Introduction

Circulating antibodies (Ab) are crucial mediators of protection against potentially pathogenic microbes, but Ab also can cause tissue damage or cellular dysfunction (1-5). In both adaptive and pathological settings, Ab functions depend on their heavy chain class, affinity, and concentration [reviewed in (6-8)]. Increased affinity after initial and repeated immunizations can be selected in several ways, one of which involves processes that occur in the micro-anatomy of germinal centers (also known as secondary follicles) [reviewed in (9-11)]. Progeny cells that survive a gauntlet of programmed death in selecting for continued clonal proliferation ultimately can generate either memory B cells (MBC) or an end-stage descendant of the B lineage, the plasma cell (PC) (11-15). The magnitude of PC output (the numbers generated after B cell activation) and their rates of immunoglobulin secretion (15-18) determine the net output and steady-state concentrations of Ab. Accordingly, understanding mechanisms that regulate the qualities or quantities of Ab, and that reduce or enhance the efficiency of generating PC, are important goals.

To elicit protective Ab, immune exposures such as vaccination activate B cell proliferation, and some progeny of such activated B cells develop into Ab-secreting PC (19-21). Specificity is fostered by a reliance on signal initiation through the BCR, the clonal Ag receptor on mature B lymphocytes (22-24). The capacity to be activated, survival efficiency, rates of proliferation, and propensity to yield GC B cells or PC depend on an array of other signal-initiating receptors, among them CD19, co-stimulators such as CD40 [reviewed in (25, 26)], the death receptor Fas (26), pattern recognition receptors in the TLR family [(27); reviewed in (28, 29)] and cytokine receptors (30-34). One consequence of activation is that B lymphoblasts dramatically increase expression of an array of nutrient transporters that include those for amino acids and for glucose [(35, 36); reviewed in (37)]. Beyond this general picture, however, the effects or outputs of receptors likely are scalar and conditional, i.e., dependent on other signals. For instance, antigen with a high affinity and high valency of the BCR epitope signals differently from a monovalent, lower-affinity antigen [(38-43), reviewed in (44))], and may influence generation of PC. Moreover, increased expression of TLR7 or type 1 interferon (IFN) receptor stimulation each reduces the capacity of B cells to maintain self-tolerance (45-48), presumably involving changes in the quantitative balance among signal transducers and transcription factors. At a later stage, the IL-21 receptor is crucial for PC differentiation (33). Moreover, IL-21 has recently been reported to promote affinity maturation of GC-derived PC (49). Little is known about the relationships between signal strength and metabolism. In vitro experiments showed that hypoxia or genetically-mediated stabilization of hypoxia-inducible transcription factors (HIFs) blunted activation-induced increases in amino acid uptake and transporter gene expression as well as GC and Ab responses (36, 50). However, it remains unclear if metabolism can modulate signaling.

Nonetheless, a substantial and increasing body of evidence indicates that nutrient supply, uptake, and intermediary metabolism influence the properties of adaptive immunity, in part by their effects on differentiation and function of cells that mediate host responses [reviewed in (31, 51-53)]. The bulk of evidence is from analyses of the differentiation and functions of T cells, macrophages, and dendritic cells and documents condition- and cell type-specificity of findings. Nonetheless, several analyses point to important effects on B cells as well. In addition to a function during B cell development, expression of the glucose transporter GLUT1 on mature B cells supported the initial GC size, the Ag-specific Ab responses derived from B cells, and affinity maturation (35, 54, 55).

Among nutrients, only the distribution and sufficiency of oxygen are readily measured, but such analyses consistently reveal evidence of functional hypoxia in follicles, most accentuated in GC (36, 50, 56, 57). Consistent with and potentially as a result of the hypoxia, GC B cells can exhibit higher rates of glucose uptake with several-fold increased glucose oxidation as well as utilization in the pentose phosphate pathway (36, 54-56). Moreover, the ability to feed intermediates into one-carbon metabolism via increased expression of the enzyme phosphoglycerate dehydrogenase was reported to be crucial for germinal center size and normal degrees of high-affinity antibody production (58). Cell type- and stage-specific inactivation of Slc2a1, the gene that encodes GLUT1, indicates that the increased glucose uptake is crucial for both GC homeostasis and the capacity of B cells to yield antibody-secreting PC (54, 55). Metabolomic analyses suggested that increases in anaplerotic conversion and usage of amino acids may be evoked in GLUT1-deficient B cells as a potential compensatory mechanism (55). In parallel, however, a remarkable degree of metabolic flexibility was suggested by the in vitro finding that although glucose is absolutely essential for PC development and Ab production in the absence of other hexoses, mannose and galactose together could substitute for glucose (54). Thus, the reduced influx and availability of glucose reduced Ab responses and affinity maturation, but the decrease in these outputs fell short of what likely could be therapeutically meaningful. Moreover, the importance of feeding the increase in B cell-intrinsic generation of serine with diversion of 3-phosphoglyerate away from generating pyruvate raised the question whether the capacity of mitochondria to use this latter metabolite was important in promoting greater quality and quantity of Ab after immunization.

To explore these issues, we used both immunizations and in vitro analyses that leveraged genetic and pharmacological approaches to test the importance of glutaminolysis - on its own or in combination with a second pathway - in Ab responses. Glutamine concentrations in vitro were found to be crucial for class switching as well as PC and Ab production, and B cell type-restricted inactivation of the Gls gene, which encodes the glutaminase expressed in post-natal cells outside the liver, impaired the response to ovalbumin in one immunization model. Surprisingly, however, the anti-NP response in a hapten-carrier model was unaffected. However, a synergistic impact resulted when GLS depletion was combined with targeted loss-of-function for either GLUT1 (Slc2a1) or an essential subunit of a mitochondrial pyruvate import carrier [Mpc2, (59, 60)]. Thus, glutaminolysis promoted the anti-NP response only if glucose intake or pyruvate flux was impaired, revealing one limit imposed on the metabolic flexibility of B cells in their mediation of Ab responses. The combined restriction of two pathways that feed mitochondrial metabolism represented what can be termed "synthetic auxotrophy" rather than synthetic lethality (61, 62), in that effects on differentiation were observed independent from death or proliferation. Gene expression signatures in GC B cells highlighted a requirement for GLS in metabolic reprogramming, i.e., both amplification of glucose-stimulated proton secretion and a major increase in mitochondrial respiration between the first and second days after B cell activation. Surprisingly, MPC2 and GLS were found to promote not only IL-21-stimulated STAT3 activation but also an interferon response signature in addition to promoting increased respiration. Together, the results reveal that an interplay between two metabolic pathways unexpectedly converged on programming not only IL-21R activation of STAT3, PC differentiation and high-affinity Ab responses, but also the tyrosyl and serine phosphorylation of STAT1 induced by type 1 and type 2 IFNs. These findings have important implications for diseases, such as systemic lupus erythematosus, in which B cells and their response to IFN are crucial elements of pathogenesis.

Results

Alpha ketoglutarate can alter threshold glutamine concentration requirements for B cell responses

B lymphoblasts subjected to persistent stabilization of the hypoxia-inducible transcription factors took up amino acids at lower rates in vitro and exhibited decreased mTORC1 activity (36). mTORC1 promotes Ab class switching by a mechanism beyond effects on the rate of proliferation (36, 63, 64), and extracellular glutamine influences class switch recombination in B cells through a mechanism involving the action of mTORC1 on proteins regulating translation (65). Other work has provided evidence that plasmablasts developing in the setting of plasmodial infection are glutamine avid (66). The admixture of red and white pulp in spleen preclude accurate measurement of amino acid concentrations surrounding B cells in lymphoid follicles. Instead, we adapted a technique used for tumor masses (67, 68) to measure glutamine in the fluid eluted from transected lymph nodes, revealing a strikingly lower concentration than in plasma (Fig 1a). This apparent glutamine gradient was not a universal feature of the lymph node, in that neither leucine nor pyruvate differed substantially between plasma and interstitium (Fig 1a). To test the dependence of B cells on extracellular glutamine concentration in vitro, we measured both class switching and plasma cell differentiation. Titrations of glutamine in media showed that PC differentiation as well as the Ab class switch to IgG1 were impaired, but with different dose-response curves for the inflection point at which impairment was observed (addition of 0.2 mM for PC, at which the %IgG1⁺ was not affected) (Fig 1b; Supplemental Fig 1a). Although not pursued herein, it is intriguing that some aspects of the immune micro-environment appear to support robust switching and PC generation at a glutamine level that would be insufficient in vitro, suggesting that the required concentration is affected by the signals or nutrients in vivo.

Figure 1

Among uses of glutamine after its uptake into cells, glutaminolysis is the first step in generation of α-ketoglutarate (αKG) from the amino acid for usage in the Krebs (TCA) cycle (69). To screen whether this process might be of functional significance in the molecular regulation of the class-switch or PC differentiation, and to test if the observed effects were purely division-related, we used dye partitioning in vitro combined with addition of dimethyl-ketoglutarate (DMK), a cell-permeable molecule that can bypass glutaminolysis yet supply αKG. The glutamine requirement for switching to IgG1 was almost completely substituted by DMK. Moreover, development of CD138⁺ cells involved a process beyond just the proliferative effect, and DMK in the same cultures effected a definite but incomplete reversion of the glutamine requirement (Fig 1c; Supplemental Fig. 1b, c). These findings suggest that although other mechanisms may also apply, the need for glutamine involves a capacity to supply sufficient αKG to the B lymphoblasts.

Glutaminolysis in B cells promotes an anti-ovalbumin Ab response

Glutamine can be taken up via many distinct transporters, and reduction in one can be compensated by increases in others (68). However, the in vitro data supported testing to what extent immunization-induced Ab responses are affected by glutaminolysis. To bypass potential developmental effects in the B lineage as well as the complex roles of glutamine in T cells and other cell types (70-72), we introgressed a transgene that encodes B lineage-specific expression of a chemically activated recombinase onto a Gls f/f background (72, 73). Inactivation of the Gls gene in mature B cells (Fig 1d; Supplemental Fig. 1d), hereafter termed Gls iB-Δ/Δ) could then be initiated at different times after establishment of a normal pre-immune B lineage. Mice (all huCD20-CreER^T2 +/-, and Gls +/+ or f/f) in which tamoxifen was administered prior to immunization were sensitized to ovalbumin followed by antigen (Ag) challenge in the airways. Lack of GLS (Supplemental Fig. 1d) reduced class-switched GC B cells in lymph nodes draining the lung (Fig 1e) as well as the populations of ovalbumin-specific Ab-secreting cells (ASCs) in the lung and bone marrow (Fig 1f-h). In line with these data, anti-ovalbumin IgG1 and IgM concentrations in sera of Gls iB-Δ/Δ mice were ∼0.25x those of CreER^T2 controls (Fig 1i; Supplemental Fig. 1e). We conclude that B cell glutaminolysis can promote the development of Ag-specific ASCs and an anti-protein Ab response.

Glutaminolysis is dispensable in an anti-NP response yet synergizes with other metabolic needs

Mitochondrial oxidative metabolism appears to promote affinity maturation in the classic hapten-carrier model of an Ab response specific for nitrophenolic (NP) adducts (74-76). In contrast to the impact of Gls inactivation on an anti-protein response (Figs 1, 2a-c), the anti-NP Ab response and affinity maturation were unaffected (Fig 2a, d-i). However, inducing B cell-specific inactivation of Gls synergized with loss of GLUT1. Thus, the combination of inducible B cell-specific gene targeting (iB) that restricted glucose import into B cells (35, 54, 55) with GLS depletion caused dramatically greater decreases in anti-Ova and anti-NP ASCs than iB inactivation of Slc2a1 alone (Fig. 2b-f). Combined Gls and Slc2a1 inactivation had a relatively modest impact on IgM anti-NP compared to iB-Slc2a1 Δ/Δ samples (Fig 2g). In contrast, a synergism was evident in the effects on circulating Ag-specific IgG1, for which iB-inactivation of both Gls and Slc2a1 dramatically compromised the humoral response, whereas GLUT1 depletion reduced anti-NP Ab to 1/2 - 1/4 levels of WT controls and a single loss-of-function mutation for Gls had no effect (Fig 2h-i). The synergism was most striking for high-affinity IgG1 (Fig 2i), though it also impacted all-affinity IgG1 (Fig 2h).

Figure 2

Glucose oxidation rates are several-fold higher in GC B cells than in their naive counterparts (54). However, the effects of eliminating GLUT1 on a GLS-deficient background could be due to glucose usage in serine generation (58), the pentose phosphate pathway (54), or glycosylation (60) rather than from reduced pyruvate flow into mitochondrial metabolism. Directly to interfere with pyruvate import into mitochondria for generation of acetyl-CoA, we used a previously characterized allele for inducible inactivation of Mpc2, a gene encoding an essential subunit of the mitochondrial pyruvate channel (Fig 3a) (59, 60). Levels of class-switched anti-NP antibodies in sera after the second immunization were modestly reduced after inducible Mpc2 targeting in B cells (Fig 3b, c), as was affinity maturation (Fig 3d). Whereas GLS depletion had no effect on its own, the combined loss-of-function synergized and dramatically reduced Ag-specific IgM (Supplemental Fig. 2a) and responses of several switched isotypes (Fig 3b, c). Moreover, affinity maturation was reduced further (Fig 3d), and high-affinity IgG2c anti-NP in mice with doubly deficient B cells was ∼ 0.02x controls while the all-affinity was ∼ 1/8^th of the control values (Fig. 3c). B cells with compound loss-of-function yielded dramatically fewer cells secreting high-affinity anti-NP Ab (Fig. 3e-g), whereas for GC- and memory-phenotype B cells (GCB, MBC) the magnitude of decrease was modest (∼ 1/2 control numbers; Supplemental Fig 2b-d) in parallel with decreased BrdU uptake into GC B cells (Supplemental Fig. 2e). In contrast to our findings with conditional targeting of Slc2a1 in B cells, in which a substantial degree of counterselection was evident (54), the mRNA encoding both Gls and Mpc2 were far lower in flow-sorted GC B cells (Supplemental Fig 2f, g).

Figure 3

Most of the splenic ASC pool at 7 d after a recall challenge and high-affinity IgG anti-NP Ab in sera is expected to derive from re-activation of memory B cells (MBCs) that directly differentiated into PCs after the second immunization (14, 77, 78; reviewed in 11, 15). We tested how the loss-of-function interventions impact the recall process by performance of a priming immunization, tamoxifen treatments only several weeks thereafter, followed by the second immunization and harvest (Fig 3h). In this setting where the primary response from a pre-immune repertoire was unperturbed, a synergistic interplay of Gls and Mpc2 was again observed (Fig 3i; Supplemental Fig. 3a-h). Specifically, there were substantially decreased frequencies of anti-NP ASCs (Supplemental Fig. 3g, h). Moreover, high-affinity anti-NP Ab of several switched isotypes were drastically reduced: IgG1 to < 1/4 of control values (Fig 3i), with IgG2c and IgA almost undetectable (Supplemental Fig 3a-c). Since high-affinity, class-switched (IgG or IgA) Ab are undetectable 1 wk after a primary immunization, we infer from these results that the rapid generation of ASCs from memory B cells required glutaminolysis and pyruvate import. Together, the findings reveal a cell-intrinsic process that promotes both affinity selection and differentiation toward a PC fate through the collaboration of pyruvate influx rate and glutaminolysis.

Synthetic auxotrophy of progressively increased respiration and PC differentiation

Glutaminolysis and mitochondrial pyruvate import each can be inhibited by a well-characterized pharmaceutical compound - CB839 (72) and UK5099 (59, 60), respectively. We tested the effect of these agents - in combination, or each one singly-on physiology and differentiation of B cells activated via CD40. CB839 reduced the production of CD138⁺ progeny (Fig 4a, b), ASC (Fig 4c, d) and Ab (Fig 4e). Consistent with its ability to inhibit glutaminase, this agent decreased intracellular glutamate and the calculated glutaminolysis rate while increasing the glutamine pool (Supplemental Fig. 4a-c, e). On its own, UK5099 had no effect on development of CD138⁺ cells, yet addition of UK5099 to CB839 substantially intensified the inhibition of ASC generation and Ab production. Of note, UK5099 reduced neither glutamine nor glutamate, yet substantially reduced intracellular alanine, which is generated in mitochondria by glutamate-pyruvate transaminase 2 (GPT2) conversion of glutamate and pyruvate to alanine and αKG (Supplemental Fig. 4b, d). Combining UK5099 with CB839 trended toward accentuation of the glutamate reduction caused by the glutaminase inhibitor on its own (Supplemental Fig. 4b, c). Further, the frequencies of ASCs and spot sizes were substantially reduced when metabolism was inhibited only during the differentiation phase, followed by washout of the inhibitors (Supplemental Fig 5a, c). These data indicate that glutaminolysis and mitochondrial pyruvate import promote the differentiation process, but results from applying inhibitors only during the ELISpot culture, i.e., after 4 d of differentiation in vitro, suggest that glutaminolysis also non-redundantly supports the synthesis and/or secretion of Ab (Supplemental Fig 5b, d).

Figure 4

Progression from B cell activation to the differentiation of plasmablasts and plasma cells requires several rounds of cell cycling and a division-counting mechanism (79). When the frequency of CD138⁺ progeny was measured as a function of division number, the results showed that CB839 reduced the capacity to become CD138-positive at equal divisions (Supplemental Fig 6a-d). Moreover, whereas UK5099 on its own did not reduce differentiatino efficiency at equal divisions, its addition intensified the inhibitory effect of CB839 (Supplemental Fig 6a-d). Recent work with profound B cell-depletion therapy in chronic autoimmune disease indicates that disease activity depends on continual production of new auto-Ab-producing cells [80, reviewed in (81)]. Hydroxychloroquine (HCQ) is the standard of care in systemic lupus erythematosus; while effective, it incompletely reduces pathogenic auto-Ab concentrations and disease activity, and often is not sufficient (82). We explored the interplay between HCQ and CB839 (glutaminase inhibition), alone or along with UK5099. Strikingly, whereas HCQ alone slightly increased the %CD138⁺ at each division and overall, addition of HCQ to metabolic inhibition caused more substantial reductions in PC differentiation (Supplemental Fig 6a-d).

Levels of Prdm1 mRNA, which encodes a transcription factor that determines PC fate and identity [reviewed in (9, 83)], were lower in inhibitor-treated cells at a time in the cultures when the CD138⁺ population just started to develop (Fig 4f). This finding is consistent with an effect on differentiation beyond simple division-counting and with the pattern of results with metabolic inhibition. Moreover, analyses of cell-type modules in RNA-seq data from B lymphoblast cultures at this inflection point yielded evidence that the main systemic impact of combined inhibition was on plasmablast - plasma cell gene expression programs (Fig 4g, h). We further found that DMK, the cell-permeable αKG precursor, partially reversed the inhibitory effects of CD839 and UK5099 on B cell differentiation to a CD138⁺ plasmablast/PC status (Fig. 4i). While not the entire mechanism, these findings indicate that αKG generation via glutaminolysis promotes PC differentiation.

The serological data highlight an impact on affinity maturation of class-switched Ab after a secondary immunization. T cell help drives greater proliferation of both extrafollicular- and GC- B cell-derived plasmablasts that bore higher-affinity BCR (49, 84), but the findings (Fig 3b-d; Supplemental Fig 2c, e) demonstrate effects on GC B cells. Since in vitro systems cannot faithfully recapitulate GC B cell physiology, we performed RNA-seq analyses of flow-purified GC B cells from the mice with induced B cell type-specific gene inactivation and matched controls (CreER^T2 transgenics injected with tamoxifen). The analyses consistently showed major reductions in the mRNA encoded by the targeted genes (Mpc2; Gls), indicating both that deletion of the flox alleles was efficient and that the GC were not dominated by selective outgrowth of the B cells that retained functional genes (Supplemental Fig 2f, g). Gene set enrichment analyses (GSEA) of the impact of GLS on the relative RNA levels in GC B cells highlighted modules for the interconnected functions of Myc-regulated genes, those connected to mTOR signaling and oxidative metabolism (oxidative phosphorylation; fatty acid oxidation) (Fig 4j-l; Supplemental Fig 7). Moreover, the abnormal metabolism led to reduced expression of RNAs linked to proliferation (E2F-responsive genes and those associated with the cell cycle G2-M transition) (Supplemental Fig 7a). These effects could be discerned when comparing Gls Δ/Δ B cells to CreER^T2, +/+ controls but effects were heightened when Mpc2 inactivation was compared to Mpc2 Δ/Δ, Gls Δ/Δ samples or in comparing doubly-deficient B cells to WT controls (Supplemental Fig 7).

The gene modules pointed to proliferation, c-Myc (a central mediator of metabolic reprogramming), and oxidative metabolism. Oxidative metabolism is vital for PC differentiation (85). Accordingly, we measured respiration by B lymphoblasts of each genotype on two successive days after activation and culture in conditions promoting development into ASC. Of note, rates of mitochondrion-dependent oxygen consumption by WT cells quadrupled between days 1 and 2 after mitogenic stimulation, (Fig 5a-d). Consistent with the GSEA data, the increased respiration was blunted in GLS-deficient B cells (Fig 5a, b). In addition, the inactivation of Mpc2, while not affecting respiration on its own, further reduced both the absolute rate and the capacity to increase this core mitochondrial function during the second day when GLS was deficient (Fig 5a, b). To explore if these effects were attributable either to abnormal naive B cells prior to stimulation, or simply to a failure to provide sufficient substrates to the mitochondria, we added CB839 or UK5099 to cultures of normal B cells after activation but performed the metabolic flux assays in the absence of inhibitors. Strikingly, the impacts on respiration caused by pharmacological inhibition initiated at the time of B cell activation matched those obtained using genetic loss-of-function (Fig 5c, d). Further consistent with altered mitochondrial metabolism, the inhibition of glutaminolysis also impacted the acetyl-carnitine /ratio that is central to the provision of substrates for fatty acid oxidation, with reduced L-carnitine and increased acetyl-carnitine as well as computationally inferred rates of fatty acid oxidation (Supplemental Fig. 4g-j).

Figure 5

Rates of extracellular acidification rates (ECAR) were measured to test if B lymphoblasts compensated for the reduced respiration by increasing "aerobic glycolysis" - the hallmark of which would be to generate and excrete lactate from pyruvate. Targeting glutaminolysis led to striking decreases in ECAR that were even greater when in concert with interference with the MPC (Fig 5e, f). These decreases indicate that the interference with glutaminolysis broadly reprograms metabolism in the activated B cells. Although activated B cells perform fatty acid oxidation [(104); reviewed in (37)] the inhibitors decreased steady-state ATP concentrations in the CD40-activated lymphoblasts (Fig 5g). Of note, the main cause of these decreases appeared to be interference with the capacity of B cells to increase their respiratory capacity during growth (Fig 5h) rather than the modest impact of acute inhibition of glutaminolysis and pyruvate import in blasts cultured in the absence of inhibitors during 2 d prior to adding inhibitors (Fig 5i). The reduced respiration, lower signals from TMRE staining (Fig 5j), together with normal signal after MitoTracker Green staining and qPCR for mitochondrial DNA content (Supplemental Fig 6e, f), indicate that decreased respiration was linked to a reduction in mitochondrial membrane potential rather than a failure to increase mitochondrial mass or DNA. This decrease in turn was associated with an increase in reactive oxygen species (ROS) (Fig 5k).

Metabolic regulation of interferon) cytokine receptor responsiveness

To gain further insight into molecular consequences of synthetic auxotrophy, we analyzed the RNA-Seq data to determine how Mpc2 gene inactivation affected GLS-depleted GC B cells. Surprisingly, the RNA-seq data indicated that a Jak-STAT3 signature was reduced not only in GC B cells deficient GLS and MPC2 but also in the comparison of Mpc2 Δ/Δ to combined loss-of-function (Fig 6a; Supplemental Fig 7a). IL-21R, acting via STAT3, is crucial for PC differentiation (33, 34, 86, 87). Cytokine receptors activate STAT3 via Jak1-mediated tyrosine phosphorylation, but ROS can alter STAT3-dependent gene transcription (88-90). We found that the tyrosyl phosphorylation of Jak1 and STAT3 stimulated by IL-21 treatment of activated B cells was attenuated in cells cultured in CB839 and UK5099 (Fig 6b). Unexpectedly, computational algorithms identified the interferon (IFN)-stimulated gene signature as the most prominently impaired pathway in the Mpc2, Gls1 double-deficient GC B cells (Fig 6c). Thus, the effects of MPC2 in the setting of GLS depletion were to promote interferon response signatures and Jak-STAT signaling as well as suites of RNA linked to respiration (Fig 6d; Supplemental Fig 7a, d). Among processes that could lead to such a strong effect on mRNA levels, a straightforward possibility was that the metabolism in B cells affects their IFN-elicited signal transduction. Strikingly, IFN-induced phosphorylation of the conserved tyrosine was attenuated when activated cells were treated with CB839 and UK5099 prior to stimulation with type 1 or 2 IFN (IFN-β and -γ, respectively) (Fig 6e, f). This effect was observed in B cells three days after initial activation via CD40 crosslinking, and inhibition was observed when the pharmacological agents were present only for the final 16 h, or when anti-CD40-activated blasts were analyzed the day after activation (Fig. 6e, f). Decreased IFN-stimulated phospho-STAT1 was also observed in comparisons between WT and Gls^Δ/Δ Mpc2^Δ/Δ B cells from tamoxifen-treated mice (Supplemental Fig. 8a-e). One consequence of STAT1 nuclear translocation and chromatin association is that the C-terminal transcription activation domain (TAD) can become phosphorylated at S727, an event that depends on the Jak/Tyk2 tyrosyl kinase action (91). Of note, IFN-induced STAT1(Y727) phosphorylation also was attenuated when GLS and MPC2 were impeded in activated B cells (Fig. 6e, f). We conclude that altered intermediary metabolism in B cells can, within less than a day, change IFN receptor induction of phospho-STAT1.

Figure 6

Finally, we tested models of mechanisms that could mediate the observed ability of metabolism (i.e., glutaminolysis and mitochondrial pyruvate import) to increase STAT1 phosphorylation in activated B cells. Flow cytometry results indicated that a simple model in which surface expression of IFN receptors was reduced is unlikely (Supplemental Fig 8f), suggesting instead an impairment of receptor signaling to STAT1. One observed effect of the combined metabolic reprogramming was to increase steady-state ROS in B cells (Fig 5k). Of note, ROS can reduce function of complex I of the electron transport chain (ETC) by modifying reactive sulfhydryl groups in the complex, and elimination of complex I from the mitochondrial electron transport chain (ETC) decreased IFN-γ activation of STAT1 in a genome-wide loss-of-function screen with an immortalized macrophage-like cell line (92). The absence of complex I was mimicked by the ETC inhibitor rotenone (92). Longer rotenone treatments were highly toxic to B cells, but we found that p-STAT1(Y701) induction in activated B cells was dramatically decreased by a remarkably short (45 min) exposure to this agent prior to IFN stimulation (Fig. 6g, h). Metformin also inhibits complex I, albeit at higher concentrations than apply clinically (93, 94). When tested with activated B cells, this pharmaceutical agent also decreased p-STAT1 elicited by IFN stimulation (Fig. 6g, h).

Dysfunction or chemical inhibition of ETC complex I causes increased mitochondrial production and export of ROS (95), so it was possible that these results were in part attributable to elevated ROS. Using an approach that had provided evidence of ROS inhibiting STING-dependent IFN-β production (96), we increased ROS in activated B cells by applying either menadione or H₂O₂ prior to interferon treatment. Menadione substantially reduced induction of P-STAT1 by receptors for both IFN-β and -γ, which was almost completely abrogated by a short-term pre-treatment with H₂O₂ (Fig 6i, j). We conclude that increased steady-state ROS and mtROS in B cells, which arise when GLS and MPC function are impaired, contribute to an unexpected defect in IFN-R signaling to STAT1. All together, the results reveal that an interplay between two metabolic pathways progressively reduced the expansion of mitochondrial respiration in B cells. A key functional effect of the combined impairment of two metabolic pathways is the establishment of a block to PC differentiation and high-affinity Ab responses. Unexpectedly, however, the mitochondrial dysfunction and elevated ROS also converged on IFN-receptor signaling and the programming interferon-response functions.

Discussion

We and others have published examples of how metabolic flexibility in B lineage cells may allow limiting or negating the effects of loss-of-function mutations in key steps of intermediary metabolism (54, 55, 97). These findings raise questions as to forms of intervention that could interfere with the ongoing production of plasma cells secreting pathogenic auto-Ab or, conversely, which pathways might need to collaborate for promoting full-strength Ab responses. The work presented here supports three central insights. First, we have identified conditions under which the anti-NP response is relatively glutaminase-independent, both in the primary phase and upon recall stimulation of memory B cells. Despite this apparent independence, this enzyme central to processing of glutamine and to anaplerosis takes on a vital role if mitochondrial pyruvate import is impeded. A related point is that the findings underscore the fallacy in stating conclusions as if universally applicable, inasmuch as the effect of B cell specific loss-of-function for GLS depended on the nature of the immunization. Thus, the anti-ovalbumin response derived from GLS-depleted B cells was substantially reduced even though the anti-NP responses after multivalent hapten-carrier immunization with NP were not similarly reduced with isolated Gls gene inactivation. Mechanistically, the findings revealed that glutaminase activity promotes a progressive increase in mitochondrial respiration in a manner enhanced by pyruvate import, an alternative pathway for providing substrates to the TCA (Krebs) Cycle. A third major finding is that the efficiency of coupling IFN receptors to the tyrosyl and serine phosphorylation of STAT1 - and IL-21 stimulation to STAT3 - depends on the programming of intermediary metabolism in a manner linked to ROS homeostasis. Taken all together, the findings reveal not only new insights into how intermediary metabolism can modulate differentiation and receptor signaling, but suggest the potential for new interventions in autoimmune diseases driven by interferon effects and auto-Ab.

A self-reinforcing network of increased interferon production and pathological levels and persistence of auto-Ab production is characteristic of several autoimmune conditions of humans. Systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) are most common and studied, but this pairing also is involved in inflammatory myopathies and primary Sjogren’s syndrome (98, 99). Effectiveness of therapeutic mAb for B cell depletion is best established in RA; unimpressive results in SLE raised concerns that disease might be perpetuated by long-lived PC that secrete auto-Ab, do not need BAFF, and are CD20^neg. However, dramatic results of testing CD19-directed chimeric Ag receptor T cells in patients with SLE and myositis showed that targeting B cells reduced clinical disease, IFN levels and those of pathologic auto-Ab [80, reviewed in (81)]. These effects imply that ongoing production of PC from B cells maintains IFN production and disease activity. Cost, production complexity, and risk profile mean that a need to identify other approaches that restrict the ongoing differentiation of B cells into PC remains, despite the dramatic results in patients with severe disease. The capacity of CB839 to collaborate with UK5099 to suppress both PC differentiation and responsiveness to IFNs suggests that this agent and glutaminase-specific inhibition in general merit further analysis. Since CB839 appeared safe in patients during phase 2 clinical testing (100), our finding of an interplay between the standard-of-care agent hydroxychloroquine and CB839 in the inhibition of plasma cell development in vitro similarly militates in favor of translational investigation. Which B cells are major sources of the PC that secrete pathological auto-Ab, and the extent to which SLE susceptibility might involve a distinct relationship between intermediary metabolism and PC differentiation, are important issues beyond the scope of the present study. Our data indicate that inhibition of the pathways impacted memory B cells that rapidly differentiated after a secondary challenge, but do not specifically test the atypical MBC that are implicated in SLE pathogenesis (101).

Another finding of the work bears both on consideration of the intersection of B cell metabolism and function after immunization, and perhaps on the relationship of our findings to SLE. Our data show that although the Ag-specific Ab response elicited by a protein, ovalbumin, was reduced by B cell-specific depletion of GLS in an allergic lung inflammation model of immunization, the anti-NP response was normal. Our finding that the metabolic requirements for Ab responses differed according to Ag or type of immunization raises questions that merit being worked out in future research. A key point is that immunization to elicit anti-NP responses usually is carried out with heavily haptenated carriers, i.e., many NP-groups on each protein molecule. In sharp contrast, a particular epitope in the carrier protein (ovalbumin, KLH, gamma globulin) will be monovalent unless part of an aggregate. Valency and the extent of BCR cross-linking influence signaling by the BCR, which in turn can affect metabolic regulators such as mTORC1 and c-Myc. A related possibility derives from the characteristics of the anti-NP response, which is dominated by a limited Ig heavy chain repertoire (102). Thus, intrinsic biochemical characteristics of V_H125:V_Lλ BCR interactions with NP might lead to programming metabolic requirements that differ from what is elicited by the BCR repertoire responding to ovalbumin. The issue of valency has a bearing on the pathological Ab in auto-immunity in that so many auto-Ab of SLE are directed against multivalent Ag such as dsDNA and ribonucleoprotein complexes (3-5, 103). Finally, the specific immunization regimen might affect the results. General issues of valency in engagement of BCR at the cell surface remain incompletely clarified despite years of papers, so extensive further studies will be needed to understand better the relationship to a specific step in intermediary metabolism. Nonetheless, a striking finding herein is that a vital requirement for glutaminolysis is cryptic in the anti-NP response until uncovered by reduced import rates of either glucose into cells or pyruvate into mitochondria.

Our findings connect to questions of metabolic flexibility, which is a central issue for the potential translatability or implications of intermediary metabolism in targeting immune responses or auto-immunity. Targeting a particular process is less likely to be efficacious to the extent that GC B cells, or B cells in general, can use alternatives and achieve normal responses or outputs. For instance, although there is evidence that the majority of energy generation by activated and GC B cells involves fatty acid (FA) oxidation, most of it in mitochondria (97, 104), elimination of a key step in mitochondrial acquisition and use of FAs had no impact on Ab responses or discernible reduction in GC or affinity maturation (97). Alternatively, restricting cell-intrinsic generation of serine from a glycolytic intermediate represents one potential point of vulnerability that was not compensated by flexibility (58). Depletion of a major glucose transporter - and inferentially of reduced supply of glucose for some combination of NADPH regeneration from NADP, serine biosynthesis, oxidation, and perhaps other use of glucose - also reduced GC size, Ab responses, and affinity maturation (54, 55). However, the effect magnitude (reduction to ∼1/4 normal) appeared less promising from the standpoint of potential treatment of autoimmunity. Metabolomic data suggested that increased anaplerosis was a compensatory mechanism that might have limited the severity of the defect in ASC generation and Ab responses of GLUT1-depleted B cells (55). Our findings support this inference in that the defects with Slc2a1 Δ/Δ or Mpc2 Δ/Δ B cells were each dramatically amplified by impedance of glutaminolysis. The size of the differences, the impact on the IgG2(a/)c isotype that is particularly important in SLE pathogenesis [reviewed in (105)], and the vital functions of STAT1 induction by IFN receptors in autoimmune models (106, 107) indicate that this synthetic auxotrophy merits translational consideration.

Materials and methods

Mice and immunizations

All animal protocols - reviewed and approved by Vanderbilt University Institutional Animal Care and Use Committee - complied with the National Institutes of Health guidelines for the Care and Use of Experimental Animals. Mice were housed in specified pathogen-free conditions. Both male and female mice, aged 6-10 weeks, were used; sex-specific subgroup analyses did not reveal any significant differences and yielded the same conclusions. To enable tamoxifen-induced, B cell type-specific inactivation of GLS, GLUT1, and/or MPC2 coding potential, Gls^f/f (72) Slc2a1^f/f (35), and Mpc2^f/f mice (60) were crossed with huCD20-CreER^T2 (73) transgenic mice; for confirmatory in vitro work with purified B cells, some experiments used Rosa26-CreER^T2 [as in (36)]. Tamoxifen was administered as reported previously (36, 54). To control for potential Cre toxicity in B cell responses or GC B cells, age-matched Gls^+/+, Slc2a1^+/+, and Mpc2^+/+, huCD20-CreER^T2 mice were co-housed with huCD20-CreER^T2 mice that had conditional alleles [Gls^f/f, Slc2a1^f/f, and/or Mpc2^f/f], alone or in combination, and used as wild-type controls.

Mice were immunized, or immunized and boosted, by one or two intraperitoneal (i.p.) injections of 100 μg of 4-hydroxy-3-nitrophenylacetyl hapten (NP) conjugated to ovalbumin (NP-ova, Biosearch Technologies, Novato, CA), emulsified in 100 μL alum [Imject^® alum (Thermo Fisher Scientific, Pittsburgh, PA), as described previously (36, 54, 63), or after this product was discontinued, Alhydrogel 2% (InvivoGen, San Diego, CA). For mucosal challenges to elicit an anti-carrier Ab response, ovalbumin (50 μg dissolved in 20 μL PBS) was instilled intranasally (i.n.) once daily across seven consecutive days starting three weeks after i.p initial sensitization of mice with NP-ova. Mice were harvested 12 h after the final inhalation. Alternatively, mice were immunized with sheep red blood cells (SRBC) as described (36). Harvested spleens in some experiments were used to purify GC B cells for RNA-seq (detailed below). To assess proliferation rates of GC B cells in vivo, intravital BrdU incorporation measured by injecting mice with 2 mg BrdU (Sigma-Aldrich, St. Louis, MO) 16 h and 4 h before harvesting spleens from SRBC-immunized mice. Single-cell suspensions, prepared as described above, were stained for surface markers (B220, IgD, GL7, CD38; dump channel markers as specified in the Flow Cytometry section) followed by cell fixation, permeabilization, and staining BrdU-containing DNA using the APC BrdU Flow Kit (B-D Pharmingen, San Jose CA) according to manufacturer’s protocol. GC B cells were defined as B220⁺ GL7⁺ IgD^neg CD38^neg, as control analyses have shown that a CD138⁺ pre-plasma cell population is < 2% of this gate.

Flow cytometry

Fluor-conjugated mAbs were purchased from BD Pharmingen (San Jose, CA), Life Technologies, eBiosciences (San Diego, CA), or Tonbo Biosciences. For detection of GC- and memory-phenotype B cells in the spleens of immunized mice, samples were stained as previously described (22). In brief, 3 x 10⁶ splenocytes were stained with anti-B220, -GL7, -Fas, -IgD, -CD38, NP-APC and a dump channel containing anti-CD11b, -CD11c, -F4/80, -Gr-1, and viability marker (7-AAD or Ghost-Brilliant Violet 510) in 1% BSA and 0.05% sodium azide in PBS. For detection of PC in products of in vitro cultures, viable cells (gated via FSC, SSC, and Ghost e450) were stained with fluorophore-conjugated anti-CD138, -B220, -CD19, or -TACI. For flow analyses of mitochondrial and total intracellular ROS, cells (1-3 x 10⁶) were washed in PBS and stained with 40 nM MitoTracker Green, 10 nM TMRE, 5 μM MitoSOX or 1.25 μM H₂DCFDA, respectively, along with Ghost-780 in PBS (20 min at 37°C), then washed again (1% BSA in PBS) and further stained with anti-B220, -CD19, -CD138, or -TACI. Samples were analyzed using a FACS Canto flow cytometer driven by BD FACS Diva software or as part of preparative flow purification with a FACS Aria flow sorter. Data were processed using Flow-Jo software (FlowJo LLC, Ashland, OR).

Cell cultures & reagents

B cells were purified from mouse spleens using negative selection with biotinylated anti-CD90 and an iMag system (BD-Pharmingen) or positive selection with anti-mouse B220 nanobeads (Miltyeni Biotec, Auburn CA). To induce plasma cell differentiation, B cells (seeded at 5 x 10⁵ / mL) were cultured with 1 μg/mL anti-CD40 (Tonbo, San Diego CA), 10 ng/mL BAFF (AdipoGen, San Diego, CA), 10 ng/mL IL-4 (Peprotech, Rocky Hill, NJ), 5 ng/mL IL-5 (Peprotech), and, if derived from inducible gene deletion models, 50 nM 4-hydroxy-tamoxifen (4-OHT) (Sigma-Aldrich, St. Louis MO) and re-fed at day 3. When cultured 2 d or less, cells were in the same conditions after plating at 2 x 10⁶/mL. Alternatively, B cells were activated with both Fab2’ anti-mouse IgM (Southern Biotech, Birmingham, AL) as well as 1 μg/mL anti-CD40 and cultured as for anti-CD40-activated cells. lutamine supplementation analyses, with or without added DMK (5 mM) were performed using glutamine-free RPMI (Thermo Fisher, Waltham MA) supplemented with 10% dialyzed Gibco FBS (Thermo-Fisher), 100 U/mL penicillin (Invitrogen, Waltham MA), 100 μg/mL streptomycin (Invitrogen), non-essential amino acids (NEAA, Invitrogen), 10 mM HEPES (Invitrogen), 0.1 mM 2-mercaptoethanol (Sigma). To test how inhibitors affected in vitro proliferation, cytokine receptor signaling, or plasma cell differentiation, cultures were supplemented with 1 µM CB839, 10 µM UK5099, and/or 3 µM hydroxychloroquine (all from Cayman Chemical Co, Ann Arbor MI), individually or in combinations, and 1 mM DMK was used to supplement some cultures subjected to glutaminolysis inhibition by CB839. To analyze proliferation in vitro, B cells (2 x 10⁶, purified as above) were labeled with 5 μM CellTrace Violet (CTV) (Invitrogen, Waltham, MA) and then activated and cultured as above. To test the impact of ROS and electron transport chain (ETC) inhibition on activation of STAT transcription factors, cultures of B cells activated with anti-CD40, BAFF, IL-4, and IL-5 were supplemented with 0.2 mM H₂0₂ (Sigma-Aldrich), 2 µM menadione (Sigma-Aldrich), 2 mM metformin (Sigma-Aldrich), or 1 µM rotenone (Calbiochem, San Diego CA).

ELISA and ELIspot

Relative NP-specific Ab concentrations in sera, and secreted Ab in culture supernatants, were measured using capture ELISA as described previously (36, 54, 63). All- or high-affinity hapten-specific Ab were measured using plates coated with NP₂₀-BSA or NP₂-PSA (Biosearch Technologies), respectively. Ovalbumin-specific antibodies were measured using serial dilutions of sera incubated on ovalbumin (1 μg/well)-coated 96-well plates (16 h at 4° C). Total secreted Ab were captured using an anti-Ig(H+L) reagent. Captured Ab retained after rinsing were detected in isotype-specific manner using HRP-conjugated anti-IgM, -IgG1, IgG2c, or -IgE incubation and colorimetric development with Ultra TMB Substrate (Thermo Fisher Scientific).

The frequencies of antibody secreting cells (ASCs) in the bone marrow, lung, or spleen after immunization were measured as described previously (54, 63). Briefly, high protein binding plates (Corning Life Science, Corning NY) were coated with NP₂₀-BSA (for all-affinity ASCs) or NP₂-PSA (for high-affinity ASCs) (Biosearch Technologies), then seeded with splenocyte suspensions [0.5 - 1ξ10⁶ cells/well (IgM), 1 – 2ξ10⁶ cells/well (IgG1), and 2ξ10⁶ cells/well (IgG2c)], cultured in medium with 10% FBS (16 hr), and probed with biotinylated anti-IgM, anti-IgG1, or anti-IgG2c antibodies (Southern Biotechnologies, Birmingham AL). For the detection of IgM or IgG1 ASC in day 5 in vitro cultures, high protein binding plates were coated with anti-Ig(H+L) (Southern Biotechnologies), seeded with suspensions of cultured cells [20 hr, with 100, 200, and 500 cells per well (IgM) or 1000, 2000, and 5000 cells/well (IgG1)], then probed with biotinylated anti-IgM or -IgG1 antibodies (Southern Biotechnologies, Birmingham AL). ASC numbers and spot sizes were quantified using an ImmunoSpot Analyzer (Cellular Technology, Shaker Heights, OH) and the densities based on dilutions with readily resolved spots.

Nucleic acid analyses

RNA was isolated from sorted B cells or in vitro-generated B lymphoblasts using TRIzol reagent following the manufacturer’s instructions (Life Technologies, Carlsbad, CA). RNA from cultured B cells purified from TriZol and processed for bulk RNA-seq by the VANTAGE Core of the Vanderbilts. To assess the extent of the loss-of-function in B cell populations, gene expression was analyzed using quantitative real-time RT-PCR (qRT²-PCR). After cDNA synthesis by reverse transcription with an AMV Reverse Transcriptase kit (Promega, Madison, WI), amplifications were performed using SYBR green Power UP PCR master-mix (ThermoFisher Scientific) and primer pairs specific for Gls (Forward 5’-GGGAATTCACTTTTGTCACGA -3’; Reverse 5’-GACTTCACCC TTTGATCACC -3’), Mpc2 (Forward 5’-TGTTGCTGCCAAAGAAATTG-3’; Reverse 5’-AGTGGACTGAGCTGTGCTGA -3’), and b-actin (Forward 5’-GGCACCACAC CTTCTACAATG-3’; Reverse 5’-GGGGTGTTGAAGGTCTCAAAC-3’). The relative level of gene expression was calculated using the 2-DDCT method after each sample was normalized to b-actin. Libraries for low-input bulk RNA-seq analyses with flow-purified GC B cells were generated and sequenced essentially as described (108) and sequenced at the Emory Integrated Genomics Core. In brief, For each sample 1,000 cells were sorted directly into RLT buffer (79216; Qiagen) containing 1% (v/v) 2-mercaptoethanol. RNA was isolated using Zymo Quick-RNA MicroPrep Kit (11-328M; Zymo Research). Synthesis of cDNA was performed using SMART-Seq v4 Ultra Low Input RNA Kit (634894; Takara Bio) kit. Final libraries were generated using 200 pg of cDNA as input for the NexteraXT kit (Illumina, FC-131-1024) with 12 cycles of PCR amplification. Final RNA-seq libraries were quantitated by QuBit (Life Technologies, Q33231), size distributions determined by bioanalyzer (Agilent 2100), pooled at equimolar ratios, and sequenced at the Emory Genomics Core on a NovaSeq6000 using a PE100 run. The RNA-seq samples of a biologically independent replication set of flow-purified GC B cells from immunized mice representing the four different genotypes were sequenced at Novogene on a NovaSeq X Plus using a PE150 runBulk RNA-seq. RNA from cultured B cells was performed by similar methods at the Vanderbilts’ VANTAGE core.

To analyze sequencer outputs of each core, Fastq files were assessed for quality using FastQC. Adapters were trimmed using TrimGalore, followed by FastQC to validate adapter removal. hisat2 was used to align fastq files to the mm10 genome downloaded from UCSC genome browser. Qualimap was used to analyze the aligned sequences for quality, Subread’s featureCounts was used to count and normalize sequence counts in groups being compared. featureCount outputs were then used for DESeq2 analysis to identify fold change and significance of observed differences in gene expression. Normalized RNA-Seq counts were analyzed using GSEA 4.3.2 (Broad Institute) and the Hallmark gene sets collection from the Molecular Signatures Database (MSigDB) to identify changes in biological pathways. Differentially expressed genes identified by DESeq2 were analyzed using the MyGeneset tool from the Immunological Genome Project (Immgen) to identify B cell subsets associated with genes effected by intervention in metabolic pathways. The web-based gene set analysis toolkit (WebGestalt) was used for gene ontology over-representation analysis of biological processes using the protein coding genome as the reference set. Differentially expressed genes identified by DESeq2 were analyzed with WebGestalt to identify biological processes implicated by the change in gene expression.

Metabolic analyses

For proton nuclear magnetic resonance spectroscopy (¹H-NMRS) quantitation of metabolites in solution, plasma and lymph node interstitial fluids were collected and processed as previously reported (67, 68). In brief, a total of 50 μL deuterated water (D₂O) and 50 μL of 0.75% sodium 3-trimethylsilyl-2,2,3,3-tetradeuteropropionate (TSP) in D₂O were added to 500 μL of diluted biological fluids and transferred to 5-mm NMR tubes (Wilmad-LabGlass, Kingsport, TN). ¹H-NMRS spectra were acquired on an Avance III 600 MHz spectrometer equipped with a Triple Resonance CryoProbe (TCI) (Bruker) at 298 K with 7500-Hz spectral width, 32,768 time domain points, 32 scans, and a relaxation delay of 2.7 seconds. The water resonance was suppressed by a gated irradiation centered on the water frequency. The spectra were phased, baseline corrected, and referenced to TSP using Chenomx NMR Suite. Spectral assignments were based on literature values.

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using a Seahorse Bioscience XFe96 extracellular flux analyzer (Agilent Technologies, Santa Clara CA). B cells (2.5 x 10⁵) activated for 2 days with anti-CD40, BAFF, IL-4, and IL-5 were seeded per well of a Cell-Tak (5 μg/mL; Corning) coated plate. Glycolytic and mitochondrial stress tests were performed as previously described (54, 63). Basal respiration, maximum respiration, and ATP production were calculated using formulas derived from the software accompanying the Agilent Seahorse platform. Steady-state ATP concentrations in activated B cells were measured using the Promega Mitochondrial ToxGlo™ Assay (Promega, Madison WI), following the manufacturer’s protocol. B cells activated and cultured 2 d in anti-CD40, BAFF, IL-4, and IL-5 after negative selection of splenocytes with anti-Thy1.2, in the presence or absence of metabolic inhibitors (CB839, UK5099, or both). Equal numbers of viable (Trypan Blue^neg), washed twice with warm, sterile phosphate-buffered saline, were resuspended in glucose-supplemented RPMI1640, with their designated drug treatment added, and cultured (10⁵ B cells in 100 μL) 90 min at 37°C. Cytotoxicity Reagent (20 µL 5x) was added to each well followed (30 min) by quantifying fluorescence (520-520 nm Em; 485nmEx/). 100 μL of ATP Detection Reagent was then added to each well followed by orbital shaking (500rpm for 5 min). Luminescence was then measured to quantify ATP present in the B cells.

For metabolomic analyses, frozen (-70 C) cell pellets were extracted in plastic vials with 40 µL of UPLC grade isopropanol, sonicated (3 cycles, 10 seconds each), and clarified by centrifugation (20 min). Total protein content was measured using Pierce BCA Protein Assay kit (Fisher-Thermo Scientific). Supernatants were analyzed using the Biocrates MxP Quant 500 XL targeted metabolomic kit (Biocrates Life Sciences AG, Innsbruck, Austria), potentially quantitating 106 small molecules and free fatty acids in chromatography mode and 913 complex lipids in flow-injection mode (FIA-MS/MS).

Metabolite identification was based on triple quadrupole ultra-high-performance liquid chromatography tandem mass spectrometry (UHPLC-MS/MS) using a Shimadzu Nexera chromatography platform (Shimadzu Corporation, Kyoto, Japan) coupled to Sciex QTRAP 7500 mass spectrometer (AB Sciex LLC, Framingham, Massachusetts, USA). Metabolic indicators (474) were then calculated from sums or ratios of relevant metabolites according to Biocrates MetaboINDICATOR formulas (110). These indicators can be regarded as physiologically relevant measures and are statistically analyzed separately from metabolites. In addition, indicators denoted as “X synthesis” were computed as a ratio of metabolite X and its main precursors as a reflection of the inferred conversion ratio. Using solvents of LC/UHPLC-MS grade, cleared extracts were transferred onto a kit plate with pre-injected internal standards, dried, derivatized with 5% phenylisothiocyanate in pyridine, ethanol and water (1:1:1 v/v/v), and subsequent extraction with 5 mM ammonium acetate in methanol. UHPLC was performed with 0.2% formic acid in acetonitrile (organic mobile phase) and 0.2% formic acid in water (inorganic mobile phase). Flow-injection analysis was performed with methanol and Biocrates MxP Quant 500 XL additive. Sample handling, randomized with stratification prior to processing to avoid potential biases, was done on dry ice to avoid multiple freeze-thaw cycles. Plates included isopropanol blanks to calculate limits of detection, repeats of a kit quality control sample to calculate concentrations and monitor the coefficient of variation, and kit calibrators for seven-point calibrations of certain compounds. Analytes (metabolic indicators) were calculated according to Biocrates MetaboINDICATOR formulas (110). Ratios with zeros in denominator were treated as missing values that were not considered in the analysis. As additional data transformations, Box-Cox transformation with R package car (111), and Tukey’s fencing (112) were applied for metabolomic data. Furthermore, the values were standardized with respect to the DMSO group to facilitate comparison of regression coefficients in the statistical analysis.

Measurement of cytokine-induced STAT phosphorylation

To measure the levels of total and phosphorylated proteins, B cells were stimulated with anti-CD40 and BAFF, cultured in the presence or absence of inhibitors as detailed in Figure 6 and its Legend, and stimulated (25 min) with IFNβ, IFNγ, or IL-21. Total cell lysates were prepared with modified RIPA lysis buffer (50 mM Tris×Cl (pH8.0), 150 mM NaCl, 1% NP-40, 0.1% SDS, 1 mM EGTA, 1 mM Na₂VO₄, 1 mM NaF, and protease inhibitor cocktail). Proteins resolved by SDS-PAGE were transferred onto Immobilon^TM Nylon membranes (Millipore) that were then incubated with anti-phospho-STAT1-Y701 Ab, anti-phospho-STAT1-S727 Ab, anti-phospho-STAT3-Y705 Ab, anti-GLS1 Ab, or anti-cyclophilin Ab followed by fluorophore-conjugated, species-specific secondary anti-Ig Abs (Rockland Immunchemicals, and LI-COR). Proteins were visualized and quantitated by Odyssey infrared imaging scanner (LI-COR).

Statistical analyses

Data are reported based on results replicated in at least three biologically independent performances. The primary analyses were conducted on pooled data points from independent samples and replicate experiments, t testing with post-test validation of its suitability. Tests used in particular settings are generally indicated in the Figure Legends. In experiments with individual analyses and comparisons of limited numbers of samples, t-tests or non-parametric formulae were applied as appropriate, based on modeling of the similarity of variances between data from the two conditions being compared, and of the estimated fit to Gaussian distributions; Mann-Whitney testing was used if statistical analysis indicated that the two samples sets probably had unequal variance. Data are displayed as mean (± SEM); conventionally, a difference is considered "statistically significant" when the p value of for the null hypothesis of a comparison was <0.05. In addition to application of this method to individual features in the metabolomic analyses, a series of multivariable linear regression models were employed, each corresponding to a specific analyte, to analyze differences in metabolomic data. In these models, the values of the analytes served as the dependent variables, while individual non-DMSO treatment groups were included as key independent variables, i.e. considering the DMSO group as the reference.

Supplemental information

Supplemental figures

Supplemental Figure 1

Supplemental Figure 2

Supplemental Figure 3

Supplemental Figure 4

Supplemental Figure 5

Supplemental Figure 6

Supplemental Figure 7

Supplemental Figure 8

Acknowledgements

Experimental work was supported by NIH grant R01 AI113292 followed by AI149722 (M. R.B.). and Pathology-Microbiology-Immunology (P. M. & I.) departmental funds. S.K. Brookens was supported by a supplement to AI113292. We thank D. Bhattacharya (U of AZ) for generously shipping Mpc2 f/f breeding stock essential for this work, X. Ye for helpful advice in RNA-seq analyses, M. A. Jones (Biocrates, Inc) for facilitating the metabolomic analysis, the Emory Integrated Genomics Core Facility (RRID:SCR_023529) for added support of the efforts, and Vanderbilt institutional cores (High-Throughput Screening; Flow Cytometry Shared Resource; Small Molecule NMR; Cell Imaging Shared Resource; VANTAGE) for equipment, expertise, and assistance. NIH Institutional Equipment (S10) grant 1S10OD018015 was instrumental in acquisition of equipment for metabolic flux analyses and flow cytometry, and scholarships via the Cancer Center Support Grant (CA068485) and Diabetes Research Center (DK0205930) helped defray costs of Vanderbilt Cores.

Additional information

Abbreviations used

• 2-DG: 2-deoxyglucose

• 4OHT: 4-hydroxytamoxifen

• αKG: α-ketoglutarate

• Ab: antibody

• Ag: antigen

• ASC: Ab-secreting cell

• BAFF: B-cell activating factor

• 2-deoxyglucose

• BCR: B cell Ag receptor

• BrdU: bromodeoxyuridine

• BSA: bovine serum albumin

• WT: wildtype

• KO: knockout

• MZB: marginal zone B cell

• PC: plasma cell

• MBC: memory B cell

• Ig: immunoglobulin

• CD: cluster of differentiation

• CoA: coenzyme A

• CTV: CellTrace Violet

• DMK: dimethyl-ketoglutarate

• DMSO: dimethylsulfoxide

• ECAR: extracellular acidification rate

• ELISA: enzyme-linked immunosorbent assay

• ELISpot: enzyme-linked immunosorbent spot

• ETC: electron transport chain

• FAO: fatty acid oxidation

• FCCP: carbonyl cyanide 4-phenylhydrazone

• GC: germinal center

• GLS: glutaminase

• GSEA: gene set enrichment analyses

• H2DCFDA: dichlorodihydrofluorescein diacetate

• HIF: hypoxia-inducible factor

• HCQ: hydroxychloroquine

• IFN: interferon

• IL-: interleukin-

• Jak: Janus kinase

• MPC: mitochondrial pyruvate channel

• NP-OVA: 3-Nitrophenylacetyl (NP)-ovalbumin (OVA)

• OCR: oxygen consumption rate

• PDH: pyruvate dehydrogenase

• PSA: porcine serum albumin

• ROS: reactive oxygen species

• mtROS: mitochondrial ROS

• SDS-PAGE: sodium dodecyl-sulfate polyacrylamide gel electrophoresis

• SEM: standard error of means

• SRBC: sheep red blood corpuscule

• STAT: signal transducer and activator of transcription

• TCA cycle: tricarboxylic acid cycle

• TLR: Toll-like receptor

• TMRE: tetramethylrhodamine ester

• Tmx: tamoxifen

Funding

National Institute of Allergy and Infectious Diseases (AI149722)

National Institute of Allergy and Infectious Diseases (AI113292)

Vanderbilt University Medical Center (007)