Choline metabolism underpins macrophage IL-4 polarization and RELMα up-regulation in helminth infection

Type 2 cytokines like IL-4 are hallmarks of helminth infection and activate macrophages to limit immunopathology and mediate helminth clearance. In addition to cytokines, nutrients and metabolites critically influence macrophage polarization. Choline is an essential nutrient known to support normal macrophage responses to lipopolysaccharide; however, its function in macrophages polarized by type 2 cytokines is unknown. Using murine IL-4-polarized macrophages, targeted lipidomics revealed significantly elevated levels of phosphatidylcholine, with select changes to other choline-containing lipid species. These changes were supported by the coordinated up-regulation of choline transport compared to naïve macrophages. Pharmacological inhibition of choline metabolism significantly suppressed several mitochondrial transcripts and dramatically inhibited select IL-4-responsive transcripts, most notably, Retnla. We further confirmed that blocking choline metabolism diminished IL-4-induced RELMα (encoded by Retnla) protein content and secretion and caused a dramatic reprogramming toward glycolytic metabolism. To better understand the physiological implications of these observations, naïve or mice infected with the intestinal helminth Heligmosomoides polygyrus were treated with the choline kinase α inhibitor, RSM-932A, to limit choline metabolism in vivo. Pharmacological inhibition of choline metabolism lowered RELMα expression across cell-types and tissues and led to the disappearance of peritoneal macrophages and B-1 lymphocytes and an influx of infiltrating monocytes. The impaired macrophage activation was associated with some loss in optimal immunity to H. polygyrus, with increased egg burden. Together, these data demonstrate that choline metabolism is required for macrophage RELMα induction, metabolic programming, and peritoneal immune homeostasis, which could have important implications in the context of other models of infection or cancer immunity.


Introduction
Macrophages represent a heterogeneous and plastic subset of innate immune cells that perform critical homeostatic, surveillance and effector tasks in almost all tissues [1].They can dynamically respond to diverse endogenous and exogenous cues, including nutrient and energy availability.Metabolic programming can directly drive macrophage function but can also be rewired in response to various stimuli [2].While we continue to expand our understanding of how energy-generating pathways and their intermediates underpin immunometabolic function, the importance of nutrients in macrophage biology remains largely underexplored.
Choline is a quaternary amine and essential nutrient that can form the main eukaryotic phospholipid class, glycerophosphocholines (the most abundant of which are phosphatidylcholines (PC)), or form acetylcholine in cholinergic tissues [3].Additionally, methyl groups from choline contribute to one-carbon metabolism after its mitochondrial oxidation.Given its positive charge, choline requires facilitated transport across cellular membranes, representing the first but least studied step in its metabolism.In immune cells, choline transporter-like protein 1 (Slc44a1) is responsible for the majority of choline uptake [4,5].Upon uptake, choline is thought to be rapidly phosphorylated by choline kinase (Chkα/Chkβ) and is then committed to the PC biosynthetic (CDP-choline/Kennedy) pathway [3,6].
Phospholipid biosynthesis is indispensable for membrane biogenesis, which plays an important role in cellular differentiation and phagocytosis [7].The first insight into the potential role of choline in macrophage function came over twenty years ago with the observation that deletion of myeloid Pcyt1a, encoding the rate-limiting enzyme in PC synthesis, diminished adaptive responses to endogenous stimuli (free cholesterol) [8] and impaired the secretion of pro-inflammatory cytokines from macrophages [9].Recent lipidomic analyses of cultured macrophages demonstrated sweeping alterations to phospholipid composition and content in response to various inflammatory stimuli [4,[10][11][12].To begin to address the specific roles of choline as a regulator of macrophage functions, we and others have shown that polarization with the bacterial endotoxin lipopolysaccharide (LPS; termed M[LPS]) in primary mouse macrophages leads to increased choline uptake and PC synthesis in an Slc44a1-dependent manner.Inhibiting choline uptake and metabolism led to more pronounced inflammatory responses in M[LPS], but also diminished inflammasome-mediated cytokine release [4,5].How choline metabolism may be regulated by or be important for macrophage phenotypes in response to Th2 cytokine exposure remains unknown.
Interleukin-4 (IL-4), along with IL-13, is a signature cytokine of type 2 inflammatory responses [13].In addition to being widely used to polarize macrophages in vitro (termed M ), secretion of these cytokines in vivo is a critical response triggered by invading macroparasites or allergens [14].Since it is now appreciated that macrophage inflammatory and metabolic programs exhibit a dichotomy in response to prototypical pro-and anti-inflammatory stimuli, we sought to characterize the interplay between choline metabolism and IL-4-mediated polarization of macrophages.Interestingly, we found that identical to LPS, choline uptake and incorporation into PC was augmented in M[IL -4].Moreover, when choline uptake and metabolism was inhibited, we observed select but striking differences in metabolic and IL-4-specific responses, underpinned by a dramatic reduction in RELMα both in vitro and in mouse parasitic helminth infection.

IL-4 increases PC content and alters choline containing phospholipid composition
To begin to interrogate the potential effect of IL-4 stimulation on choline metabolic pathways, we performed targeted LC-MS/MS analysis of choline-containing phospholipids in naïve (M [0]), M[IL-4] and M[LPS] (Fig 1A).Surprisingly, the typically divergent polarization states led to a symmetrical increase in total cellular PC levels (Fig 1B ) with a significant increase predominantly in PCs containing unsaturated and monounsaturated fatty acids (Fig 1C).With respect to other choline-containing lipids, no differences were found in the level of sphingomyelin (Fig 1D).Interestingly, there were also changes within the choline-containing lipidome in response to either IL-4 or LPS.The plasmalogen forms of PC, which have been demonstrated to be important for downstream lipid signaling, showed divergent changes.Levels of plasmenyl PCs (PC(P)), structurally known as 1-alkenyl,2-acylglycerophosphocholines, as well as their immediate precursors (LPC(P)) were reduced in response to both stimuli, while levels of plasmanyl PCs (PC(O)), known as 1-alkyl,2-acylglycerophosphocholines remained unchanged (Figs 1E, 1F, and S1A).The levels of platelet activating factors (PAF), their immediate precursor the lyso-PAFs (LPC(O)) (S1B and S1C

IL-4 up-regulates choline metabolism in macrophages
In keeping with higher levels of total cellular PC, polarization with IL-4 increased the incorporation of choline into PC ( Slc44a1 protein expression was significantly augmented, suggesting post-transcriptional mechanisms.There were no changes in Slc44a2 protein levels.Taken together, these data suggest a coordinated up-regulation of choline uptake and subsequent incorporation into PC in response to IL-4 that is comparable in magnitude to LPS stimulation, though subtly different in terms of mechanism.

Inhibiting choline metabolism selectively impairs IL-4 polarization
Characteristically, IL-4 polarization induces the expression of genes involved in dampening inflammation and promoting tissue repair.We next sought to understand how limiting choline availability and metabolism may affect these well-known responses.We used an inhibitor of extracellular choline uptake (hemicholinium-3; HC3) [16] and an inhibitor of choline incorporation into PC, the choline kinase inhibitor RSM-932a (RSM) [17].BMDM incubated with HC3 or RSM for 24 h prior to polarization had little effect on most hallmark M[IL-4] genes, such as Arg1, Mrc1 and Chil3 (Fig 2A -C); however, completely suppressed IL-4-mediated induction of Retnla (Figs 2D and S3A-S3C).Interestingly, shorter (6 h) treatment with HC3 was insufficient to affect Retnla expression (S3D Fig) .To ascertain a better understanding of the kinetics of Retnla inhibition, we conducted a simple experiment in which cells were treated with inhibitors for 24 h, followed by IL-4 for a further 24 h (labeled POST), 24 h of concurrent inhibitor and IL-4 treatment (labeled as CONC) and 24 h of IL-4 polarization, followed by 24 h of inhibition (labeled as PRE).Inhibiting choline metabolism prior to IL-4 polarization flattened any induction of Retnla.Concurrent inhibition of choline uptake and metabolism along side polarization also significantly limited Retnla up-regulation.Finally, while inhibition of choline uptake with HC3 in already polarized macrophages tended to diminish expression, blocking choline metabolism with RSM in polarized cells was sufficient to completely suppress Retnla expression (Fig 2E).To begin to explain the select effect of choline inhibition in M [IL-4], we assessed the activation of pSTAT6-Y641 after IL-4 stimulation, which was significantly blunted by treatment with HC3 or RSM after 15 minutes (Fig 2F).There were no changes in levels of Il4ra transcript (S3E Fig), and pSTAT6-Y641 phosphorylation was further diminished over time (S3F Fig) .Together, this suggests that disruptions of choline metabolism in macrophages can blunt STAT6 signaling, though it is likely that ligand recognition through IL-4R is not impaired [18].
To further probe how limiting choline metabolism alters macrophage IL-4 polarization, we used CD38 and Egr2, which have been identified as effective discriminators of M [LPS] vs. M [IL-4], respectively [19].Blocking choline uptake did not impair the induction of Egr2, nor did it promote CD38 expression in M [IL-4] (S3G Fig), despite Egr2 being induced in an IL-4/ STAT6-dependent manner [20].We next assessed the expression of the checkpoint ligands PD-L1 and PD-L2, which have also been used discriminate between M [LPS] and M [IL-4] polarization [21].Inhibiting choline metabolism with either HC3 or RSM led to consistent PD-L1 up-regulation in M[0] and M[IL-4], but not in M[LPS] where expression was already maximal (Fig 2G).In contrast, inhibiting choline metabolism reduced PD-L2 expression in M [IL-4] (Fig 2G).Therefore, by limiting uptake and phosphorylation of choline, there were pointed changes to traditional IL-4-induced cytokines (e.g., Retnla) and a skewing toward a M [LPS]-like profile.

Choline metabolism inhibition impairs macrophage function
The secreted cytokine resistin-like molecule alpha (RELMα) is encoded by Retnla and has pleiotropic functions, including suppression of Th2 responses and promotion of wound healing [22].In IL-4-polarized macrophages, blocking choline uptake significantly reduced intracellular RELMα levels (Fig 2H).Furthermore, to rule out off-target pharmacological effects, culture of M [IL-4] in choline-deficient media also curtailed intracellular RELMα to the same extent as HC3 and RSM (Fig 2H).Finally, we confirmed that secreted RELMα in M[IL-4] conditioned media was significantly lower when choline uptake was inhibited (Fig 2I).Functionally, given that RELMα levels were dramatically reduced in response to altered choline uptake and metabolism, we sought to assess the in vitro wound healing capacity of macrophages.A "scratch wound" was created among 3T3-L1 fibroblasts grown to confluency and when exposed to conditioned media from HC-3-treated M[IL-4], we observed a significantly impaired rate of wound healing (S3H Fig) .PLOS PATHOGENS

Choline metabolism supports naïve and M[IL-4] transcriptional programs
To gain an unbiased understanding of the transcriptional changes in response to blocking choline uptake and metabolism, we next performed RNA transcriptomic analysis.Naïve and IL-4-polarized macrophages that had been pretreated with either HC3 or RSM prior to polarization showed profound and sweeping differences in transcript expression patterns.Reactome pathway analysis revealed the up-regulation of extracellular matrix organization genes by choline inhibition, and down-regulation of genes involved in GPCR signalling (M[0]) and cell cycle (M[IL-4]) (S4 Fig) .Importantly, we confirmed that Retnla was only expressed with IL-4 polarization and was one of the most strongly down-regulated gene by fold change compared to vehicle for both HC3 and RSM treatments (Fig 3A).We next performed Enrichr-KG analysis [23] on the top 100 shared genes up-and down-regulated by HC3 and RSM in both M[0] and M[IL-4].These analyses indicated a conserved induction of unfolded protein response and ER stress pathway genes, such as Trib3, Ern1, Ddit3 (Fig 3B and 3C).In contrast, both drugs downregulated genes involved in tyrosine phosphorylation and DNA replication (Fig 3D).

Choline metabolism inhibition profoundly alters mitochondrial morphology and function
Macrophage metabolism is highly responsive to exogenous stimuli [24].In response to IL-4, glycolysis, glutaminolysis, and oxidative phosphorylation are up-regulated [25].The inhibition of choline metabolism led to a dramatic decrease in select electron transport chain (ETC) genes encoded in the mitochondrial genome, with concomitant up-regulation of two mitochondrial ribosomal genes (mt-Rnr1/mt-Rnr2) (Fig 4A).Immunoblot analysis of ETC complexes showed a strong trend toward reductions in complexes I-IV with HC3 and RSM (Fig 4B ), though only changes in complex I was significantly different (S5A Fig) .To interrogate further, we performed extracellular flux analysis and as expected, oxygen consumption rate (OCR) and spare respiratory capacity (SRC) were increased in M[IL-4] compared to M[0] (Fig 4C -I).Treatment with HC3 or RSM resulted in a potent reduction of SRC (Fig 4H).Presumably as a compensatory mechanism, glycolysis as measured by extracellular acidification rate (ECAR) was modestly elevated by HC3 and highly increased by RSM (Fig 4D and 4F), effects reminiscent of complex I inhibition.We used a recently refined method to calculate energy derived from glycolysis (ATP Glyco ) or oxidative phosphorylation (ATP OXPHOS ) using extracellular flux assay values [26].This showed that although ATP OXPHOS was diminished with HC3 or RSM, overall ATP production remained relatively stable (Fig 4J).To understand the kinetics of these effects, we next assessed the time needed for metabolic changes in response to choline kinase inhibition.Prior to washing the assay plate, just 5 minutes of RSM treatment was sufficient to partially reduce ATP OXPHOS production, while 1 hour of treatment strongly reduced ATP OXPHOS (S5B Fig).However, direct injection of RSM instead of rotenone/antimycin A failed to reduce OCR (S5C Fig).This suggests that while rapid, the inhibitory effect on oxidative metabolism is indirect by way of general mitochondrial disruption rather than a direct offtarget activity on ETC complexes.

Choline kinase inhibition in vivo reduces tissue-resident macrophage RELMα
RELMα is strongly up-regulated upon IL-4 treatment in vitro but is also expressed endogenously in tissue-resident macrophages throughout the body, especially serosal cavity macrophages [27,28].We next sought to pharmacologically block choline metabolism in vivo to interrogate whether this could mimic changes in IL-4 responses observed in cultured cells.Mice were dosed with RSM (3 mg/kg i.p.) every other day for seven days (Fig 5A).Previously, long-term RSM administration in preclinical tumour models resulted in significant reduction in tumour growth [17], highlighting its low toxicity and therapeutic potential.Surprisingly, RSM treatment resulted in drastic remodelling of peritoneal exudate cell (PEC) populations (Fig 5B).Uniform manifold approximation and projection plots show a reduction in B cells, macrophages and eosinophils, and an increase in Gr-1 + polymorphonuclear cells (PMNs) and monocytes.Furthermore, intracellular RELMα levels were significantly decreased in large (CD11b + F4/80 hi ), and small (CD11b + F4/80 lo ) PEC macrophages (LPM and SPM) taken from RSM-treated mice compared to vehicle control (Figs 5C and gating in S6A).Moreover, expression of PD-L1 was also significantly increased and PD-L2 reduced compared to vehicle-treated mice, which mirrored in vitro results (Fig 5D and 5E).The suppressive effect of RSM on LPM and SPM, which contained moderate or high levels of intracellular RELMα was evident following just two doses, separated by 48 h (S6B and S6C Fig) .Inflammation, infection, or sterile injury of serosal (peritoneal and pleural) cavities cause the disappearance of LPM from serosal fluids (termed macrophage disappearance reaction or macrophage disturbance of homeostasis reaction (MDHR) [29].Several recent studies have shown that these macrophages are primarily forming aggregates within the cavity or localizing to the omentum [30].We adapted a recently published comprehensive methodology [30,31] to address MDHR in our studies by analyzing potential sites of LPM egress: floating and adherent cellular aggregates, homing to the omentum, and migration to draining lymphatics.We found that aggregated MHCII lo LPM and MHCII + SPM in RSM-treated mice expressed nearly no RELMα, similar to suspended lavage cells (Fig 5F).Thus, in vivo inhibition of choline metabolism restructures endogenous tissue-resident macrophage phenotypes and supresses RELMα in multiple tissue compartments.
RELMα is further found in white adipose tissue (WAT) and lung, and though previously thought to be an adipokine [32], RELMα is mainly produced by WAT macrophages.

Inhibiting choline metabolism with RSM improves measures of chemicalinduced colitis
The link between choline metabolism and RELMα was unexpected.To further probe the functional consequences of altering RELMα levels under pathophysiological conditions, we used dextran sodium sulfate (DSS) to chemically induce a state of colitis in female C57BL/6J mice over a 7-day protocol, where mice were given vehicle or RSM (3 mg/kg i.p.) bi-daily (S6H Fig) .While there were no significant differences in body weight (S6I Fig), colon length was significantly longer and hemoccult was significantly lower in RSM-treated mice compared to control, respectively (S6J and S6K Fig).This is entirely consistent with previous reports where Retnladeficient mice were protected against DSS-induced colitis [33].

Choline inhibition shifts peritoneal immunity and impairs M[IL-4] polarization to multiple helminth infections
Our data suggest a role for choline metabolism in supporting IL-4-induced (Th2) responses in vitro and in vivo.To interrogate choline metabolism function in the physiologic setting of a Th2-skewed environment driven by helminth infection, we infected female mice with Heligmosomoides polygyrus (Hp), a parasitic nematode that colonizes the small intestine and drives protective M[IL-4]-like responses [34].Intestinal Hp infection induces significant peritoneal cavity inflammation and macrophage M[IL-4] responses [35].The striking RSM-induced changes in the peritoneal immune landscape of naïve mice was exacerbated during Hp infection (Figs 6A and S7A).We detected more than a 10-fold increase in peritoneal cells compared to naïve mice and during infection, there was a significantly lower number total cells when mice were treated with RSM (Fig 6B).There was a substantial drop in eosinophils and B-1 cells in RSM-treated mice, accompanied by a rise in monocytes and neutrophils, both by frequencies and total cell numbers (Fig 6C -6F).Peritoneal macrophages were also decreased by RSM treatment in both naïve and infected mice (  Taken together, these data support a major role for choline metabolism in sustaining RELMα expression and shaping the composition and phenotype of immune populations in vivo.

Choline kinase inhibition does not affect helminth burden despite dramatic blunting of RELMα
Given the dramatic shift in peritoneal immune landscape, we next aimed to assess how markers of parasite burden may be affected.Mice were orally infected with Hp larvae and left for 8 days to allow parasites to develop into adults within the intestinal lumen, followed by treatment with RSM or vehicle every two days.Naïve mice were also treated with RSM at the same dose and interval to serve as controls for infection (Fig 7A).Treatment with RSM had modest effects on the weight of naïve or Hp-infected mice, with the trend of decreasing weight gain but no change in final weights between groups (Fig 7B and 7C).IL-4-polarized macrophages have been shown to play an important role in reducing helminth parasite burden [36]; yet, neither egg nor worm burden were affected following RSM treatment (Fig 7D and 7E).Hp infection significantly increased RELMα levels in the peritoneal fluid and serum; nevertheless, consistent with in vitro and in vivo observations, RELMα levels were significantly decreased with RSM treatment (Fig 7F and 7G).Furthermore, the presence of CD206 + and RELMα + cells in intestinal tissue, the primary site of infection, was also reduced (Fig 7H -7J).To determine whether macrophage derived RELMα was specifically affected by RSM treatment, intestinal tissue sections were co-stained with F4/80 and RELMα and double positive cells quantified.This analysis revealed that Hp infection strongly induced F4/80 + RELMα + cells in the intestine, but these were significantly reduced by RSM treatment (Fig 7K and 7L).
We next probed the effect of chronic RSM treatment in a vaccination model for Hp involving secondary infection (S9A Fig) , where IL-4-polarized macrophages are critical for protection [36].Over the course of the 10-week infection, RSM treatment resulted in significantly increased fecal egg burden specifically at week 8 but showed a trend at all other time points (S9B Fig) .Taking the area under the curve for egg burden, RSM treated mice had significantly higher total egg burden (S9C Fig), suggestive of some deficiency in optimal immunity to the parasitic worm.However, there were no significant differences in the adult worm burden after secondary challenge (S9D Fig) .Remarkably, consistent with the acute regiment, RSM treatment led to a significant drop in peritoneal and serum RELMα (S9E and S9F Fig), as well as down-regulated CD206 and up-regulated CD86 expression in the intestinal cells of infected mice (S9G-S9I Fig) .We conclude that choline metabolism inhibition significantly reduces RELMα but this is not sufficient to fully impact worm burden, however, can impact worm fertility.This may be partly explained by the fact that RSM does not inhibit other M2 macrophage effectors such as Arg1.Together, this data indicates the importance of choline metabolism and choline kinase signaling in promoting M[IL-4] polarization and RELMα expression in intestinal helminth infection.

Discussion
Upon activation, macrophages up-regulate choline metabolism to fuel PC synthesis.LPS and IL-4 both induce expansion of organellar membranes including ER, mitochondria, and endolysosomes [37][38][39][40], in addition to a general enlargement of the cell itself [41].Building on our understanding of macrophage choline metabolism during LPS polarization, we show here that stimulation of divergent signaling pathways (i.e., LPS-TLR4 vs. IL-4-JAK/STAT6), both lead to more choline taken into the cell with an associated up-regulation of Slc44a1, which likely facilitates increased flux through the PC synthetic pathway and more PC levels for membrane biogenesis.Recently, another putative choline transporter has been described, FLVCR1 [42].Flvcr1-knockout mice are embryonic lethal, but little has been reported as to the role of this gene in immune cells.Interestingly, cells from Flvcr1-knockout embryos had major defects in mitochondrial structure and morphology.Our RNA-seq analysis revealed robust expression of Flvcr1 in M[0] and M[IL-4], which was up-regulated by choline metabolism inhibition.In contrast, transcript expression of Slc44a1 was inhibited by choline metabolism inhibition.The physiological role and individual contributions of choline transporters such as Slc44a1, Slc44a2 and Flvcr1 to choline transport and metabolism remains unexplored but of significant priority.
How phospholipid composition is affected by macrophage polarization has been addressed in studies using similar experimental designs.While the increased accumulation of PC in IL-4/ IL13-polarized macrophages (RAW264.7 cells) was shown in a comprehensive assessment of phospholipid content after long term exposure to polarizing stimuli, this was not seen in human THP-1-derived macrophages [43], nor were increases observed before 18 h [38].Coupled with an increase in total PC content, we observed an increase in saturated and monounsaturated fatty acyl chains in PCs elevated in both M[IL-4] and M[LPS].We and others have previously shown that M[LPS] up-regulate de novo lipogenesis to supply new fatty acids for phospholipid synthesis [4,10].There is now evidence to support a role for both exogenous fatty acid uptake [11,12], as well as de novo lipogenesis in M[IL-4] [44].Despite the potentially divergent sources of FA, differences in the side chain composition of choline-containing phospholipids between macrophage polarizations may fuel distinct lipid signaling pathways following phospholipase-specific processing.
A handful of studies have now interrogated the importance of choline metabolism in macrophage biology, though all have been in the context of LPS-induced polarization and inflammation [4,5,9,45,46].Conversely, while macrophage IL-4 polarization is widely used in vitro and typically yields a reproducible response, how this might be affected by or might affect choline metabolism remained unclear.By blocking choline uptake and/or subsequent metabolism, the transcriptional induction of Mrc1, Chil3, Arg1 were not changed.However, limiting the availability of choline in the media, or pharmacologically blocking the uptake/phosphorylation of choline caused the targeted reduction in IL-4-induced Retnla expression and RELMα secretion.Early studies indicated that signaling through STAT6 is essential for proper induction of Th2 cytokines and IL-4-induced RELMα expression [47].In response to limiting choline availability and PC synthesis, STAT6 phosphorylation, as a surrogate for IL-4R signaling efficacy, was decreased.Though this supports the robust down-regulation of Retnla, similarly regulated transcripts such as Arg1 and Chil3 were not affected or mildly increased.Curiously, Chil3 has even been suggested to augment Retnla expression [48].The mechanism(s) by which choline availability regulates Retnla are likely transcriptional; however, the exact mechanism remains unclear.Interestingly, transient expression of Retnla was recently shown in nearly all tissueresident macrophages and its induction in PEC macrophages was IL-4R-and STAT6-independent [49,50].Consistent with this, other studies report that RELMα is induced by signals other than Th2 cytokines, such as hypoxia or phagocytosis of apoptotic cells [51,52].Furthermore, a reporter mouse line showed RELMα expression mainly in WAT, intestine, and peritoneal macrophages [50], as well as peritoneal eosinophils and type II alveolar epithelial cells.
Our findings support the differential induction of canonical M[IL-4] genes and identify choline metabolism as a specific inducer of Retnla.In our study, we found that Chkα inhibition resulted in major changes to PEC populations, akin to the macrophage disappearance reaction upon inflammatory stimuli [53], pointing to a possible connection between macrophage tissue residence and choline metabolism.In addition, B cells in the peritoneum were drastically reduced, which corroborates previous observations demonstrating that RSM and another Chkα inhibitor (MN58B) dramatically lowered splenic B cell populations upon systemic delivery [54].How this altered peritoneal immune profile may influence RELMα levels remains to be determined; however, since lung B cells express similar amounts of RELMα during Nippostrongylus brasiliensis infection as alveolar macrophages [55], the effect may not be exclusive to macrophages.During Nb infection, lung-resident interstitial macrophages up-regulate RELMα in a STAT6-dependent manner [50].Separately, a subset of lung interstitial macrophages identified as CD169 + nerve-and airway-associated macrophages (NAMs) were found to express high levels of RELMα [56].Depletion of all interstitial macrophages impaired control of helminth infection [50], while depletion of NAMs exacerbated lung inflammation but had no impact on control of influenza infection [56].
Macrophage-derived RELMα has been shown to promote profibrotic collagen cross-linking in fibroblasts [57], while also contributing to pro-inflammatory processes in intestinal inflammation and infection [33,58,59].Given that limiting choline metabolism in M[IL-4] markedly depleted RELMα levels and that in vivo RSM treatment recapitulated this effect, we reasoned that blocking choline metabolism during helminth infection may phenocopy models of RELMα deficiency, where enhanced resistance to parasites is observed [60][61][62].Contrary to our hypothesis, RSM-mediated choline kinase inhibition had minimal impact on indices of parasite burden in the helminth models we assessed, despite profound reductions in local and systemic levels of RELMα (Figs 6, S5, and S7).While the explanation for this remains unclear, there may be compensatory changes in RSM-treated mice, such as the altered immune profile, differential activation of macrophages, or direct effects of RSM on the pathogen that could confound the interpretation.The importance of macrophage-specific metabolism for M  activation in the context of helminth infection has previously been shown.Inhibition of lysosomal lipolysis in H. polygyrus-infected mice reduced M[IL-4] polarization and RELMα expression, leading to increased parasite egg burden [63].Similarly, our studies indicate that choline metabolism is also important in optimal M[IL-4] polarization in response to H. polygyrus infection, but with slight effects on egg burden in chronic studies and no effects on parasite burden.Future work could involve genetic and tissue-specific models to directly address the relevance of macrophage choline metabolism in controlling helminth infections.
Metabolic reprogramming that occurs in response to macrophage polarization represents one of the fundamental observations of immunometabolism.Our RNAseq analyses revealed intriguing changes in mitochondrial transcripts (ribosomal genes up-regulated and electron transport chain component transcripts down-regulated).The extent to which mitochondrial morphology and metabolism were altered was intriguing.M[IL-4] up-regulate oxidative and glycolytic programs, with key signals coming from metabolic intermediates such as α-ketoglutarate and succinate [64,65].When choline cannot enter the cell or be used in the synthesis of PC, cellular organelles are swiftly affected.The capacity to continuously provide PC, from ER to mitochondria is critically important and when disrupted, results in clear destruction of the mitochondrial cristae, results consistent with Flvcr1-and SLC44A1-deficient cells [5,66].As a result, macrophages, independent of IL-4 stimulation, must rely on anaerobic means to generate cellular ATP.Importantly, we observed a dramatic difference in the potency of inhibiting choline uptake via HC3 or choline kinase activity via RSM, whereby the latter caused the most dramatic effect.We reason that the inhibition of choline uptake, while limiting supply of exogenous choline, would continue to allow for the recycling and modification of existing phospholipids, which lessens the impact of the treatment.Choline kinase inhibition; however, would almost completely limit PC supply.While inhibiting macrophage PC synthesis was shown to lead to the loss of mitochondrial membrane potential, leading to mitophagy and a protection against inflammasome activation, future work will be required to identify whether metabolic reprogramming is mechanistically linked to Retnla down-regulation in M[IL-4] cells.
In addition to the striking effect on RELMα, we found the costimulatory checkpoint ligands PD-L1 and PD-L2 were coordinately regulated by choline metabolism, which is contrary to conventional IL-4-polarization.In tumour settings, nutrient availability may be heterogenous, especially for infiltrating immune cells, which can explain the wide range of PD-L1 positivity among tumour samples and the relatively poor response to anti-PD-1/PD-L1 immune-checkpoint inhibitor therapies [67].Up-regulated choline metabolism is a recognized hallmark of certain cancers [68,69] and targeting Chkα can both inhibit cancer cell growth and render cell death-resistant cancers more susceptible to immunotherapy [70].Similarly, RSM-932A increased CD86 expression in the intestine in the secondary H. polygyrus infection, which is a hallmark of M1-polarized macrophages [71].CD86 in antigen-presenting cells provides costimulatory signals to activate T cells through CD28 and CTLA4 binding [72,73].Previous studies reported that CD86 was also up-regulated by helminth-derived peptide GK-1 in BMDM [74].These changes in costimulatory and checkpoint proteins supports a potential immunoregulatory role of choline metabolism that may represent an underappreciated therapeutic potential in certain disease settings.
Macrophages up-regulate choline uptake and incorporation in response to IL-4 stimulation.Reciprocally, when the availability of choline is compromised, normal IL-4 responses are affected.We used two pharmacological approaches to limit choline availability and nutrient deprivation in the media.These approaches carry numerous caveats, including dose, timing, route, and potential off-target effects; however, our in vitro pharmacological approach targeted distinct steps in the CDP-choline pathway.Separately, other fates of choline may be modified by or play a role in mediating the alteration of immune cell function.For instance, phosphocholine may post-translationally modify proteins [75,76], and macrophages and other immune cells have been shown to produce acetylcholine [77].Conversion of choline to betaine may also support osmoregulation [78] or contribute to histone methylation [79], but these latter mechanisms require deeper investigation in immune contexts and remain speculative.Future work using genetic models to solidify the importance of choline uptake and subsequent metabolism are warranted.
In summary, we describe a critical role for choline metabolism in the mediating the full potential of macrophage M[IL-4] polarization and metabolic programming.Moreover, systemic pharmacological inhibition of choline metabolism in mice closely mirrors the dramatic down-regulation of RELMα but fails to affect the pathology of intestinal or lung helminth infection models.This work also highlights that the inhibition of choline metabolism via RSM caused a dramatic shift in immune cell profile and potentially polarization, which may be therapeutically beneficial in other disease settings such as peritoneal metastases or fibrosis.

Ethics statement
All animal procedures were approved by the University of Ottawa Animal Care Committee (BMI-1863) and the University of California Riverside Institutional Animal Care and Use Committee (protocol A-20210017).

Animals
Mice (C57BL/6J) were originally purchased from Jackson Laboratories (Stock no.00064) and bred in a pathogen-free facility in the University of Ottawa animal facility or acclimated for 7 days prior to H. polygyrus infection at the University of California Riverside.Mice were maintained on a 12-hour light dark cycle (lights on at 7:00 am) and housed at 23˚C with bedding enrichment.Male and female mice ages 8-16 weeks were used for the generation of primary macrophages as described below.

Choline uptake, incorporation, uptake inhibition and degradation
The rates of choline uptake, incorporation into PC, uptake inhibition and degradation were determined in M[0] and M[IL-4] using 3 H-choline chloride (Perkin Elmer) as previously described [4].

High-performance liquid chromatography electrospray ionization tandem mass spectrometry (LC-ESI-MS/MS) Lipidomics
Cells were counted, pelleted, and frozen at -80˚C in 15 mL Falcon tubes for 24 h.Pellets were resuspended in 1 mL sodium acetate using a syringe pipette and transferred to Kimble tubes.Tubes were washed two times in 1 mL and 1.2 mL sodium acetate, respectively, to recover all material and these volumes were also transferred to the corresponding Kimble tube.All pipette tips were glass.MS-grade lipid standards (all from Avanti Polar Lipids) consisting of 90.7 ng LPC (13:0/0:0), 100 ng PC(12:0/13:0), and 75 ng SM(d18:1/18:1-D9), were added to the samples at time of extraction.Lipids were extracted using a modified Bligh and Dyer [82] protocol at a final ratio of acidified methanol/chloroform/sodium acetate of 2:1.9:1.6 as previously described [83,84].The organic chloroform phase was retained, and the aqueous phase was successively back-extracted using chloroform three times.The four chloroform extracts were combined and evaporated at room temperature under a constant stream of nitrogen gas.Final extracts were solubilized in 300 μL of anhydrous ethanol (Commercial Alcohols) and stored under nitrogen gas at -80˚C in amber vials (BioLynx).
Lipid quantification was performed on an Agilent 1290 Infinity II liquid chromatography system coupled to a QTRAP 5500 triple quadrupole-linear ion trap mass spectrometer using a Turbo V ion source (AB SCIEX) in positive ion mode.Reverse phase chromatography was performed using a binary solvent system composed of solvent A (water with 0.1% formic acid (Fluka) and 10 mM ammonium acetate (OmniPur)) and solvent B (acetonitrile (J.T. Baker) and isopropanol (Fisher) at a ratio of 5:2 v/v with 10 mM ammonium acetate and 0.1% formic acid) and a nanobore 100 mm x 250 μm (i.d.) capillary column packed in-house with Repro-Sil-Pur 200 C18 (particle size of 3 μm and pore size of 200 Å; Dr. A. Maisch, Ammerbruch).Three μL of sample at 4˚C was injected with the system operating at 10 μL/min and using a gradient that started at 30% solvent B, reached 100% solvent B at 8 minutes, and remained at 100% solvent B for 45 min.At 45 min, composition was returned to 30% solvent B and the column was regenerated for 15 min.Samples were prepared for HPLC injection by mixing 5 μL of lipid extract with 2.5 μL of an internal standard mixture (all from Cayman Chemicals) consisting of 2.5 ng LPC(O-16:0-D4/0:0), 2.5 ng LPC(O-18:0-D4/0:0), 1.25 ng PC(O-16:0-D4/2:0) and 1.25 ng PC(O-18:0-D4/2:0) in EtOH, and 16 μL of Solvent A. Transition lists were established using precursor ion scan interrogating a diagnostic fragment ion at m/z 184.1, corresponding to the phosphorylcholine headgroup of the protonated molecular ion ([M+H] + ).Once all detectable species were established in precursor ion discovery mode, targeted lipid quantification was performed using selected reaction monitoring (SRM), monitoring transitions from [M+H] + with the product ion 184.1.ESI-MS/MS acquisition and instrument control were performed using Analyst software (version 1.6.3,SCIEX).The ion source operated at 5500 V and 0˚C.Nebulizer/heated gas (GS1/GS2), curtain gas, and collision gas (all nitrogen) were set to 20/0 psi, 20 psi, and medium, respectively.Compound parameters (de-clustering potential, entrance potential, collision energy, and collision cell exit potential) were individually optimized for each transition.MultiQuant software (version 3.0.2AB SCIEX) was used for peak picking and processing quantitative SRM data.Bayesian Annotations for Targeted Lipidomics (BATL v2.7, https://www.complimet.ca/shiny/batl/) was used to assign peaks [84].Peak areas were normalized to both cell number and either LPC(13:0/0:0) or SM(d18:1/ 18:1-D9) for PCs and SMs, respectively to account for extraction efficiency and instrument response.Data are expressed as nmol or pmol equivalents per 10 6 cells.
The identities of all lipid species in our samples were structurally determined from quality control samples composed of an equi-volume pool of all samples analyzed.SRM was used as a survey scan to trigger information dependent acquisition of EPI spectra in the linear ion trap.After acquisition, the EPI fragment spectra were analyzed to determine structural identities of the lipids.

Wound healing
3T3-L1 fibroblasts were passaged in complete DMEM, plated in ImageLock plates (Essen Bioscience) and grown to confluency.Macrophages were pre-treated with vehicle or HC3 for 24 h, washed with PBS, and supernatants were collected after a further 24 h in complete DMEM ± 20 ng/mL IL-4.A scratch wound was made in each well of 3T3-L1 fibroblasts, wells were gently washed with warm PBS, and supernatants were added to 4 or 5 replicate scratch wound wells.Wound healing was monitored hourly for >24 h on an IncuCyte ZOOM system (Essen Bioscience).

Transcript and protein expression
Total RNA was isolated from BMDM using the TriPure reagent protocol (Roche Life Sciences).Isolated RNA was re-suspended in RNase/DNase-free water (Wisent).QuantiNova™ reverse transcription kit (Qiagen) or ABM All-in-one RT kit (ABM) was used to synthesize cDNA according to manufacturer's instructions.To determine transcript expression, the Quanti-Nova™ Probe PCR kit (Qiagen) was used in combination with hydrolysis primer-probe sets (ThermoFisher) or BrightGreen Express MasterMix (ABM) with custom-designed primers (ThermoFisher).Comparative qPCR reactions were run on the Roto-Gene Q (Qiagen) or Stratagene MX3005p (Agilent).Relative transcript expression was determined using the delta-delta C q method and normalized to Actb [85].Following treatments, macrophage protein expression and phosphorylation status were assessed by immunoblotting as previously described [4].
For RNA-seq analysis, total RNA was isolated from BMDM using Tripure reagent and chloroform and subsequently purified using spin columns (PureLink RNA Mini, ThermoFisher) according to manufacturer instructions.Messenger RNA was purified from total RNA using poly-T oligo-attached magnetic beads.After fragmentation, the first strand cDNA was synthesized using random hexamer primers followed by the second strand cDNA synthesis.The library was ready after end repair, A-tailing, adapter ligation, size selection, amplification, and purification.The minimum Q30 score was 90.96%.

Dextran sodium sulfate (DSS)-induced colitis
Mice were provided 2-2.5% dextran sodium sulfate (MW 40-50 kDa, MP Biomedical) in drinking water ad libitum and treated intraperitoneally with vehicle (40% DMSO in PBS) or RSM-932a (3 mg/kg) every other day for 6 days and sacrificed at day 7. Colons were dissected out and measured with uniform tension by hanging the caecum from a clamp and applying a weight to the distal end.Liver, white adipose tissue, and 0.5-cm pieces of colon were snap-frozen in liquid N 2 or fixed in 10% neutral-buffered formalin (Sigma-Aldrich).Feces was collected daily and hemoccult was assessed by luminescence with ECL substrate (BioRad).

Heligmosomoides polygyrus (Hp) infection model
Hp life cycle was maintained in mice, as previously described [86,87].Mice were orally gavaged with 200 Hp L3 stage larvae or with PBS (naive group).Egg burden was measured at

PLOS PATHOGENS
Choline sustains M2 macrophage responses indicated day post-infection by counting eggs in feces with McMaster slide.To enumerate L5 stage Hp, the small intestines of infected mice were cut longitudinally, and larvae were isolated by fine forceps.Isolated larvae were counted on the petri dish with grids under the dissection microscope.Hp infection experiments were conducted at least twice and as indicated in the figure legends.

RSM-932A
RSM-932A (Cayman Chemical, Ann Arbor, MI) was reconstituted to 30 mg/mL in DMSO.For in vivo injection, RSM-932A was further diluted with DMSO and PBS to 1.5 mg/mL.Mice were injected intraperitoneally with 3 mg/kg of RSM-932A or 40% DMSO in PBS (vehicle group).In vivo injection experiments were conducted at least 3 independent times and as indicated in the figure legends.

RNA-seq data analysis
Raw paired-end reads were trimmed of adapters using fastp [88] before being aligned on the mouse genome (assembly GRCm38) using STAR [89] in local mode and retaining only read pairs with mapq score of 40.Duplicate reads were marked but retained using Picard [90].Reads aligning to exons were counted and summarized to genes using Subread featureCounts [91,92].Only genes with at least 10 counts in at least two samples were retained for further analysis.Differential expression testing was performed in R/Bioconductor [93] using edgeR and the glmQLFTest function [94].For differential expression calls, a log2 fold-change value of 1 or -1 combined with a Benjamini-Hochberg-calculated FDR value less than 0.05 was judged significant [95].Differentially expressed gene sets (up-or down-regulated genes with log2 fold-change of at least 1 or -1 and FDR lower than 0.05) were searched for potential enrichment in gene functional terms.The R/Bioconductor packages ReactomePA [96] and clusterProfiler [97] were used to search for enrichment in Reactome terms [98].In these analyses, the search universe was limited to genes filtered for minimum expression.

Extracellular flux analysis
BMDM were plated in 96-well culture plates (Agilent) at 7.5x10 4 cells/well in complete DMEM.Cells were treated with DMSO, HC3, or RSM-932a at different concentrations ± 20 ng/mL IL-4 for 24 h.Extracellular flux analysis was performed on XFe96 (Agilent) using an adapted MitoStress Test Kit (Agilent).Cartridge ports were loaded with 10x concentrations of drugs: 15 μM oligomycin, 140 μM BAM15 (Cayman Chemicals), and 10μM rotenone/10 μM antimycin A/20 μM Hoechst 33342.BAM15 is a mitochondrial uncoupler with improved induction of maximal respiration compared to FCCP [99].RSM932a was loaded in place of rotenone/antimycin A in some wells.Data were normalized by cell counts obtained by nuclear Hoechst 33342 quantification on an EVOS FL Auto 2 (ThermoFisher) microscope and Qupath (University of Edinburgh).Bioenergetics were calculated using a worksheet template [26].

Transmission electron microscopy
BMDM were plated in 6-well plates and treated with DMSO or RSM-932a ± 20 ng/mL IL-4 for 24 h.Cells were lifted by gentle scraping in 10mM EDTA in PBS and fixed in 4% paraformaldehyde/3.5% glutaraldehyde overnight.Cell pellets were embedded in agarose and blocks were sectioned at 50-μm thickness using a Lecia VT1000S vibratome.Resin blocks were sectioned at 80-nm thickness using a Leica EM UC6 ultramicrotome and placed on microgrids.Sections Fig), as well as total lysophosphatidylcholines (LPCs) (Fig 1G) did not change in response to IL-4 but were significantly up-regulated in M[LPS].The fatty acid composition of choline-containing subspecies was indistinguishable between LPS and IL-4 polarization, except for a divergence in total LPCs (S1D Fig).
Fig 1H), as well as the rate of choline uptake in bone marrowderived macrophages (BMDM) (Fig 1I), without changing choline transport affinity or sensitivity to inhibition (S2A and S2B Fig).Moreover, while select phospholipases are up-regulated in M[IL-4] [15] we observed no difference in the rate of PC degradation (S2C Fig), which completely mirrored the effects of LPS polarization [4].We and others have shown that macrophages express two choline transporters, Slc44a1 and Slc44a2.IL-4-treated cells did not show increased Slc44a1 or Slc44a2 mRNA (Fig 1J); however, total (Fig 1K) and surface (Fig 1L)

Fig 3 .
Fig 3. Inhibiting choline metabolism drives pronounced changes in gene transcription in naïve and IL-4-polarized macrophages.A) Schematic of bulk RNA-seq sample preparation.Heatmap of z-scores for select genes of interest.n = 3. B) Volcano plots of comparisons between Veh and HC3 or RSM in M [0] or M[IL-4].Red dots represent genes with false discovery rate (FDR) < 0.05 and log 2 fold change (FC) < -1 or > 1. Select genes of interest are annotated.C-D) The top significantly up-or down-regulated genes shared in all comparisons in B) were analyzed by Enrichr-KG to identify common pathways.Schematics were created using BioRender.https://doi.org/10.1371/journal.ppat.1011658.g003
Fig 6G).Bulk peritoneal macrophages showed differences in CD206, CD86, and PD-L1 in infected mice treated with RSM (Fig 6H-6J), similar to immunofluorescent staining and in naïve mice (Fig 5D).For acute Hp-infected mice, we further refined peritoneal macrophage classification into MHCII hi SPM and MHCII lo LPM.The numbers and frequencies of MHCII hi and MHCII lo peritoneal macrophages were decreased upon RSM treatment (S7B Fig).CD86 and PD-L1 were increased whereas CD206 was up-regulated in both populations, with a stronger effect seen in MHCII lo macrophages (S7C-S7E Fig).In contrast, PD-L2 was decreased only in MHCII hi macrophages in RSMtreated mice (S7F Fig).These results were directly recapitulated during chronic Hp infection and RSM treatment, where we observed differences in resident peritoneal immune

Fig 7 .
Fig 7.In vivo choline kinase inhibition during primary infection impairs systemic and intestinal RELMα and intestinal macrophage alternative activation.A) Schematic of primary infection with 200 H. polygyrus (Hp) L3 larvae through oral gavage.Mice were treated intraperitoneally with vehicle (40% DMSO in PBS) or RSM-932a (3 mg/kg) every other day for 8 days starting on day 8 and sacrificed on day 17.B-C) Percentage of weights at start of vehicle or RSM injections at 8 DPI with H. polygyrus (B) or final weights at day 17 (C).Mixed-effects analysis with Tukey's test for multiple comparisons (* p < 0.05).D-E) Eggs in feces were counted at multiple time points after infection, and E) adult worms were isolated from the small intestine and enumerated on the day of sacrifice.Values represent means ± SEM (n = 3-5 mice per group), representative of 3 experiments.Two-way ANOVA with S ˇı ´da ´k's test for multiple comparisons and unpaired t test (ns).F-G) Detection of serum and G) peritoneal fluid RELMα by ELISA in naïve and H. polygyrus-infected mice.n = 3-5 per group.Two-way ANOVA with S ˇı ´da ´k's test for multiple comparisons (* p < 0.05, *** p < 0.001 for differences between treatment and ## p < 0.01 for differences between naïve and infected vehicle-treated mice).H-L) Immunofluorescent staining of intestinal tissue for CD206 and RELMα against DAPI counterstain.Scale bar 50 μm.Quantification of CD206 + (I) or RELMα + (J) per DAPI + cell.Unpaired t test (* p < 0.05).Quantification of F4/80 + (K) and RELMα + F4/80 + (L) per DAPI + cell.n = 6.Two-way ANOVA with S ˇı ´da ´k's test for multiple comparisons (**** p < 0.0001).Schematics were created using BioRender.https://doi.org/10.1371/journal.ppat.1011658.g007