover, numbers and frequencies of splenic GL7-expressing Teff in Yaa DKO mice were unaltered with 2DG treatment (Supplemental Figure 5C ). Together, these data suggest that short-term 2DG exposure selectively induces apoptosis in autoreactive GCB.

Apoptosis can be induced by endoplasmic reticulum (ER) stress and the unfolded protein response (UPR) 32,33 . The inositol-requiring enzyme 1/X-box-binding protein 1 (XBP1) pathway is a potent UPR signaling pathway in mammalian cells, promoting ER-induced apoptosis. To ascertain whether UPR signaling is elevated with 2DG treatment, we assessed mRNA expression of Xbp1 and its spliced form, Xbp1s, as markers of ER stress 33,34 . Xbp1 expression in both autoreactive GCB and Tfh were unchanged (Figure 5C ), whereas expression of Xbp1s was decreased in Tfh with 2DG-treatment, suggesting that neither GCB nor Tfh experienced ER stress in situ following transient 2DG. Glucose deprivation may reduce cellular ATP production and increase reactive oxygen species (ROS) levels, leading to oxidative stress and the subsequent death of tumor cells 35,36 . Hence, we interrogated 2DG-treated cells and found that the overall ATP production rate was reduced in autoreactive GCB but not in Tfh, compared with those 13 isolated from untreated mice ( Figure 5D ). This was further confirmed with lower levels of total ATP measured in GCB treated with 2DG compared to untreated controls ( Figure 5E ). We therefore examined the accumulation of mitochondrial ROS as an indicator of oxidative stress 35 , in Yaa DKO mice. Following 2DG treatment, ROS production was significantly elevated in GCB, but was unchanged in Tfh ( Figure 5F ), non-GCB, and Teff (Supplemental Figure 5D) ; whereas neither Tfh nor GCB experienced changes in mitochondrial mass (Supplemental Figure 5E ). Excessive ROS production in GCB is also linked to increased mitochondrial membrane potential (Supplemental Figure 5F ). Taken together, these data suggest that the underlying mechanism of selective autoreactive GCB reduction with short-term 2DG exposure was a pro-apoptotic effect from energy deprivation through increased mitochondrial oxidative stress.

Tfh are vital to the formation and stability of the GC, as they provide survival and differentiation signals to GCB 2,9 . Although numbers of autoreactive Tfh were unaffected by short-term glycolytic inhibition, it remained possible that this treatment impaired Tfh functionality and thus contributed to GCB apoptosis. To test this possible mechanism of reduction, splenic Tfh were isolated from Yaa DKO mice following 1-week 2DG treatment and examined for functional markers. Tfh-derived IL-21 is thought to play a pivotal role in lupus pathogenesis and is the major cytokine inducing B-cell differentiation 10,11,18 . Notably, neither the frequency of IL-21producing Tfh nor IL-21 expression differ following this treatment ( Figure 6A ). The transcription factor Bcl6 is essential for development of both Tfh and GCB 37,38 . We found that Bcl6 protein expression was not impacted by 2DG in either Tfh or GCB ( Figure 6B ). In addition, the expression and frequency of Ki-67, a proliferation marker, in both Tfh and GCB were unaffected by 2DG ( Figure 6C) . Furthermore, no treatment-induced reduction in the frequency or expression of the GCB-Tfh interaction molecules ICOS-L, CD40, CD80 and CD86 on GCB (Supplemental Figure 6A) , or ICOS, CD40L and CD28 on Tfh was observed (Supplemental Figure 6B) . The basal OCR of sorted Tfh was significantly elevated in mice treated with 2DG compared with untreated mice (Figure 6D) , suggesting Tfh can better adapt their metabolism to survive glycolytic inhibition. Although the basal ECAR and OCR of GCB were unaffected (Figure 6D ), the reliance of remaining GCB from 2DG-treated mice on glycolysis became negligible ( Figure 6E ). Together, these results suggest that Tfh and GCB remain functionally intact following short-term glycolytic inhibition, and that this inhibition specifically affects hyperactive GCB with high glycolytic dependency. Overall, these data indicate that the effect of 2DG on the numerical reduction of GCB is direct and not mediated by Tfh.

BCMA is a transmembrane glycoprotein exclusively expressed on activated B and plasma cells 39 , and is associated with disease activity in human lupus 5,6 . Numbers of total peripheral B cells expressing BCMA were elevated in symptomatic Yaa DKO mice compared to pre-symptomatic or healthy mice ( Figure 7A) . Importantly, the frequencies and numbers of peripheral BCMA + GL7 + B cells were significantly higher in symptomatic than in pre-symptomatic mice ( Figure   7B ). Following short-term 2DG treatment, numbers of both circulating BCMA + GL7 + B cells and splenic BCMA + GCB from symptomatic Yaa DKO mice decreased significantly (Figure 7 , C and D). We found that, compared to BCMA -GCB, BCMA + GCB showed significantly higher surface expression of Glut1 (Figure 7E) , and significantly greater 2-NBDG uptake ( Figure 7F ).

Moreover, BCMA + GCB were highly bioenergetic, exhibiting significantly higher basal ECAR and OCR rates compared to BCMA -GCB (Figure 7, G and H) . While both populations showed reduced MRC with 2DG treatment, the reduction was significantly greater in BCMA + GCB than Figure 7L ). Together, these data support the significant involvement of highly glycolytic BCMA + GCB in lupus pathogenesis and suggest that activated BCMA-expressing B cells can be targeted to treat lupus.

but not the activated T cell signature from lupus patients.

To evaluate whether there is elevated glycolysis in the B-cell over T-cell compartment in human SLE, we examined the transcriptomic profile of peripheral CD19 + B and CD4 + T cells sorted from lupus patients. Linear regression analyses indicated that the gene signature of activated B cells had a strong, positive correlation with glycolysis, significant but weak positive correlations with oxidative phosphorylation, pentose phosphate pathway, tricarboxylic acid cycle and glutamine metabolism, and no correlation with fatty acid oxidation (Figure 8) . Conversely, the activated T-cell gene signature was only weakly correlated with the pentose phosphate pathway and glutamine metabolism with negative and positive correlations, respectively, but not with glycolysis. When comparing human results to mice, it was noted that peripheral activated B cells from lupus patients and splenic autoreactive B cells of lupus prone mice ( Figure 2B ) were both positively correlated with glycolysis, suggesting that short-term glycolytic blockade may also be efficacious in preferentially targeting pathogenic B cells in human SLE.

Alterations in glucose metabolism within activated immune/inflammatory cells are increasingly appreciated to underlie lupus and related autoimmune disorders 40-42 , but how such metabolic modifications are differentially manifested by the various cell types that contribute to lupus pathogenesis remains obscure. Here, we profiled key autoimmune populations in spontaneous lupus models and their response to glycolytic inhibition. We demonstrated a greater glycolytic dependency of GCB compared to that of Tfh. Gene expression signatures for activated B cells were significantly reduced in lupus-prone mice treated with 2DG, whereas signatures for activated T cells remained unaffected. Moreover, glycolysis pathway-related genes were highly correlated with the murine activated B cell signature, but this was weaker with the activated T cell signature. A similar correlation with the glycolysis gene signature was also observed in peripheral activated B cells from human patients with lupus. We found that murine autoreactive B cells, closely resembling GCB, were highly glycolytic and strongly dependent on glycolysis for their survival, but that Tfh were far more metabolically flexible for their energetic requirements. This glycolytic dependency translated into an exploitable weakness whereby pathogenic GCB were preferentially targetable via short-term 2DG treatment. This depletion in lupus-prone mice with advanced clinical disease led to significantly ameliorated renal damage and increased lifespan. Furthermore, we interrogated this lineage in detail, demonstrating that highly activated, BCMA-expressing GCB exhibit heightened glucose metabolism and are exquisitely vulnerable to short-term glycolytic inhibition. Targeting these cells with BCMAspecific CAR-T cells induced lifespan extension equal to that seen in mice treated with 2DG.

These findings suggest that glycolytic requirements between autoreactive GCB and Tfh differ 18 and that pathogenic GCB can be selective targets of anti-glycolytic therapy, providing a novel metabolic niche for lupus treatment.

GC formation requires sustained T cell-B cell interaction 1,9 and has long been associated with numerous autoimmune conditions 2 . Long-term treatment of pre-symptomatic Yaa DKO lupusprone mice with 2DG resulted in attenuated cellular disease phenotypes in autoreactive T cells, including Tfh, and markedly reduced GCB and plasma cells/plasmablasts. Moreover, 2DG conferred complete survival for the duration of the treatment in this acute model of lupus. It is evident that long-term glycolytic inhibition via 2DG has a non-specific suppressive effect on a number of cell types, increasing the risk of adverse effects and potentially limiting clinical translatability. Moreover, the broad effects resulting from persistent 2DG treatment in vivo make it difficult to identify a key pathogenic cell type that is primarily responsible for reduced lupus manifestations induced by glycolytic inhibition. To elucidate this cell population, we analyzed the metabolic states of distinct autoreactive T and B cells. Our data show that GCB display heightened glycolytic activity (ECAR) over Tfh. Concordantly, GCB exhibited higher Glut1 expression and a correspondingly-elevated glucose uptake compared to Tfh. Elevated Ki-67 expression in GCB over Tfh indicates that these GCB are highly proliferative, which further implicates higher glycolytic rates 13-15 , directly supporting previous inferences 41-43 . Current standard treatment for lupus is non-specific immunosuppression 40,44 , leaving patients with increased risk of infections and cancers. Preventative effect of long-term 2DG treatment on the activation of numerous autoimmune cell types was previously reported in multiple chronic lupus-prone mice 16,17 . Recent efforts have focused on targeted therapies for lupus to improve efficacy with fewer adverse effects. For example, belimumab neutralizes the B-cell survival factor TNFSF13B, thereby reducing overall B cell numbers and providing clinical benefit in human lupus 7 . Here we demonstrate that a 1-week short-term 2DG treatment selectively controls pathogenic GCB in clinically diseased Yaa DKO and NZBWF1 mice. Moreover, this GCB reduction served to reverse lupus disease pathology and significantly extend the lifespan of treated acute and chronic models of lupus. Our study also provides the mechanistic explanation by which 2DG reversed abnormal immunophenotypes in vivo by demonstrating that glycolytic inhibition caused deficiency of ATP production in pathogenic GCB and a subsequent increase in mitochondrial ROS-mediated metabolic oxidative stress resulting in apoptosis. The increased glycolytic rate in autoreactive GCB may be an adaptation to compensate for excessive ROS production 45 . Treatment with 2DG impairs mitochondrial function in autoreactive GCB that lack the metabolic flexibility to oxidize other fuels in compensation for glycolytic blockade.

Conversely, Tfh exhibit a greater degree of metabolic flexibility to compensate for fuel oxidation during glycolytic restriction, which is likely the reason that they are able to maintain ATP production, survive, and have unimpeded functionality. Furthermore, the lack of 2DG-sensitivity in GCB taken from mice treated with transient 2DG points to the targeted reduction of the pathogenic GCB populations with higher glycolytic dependency, sparing those with lesser glycolytic demand. The evidence of prolonged lifespan of lupus-prone mice with this short treatment 2DG course indicates that glycolysis-dependent autoreactive GCB are pathogenic and 2DG can target these pathogenic GCB populations, thereby avoiding a pan reduction of GCB and other immune cells that would leave the immune system compromised. In addition, the lack of a strong effect on the activated T-cell gene signature in long-term 2DG-treated mice begs the question of the primary mechanism of this glycolytic restriction; especially in view of the 20 interdependency of Tfh and GCB 2, 9 . This presents the possibility that observed Tfh number reductions stem, not from a direct effect of 2DG on Tfh, but on a lack of costimulatory signaling from autoreactive GCB after their depletion via 2DG.

Our findings that autoimmune GCB rely heavily on glucose metabolism differ from those in an immunization paradigm, in which GCB isolated from NP-CGG immunized mice exhibit very low rates of glycolysis and instead rely mainly on FAO to meet their energetic demands 29 . We observed higher OXPHOS rates in autoreactive GCB over Tfh; however, GCB from lupus mice displayed significantly increased glucose uptake, paired with elevated rates of glycolysis. One reason behind this discrepancy might be that chronic autoantigen stimulation in lupus could cause immune cells to become increasingly activated as disease progresses, versus the more acute stimulation resulting from exposure to a foreign antigen. Moreover, a previous study demonstrated that autoimmune mice treated with 2DG for 9 weeks, and immunized with NP-KLH, maintained a robust GC response against the exogenous antigen, equal to that of untreated, immunized mice, while still exhibiting a reduction in autoreactive GCB numbers 17 . This suggests that the response of GCB to foreign antigens is refractory to conditions that induce apoptosis in autoreactive GCB, indicating a major disparity in energy requirements between the two GCB types. Further studies are needed to understand this metabolic divergence between autoimmune-and immunization-induced GCB.

Our data has shown highly activated GCB (GL7 + B cells) to be the central driver of disease pathogenesis as well as the prime target of anti-glycolytic therapy. Therefore, we further asked the question of what specific cell subtype within this lineage is the main target of 2DG, and 21 found that the subset of BCMA-expressing GCB in lupus-prone mice displayed an extremely high reliance on glycolysis and sensitivity to 2DG. It has been reported that BCMA is overexpressed on both human and murine circulating B cells during lupus progression 5,46 .

Ligation of BCMA is involved in increased expression of the anti-apoptotic gene, Mcl-1, in B cells committed to differentiation into long-lived plasma cells 47,48 . This sustained survival and resistance to apoptosis are likely to be detrimental in lupus. Here, we found elevated BCMA expression on GCB in Yaa DKO mice, which was reduced after short-term 2DG treatment, was In the future, it would be beneficial to interrogate the effect of 2DG treatment in multiple tissues in lupus-prone mice during varied disease stages to understand tissue-specific effects,

BXSB.Cg-Cd8a tm1Mak Il15 tm1Imx /Dcr Yaa (Yaa DKO), NZBWF1/J (JAX#100008), BXSB.B6-Yaa + /MobJ (JAX#001925), and C57BL/6J (JAX#000664) mice were bred and housed at The Jackson Laboratory (Bar Harbor, Maine). Mice were provided 10% fat food JL Mouse Breeder/Auto (LabDiet ® 5K20) and water ad libitum; they were housed on a 14-hour light, 10hour dark cycle in a specific pathogen-free room. Symptomatic male Yaa DKO mice had visibly swollen lymph nodes and displayed elevated cellular disease biomarkers (often found at age 10-12 weeks old); whereas pre-symptomatic males (6-7 weeks old) had yet to develop these manifestations. Female littermate mice lacking the Yaa-containing Y chromosome are immunologically normal and were used as healthy controls 18 . Before experiments, mice were bled to verify disease state. Symptomatic NZBWF1 mice, were 36-40-week-old females with elevated protein urea. Experiments were performed on symptomatic mice unless otherwise stated. Mice in survival experiments were euthanized when classified as moribund, and mice euthanized before this point or for non-lupus-related issues were excluded from survival data.

For in vivo metabolic treatments, 2DG (Thermo Fisher, 6 g/L) was dissolved in drinking water and was provided to mice ad libitum. Animal studies were approved by Institutional Animal

Care and Use Committee at The Jackson Laboratory.

Single-cell suspensions from peripheral blood, spleens, or lymph nodes were blocked with anti-CD16/32 and stained with antibody cocktails on ice after red blood cell lysis. Flow data was acquired on an LSR II flow cytometer and was analyzed using FlowJo software version 10.6.1. 

Yaa DKO splenocytes or healthy human PBMCs (Precision for Medicine, MD) were cultured at 4x10 6 per ml of R10 medium (RPMI plus 10% FBS, 2 mM L-glutamine, 50 μM 2-ME, 100 U/ml penicillin, and 100 μg/ml streptomycin) with 0.5 μg/ml of anti-CD3/28 antibodies, 100 U/ml IL-2 and 30 ng/ml of the TLR7 agonist R848 (Adipogen) for 2 days to achieve cellular activation 53 .

On day 2, the cells were harvested and plated at 2.5x10 6 cells per ml in R10 medium with 0, 0.2, or 1 mM 2DG onto a 96-well plate at 200 μl per well. Cells were stained with antibody cocktail and analyzed by flow cytometry to ascertain numbers of live cells of each type by PI staining.

For measuring viability of autoreactive cells, indicated cells were sorted from the splenocytes of symptomatic Yaa DKO mice and were plated at 1x10 5 cells per ml in R10 media with or without indicated glycolysis inhibitors for 1-2 days. Cells were harvested and analyzed by flow cytometry to ascertain live cell numbers.

Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were measured in (indicating no enrichment) and +1 (indicating enrichment). Significant differences in enrichment 28 between 6wk, 10wk, and 10wk+2DG cohorts were calculated using Brown-Forsythe and Welch ANOVA tests with Dunnett T3 test for multiple comparisons with an alpha of 0.05.

Mouse gene sets used as input for GSVA are listed in Supplemental Tables 2 and 3 Cells were activated with anti-CD3/28 antibodies for 1 day and exposed to viral supernatant for 90 minutes (at 2,500 rpm, 30°C) in medium containing hexadimethrine bromide (8 μg/mL), HEPES (1 mM) and IL-2 (100 U/mL). Cells were cultured for an additional 2 days before sorting for GFP-expressing cells, which was the marker for retroviral expression.

For adoptive transfer of autoreactive GCB, symptomatic Yaa DKO mice were pre-treated with 2DG for 1 week and were then removed from treatment. Two days after removal, each mouse was injected intravenously with 1.5 x10 6 GCB or non-GCB cells sorted from untreated, symptomatic Yaa DKO mice. For adoptive transfer of APRIL-based CAR-T cells, 1x10 6 transduced T cells were injected intravenously into symptomatic Yaa DKO mice.

Comparisons for two groups were calculated by using an unpaired, two-tailed Student's t-test.

Comparisons for more than two groups were calculated using a one-way ANOVA followed by Correlations with p<0.05 were considered significant. TCA, tricarboxylic acid.

All antibodies were purchased from BioLegend or eBioscience. The following markers were used to identify each ex vivo population: B220 + CD4 -(total B cells), B220 + CD4 -CD138 -GL7 + (GCB/activated GL7 + B cells), B220 + CD4 -CD138 -GL7 -(non-GCB), B220 + CD4 -CD138 + GL7 -(plasmablasts/plasma cells), B220 -CD4 + (total CD4 T cells), B220 -CD4 + CXCR5 + PD1 + (Tfh), B220 -CD4 + CD62L -CD44 + (Teff), B220 -CD4 + CD44 -CD62L + (naïve T cells), B220 - 

Germinal center B and follicular helper T cells: siblings, cousins or just good friends?

Spontaneous germinal centers and autoimmunity

pentose phosphate pathway; OXPHOS, oxidative phosphorylation pathway; FAO, fatty acid oxidation pathway