Classically activated macrophages undergo functionally significant nucleotide metabolism remodeling driven by nitric oxide

Abstract

During an immune response, macrophages specifically reprogram their metabolism to support functional changes. Here we revealed that nucleotide metabolism is one of the most significantly reprogrammed pathways upon classical activation. Specifically, de novo synthesis of pyrimidines is maintained up to UMP, but blocked at CTP and dTMP synthesis; de novo synthesis of purines is shut off at the last step (catalyzed by ATIC), and cells switch to increased purine salvage; nucleotide degradation to nitrogenous bases is upregulated, but complete oxidation of purine bases (catalyzed by XOR) is inhibited, diverting flux into salvage. Mechanistically, nitric oxide was identified as a major regulator of nucleotide metabolism, simultaneously driving multiple key changes, including the transcriptional downregulation of Tyms and profound inhibition of ATIC and XOR. Inhibiting purine salvage by HPRT knockout or inhibition alters the expression of many stimulation-induced genes, suppresses macrophage migration and phagocytosis, and increases the proliferation of the intracellular parasite Toxoplasma gondii. Together, these results thoroughly uncover the dynamic reprogramming of macrophage nucleotide metabolism upon classical activation and elucidate the regulatory mechanisms and the functional significance of such reprogramming.

Introduction

Macrophages are crucial cells of the innate immune system that can dynamically take on a spectrum of functional states. For instance, upon sensing classical activation signals associated with infection such as lipopolysaccharide (LPS, a bacterial cell wall component) and interferon-γ, macrophages first turn on functions that are pro-inflammatory and help eliminate pathogens. Then over time, these responses subside, and macrophages develop phenotypes that help promote healing and the resolution of inflammation.^1–4 Dynamic reprogramming of metabolism is often coupled to these functional transitions and plays a critical part in macrophage immune responses.

Over the past decade, growing efforts have sought to identify the distinct metabolic characteristics associated with specific macrophage functional states, and to elucidate the metabolic tipping points where alteration of key metabolic activities can orchestrate macrophage functions. Understanding the metabolic reprogramming coupled to immune responses and macrophage function has broad significance, especially given macrophages’ wide-ranging involvement in many health conditions. One example of such functionally important metabolic reprogramming, highlighted by recent studies, is that of the citrate cycle.^1–5 However, beyond central carbohydrate and lipid metabolic pathways,⁶ macrophage metabolism across broader metabolic networks remains to be better understood.

Comprehensive understanding of the metabolic reprogramming during an immune response requires answering three major questions: (1) What metabolic pathways are substantially reprogrammed (i.e., the metabolic characteristics associated with specific activation states)? (2) How do the specific metabolic changes occur (i.e., the regulatory mechanism)? and (3) Why does such metabolic remodeling happen during an immune response, and how does it affect macrophage functions (i.e., the functional significance)? Here, we started with an unbiased multi-omics analysis to characterize the dynamic metabolic reprogramming in macrophages upon classical LPS+IFNγ stimulation, and identified nucleotide metabolism, a fundamental yet largely understudied pathway in macrophages, as one of the most substantially remodeled metabolic pathways. Then through a series of isotopic tracing studies, we revealed the profound changes in activities through purine and pyrimidine de novo synthesis, degradation, and salvage pathways, and pinpointed several key regulation points which drive the systemic reprogramming of nucleotide metabolism. We further elucidated the mechanisms regulating these key reactions. Finally, we examined the impacts of such nucleotide metabolism remodeling (specifically the switch from purine de novo synthesis to salvage upon activation) on macrophage functions. The discoveries here have broad significance both in the relevance to health conditions involving macrophage responses, and in basic science questions regarding the fundamental process of nucleotide metabolism regulation.

Results

Systematic reprogramming of nucleotide metabolism in macrophages upon classical activation

Multi-omics analysis identifies nucleotide metabolism as a substantially rewired pathway

To systematically identify the metabolic pathways that are reprogrammed in response to classical activation, we first performed a metabolomics analysis comparing murine bone marrow-derived macrophages (BMDMs) stimulated with lipopolysaccharide and interferon-γ (LPS+IFNγ) for 24h to unstimulated BMDMs. Pathway enrichment analysis⁷ highlighted the most significantly and profoundly impacted pathways as shown in Fig. 1a, which include several groups: (1) the arginine metabolism pathway, whose remodeling is one of the classic markers of macrophage polarization;⁸ (2) the citrate cycle, which has been underscored by more recent studies for its key role in orchestrating macrophage immune responses;^1–5,9–12 (3) purine metabolism, pyrimidine metabolism, and the pentose phosphate pathway (PPP), which are all parts of nucleotide metabolism; and (4) aspartate, alanine, and glutamine metabolism, which is closely connected to both the citrate cycle and nucleotide metabolism (Fig. 1b). Overall, this analysis not only is consistent with prior knowledge in identifying the significance of arginine metabolism and the citrate cycle, but also identified nucleotide metabolism, which has been underappreciated in macrophages, as a metabolic pathway whose substantial rewiring is strongly associated with classical activation.

Figure 1: Multi-omics analysis identifies nucleotide metabolism as a substantially reprogramed metabolic pathway upon classical stimulation.

a) Pathway enrichment analysis comparing metabolomic changes in unstimulated and 24h continually stimulated BMDMs. b) Several pathways identified in Fig. 1a are directly connected to the nucleotide metabolism process. c) Metabolomic changes in pentose phosphates, nucleotides, nucleosides, and nitrogenous bases in BMDMs over a timecourse of continual or acute stimulation. d) Principal component analysis of transcriptomic changes in BMDMs over a timecourse of acute stimulation. e) Transcriptional changes in nucleotide metabolism in BMDMs over a timecourse of acute stimulation. Genes involved in nucleotide de novo synthesis (blue) or degradation (green) are indicated by color codes of gene names. c, e) Relative (c) metabolite abundances or (e) gene expression levels are compared to unstimulated cells (0h) and displayed on a log2 scale as a heatmap, with saturating color representing 10-fold change (or 3.32 on log2 scale). Each box represents the mean of n=3 independent samples.

Given the growing recognition for the highly dynamic nature of metabolic reprogramming during macrophages’ immune responses,^5,13 we further profiled the changes in nucleotide metabolism over a timecourse upon either continual or acute stimulation with classical activation signals. (For continual stimulation, macrophages were exposed to LPS+IFNγ over several days. For acute stimulation, macrophages were stimulated with LPS+IFNγ for only 2h, then the stimulants were removed, and cells were cultured in standard conditions.) Dynamic changes were observed throughout the nucleotide metabolism network. Specifically, pentose phosphates accumulated substantially upon stimulation, which is reversible upon acute stimulation (peaking at 24h) but maintained upon continual stimulation (Fig. 1c). The majority of nucleotides remained largely steady upon both continual and acute stimulation, with two notable exceptions: (1) Inosine monophosphate (IMP), the hub for both de novo synthesis and salvage of purine nucleotides, built up over 10-fold after 48–72h of continual stimulation; (2) dTTP depleted over time to undetectable levels after 48h upon continual stimulation, and transiently depleted 24–72h after acute stimulation, then began to recover at 96h (Fig. 1c). Most nucleosides and nitrogenous bases increased substantially and remained highly elevated (up over 10-fold) after 24h continual stimulation (Fig. 1c). Upon acute stimulation, accumulation was also observed in inosine, hypoxanthine, and xanthine, peaking at 24–48h post-stimulation (Fig. 1c). (Many other nucleosides were not well detected upon acute stimulation due to their low baseline abundance.) Overall, the metabolomics analysis revealed that both continual and acute stimulation induce substantial reprogramming of macrophage nucleotide metabolism after 24h. The changes are generally stronger when the cells are continually exposed to stimuli, and many changes are reversible upon acute stimulation after a long period (72–96h).

Substantial reprogramming of nucleotide metabolism is often associated with changes in cell cycle. However, it is worth noting that BMDMs in this setting, whether stimulated or not, do not proliferate and remain highly viable over the timecourse of both continual and acute stimulation (Extended Data Fig. 1a–b). Interestingly, we also observed similar changes in nucleotide metabolism compounds upon continual LPS+IFNγ stimulation in another widely used macrophage cell model, the RAW 264.7 cell line. This includes the depletion of dTTP, and the strong accumulation of IMP, pentose phosphates, nucleosides, and nitrogenous bases (Extended Data Fig. 1c). However, unlike terminally differentiated BMDMs, RAW 264.7 cells do proliferate in the unstimulated state but stop proliferation upon stimulation (Extended Data Fig. 1d). These results demonstrate that the dynamic rewiring of nucleotide metabolism is consistently associated with the macrophage response to classical stimulation across cell models. Although such nucleotide metabolism reprogramming is not due to changes in proliferation state, it may contribute to growth arrest, as observed in RAW 264.7 cells.

We further conducted a transcriptomic analysis in BMDMs over a timecourse upon acute stimulation. Principal component analysis reveals that the gene expression profile evolved over the timecourse and returned to near baseline at 96h post-stimulation (Fig. 1d). Among the top 10 genes most aligned with the major changes represented by principal component 1 (PC1) are many immune response genes (Cxcl10, Il12b etc.). Also included are important metabolic genes including Nos2 (the well-recognized metabolic marker for classical activation which functions in arginine metabolism), Irg1 (an enzyme branching out of the citrate cycle), and the nucleotide degradation gene Upp1 (uridine phosphorylase) (Extended Data Fig. 1e). This highlights the temporal correlation between functional changes and metabolic rewiring across the course of the immune response. Many other nucleotide degradation genes, such as Pnp1 and Pnp2 (purine nucleoside phosphorylase) showed similar dynamic trends as Upp1—a large increase in expression upon acute stimulation that returned to baseline over time. Conversely, many genes in nucleotide de novo synthesis and the conversion of ribonucleotides to deoxyribonucleotides (Rrm1 and Rrm2) showed the opposite trend (Fig. 1e). The GO term “regulation of nucleobase-containing compound metabolic process” is highly enriched for its dynamic changes along the first principal component (adjusted p=1.47 × 10⁻⁹). Taken together, the transcriptomic and metabolomic results are highly consistent and show that nucleotide metabolism undergoes substantial reprogramming upon classical stimulation.

Nucleotide de novo synthesis is shut off upon LPS+IFNy stimulation

We next sought to understand the changes in metabolic fluxes through nucleotide metabolism pathways that underly the accumulation and depletion of nucleotide metabolites. First, to examine nucleotide de novo synthesis, we performed kinetic isotopic tracing in stimulated BMDMs (continually stimulated with LPS+IFNγ for 48h) or unstimulated BMDMs (cultured for the same duration without stimulation) using a γ−¹⁵N-glutamine tracer. Because glutamine is a main nitrogen donor for both purine and pyrimidine de novo synthesis, the rates of labeled nitrogen incorporation from glutamine into downstream nucleotide synthesis intermediates (quantified by LC-MS, see methods for detail), reflects the activities of different purine and pyrimidine synthesis reactions (Fig. 2a). Within 2h after switching cells to media containing labeled glutamine, intracellular glutamine becomes fully labeled in both stimulated and unstimulated macrophages (Fig. 2b). Likewise, the pyrimidine synthesis intermediates carbamoyl aspartate and orotate quickly become nearly fully labeled (Fig. 2c).

Figure 2: Kinetic glutamine labeling reveals blockages in nucleotide de novo synthesis.

a) Schematic showing labeling incorporation from γ−¹⁵N-glutamine into nucleotide metabolism. b–j) Labeling kinetics of indicated intracellular metabolites of nucleotide metabolism in unstimulated or 48h continually stimulated BMDMs. Cells were labeled with γ−¹⁵N-glutamine for 0–8 hours as indicated on x-axis. Mean ± SD (n=3 independent samples). Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated and stimulated cells at the 8h time point. ns indicates not significant (p>0.01).

Labeling rates into downstream pyrimidine nucleotides were notably different in each stimulation state: labeling into UMP (Fig. 2d), and similarly UDP and UTP (Extended Data Fig. 2a), are significantly faster in stimulated macrophages, while the total levels of UMP are similar (Extended Data Fig. 2b), suggesting stimulation induces a slight increase in pyrimidine synthesis up to UMP. In striking contrast, labeling incorporation into nucleotides downstream of UMP only occurs in unstimulated macrophages but is completely abolished in stimulated macrophages (Fig. 2e–f, Extended Data Fig. 2c). These pyrimidine nucleotides include dTMP, which is synthesized from labeled UMP via thymidylate synthase (TYMS) and thus labeled with one 15N (denoted as M+1), and CTP and dCTP (M+1 or M+2), both of which are synthesized from UTP via CTP synthase (CTPS). We also conducted a similar experiment in unstimulated and stimulated RAW 264.7 macrophages and observed consistent results: labeling from γ−¹⁵N-glutamine was actively incorporated into pyrimidine de novo synthesis intermediates up to UMP in both stimulated and unstimulated macrophages (Extended Data Fig. 2d). However, the further incorporation into dTMP and CTP was only highly active in unstimulated cells, whereas it was completely lost in stimulated cells (Extended Data Fig. 2e). These results clearly demonstrated that while macrophages in both stimulated and unstimulated states actively perform pyrimidine nucleotide de novo synthesis up to uridine nucleotides, the synthesis of cytidine and thymidine nucleotides, through CTPS and TYMS respectively, is shut off upon LPS+IFNγ stimulation.

Similarly, we found drastic changes in purine de novo synthesis upon stimulation. 5-Aminoimidazole-4-carboxamide ribonucleotide (AICAR), the last purine de novo synthesis intermediate before the synthesis of IMP, rapidly became fully labeled (M+2) from γ−¹⁵N-glutamine in both stimulated and unstimulated states (Fig. 2g). Unstimulated BMDMs continue to incorporate the labeling through into IMP (Fig. 2h), as well as nucleotides downstream of IMP, such as adenine nucleotides (Fig. 2i, Extended Data Fig. 2f). In sharp contrast, such labeling is completely lost in stimulated BMDMs. Consistently, in RAW 264.7 macrophages, even though labeling up to AICAR was preserved, labeling into IMP, AMP, and XMP was profoundly inhibited upon stimulation (Extended Data Fig. 2g). These results demonstrate that the last step of purine de novo synthesis, which converts AICAR to IMP, catalyzed by the bi-functional enzyme AICAR transformylase/IMP cyclohydrolase (ATIC), is blocked upon stimulation.

Macrophages switch to nucleotide salvage

While stimulation causes blockages in both pyrimidine and purine de novo synthesis, we observed that only dTTP is depleted upon stimulation, as expected from blocked TYMS flux. Otherwise, the abundances of most purine nucleotides were largely maintained, and IMP even accumulated (Fig. 1c). This suggests cells switch to other means to obtain purine nucleotides upon stimulation. To this end, we made use of the fact that with γ−¹⁵N-glutamine tracing, different pathways generate different labeled forms of GMP: de novo synthesis generates GMP with 3 labeled nitrogen atoms (M+3) whereas when cells salvage unlabeled hypoxanthine, xanthine, or inosine to produce GMP, it generates GMP with only one labeled nitrogen (M+1), which is incorporated by the reaction of GMP synthase (Fig. 2a). Kinetic tracing with γ−¹⁵N-glutamine revealed that even though the generation of M+3 GMP through de novo synthesis was completely suppressed in stimulated BMDMs, the rate of M+1 GMP production increased substantially (Fig. 2j), suggesting stimulated macrophages shift to purine nucleotide salvage.

We next quantified how the contributions of salvage and de novo synthesis change over the timecourse of continual or acute stimulation by pseudo-steady state γ−¹⁵N-glutamine tracing in BMDMs stimulated for various durations. Upon continual stimulation, the fraction of M+1 GMP (indicating GMP produced by salvage) increased gradually over time; the fraction of M+3 GMP (indicating de novo synthesis) decreased to near zero by 48h (Fig. 3a). Similarly, upon acute stimulation, the fraction of M+1 GMP increased, and the fraction of M+3 GMP decreased to near zero by 48–72h. However, it starts to recover at 96h (Fig. 3a). Together, these data demonstrate that cells shift from purine de novo synthesis towards salvage upon stimulation. Acute stimulation is sufficient to induce the substantial switch, and the effect can persist for several days but eventually can recover.

Figure 3: Macrophages shift to purine salvage upon stimulation.

a) Labeling pattern of GMP in BMDMs over a timecourse of continual or acute stimulation. Cells were labeled with γ−¹⁵N-glutamine for 24h before analysis. b–e) Total labeled fraction of (b) PRPP, (c) IMP, (d) UTP and (e) dTTP in BMDMs over a timecourse of continual stimulation. Cells were labeled with U-¹³C-glucose for 24h before analysis. f) Labeling kinetics of intracellular inosine (M+4) in unstimulated or stimulated BMDMs. Cells were supplemented with 50 μM U-¹⁵N-inosine for 0–120 minutes as indicated on x-axis. g) Concentration of inosine remaining in media after unstimulated or stimulated BMDMs were incubated with media supplemented with 50 μM inosine for 0–30 minutes. h) Fold change of intracellular inosine in unstimulated or stimulated BMDMs after supplementing with 50 μM U-¹⁵N-inosine for 30 minutes, compared to cells of the same stimulation state without inosine supplementation. i-j) Labeled fraction of intracellular IMP (M+4) and fold change of total IMP level in unstimulated or stimulated BMDMs, after supplementing with 50 μM U-¹⁵N-inosine for 30 minutes. k) Relative abundance of intracellular IMP in unstimulated or stimulated BMDMs, with or without supplementation of 50 μM hypoxanthine for 48 hours. Data are presented relative to abundance in unstimulated, unsupplemented cells. l) Fold change of intracellular IMP in wildtype or Hprt KO RAW 264.7 cells, either unstimulated or stimulated. f–l) Stimulated cells are continually stimulated for 48h. a–l) Mean ± SD (n=3 independent samples). n.d. indicates not detected. a–e) Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells (0h). ns indicates not significant (p>0.01). f–j) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated and stimulated cells. The analysis for f) was performed at the final timepoint. ns indicates not significant (p>0.01). k–l) Statistical analysis was performed using one-way ANOVA followed by post hoc Tukey’s test. Bars with different lower-case letters (a, b, c, or d) indicate a statistically significant difference with p<0.05.

Both nucleotide de novo synthesis and salvage allow macrophages to turn over their nucleotide pools. To examine general nucleotide turnover rate, we also performed tracing with U-¹³C-glucose. U-¹³C-glucose can be incorporated into the ribose moiety of nucleotides through both de novo synthesis and the salvage of nitrogenous bases via the common precursor for purine and pyrimidine nucleotides, PRPP. Therefore, the nucleotide turnover rate (combination of de novo synthesis and salvage) can be probed by measuring the labeling incorporation from PRPP to nucleotides. PRPP labeled completely (M+5) from U-¹³C-glucose in both unstimulated and stimulated BMDMs (Fig. 3b). The labeling of IMP is generally maintained at a high fraction throughout the stimulation timecourse (Fig. 3c), suggesting overall purine nucleotide turnover remained highly active upon stimulation despite de novo synthesis of IMP shutting off. For pyrimidine nucleotides, the turnover of UTP is maintained (Fig. 3d). However, labeling into dTTP is no longer detected after 24h of stimulation (Fig. 3e), suggesting stimulated macrophages do not significantly salvage thymine by attaching a new ribose moiety. Furthermore, we found expression of both thymidine kinase 1 and 2, which are required for thymidine salvage, are greatly reduced upon stimulation (Supplementary Fig. 1). The lack of both de novo synthesis and salvage can cause the unique depletion of dTTP among all nucleotides (Fig. 1c).

As the results above showed that stimulated macrophages switch to relying on the salvage pathway to maintain active turnover of purine nucleotides, we next directly assessed the changes in cells’ capability to salvage purine nucleosides and bases. When supplemented with U-¹⁵N-inosine, BMDMs rapidly take it up from media, and intracellular inosine quickly becomes fully labeled (Fig. 3f–g). Compared to unstimulated BMDMs, stimulated BMDMs uptake inosine at a significantly faster rate (Fig. 3g) and have a ~10-fold higher buildup of intracellular inosine (Fig. 3h), suggesting stimulated macrophages have a greater potential to utilize extracellular inosine. Both unstimulated and stimulated macrophages can salvage inosine to produce IMP, as indicated by the generation of labeled IMP from U-¹⁵N-inosine. Noticeably, both the fraction of labeled IMP and the increase in IMP total abundance upon inosine supplementation are significantly higher in stimulated macrophages (Fig. 3i–j), which is consistent with their greater capability to salvage purine nucleosides. Similarly, when cells are supplemented with hypoxanthine, IMP accumulates, and the accumulation is greater in stimulated macrophages (Fig. 3k), showing stimulated macrophages can also actively salvage purine bases via hypoxanthine-guanine phosphoribosyltransferase (HPRT).

We hypothesized that the increased ability to salvage purine bases (via HPRT or adenine phosphoribosyltransferase (APRT)) and nucleosides (either directly by nucleoside kinases, or indirectly by first breaking the nucleosides down to bases then salvage via HPRT or APRT) is the reason that total intracellular IMP level is increased in stimulated macrophages, despite that IMP de novo synthesis is shut off. To test this hypothesis, we generated two Hprt knockout lines in RAW 264.7 cells, and found knocking out Hprt nearly completely prevented the accumulation of IMP in stimulated macrophages (Fig. 3l), underscoring the importance of salvage via HPRT upon stimulation.

Nucleotide degradation is increased but complete oxidation of purine bases is blocked upon LPS+IFNγ stimulation

The metabolomics results highlighted that among the most significant stimulation-induced changes are the buildup of nucleosides and nitrogenous bases (Fig. 1c). These metabolites are the products of nucleotide degradation and can serve as substrates for salvage. Moreover, the RNAseq analysis revealed several important nucleotide degradation enzymes, including Pnp1, Pnp2, Upp1, and the multifunctional enzyme Lacc1 which can function in nucleotide degradation,^14,15 increased substantially upon stimulation (Fig. 1e). Therefore, we hypothesized that nucleotide degradation is significantly remodeled in response to LPS+IFNγ stimulation. To probe nucleotide degradation, we first performed kinetic U-¹³C-glucose tracing and measured the labeling incorporation from nucleotide to nucleoside. U-¹³C-glucose is incorporated into nucleotides like IMP quickly in both unstimulated and stimulated BMDMs (Extended Data Fig. 3a). The kinetic delay of inosine labeling compared to IMP labeling indicates IMP degradation activity. We found that compared to unstimulated BMDMs, stimulated BMDMs incorporate labeling from IMP to inosine much faster (Extended Data Fig. 3b), indicating increased IMP degradation, which likely contributes to the observed accumulation of inosine upon stimulation.

Nucleosides, like inosine, can be further broken down to corresponding bases, and both nucleosides and bases can be released extracellularly by macrophages.¹⁶ We found that not only did the intracellular levels of nucleosides and nitrogenous bases accumulate upon stimulation (Fig. 1c), the release of most detected nucleosides and bases into the media also increased greatly upon stimulation (Extended Data Fig. 3c). These results, consistent with increased Pnp1/2, Lacc1, and Upp1 expression, suggest that in addition to increased nucleotide degradation to nucleosides, the degradation of nucleosides to nitrogenous bases likely also increased upon stimulation.

To directly measure purine nucleoside degradation capability, we supplemented BMDMs with U-¹⁵N-inosine and followed the metabolic fate of inosine. In both stimulated and unstimulated states, macrophages rapidly convert the inosine to hypoxanthine, which became nearly fully labeled (M+4) and accumulated greatly within 30 minutes (Fig. 4a). Noticeably, stimulated macrophages built up ~5-fold more hypoxanthine from inosine supplementation. The hypoxanthine generated was further released into media, with a release rate 2–3-fold faster in stimulated BMDMs (Fig. 4b). Both observations are consistent with stimulation increasing inosine degradation activity. Hypoxanthine can be further degraded to xanthine and then urate via the action of xanthine oxidoreductase (XOR). In striking contrast to higher labeled hypoxanthine level, we found stimulated macrophages accumulated much less labeled intracellular xanthine (M+4) from inosine supplementation—despite a higher baseline xanthine level (M+0) (Fig. 4c). The release of inosine-derived xanthine into media is much slower in stimulated macrophages as well (Fig. 4d). Even more striking differences were observed in the final degradation product urate— inosine supplementation resulted in urate released into media by unstimulated but not stimulated macrophages (Fig. 4e), and a much lower total intracellular urate level in stimulated cells compared to unstimulated cells (Fig. 4f). Together, these results revealed that an increase in PNP upon stimulation supports increased hypoxanthine production, but the activity of XOR to completely degrade hypoxanthine to urate is strongly decreased. This can contribute to the observed accumulation of hypoxanthine, diverting flux away from its complete degradation, and instead into its salvage to IMP via HPRT.

Figure 4: Alterations in nucleotide degradation upon stimulation.

a) Relative intracellular abundance of M+0 and M+4 hypoxanthine in unstimulated or stimulated BMDMs supplemented with 50 μM U-¹⁵N-inosine for 30 minutes. Data are presented relative to the level of M+0 hypoxanthine in unstimulated BMDMs. b) Release of hypoxanthine into media over time by unstimulated or stimulated BMDMs, supplemented with 50 μM inosine. c) Relative intracellular abundance of M+0 and M+4 xanthine in unstimulated or stimulated BMDMs supplemented with 50 μM U-¹⁵N-inosine for 30 minutes. Data are presented relative to the level of M+0 xanthine in unstimulated BMDMs. d–e) Release of (d) xanthine and (e) urate into media over time by unstimulated or stimulated BMDMs, supplemented with 50 μM inosine. f) Relative intracellular abundance of urate in unstimulated or stimulated BMDMs, supplemented with 50 μM inosine for 10 minutes. Data are presented relative to abundance in unstimulated cells. a–f) Stimulated BMDMs are continually stimulated for 48h. g–h) Relative intracellular abundance of (g) hypoxanthine, and (h) IMP in BMDMs over a timecourse of continual stimulation, with or without treatment of 10 μM forodesine (FD) throughout the timecourse. Data are presented relative to abundance in unstimulated, untreated cells. a–h) Mean ± SD (n=3 independent samples). b, d–f) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated to stimulated cells at each time point. g–h) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing untreated to FD treated cells at each time point. ns indicates not significant (p>0.01).

To further validate the role of PNP in generating hypoxanthine accumulation and providing substrate for salvage, we treated macrophages with the PNP inhibitor forodesine. Forodesine treatment caused the accumulation of PNP’s substrate, inosine, as expected (Extended Data Fig. 3d), and completely prevented the stimulation-induced accumulation of hypoxanthine (Fig. 4g). Furthermore, forodesine treatment completely prevented the stimulation-induced accumulation of IMP (Fig. 4h). This agrees with the findings in Hprt KO cells (Fig. 3l), demonstrating salvage of PNP-produced hypoxanthine via HPRT is the main source of IMP buildup in stimulated macrophages. Consistently, by limiting hypoxanthine availability, PNP inhibition significantly reduced the fraction of GMP that resulted from salvage, as indicated by a lower M+1 fraction from γ−¹⁵N-glutamine tracing (Extended Data Fig. 3e), and caused further buildup of the other HPRT substrate, PRPP, in stimulated macrophages (Extended Data Fig. 3f).

Mechanisms driving the reprogramming of nucleotide metabolism

The studies above revealed systematic rewiring of nucleotide metabolism in macrophages upon LPS+IFNγ stimulation and identified a series of important regulation points, as summarized in Fig. 5a. These include: (1) De novo synthesis of purine nucleotides is profoundly inhibited, and the blockage particularly occurs at the step of IMP synthesis from AICAR, catalyzed by ATIC; (2) De novo synthesis of pyrimidine nucleotides is also profoundly inhibited, and the blockage occurs at CTP and dTMP production from UMP, catalyzed by CTPS and TYMS, respectively; (3) Cells switch to salvage to maintain many nucleotides; (4) Nucleotide degradation to nucleosides and nitrogenous bases is generally increased, but full degradation of purine bases through the reaction catalyzed by XOR is reduced, shunting flux into purine salvage. We next sought to elucidate the mechanisms driving these key changes in nucleotide metabolism reactions.

Figure 5: NO plays a key role in nucleotide metabolism reprogramming.

a) Summary of nucleotide metabolism rewiring in macrophages upon stimulation. Changes in metabolite levels over time are presented by color code in heatmap bars compared to unstimulated cells. All the significant changes in transcript level at 24h after acute stimulation are presented as color codes of the gene name. Blockages of reaction activity are indicated by blue X and increases of reaction activity are indicated by red arrows. b–c) Changes in iNOS protein and Tyms expression in BMDMs over a timecourse of continual stimulation. d–g) Labeled fraction of ATP (M+2), GTP (M+1 and M+3), UTP (M+1), and the ratio of labeled CTP (M+2) to labeled UTP (M+1), after 24h γ−¹⁵N-glutamine labeling in unstimulated or stimulated wildtype or Nos2^−/− BMDMs, with or without treatment of 200 μM DETA-NONOate. n.d. indicates not detected. h) Left: Relative intracellular dTTP abundance in unstimulated or stimulated wildtype or Nos2^−/− BMDMs. Right: Relative intracellular dTMP abundance in unstimulated or stimulated wildtype or Nos2^−/− RAW 264.7 cells. i) Expression of selected genes in nucleotide metabolism, probed by qPCR in unstimulated or stimulated wildtype or Nos2^−/− BMDMs, with or without treatment of 200 μM DETA-NONOate. Data are presented relative to expression in unstimulated, untreated wildtype cells and displayed on a heatmap. Each box represents the mean of n=3 independent samples. j–l) Relative (j) Tyms expression and relative intracellular abundance of (k) hypoxanthine, and (l) IMP in unstimulated or stimulated wildtype or Nos2^−/− BMDMs, with or without treatment of 200 μM DETA-NONOate throughout the timecourse. d–l) Stimulated cells are continually stimulated for 48h. c–h, j–l) Mean ± SD (n=3 independent samples). c) Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells (0h). d, f–g, j–l) Statistical analysis was performed using one-way ANOVA followed by post hoc Tukey’s test. Bars with different lower-case letters (a, b, c, or d) indicate a statistically significant difference with p<0.05. h) Statistical analysis was performed using unpaired two-tailed student’s t-test comparing the two indicated groups.

One-carbon metabolism supplements do not rescue the inhibition of nucleotide de novo synthesis

Both the reactions catalyzed by ATIC and TYMS require one-carbon units supplied by the folate pathway (10-formyl-THF and 5,10-methylene-THF, respectively) as substrates to support the production of IMP and dTMP. Moreover, transcriptomic analysis revealed stimulation causes a significant reduction in the expression of some folate pathway enzymes (Shmt1, Shmt2, Dhfr, Mthfd1, etc.) and the rate-limiting enzyme in the serine de novo synthesis pathway, Phgdh, which feeds into one-carbon metabolism (Extended Data Fig. 4a). Therefore, we tested the possibility that a lack of one-carbon units causes the reduced flux through ATIC and TYMS.

First, we supplemented BMDMs with 1mM ¹³C-formate, an established method to increase the supply of one-carbon units.^17–21 With this supplement, AICAR became nearly fully labeled (M+1) in both unstimulated and stimulated cells (Extended Data Fig. 4b), demonstrating ¹³C-formate is sufficiently metabolized into available one-carbon units, which is then used to supply AICAR synthesis. In unstimulated macrophages, the action of ATIC further adds another labeled carbon from 10-formyl-THF to AICAR, generating M+2 labeled IMP (Extended Data Fig. 4c), as well as labeled downstream nucleotides such as ATP and GTP (Extended Data Fig. 4d–e). However, M+2 labeled IMP, ATP, or GTP was completely absent in stimulated cells (Extended Data Fig. 4c–e), indicating that formate supplementation did not rescue the blockage at ATIC upon stimulation. Nor did it rescue dTTP synthesis— no labeled dTTP was detected from ¹³C-formate, and the dTTP depletion upon stimulation persisted despite formate supplementation (Extended Data Fig. 4f).

It is possible that insufficient levels of the THF cofactor, rather than one-carbon units (which would be supplied by formate), can limit one-carbon metabolism and thus further limit ATIC and TYMS flux. To test this possibility, we supplemented RAW 264.7 cells with folinate, which has been shown effective in rescuing limited folate availability, such as that caused by treatment with the anti-folate drug methotrexate.^22,23 Indeed, treating unstimulated cells with methotrexate to inhibit folate metabolism can strongly inhibit IMP and dTTP de novo synthesis, as evidenced here by the substantially decreased labeling into these compounds from the γ−¹⁵N-glutamine tracer (Extended Data Fig. 4g–h), providing a positive control. However, there is an important difference between methotrexate-induced and stimulation-induced inhibition of nucleotide de novo synthesis: while the former can be fully rescued by folinate supplementation, stimulation-induced inhibition is not affected by folinate (Extended Data Fig. 4g–h). Consequentially, the depletion of dTTP caused by methotrexate treatment can be rescued by folinate supplementation, but the depletion of dTTP induced by stimulation cannot (Extended Data Fig. 4i). Together, these experiments show strong evidence that mechanisms other than insufficiency of THF or one-carbon units are primarily responsible for the inhibition of nucleotide de novo synthesis upon macrophage stimulation.

Rescuing NAD/NADH ratio and aspartate level does not revert stimulation-induced blockages in nucleotide synthesis

When macrophages are stimulated with LPS+IFNγ, respiration rate (Extended Data Fig. 5a) and citrate cycle flux reduce significantly.^1–5,9–12 Related to these changes, we observed the NAD/NADH ratio decreased significantly (Extended Data Fig. 5b), and aspartate, which is synthesized from a citrate cycle intermediate, decreased gradually over the continual stimulation timecourse (Extended Data Fig. 5c). Several previous studies have demonstrated that a decreased NAD/NADH redox ratio can limit the synthesis of nucleotides in cancer cells, because nucleotide synthesis requires electron acceptors to support the synthesis of various precursors, including aspartate.^24–29 To test whether reductive stress or the decrease in aspartate are the limiting factors for nucleotide synthesis in stimulated macrophages, we supplemented cells with pyruvate, which can act as an electron acceptor, or aspartate, as these perturbations have been shown to effectively rescue nucleotide synthesis constrained by a decreased NAD/NADH ratio in cancer cells. Pyruvate supplementation effectively rescued the decrease in cellular NAD/NADH ratio in stimulated macrophages (Extended Data Fig. 5b), and both pyruvate and aspartate supplementation completely reversed the decrease of intracellular aspartate level in stimulated macrophages (Extended Data Fig. 5d), indicating the success of these supplementations. However, neither pyruvate nor aspartate rescued the inhibition of purine de novo synthesis in stimulated macrophages, as evidenced by the persistent complete loss of 2-labeled IMP or AMP when tracing with γ−¹⁵N-glutamine (Extended Data Fig. 5e–f). Purine nucleotides continue to be mainly supplied by salvage upon stimulation, as indicated by an elevated M+1 fraction of GMP and a complete lack of M+3 fraction (Extended Data Fig. 5g). Similarly, neither pyruvate nor aspartate supplementation rescued the inhibition of CDP synthesis in stimulated macrophages, as indicated by the complete absence of both M+1 and M+2 fractions of CDP from γ−¹⁵N-glutamine labeling (Extended Data Fig. 5h), despite UMP being labeled (Extended Data Fig. 5i). These results are consistent with our finding that the inhibition of pyrimidine synthesis does not occur upstream of UMP synthesis, where aspartate is required as a substrate. Together, these results showed that the inhibition of de novo purine and pyrimidine synthesis in stimulated macrophages cannot be sufficiently explained by the changes in NAD/NADH ratio or aspartate level.

Nitric oxide drives the decrease of nucleotide de novo synthesis

To identify the regulators that may be responsible for the reprogramming of nucleotide metabolism, we searched for changes that temporally correlated with the changes in nucleotide metabolism activities. We found that: (1) The inhibition of nucleotide de novo synthesis occurs soon after the strong induction of inducible nitric oxide synthase (iNOS, encoded by the gene Nos2), which can be detected at the protein level (Fig. 5b), and consistently by the accumulation of its product citrulline (Extended Data Fig. 6a), as soon as 4–6h, and continues to increase over time. (2) Upon stimulation, the expression of many nucleotide de novo synthesis genes decreases (Fig. 1e). Most noticeably, Tyms expression decreases substantially upon both acute stimulation (Fig. 1e) and continual stimulation (Fig. 5c) before dTTP depletes. This Tyms decrease follows the strong transcriptional induction of Nos2 (Extended Data Fig. 6b). Therefore, we investigated the role of NO in driving the reprogramming of nucleotide metabolism.

To perturb NO level in macrophages, we took two approaches. To limit endogenous NO production in macrophages upon stimulation, we used BMDMs isolated from Nos2^−/− mice.³⁰ The efficacy is confirmed by the lack of citrulline accumulation over time after stimulation (Extended Data Fig. 6c). To increase cellular NO level, we treated unstimulated macrophages or Nos2^−/− macrophages with the NO donor DETA-NONOate to recapitulate the increase in NO upon stimulation in wildtype macrophages. To examine nucleotide de novo synthesis and salvage activity, we performed pseudo-steady state γ−¹⁵N-glutamine tracing, as illustrated in Fig. 2a. Across all the conditions, intracellular glutamine remained nearly fully labeled (Extended Data Fig. 6d). Strikingly, the loss of 2-labeled ATP (Fig. 5d) and 3-labeled GTP (Fig. 5e) in stimulated wildtype BMDMs, which indicates the stimulation-induced blockage of purine de novo synthesis, was rescued in stimulated Nos2^−/− BMDMs. Conversely, treating unstimulated macrophages with exogenous NO donor is sufficient to cause the complete loss of de novo purine synthesis, like that observed in stimulated macrophages (as indicated by the loss of 2-labeled ATP and 3-labeled GTP). Moreover, treating Nos2^−/− macrophages with NO donor reversed the rescue of purine de novo synthesis upon stimulation (Fig. 5d–e). These results show strong evidence that the shutdown of purine de novo synthesis upon stimulation is driven by NO production.

Mirroring the effects on purine de novo synthesis, we found the contribution of salvage to GTP production, as indicated by the fraction of M+1 labeled GTP, increases greatly in conditions where cellular NO level is high (in wildtype cells upon stimulation or in cells treated with exogenous NO donor). This finding demonstrates that NO drives the switch from de novo synthesis to salvage for purine turnover. Interestingly, overall GTP labeling turnover (i.e., de novo-derived M+3 and salvaged-derived M+1 GTP combined) increases upon stimulation. This stimulation-induced increase is independent of NO status, as it stayed the same regardless of Nos2 genotype or NO donor treatment (Fig. 5e).

As for pyrimidine de novo synthesis, pseudo-steady state γ−¹⁵N-glutamine tracing showed that UTP remained ~50% M+1 labeled in all conditions, suggesting pyrimidine de novo synthesis up to UTP is largely not altered by NO, while the labeling is slightly higher in stimulated conditions (Fig. 5f). We have demonstrated that upon stimulation, pyrimidine de novo synthesis is maintained up to UTP, but blocked at the steps of CTP and dTMP synthesis. The synthesis of CTP by CTPS is indicated by the conversion of M+1 labeled UTP to M+2 labeled CTP (Fig. 2a). We found the loss of M+2 labeled CTP upon stimulation is partially rescued in Nos2^−/− BMDMs (Fig. 5g). Furthermore, treatment with NO donor significantly reduced the ratio of M+2 CTP to M+1 UTP in unstimulated macrophages, partially recapitulating the decrease of CTPS flux upon stimulation. Additionally, treating stimulated Nos2^−/− macrophages with NO donor reversed the rescue of CTP synthesis (Fig. 5g). These results demonstrate that NO contributes to the blockage of CTPS upon stimulation. Nos2^−/− also rescued the depletion of dTTP upon stimulation in BMDMs, and to a greater extent, rescued the dTMP depletion in stimulated RAW 264.7 cells (Fig. 5h), indicating NO plays a significant role in inhibiting dTMP synthesis.

As the results above suggested that NO is a key driver in the inhibition of nucleotide de novo synthesis, we next examined whether NO’s effect occurs through, or independent of, transcriptional changes. Consistent with the transcriptomic analysis upon acute stimulation, qPCR analysis confirmed the expression of many purine de novo synthesis genes decreased upon continual stimulation (Fig. 5i). We found most of these changes are induced by stimulation in a NO-independent manner, as Nos2^−/− or treatment with NO donor had no major consistent effect on their expression (Fig. 5i). The decrease of these genes, such as Gart and Pfas, is unlikely to be strong enough to cause the complete block of overall purine de novo synthesis flux upon stimulation, because we showed flux is mostly maintained up to AICAR production. Furthermore, Nos2^−/− can nearly completely rescue purine de novo synthesis in stimulated cells despite not rescuing the expression of these upstream purine de novo synthesis genes. Nonetheless, NO-independent transcriptional downregulation of enzymes upstream of ATIC can have a minor role in regulating purine synthesis. Particularly, it may prevent the overproduction of purine synthesis precursors when the last step catalyzed by ATIC is blocked, especially at later time points. Consistently, we found over the continual stimulation timecourse that AICAR level initially accumulates strongly (peaks at 12h), which is likely due to the early induction of NO production rapidly inhibiting ATIC and cutting off AICAR consumption flux. Then AICAR decreased at later time points, likely due to the transcriptional downregulation of upstream enzymes taking effect at this time, limiting the production of AICAR (Extended Data Fig. 6e).

Regarding pyrimidine de novo synthesis, we found both Ctps1 and Ctps2 expressions decreased significantly upon stimulation (Fig. 5i), and this decrease is largely NO independent, as the expression does not vary in correlation with NO level across Nos2 genotypes and NO donor treatment (Fig. 5i). Such NO-independent decreases in Ctps1 and Ctps2 expression can act together with the NO-dependent inhibition of CTPS activity to block CTP synthesis flux from UTP upon stimulation, explaining the portion of CTP synthesis inhibition that is not fully rescued in stimulated Nos2^−/− BMDMs (Fig. 5g). Interestingly, the transcriptional downregulation of Tyms in stimulated macrophages is significantly rescued by Nos2^−/− (Fig. 5j). Additionally, NO donor treatment decreased Tyms expression in unstimulated macrophages, fully recapitulating the stimulation-induced effect, and NO donor treatment reversed the rescue of Tyms expression in stimulated Nos2^−/− macrophages. These results suggest that NO drives the transcriptional downregulation of Tyms (which is one of the broader transcriptional effects of NO); this in turn shuts down dTTP synthesis and causes dTTP depletion upon stimulation.

NO-driven inhibition of XOR diverts flux to purine nucleotide salvage

We found key regulation points in nucleotide degradation are also regulated by a combination of NO-independent transcriptional changes and NO-dependent inhibition. We showed above that upon stimulation, nucleotide degradation to nitrogenous bases is increased. Indeed, nucleotide degradation genes such as Nt5c3 and Pnp1 increase expression by many folds, and this increase is not affected by Nos2 (Fig. 5i), suggesting NO-independent transcriptional upregulation contributes to increased nucleotide degradation to nitrogenous bases.

Downstream of this, NO was found to play a critical role in the stimulation-induced inhibition of purine base degradation catalyzed by XOR. Knocking out Nos2 completely prevented the accumulation of hypoxanthine upon stimulation (Fig. 5k) and similarly prevented the accumulation of IMP upon stimulation (Fig. 5l), which, as we showed above, mainly results from increased hypoxanthine salvage. These effects can be reversed by treating Nos2^−/− BMDMs with NO donor (Fig. 5k–l). Moreover, treating unstimulated macrophages with NO donor is sufficient to cause the accumulation of hypoxanthine and IMP, recapitulating the stimulation-induced changes. These results provide evidence that NO drives the inhibition of XOR and diverts flux into purine salvage upon stimulation.

NO causes depletion of TYMS protein and inhibits ATIC and XOR at enzyme activity level

As the results above showed iNOS induction cause reduced fluxes through purine de novo synthesis (particularly at the step of ATIC), complete degradation of purine bases to urate (catalyzed by XOR), and synthesis of dTMP (catalyzed by TYMS), we next sought to understand the more detailed mechanism. The mTOR pathway has been well established as an important regulator of nucleotide metabolism, and it can sense cellular arginine level.³¹ It is possible that iNOS induction can rapidly consume arginine, and the depletion of arginine (or other programmed changes induced by stimulation) can lead to mTORC1 inhibition, leading to altered de novo purine and pyrimidine synthesis. We therefore measured the level of arginine and phosphorylation of the mTORC1 substrates 4E-BP1 and p70S6K, as a readout of mTORC1 activity. Intracellular arginine level elevated 24h and onward upon continual stimulation, and transiently increased 12–24h upon acute stimulation before returning to baseline (Fig. 6a). The increase in intracellular arginine, despite increased arginine consumption by iNOS, is likely due to greatly increased arginine uptake from media, which shows similar dynamic changes (Extended Data Fig. 7a). The phosphorylation of p70S6K and 4E-BP1 relative to total p70S6K and total 4E-BP1 did not decrease upon either continual or acute stimulation (Fig. 6b), suggesting mTORC1 is not inhibited upon classical activation. To test whether treatment of exogenous NO donor, which caused the inhibition of purine and pyrimidine synthesis (Fig. 5), causes inhibition of mTORC1, we treated unstimulated Nos2^−/− cells with the same dose of NO donor, and found such treatment did not cause notable decreases of mTOR, 4E-BP1, or p70S6K phosphorylation (Extended Data Fig. 7b). Furthermore, comparing between wildtype and Nos2^−/− cells, while Nos2 knockout rescued the classical activation induced-inhibition of purine and pyrimidine synthesis, 4E-BP1 and p70S6K phosphorylation remained the same (Extended Data Fig. 7c). These results suggest the inhibition of nucleotide de novo synthesis is not driven by mTORC1 inhibition.

Figure 6: NO causes decrease of TYMS and inhibition of ATIC and XOR at enzyme activity level.

a) Relative intracellular arginine level in BMDMs over a timecourse of continual and acute stimulation. Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells (0h). ns indicates not significant (p>0.05). b) p-p70S6K, p70S6K, p-4E-BP1, 4E-BP1, XOR, and ATIC protein levels in BMDMs over a timecourse of continual and acute stimulation. c) TYMS, XOR, and ATIC protein levels in unstimulated Nos2^−/− BMDMs over a timecourse of 200μM DETA-NONOate treatment. d) XOR and ATIC protein levels in wildtype or Nos2^−/− BMDMs that are unstimulated or stimulated, with or without treatment of 200 μM NO donor DETA-NONOate for 48h. e) Activity of purified ATIC incubated with NO donor DPTA-NONOate or buffer control for 3h. Statistical analysis was performed using unpaired two-tailed student’s t-test comparing ATIC control to ATIC + DPTA-NONOate at each time point. f) Total XOR activity (including XDH and XO activities) in lysates from wildtype or Nos2^−/− RAW 264.7 cells that are unstimulated or stimulated, with or without treatment of 200 μM NO donor DETA-NONOate for 48h. g–k) Relative intracellular abundance of (g) dTTP, (h) AICAR, (i) hypoxanthine, (j) xanthine, and (k) IMP in RAW 264.7 cells over a timecourse of continual LPS+IFNγ (red) or IL-4+IL-13 (blue) stimulation. Data are presented relative to abundance in unstimulated (0h) cells. a, e, g–k) Mean ± SD (n=3 independent samples). g–k) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing stimulated groups at each time point. l) Changes in nucleotides, nucleosides, and nitrogenous bases in BMDMs over a timecourse of continual IL-4 and IL-13 stimulation. Relative metabolite abundances are compared to unstimulated cells (0h) and displayed on a log2 scale as a heatmap. Each box represents the mean of n=3 independent samples.

We next measured the impact of NO on the protein levels of TYMS, ATIC and XOR by treating unstimulated Nos2^−/− macrophages with exogenous NO donor (Fig. 6c). With exposure to exogenous NO, TYMS decreased gradually, which is appreciable by 24h and became very profound by 48h. This dynamic at the protein level is consistent with the measured dynamics at the transcript level (downregulation of Tyms starts right after Nos2 induction and occurs before 24h, Fig. 5c, Extended Data Fig. 6b) and metabolite level (dTTP depletion peaks at 48h, Fig. 1c), These data support that NO production causes transcriptional downregulation of Tyms, then the decrease of TYMS, which then drives the depletion of dTTP.

Unlike TYMS, NO treatment did not cause noticeable changes in total ATIC or XOR protein levels (Fig. 6c). Confirming this, we found ATIC and XOR levels are similar across unstimulated or stimulated wildtype or Nos2^−/− BMDMs, with or without NO donor treatment (Fig. 6d), suggesting endogenous or exogenous NO do not have major effects on ATIC or XOR protein expression or degradation. In contrast, across these eight conditions, intracellular ATIC flux depletes with the presence of either endogenous or exogenous NO (Fig. 5d–e), suggesting NO inhibits ATIC at the enzyme activity level. To directly test this, we purified ATIC enzyme and incubated it with or without NO donor in vitro. After 3h of incubation with NO donor, ATIC enzyme activity is nearly completely lost (Fig. 6e), supporting NO can strongly inhibit ATIC activity.

Similarly, given that NO does not have noticeable effects on XOR protein level but the presence of endogenous or exogenous NO in cells is associated with strong buildup of the XOR substrate hypoxanthine (Fig. 5k), we hypothesized NO inhibits the purine degradation activity of XOR, which includes: (1) xanthine dehydrogenase (XDH) activity, that oxidizes hypoxanthine and xanthine to urate, coupled to NADH production, and (2) xanthine oxidase (XO) activity, that oxidizes hypoxanthine and xanthine to urate, coupled to ROS production. (Note: XOR is known to have other enzymatic activities besides degradation of hypoxanthine and xanthine, which is not within the focus of this study). We measured XDH and XO activity in cell lysates of macrophages that are wildtype or Nos2^−/−, unstimulated or stimulated with LPS+IFNγ for 48h, with or without treatment of exogenous NO donor. Across these eight conditions, the majority of XOR activity is XDH. Importantly, both XDH and XO activities are strongly inhibited by the production of endogenous NO (in wildtype cells upon stimulation), or by the addition of exogenous NO, and Nos2 knockout rescues the stimulation-induced loss of XOR activity (Fig. 6f), suggesting NO inhibits XOR enzyme activity.

Remodeling of nucleotide metabolism is a distinct feature upon classical activation compared to alternative activation

Interestingly, the results above showed that the remodeling of nucleotide metabolism, which we revealed as one of the most significant metabolic responses upon classical stimulation (Fig. 1a), is mechanistically closely connected to the rewiring of arginine metabolism and production of NO, which is the most recognized metabolic maker of classically activated macrophages.⁸ We further examined whether the remodeling of nucleotide metabolism is specific to classical activation as opposed to alternative activation. We compared the dynamic metabolomic changes in RAW 264.7 cells upon continual stimulation with IL-4 and IL-13, which induce alternative activation, to the metabolomic changes induced by LPS+IFNγ. It is well-established that alternatively activated macrophages preferentially metabolize arginine to ornithine by increased expression of arginase, in contrast to metabolizing arginine to NO and citrulline by expressing iNOS as in classically activated macrophages (Extended Data Fig. 7d). Consistent with this, we found that upon stimulation with IL-4 and IL-13, ornithine rather than citrulline accumulated over time (Extended Data Fig. 7e), confirming the lack of NO production. Nucleotide metabolites also show stark differences. The depletion of dTTP does not occur upon alternative activation (Fig. 6g), consistent with NO causing the transcriptional downregulation of Tyms and dTTP depletion. In contrast to the AICAR buildup upon classical activation due to ATIC inhibition, AICAR level stays steady upon alternative activation (Fig. 6h), consistent with NO causing the inhibition of ATIC. Finally, in contrast to classical activation, where hypoxanthine and xanthine accumulate because of NO-driven XOR inhibition, and IMP increases as a result of XOR inhibition shunting flux into purine salvage, upon alternative activation, there is no substantial increase of hypoxanthine, xanthine, and IMP (Fig. 6i–k). Consistently, in BMDMs, we also observed alternative activation caused distinct changes in nucleotide metabolism compared to classical activation. Most notably, hypoxanthine and xanthine deplete over time instead of accumulating, and dTTP is slightly elevated instead of depleting (Fig. 6l). These results demonstrated that remodeling of nucleotide metabolism is a previously unappreciated metabolic characteristic that distinguishes classically activated macrophages from alternatively activated macrophages, and further validated NO’s key role in mediating the classical activation-specific reprogramming of nucleotide metabolism.

Remodeling of nucleotide metabolism in peritoneal macrophages stimulated in vivo and in vitro

To further examine NO’s role in nucleotide metabolism remodeling in vivo, we peritoneally injected LPS+IFNγ into wildtype or Nos2^−/− mice, and harvested CD11b⁺F4/80⁺ cells (Extended Data Fig. 7f, Supplementary Fig. 2). Like observed in BMDMs, we found that in CD11b⁺F4/80⁺ cells from LPS+IFNγ injected Nos2^−/− mice, the relative level of dTTP is higher and IMP is lower compared to wildtype (Extended data Fig. 7g–h). We acknowledge that many factors can be different in vivo, including the peritoneum being a physiologically hypoxic environment. Therefore, we also isolated primary peritoneal macrophages from wildtype or Nos2^−/− mice and stimulated them in vitro in a controlled, non-hypoxic culture environment. Similar to the observations from the in vivo peritoneal injection experiment as well as in vitro cultured BMDMs and RAW 264.7 cells, knockout of Nos2 significantly reduced the classical stimulation-induced depletion of dTTP and accumulation of IMP as observed in wildtype cells (Extended Data Fig. 7i–j). These results are consistent with NO playing a robust regulatory role in nucleotide remodeling upon classical activation.

Alteration of nucleotide metabolism impacts macrophage function and pathogen growth

Given the remarkable switch that macrophages make upon stimulation to relying on salvage instead of de novo synthesis for maintaining purine nucleotides, we next investigated the functional significance of such a switch.

To this end, we used genetic and pharmacological approaches to perturb HPRT, a key enzyme in purine salvage that is crucial for nucleotide metabolism especially when de novo synthesis is limited.^29,32 We generated two Hprt knockout clones in RAW 264.7 cells, both of which were confirmed by sequencing (Supplementary Fig. 3) and remain highly viable (Extended Data Fig. 8a). Tracing with γ−¹⁵N-glutamine was performed to validate the metabolic impact of the knockout. As expected, Hprt knockout significantly reduced the stimulation-induced increase in purine salvage, as well as the baseline contribution of purine salvage (as indicated by the M+1 labeled GTP fraction, Extended Data Fig. 8b). But the stimulation-induced increase in purine salvage is not completely abolished, likely because there is still expression of other purine salvage enzymes such as Aprt (Extended Data Fig. 8c). With the significant reduction of purine salvage, which is particularly prominent upon classical activation, stimulated Hprt knockout cells compared to stimulated wildtype showed a 4-fold greater accumulation in hypoxanthine and PRPP, the substrates of HPRT, while IMP, the product of HPRT, is lower. The levels of many other purine nucleotides downstream of IMP are also reduced in Hprt knockouts (Extended Data Fig. 8d). In primary BMDMs, due to technical difficulties in applying genetic knockout with clonal selection, we assessed the impact of HPRT using a HPRT inhibitor, 6-mercaptopurine (6-MP).³³ Consistently, 6-MP treatment significantly reduced the M+1 labeled GTP fraction in γ−¹⁵N-glutamine tracing (Extended Data Fig. 8e). These results validate the effective perturbation of purine salvage.

Purine salvage by HPRT impacts macrophage gene expression

To characterize the functional impact of HPRT, we first performed RNAseq analysis in unstimulated and stimulated Hprt KO macrophages and compared the transcriptomic profiles to their wildtype counterparts. Hprt KO had a greater impact on global gene expression in the stimulated state compared to the unstimulated state (Extended Data Fig. 8f–g), consistent with cells increasing their reliance on HPRT-dependent salvage upon stimulation. Many macrophage response genes that are strongly induced upon classical activation were dampened by Hprt knockout. Among the most significant changes are Ptgs2, an enzyme that produces prostaglandins in classically activated macrophages, and Mmp14, a matrix metalloproteinase (MMP) important for macrophages’ remodeling of extracellular matrix (Fig. 7a). We then specifically validated them in BMDMs using 6-MP treatment, and found that like in Hprt knockout RAW 264.7 cells, HPRT inhibition in BMDMs significantly reduced the stimulation-induced Ptgs2 and Mmp14 expression (Fig. 7b). Similarly, expression of Il10, a cytokine produced by macrophages that is particularly important at later timepoints upon stimulation, was also dampened by Hprt KO (Fig. 7c). ELISA of spent media further independently validated that consistent with the transcript level changes, Hprt knockout in RAW 264.7 cells or 6-MP treatment of BMDMs greatly suppressed the release of IL-10 upon classical activation (Fig. 7d).

Figure 7: Alteration of nucleotide metabolism impacts macrophage function and pathogen proliferation.

a) Transcript level of Ptgs2 and Mmp14 in unstimulated or stimulated wildtype or Hprt KO RAW 264.7 cells. b) Expression of Ptgs2 and Mmp14 probed by qPCR in unstimulated or stimulated wildtype BMDMs, with or without treatment of 20 μM 6-mercaptopurine (6-MP) throughout the timecourse. Data are presented relative to expression in unstimulated, untreated cells. c) Transcript level of Il10 in unstimulated or stimulated wildtype or Hprt KO RAW 264.7 cells. d) Release of IL-10 by unstimulated or stimulated wildtype or Hprt KO (2 clones) RAW 264.7 cells (left), and wildtype BMDMs with or without 6-mercaptopurine treatment (right). c–d) n.d. indicates not detected. a–d) Stimulated cells are continually stimulated for 48h. Mean ± SD (n=3 independent samples). e) Migration of wildtype and Hprt KO (2 clones) RAW 264.7 cells across a Matrigel-coated transwell towards LPS. Mean ± SD (n=5 independent samples). f–g) Phagocytosis of pHrodo-labeled Salmonella enterica particle by (f) wildtype or Hprt KO RAW 264.7 cells and (g) BMDMs treated with or without 6-mercaptopurine. Mean ± SD (n=10 independent samples). h) Total fluorescence signal from T. gondii at indicated time points after wildtype or Hprt KO RAW 264.7 macrophages were infected with mCherry ME49 T. gondii. Mean ± SD (n=4 independent samples). a–d) Statistical analysis was performed using unpaired two-tailed student’s t-test comparing wildtype to Hprt KO or untreated to treated. e–h) Statistical analysis was performed using unpaired two-tailed student’s t-test comparing area under curve of wildtype to Hprt KO or untreated to treated. i) Representative immunofluorescence images of wildtype (left) and Hprt KO (right) RAW 264.7 macrophages infected with ME49 T. gondii for 24h. Bottom: percentage of vacuoles with 1, 2, 4, or 8 parasites per vacuole, quantified from 10 random fields of view per coverslip. j) Schematic summarizing the main findings.

HPRT is required for efficient macrophage migration and phagocytosis

GO enrichment of differentially expressed genes in Hprt KO compared to wildtype macrophages revealed many gene classes important for macrophage functions are impacted. Especially, the top process that is downregulated in Hprt KO cells (in both the unstimulated, and more profoundly, the stimulated state) is cell migration (Extended Data Fig. 9a). Upon sensing signals associated with infection such as LPS, macrophages migrate to the site of infection. To directly validate the impact of purine salvage capacity on macrophage migration and invasion, wildtype or Hprt KO macrophages were seeded in transwells coated with Matrigel and migration was induced by LPS in the bottom chamber (Extended Data Fig. 9b). Consistent with the transcriptional changes, Hprt KO macrophages showed substantially reduced transwell migration towards LPS (Fig. 7e), and similarly, towards another chemoattractant C5a (Extended Data Fig. 9c), demonstrating the importance of purine salvage in macrophage migration.

The mechanism connecting purine metabolism to cell migration is of interest. Previous studies in cancer have showed purine metabolism influences migration by stimulating serine synthesis,³⁴ so we further investigated serine synthesis in macrophages. To measure serine synthesis activity, we applied U-¹³C-glucose tracing. U-¹³C-glucose nearly completely labeled the precursor of serine synthesis, 3-phosphoglycerate (3PG). In contrast, only a very minor fraction of intracellular serine (<2% in unstimulated wildtype cells) was derived from 3PG, and this labeled fraction decreases greatly upon stimulation, to nearly undetectable levels by 48h (Extended Data Fig. 9d), suggesting relative serine de novo synthesis flux decreases with classical activation. In Hprt KO cells, the contribution of de novo synthesis to cellular serine pool is significantly increased at baseline. Nonetheless, the overall contribution remains very minor (3% or less), and classical stimulation induces a sharp decrease in serine synthesis in Hprt KO macrophages as well (Extended Data Fig. 9d). Consistent with the decreased labeling incorporation, 3-phosphoserine, an intermediate in serine synthesis, decreases upon classical activation in both wildtype and Hprt KO macrophages (Extended Data Fig. 9e). This stimulation-induced decrease in serine synthesis is likely due to transcriptional downregulation of the serine synthesis genes Phgdh, Psat1, and Psph (Extended Data Fig. 9f). Downstream of serine synthesis, enzymes that transfer one-carbon units from serine, Shmt1 and Shmt2, are significantly downregulated upon classical activation as well (Extended Data Fig. 9g). Knockout of Hprt slightly increased many of these serine synthesis and one-carbon metabolism genes, but does not prevent their stimulation-induced downregulation (Extended Data Fig. 9f–g). These results, though insufficient to explain the mechanistic connection between purine metabolism and cell migration, extend our discovery into how serine synthesis and one-carbon metabolism - two pathways connected to nucleotide metabolism-- are altered upon classical activation and by Hprt knockout.

Another core function of macrophages is phagocytosis. To assess this, we measured the phagocytosis rate of pHrodo dye labeled inactivated Salmonella enterica, a type of gram-negative bacteria, by imaging analysis. Both genetic knock out of Hprt in RAW 264.7 macrophages and 6-MP treatment of BMDMs significantly suppressed the phagocytosis of S. enterica particles (Fig. 7f–g), demonstrating the impact of HPRT in this process.

Purine metabolism impacts T. gondii proliferation in macrophages

Finally, we examined the significance of purine salvage in the context of metabolic competition between macrophages and pathogens. As one of the first lines of defense against foreign invaders, macrophages are vital for the proper clearance of pathogens. However, they can also be hijacked by some pathogens for replication and dissemination. Toxoplasma gondii is one type of intracellular parasite that invades macrophages and utilizes them as a major means for dissemination across the body,³⁵ and Type II strains of T. gondii cause macrophages to become classically activated.³⁶ Interestingly, T. gondii is purine auxotrophic;³⁷ it lacks essential purine de novo synthesis enzymes and instead acquires the purine nucleotides necessary for its replication by salvaging nucleosides and bases from its host cell (Extended Data Fig. 10a). Therefore, we hypothesized that altered host purine metabolism, particularly, host cells’ ability to salvage purine bases, can impact T. gondii replication. To test this hypothesis, we infected wildtype or Hprt KO macrophages with a type II strain of T. gondii (ME49) using its replicating tachyzoite form. The replication of T. gondii was significantly faster in Hprt KO macrophages than in wildtype (Fig. 7h). To further dissect whether the increased overall pathogen growth in the Hprt KO is due to more effective initial invasion or increased parasite replication rate, we analyzed the number of parasites per vacuole 24h after infection using microscopy. Each parasite invasion creates an intracellular niche in host macrophages called parasitophorous vacuoles, in which they will replicate to two, four, or up to eight parasites during the initial 24 hours of infection. We observed more vacuoles containing two or more parasites at 24h in Hprt KO macrophages (Fig. 7i). This result shows that T. gondii replicates faster when host Hprt is lost, likely because host HPRT activity can compete with the parasites for purine substrates, ultimately limiting parasite replication. It suggests that beyond impacting host functions, the rewiring of macrophage nucleotide metabolism can allow cells to reroute key metabolites that could otherwise be used by pathogens, impacting host–pathogen interactions.

Discussion

This study revealed that upon stimulation with classical activation signals, macrophages undergo substantial, systematic reprogramming of nucleotide metabolism. Specifically, de novo synthesis of purine and most pyrimidine nucleotides (except for UMP) is shut down, nucleotide degradation is increased, and cells switch to salvage to maintain most nucleotides (Fig. 7j).^29,32 Reprogramming of nucleotide metabolism, a fundamental metabolic pathway, is of great interest and has been mostly studied in the context of cancer and adaptive immune cells. This work, together with a recent study in dendritic cells,¹⁵ start to elucidate its importance in innate immune functions. Unlike in T-cells, where the remodeling of nucleotide metabolism is closely related to the increased metabolic demands associated with their rapid expansion upon immune activation,^20,38–43 bone marrow-derived macrophages (in the context of the current study) do not proliferate regardless of stimulation state (Extended Data Fig. 1a). Thus, there is no net nucleotide requirement for biomass expansion. Nonetheless, our data showed that general nucleotide turnover is active in both stimulated and unstimulated macrophages, and the rewiring of nucleotide metabolism upon stimulation is substantial and functionally relevant. Therefore, this work demonstrates a different type of nucleotide metabolism reprogramming that is not coupled to cell cycle but rather changes in functional state. A question from the flux balance perspective is: what do non-proliferating immune cells do with the nucleotides? And related, how would switching the specific ways of handling nucleotides be beneficial in macrophages upon activation?

In nonproliferating cells, the fates of nucleotides may include RNA and mitochondrial DNA turnover, degradation and release into the extracellular space, and serving as precursors for the synthesis of other key metabolites such as NAD+ and SAM. Many of these are important for macrophage functions and are dynamically altered upon immune activation.

To measure RNA turnover rate, we labeled cells with U-¹³C-glucose, which generates labeled nucleotides via either de novo synthesis or salvage, and traced the labeling incorporation from nucleotides to RNA. The labeling in RNA is reduced in stimulated cells compared to unstimulated, even though the labeling in nucleotides is comparable (Extended Data Fig. 10b–c), suggesting that overall RNA turnover rate is lower. In this way, the shutdown of nucleotide de novo synthesis upon classical activation is unlikely to cause a short supply of nucleotides for RNA synthesis, especially given nucleotide salvage is very active.

Another fate of nucleotides is degradation, and we found purine and pyrimidine nucleotide degradation to their corresponding bases is increased upon classical activation, enabled by the transcriptional upregulation of PNP and UPP. In situations like infection and inflammation, macrophages perform essential functions in phagocytosing and degrading pathogens, as well as clearing dead host cells and neutrophil extracellular traps (NETs, which are a network of cellular DNA mixed with anti-bacterial proteins released by activated neutrophils as a means of immune defense). All these components have high nucleic acid content. In this case, increased nucleotide degradation, nucleoside uptake, and salvage capacity could prepare macrophages for the task, while the shutdown of nucleotide de novo synthesis can be beneficial to avoid nucleotide overload. Indeed, it has been recently reported that classically activated macrophages have a greater capability to clear neutrophil extracellular traps than unstimulated macrophages,⁴⁴ which is likely related to the activation-induced changes in nucleotide metabolism we revealed here. Also consistent with this idea, here we showed that knocking out or inhibiting HPRT significantly slowed down the phagocytosis of inactivated bacteria by macrophages (Fig. 7f–g). A recent study showed that macrophages can recycle and utilize the components of phagocytosed bacteria for their own metabolism,⁴⁵ so it is likely that increased nucleotide salvage can facilitate this recycling. More dedicated future studies are merited to get a mechanistic understanding of how macrophages’ capacity to degrade and salvage nucleotides impacts their ability to phagocytose bacteria and apoptotic cells or breakdown NETs, and how the downregulation of nucleotide de novo synthesis regulates nucleotide homeostasis in this setting. On the other hand, unlike some pathogens which are phagocytosed and eliminated by macrophages, other pathogens like T. gondii can infect macrophages and hijack them for replication and dissemination. Here we found knocking out host HPRT greatly increased the replication of the type II strain of T. gondii, suggesting the remodeling of host nucleotide metabolism can also impact the metabolic competition with the pathogen. T. gondii is just one example of many intracellular protozoan parasites that are purine auxotrophic; other examples include the pathogens causing leishmaniasis and malaria. The impact of nucleotide metabolism on macrophage-pathogen metabolic competition is likely to have broader health relevance and merits future studies. It would also be valuable to examine and compare the impact of nucleotide metabolism on the replication of other types of pathogens that can grow within macrophages but are not purine auxotrophic.

Another fate and source of cellular nucleotides and nucleosides is the exchange with the extracellular environment. Proper control of nucleotides in the microenvironment is critical for macrophages’ function in inflammation regulation, particularly because many nucleotides and nucleosides, especially purines, such as ATP, adenosine, and inosine, are important immunoregulatory compounds that act through specific purinergic receptors.^46–48 Here we showed classical activation induced a significant change in the exchange rate of many nucleosides and bases. This can have great relevance as nucleotide metabolism can orchestrate immune responses through autocrine and paracrine signaling. Moreover, the release of nucleosides and bases can have profound effects on the metabolism of other cells in the environment. For example, some tumors will actively engage in nucleotide salvage at baseline.³² Additionally, in response to the inhibition of nucleotide de novo synthesis, cancer cells can import and salvage nucleotide metabolites to facilitate their survival,²⁹ including nucleotide metabolites released by macrophages.¹⁶ Therefore, the reprograming of nucleotide metabolism in macrophages can also have importance in macrophages’ interactions with other cells in the microenvironment, which merits further study.

Besides macrophages’ interactions with pathogens, other immune cells, and cancer cells within the microenvironment, nucleotide metabolism can also have important cell-intrinsic effects, both through nucleotides’ signaling role and their role as important metabolic precursors. Nucleotides are precursors of essential cofactors such as SAM and NAD, whose metabolic rewiring has been shown important for macrophage immune response and function.^49–51 To measure how classical activation impacts nucleotides’ fate in NAD(P) synthesis, we traced the labeling of NAD and NADP from ATP by supplementing cells with U-¹⁵N-inosine. Classical activation significantly increased the rate of NAD and NADP labeling, suggesting increased flux from ATP was shunted into NAD and NADP (Extended Data Fig. 10d–e). Furthermore, although non-proliferating cells do not have a net nucleotide demand for synthesizing nuclear DNA, mitochondria replication and turnover are still active, which require dNTPs. Mitochondrial DNA synthesis and fragmentation have been found important for inflammasome activation in macrophages.^52–55 Interestingly, we found thymidine nucleotides, the only nucleotide type needed for DNA but not for RNA, is particularly depleted. It is possible the nucleotide imbalance has important effects in intracellular signaling and inflammasome activation.

As nucleotide metabolism is closely connected to many other metabolic pathways and a range of cellular processes, it makes sense that it is dynamically regulated upon classical activation, and that its reprogramming can impact macrophage functions. Indeed, here we showed that purine salvage influences the expression of stimulation-induced genes, phagocytosis, and migration in macrophages. The exact mechanisms by which HPRT affects macrophage functions such as migration require more research. Here we explored the potential involvement of serine de novo synthesis, inspired by work in cancer cells.³⁴ While this study revealed novel findings about serine synthesis remodeling, given the relative contribution of de novo serine synthesis is very low in macrophages, it is unlikely to play a significant role in affecting migration. We speculate that signaling pathways which can sense purines and regulate cell migration, for instance, AMPK signaling (which responds to intracellular AICAR) and purinergic signaling,^56–61 may play a role.

Regarding the mechanisms that drive nucleotide metabolism reprogramming, an important advancement made in this study is that we identified NO as a previously unknown master regulator of nucleotide metabolism. Recent studies have highlighted the crucial role of NO and related reactive nitrogen species in the regulation of mitochondrial metabolism.^30,62–64 This study adds to the growing body of work, further establishing NO as a key driver of metabolic reprogramming in murine macrophages. NO coordinates the systematic reprogramming of nucleotide metabolism upon classical activation by simultaneously targeting several important steps (Fig. 7j): shutting off ATIC to inhibit purine de novo synthesis, inhibiting XOR to divert flux away from complete purine degradation to support salvage, and causing transcriptional downregulation of Tyms to inhibit dTMP synthesis. Given the ubiquitous significance and the central place of nucleotide metabolism across myriad cell types, this discovery likely has broader significance in other biological contexts beyond immune responses and macrophages. For example, NO is also produced by endothelial nitric oxide synthase (eNOS) for vascular regulation and by neuronal nitric oxide synthase (nNOS). More work is needed to evaluate the implications in these contexts.

Overall, this study thoroughly revealed the substantial rewiring of nucleotide metabolism in macrophages upon classical activation, elucidated its underlying mechanisms, and demonstrated its functional significance. It further maps out avenues for many future discoveries in both the realms of molecular regulation mechanisms and of health implications.

Methods

Cell Culture

Murine bone marrow-derived macrophages (BMDMs) were isolated from male and female wildtype C57BL/6J mice or B6.129P2-Nos2tm1Lau/J (Nos2^−/−) mice (Jackson Lab). Mice were bred and maintained according to protocols approved by the University of Wisconsin–Madison Institutional Animal Care and Use Committee. Laboratory animals were group-housed on a 12h light/dark cycle. The environmental conditions were maintained thermostatically between 18–23°C with 40–60% humidity. Other than a 24h fast in the experiment where mice received an IP injection of LPS+IFNγ or PBS control, as indicated in the method, all mice were fed ad libitum and had free access to drinking water.

Bone marrow cells were isolated and differentiated as previously described⁶⁵ from 6–12-week-old males or females. BMDMs were cultured in RPMI 1640 (VWR: 16750–084) with 10% dialyzed fetal bovine serum (VWR: 16777–212), 25mM HEPES, 2 mM glutamine, 1% penicillin/streptomycin (Thermo Fisher Scientific: 15140122– 100 IU/mL penicillin and 100μg/mL streptomycin) and 20 ng/mL M-CSF (R&D Systems: 416-ML) at 37°C and 5% CO₂. To classically activate BMDMs, cells were incubated with 50 ng/mL LPS (E. coli O111:B4, Sigma: L3024) and 30 ng/mL recombinant mouse IFNγ (R&D Systems: 485-MI). To alternatively activate BMDMs, cells were incubated with 20 ng/mL IL-4 (R&D Systems: 404-ML) and 20 ng/mL IL-13 (R&D Systems: 413-ML-005).

Wildtype RAW 264.7 cells (ATCC^® TIB-71^™) or knockouts were cultured in RPMI 1640 with 10% dialyzed fetal bovine serum, 25mM HEPES, 2mM glutamine, 1% penicillin/streptomycin at 37°C and 5% CO₂. To stimulate the cells, 50 ng/mL LPS and 10 ng/mL recombinant mouse IFNγ were added to the media. To alternatively activate cells, cells were incubated with 20 ng/mL IL-4 and 20 ng/mL IL-13.

Peritoneal macrophages were isolated from 8–12-week-old wildtype and Nos2^−/− mice. Mice were intraperitoneally injected with 500μL of 4% Brewer’s thioglycollate broth (Fisher Scientific: DF0256-17-2) to recruit macrophages. 72h later, the peritoneal cavity was lavaged with 6mL of cold RPMI containing 2% dialyzed fetal bovine serum and 2 mM of EDTA (Promega: V4231) using a 23G needle (BD: 305145). Lavaged cells were spun down and resuspended in RPMI 1640 with 10% dialyzed fetal bovine serum, 25 mM HEPES, 2 mM glutamine, 1% penicillin/streptomycin, and allowed to adhere to 10 cm plates for 3h to select for peritoneal macrophages. After 3h, media was aspirated, and adhered peritoneal macrophages were detached with trypsin-EDTA (Sigma Aldrich: T4049). Cells in Trypsin-EDTA were spun down and resuspended in RPMI 1640 with 10% dialyzed fetal bovine serum, 25 mM HEPES, 2 mM glutamine, 1% penicillin/streptomycin, counted, and plated at 600,000 cells per well of a 6-well plate. To classically activate peritoneal macrophages, cells were incubated with 50 ng/mL LPS and 30 ng/mL recombinant mouse IFNγ.

During cell culture experiments, media was refreshed daily to avoid the depletion of nutrients and to maintain LPS+IFNγ stimulation. Media was also refreshed 2–4h before collections. For time course experiments, the time of starting stimulation was staggered to allow all treatment groups to be collected at the same time and control for the potential impact of culturing cells for different durations. For continual stimulation experiments, LPS and IFNγ were maintained in the media throughout. For acute stimulation experiments, cells were cultured with LPS and IFNγ for 2h, after which stimuli were removed, cells were washed with DPBS, and cell culture was continued in fresh media without stimuli for the remaining timecourse.

Cell number and viability were measured using two assays. When plating cells for experiments, RAW 264.7 and BMDMs were detached using Accutase solution (Sigma-Aldrich: A6964), pelleted and resuspended in media. Then an aliquot of the cell suspension was taken and stained with an equal volume of Trypan Blue (Gibco: 15250–061), and live/dead cell number was assessed using a Countess II Automated Cell Counter (Thermo Fisher). To measure cytotoxicity in live culture, 20–40,000 cells were plated per well of a 96-well plate, stained with 250 nM Cytotox green (Sartorius: 4633), and imaged using an Incucyte S3 Live-Cell Analysis System (Sartorius).

Additional reagents used can be found in Supplementary Table 1. Untreated cells were supplemented with an equivalent volume of vehicle.

Metabolomics analysis

To measure intracellular metabolites, cells were washed twice with DPBS and metabolites were extracted with −80°C LC-MS grade 80:20 MeOH:H₂O (v:v) (Fisher Scientific: A456–4; W64). The extract was transferred to a tube, vortexed, and centrifuged at max speed. Then the supernatant was transferred to a new tube. The remaining pellet was extracted again with 80:20 MeOH:H₂O and both supernatants were combined and dried under nitrogen stream. Dried samples were resuspended in LC-MS grade H₂O or LC-MS grade 40:40:20 ACN:MeOH:H₂O (v:v:v) (ACN- Fisher Scientific: A955–4) as loading buffer depending on the LC-MS method.

To measure extracellular metabolites, spent media samples were extracted with 4x volume of LC-MS grade MeOH and vortexed. The extract was centrifuged at max speed. The supernatant was further diluted 1:10 with either LC-MS grade H₂O or LC-MS grade 40:40:20 ACN:MeOH:H₂O depending on the LC-MS method.

Samples were analyzed using a Thermo Q-Exactive mass spectrometer coupled to a Vanquish Horizon UHPLC. XCalibur 4.0 software was used for data acquisition. Two methods were used for optimal quantification.

In the first method, samples resuspended in LC-MS grade H₂O were separated on a 100 × 2.1 mm, 1.7 μM Acquity UPLC BEH C18 Column (Waters) with a gradient of solvent A (97:3 H₂O:MeOH, 10 mM TBA (Sigma Aldrich: 90781), 9 mM acetic acid (Fisher Scientific: A11350), pH 8.2) and solvent B (100% MeOH). The gradient is: 0 min, 5% B; 2.5 min, 5% B; 17 min, 95% B; 21 min, 95% B; 21.5 min, 5% B. The flow rate is 0.2 mL/min with a 30°C column temperature. Data were collected on a full scan negative mode at a resolution of 70K. Settings for the ion source were: 10 aux gas flow rate, 35 sheath gas flow rate, 2 sweep gas flow rate, 3.2kV spray voltage, 320°C capillary temperature and 300°C heater temperature.

In the second method, samples resuspended in LC-MS grade 40:40:20 ACN:MeOH:H₂O were separated on a 2.1 × 150mm XBridge BEH Amide (2.5 μM) Column (Waters) using a gradient of solvent A (95% H₂O, 5% ACN, 20 mM ammonium acetate, 20 mM NH₄OH) and solvent B (20% H₂O, 80% ACN, 20 mM ammonium acetate, 20 mM NH₄OH). The gradient is: 0 min, 100% B; 3 min, 100% B; 3.2 min, 90% B; 6.2 min, 90% B; 6.5 min, 80% B; 10.5 min, 80% B; 10.7 min, 70% B; 13.5 min, 70% B; 13.7 min, 45% B; 16 min, 45% B; 16.5 min, 100% B; 22 min, 100% B. The flow rate is 0.3 mL/min with a 30°C column temperature. Data were collected on a full scan positive or negative mode.

Metabolites reported were identified based on exact m/z and retention times determined with standards. Data were analyzed with MAVEN (v.2011.6.17).^66,67 Metabolite levels were normalized to either protein content or cell counts. To quantify the release of extracellular metabolites, absolute concentrations were determined using calibration curves generated with standards of the corresponding metabolites.

Metabolic pathway enrichment analysis

Pathway enrichment analysis was performed using MetaboAnalyst.⁷ Normalized metabolomics data from unstimulated and stimulated (continually stimulated with LPS+IFNγ for 24h) BMDMs were inputted into pathway analysis, log transformed, enriched using global test, and analyzed with the out-degree centrality topology.

Expression analysis by qPCR

For RNA collection, cells were washed twice with DPBS and then lysed with RNA STAT 60 (Fisher Scientific: NC9884083). Chloroform (Sigma-Aldrich: 650498) was used for phase separation. RNA was precipitated with isopropanol (Fisher Scientific: AC327272500), washed with 75% ethanol (VWR: 71006–012), and solubilized in DEPC H₂O (VWR: 101076–146). RNA samples were quantified using a NanoDrop2000c Spectrophotometer. RNA samples were treated with RQ1 DNase (Promega: M6101) and cDNA was synthesized using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific: 18080093) according to the manufacturer’s instructions. Quantitative PCR was performed on a Roche LightCycler 480 (v.1.5) using KAPA SYBR FAST qPCR MasterMix (Roche: KK4611).

Relative gene expression level was normalized to Cyclob. The primer sequences (IDT) can be found in Supplementary Table 2.

RNAseq Analysis

RNA sequencing was performed by Novogene, utilizing Illumina’s Sequencing by Synthesis (SBS) technology. RNA samples were collected from BMDMs and RAW 264.7 cells by the same collection method used for qPCR analysis, then submitted for sequencing analysis. Quality control was done with base recognition using CASAVA, followed by error rate assessment and GC content analysis. The data were then subjected to stringent filtration, removing adapter contamination, reads with over 10% uncertain nucleotides, and reads where low-quality bases exceeded 50% of the sequence. Post-filtration, the data were aligned to the mm10 mouse genome using the STAR software. For gene expression analysis, transcripts were quantified using the FPKM (Fragments Per Kilobase of transcript per Million mapped reads) metric.

To prepare for principal component analysis (PCA) of the time course data, the FPKM table was processed with the scanpy Python library. This involved filtering genes present in less than one cell (sc.pp.filter_genes(adata, min_cells=1)) and transforming the data via natural logarithm (sc.pp.log1p(adata)). PCA was then conducted to identify key components of variation, extracting and sorting the top loadings for each principal component based on both absolute and actual values. The final gene loadings and gene names were saved in CSV format, allowing for detailed downstream analysis. The topmost highly loaded 400 genes from PC1 were then subjected to GO analysis using g:Profiler (https://biit.cs.ut.ee/gprofiler/gost).

To analyze the enrichment of differential expressed genes in wildtype compared to Hprt KO RAW 264.7 cells, Novogene performed differential expression analysis of the two conditions by using the DESeq2 R package, with the significance criteria of padj<0.05 and log2FoldChange|>1. The gene lists of significance were uploaded to DAVID and only GO terms at the most specific resolution (GO-BP,MF,CC at level 5) were considered for annotation. The Benjamini metric (adjusted p-value for controlling false discovery rate) for the top 10 terms was projected into negative log10 space using numpy for the log transformation and matplotlib was used for the visualization.

Isotopic labeling

Isotopic tracing experiments were performed to probe the fluxes through pathways of interest by giving cells stable isotope-labeled nutrients that feed into these pathways. The tracers utilized here include: (1) γ−¹⁵N-glutamine (Cambridge Isotope Laboratories: NLM-557–1), which is a main nitrogen donor for purine and pyrimidine nucleotides and was used to probe the flux through nucleotide de novo synthesis as well as purine salvage (salvage generates a distinct labeled form from de novo synthesis as shown in Fig 2a). (2) U-¹³C-glucose (Cambridge Isotope Laboratories: CLM-1396–5), which generates labeled PRPP and 3-phosphoglycerate, and was used to probe general nucleotide turnover (as PRPP supplies the ribose moiety in nucleotides for both nucleotide de novo synthesis and salvage), the RNA turnover rate (reflected by the downstream labeling incorporation from nucleotides to RNA), and serine de novo synthesis (reflected by the labeling incorporation from 3-phosphoglycerate to serine). (3) U-¹⁵N-inosine (Cambridge Isotope Laboratories: NLM-4264–0.05), which was used to probe purine salvage activity, (4) ¹³C formate (Sigma-Aldrich: 279412), which generates labeled one-carbon units and was used to examine the incorporation from one-carbon units to nucleotides through nucleotide de novo synthesis.

For stable isotope labeling with nutrients that are contained in regular RPMI culture media (glutamine or glucose), cells were incubated with media containing the designated tracer which replaces the unlabeled glutamine or glucose in regular media at the same concentration. For stable isotope labeling with supplemented labeled inosine or formate, 50 μM U-¹⁵N-inosine or 1mM ¹³C formate was added to the regular media. The pathway activity was examined by measuring the labeling incorporation from labeled substrates to downstream metabolites. The labeling in small metabolites was quantified by quenching metabolism in cells at designated time points (as indicated in figure legends), extracting metabolites, and analyzing the metabolites, including quantifying all their isotopic labeled forms by LC-MS, using analytical methods as described above in “Metabolomics analysis”. To measure the labeling in RNA, cells (unstimulated or stimulated) were cultured in media containing U-¹³C-glucose for 24h, then washed twice with DPBS and collected with 400μL of RNA STAT 60. Chloroform was used for phase separation. RNA was precipitated with isopropanol, washed with 75% ethanol, and solubilized in DEPC H₂O. RNA samples were then digested for 1h (New England Biolabs: M0649S) to single nucleotides for labeling quantification by LC-MS as described in the metabolite analysis section. Data from labeling experiments was adjusted for natural abundance of ¹⁵N and ¹³C.

Immunoblotting

For protein collection, cells were washed twice with DPBS and lysed using RIPA buffer containing phosphatase and protease inhibitors (Thermo Fisher Scientific: A32957; A32953). Whole cell lysate was passed through a 27G needle (BD Biosciences: 305109) and centrifuged at 12000xg at 4°C. Protein content in the supernatant was quantified using a BCA assay kit (Thermo Fisher Scientific: 23223) using a Biotek EPOCH2 microplate reader (Gen5 TS 2.09 software). Protein samples were heat denatured and separated on 4–12% Bolt Bis-Tris gels (Thermo Fisher Scientific: NW04125BOX). The protein was then transferred onto a 0.2 μM nitrocellulose membrane (VWR: 27376–991). Membranes were blocked in 5% BSA (Millipore Sigma: A7030) in TBS-T and probed with primary antibodies overnight. The information for primary antibodies can be found in Supplementary Table 3.

The following day, membranes were washed with TBS-T and then probed with Li-Cor secondary antibodies (1:10000 dilution) (goat anti-rabbit 800CW, goat anti-mouse 680RD [Li-Cor Biosciences: 926–32211; 926–68070]). Membranes were imaged with a Li-Cor Odyssey ClX and quantified using ImageStudio Lite (v.5.2) software.

Generation of knock out cell lines

Genetic knockouts were generated using the Alt-R RNP system in low passage wildtype RAW 264.7 cells following the manufacturer’s instructions. The system consists of Alt-R S.p. HiFi Cas9 Nuclease V3 (IDT: 1081060), Alt-R CRISPR–Cas9 gRNA (IDT), and ATTO 550 (IDT: 1075927).

The following gRNA sequences were used:

HPRT: ACGGGGGACATAAAAGTTAT

Nos2^−/− RAW 264.7 cells that were used were previously generated as described.³⁰

After the RNP complexes were formed, cells were transfected by electroporation using the Amaxa Nucleofector system Kit V (Lonza: VCA-1003). The following day, cells positive for fluorescent tracrRNA were single-cell sorted into individual wells of a 96-well plate containing RPMI, 25 mM HEPES, 2 mM glutamine, 20% fetal bovine serum. Single-cell colonies were subsequently expanded. To validate gene knockout from single-cell sorted colonies, genomic DNA was extracted using the QIAamp DNA Mini Kit (Qiagen: 51304) and a region around the CRISPR cut site target was amplified using the following primers containing M13 tags for sequencing.

Hprt	Fwd	5’-TGTAAAACGACGGCCAGTACTCACAGGAATGGTTCAGGG-3’
	Rev	5’-CAGGAAACAGCTATGACCAGCAATATGGACTGTGAGGGTT-3’

Knockouts were verified by Sanger sequencing (Genewiz) of the PCR product after column purification (Thermo Fisher Scientific: K0701). The sequencing results are shown in Supplementary Fig. 3.

Oxygen consumption rate measurement

Oxygen consumption rates (OCR) were measured using an Agilent XF Seahorse Extracellular Flux Analyzer (v.1.4.2.3) according to the manufacturer’s instructions. Cells were plated at a seeding density of 6×10⁴ cells per well in normal cell culture media. Before measuring OCR, cells were switched over to XF DMEM Medium, pH 7.4 (Agilent: 103193–100) containing 11.11 mM glucose, 1% penicillin/streptomycin, 5 mM HEPES, 2 mM glutamine, and 20 ng/mL M-CSF, without phenol red, sodium bicarbonate, or sodium pyruvate. Data was blanked from wells containing only media.

XOR activity assay

XOR activity was measured in cell lysates. For each condition, 2 million WT or Nos2^−/− RAW 264.7 cells were unstimulated or stimulated for 48h and treated with or without DETA-NONOate throughout the timecourse. After two days, media were aspirated, cells were washed twice with DPBS, and cells were collected by scraping with DPBS. The cells were pelleted, DPBS was aspirated, and the cells were lysed by storage at −80C. The next day, the lysed cells were resuspended in 100 μL assay buffer containing 100 μM EDTA with 50mM potassium phosphate at pH 7.8, and total protein content was quantified using BCA assay kit.

Xanthine was made at 200 μM in assay buffer with heating. NAD⁺ was dissolved at 5mM in assay buffer the day of activity assay. To perform the activity assay, 20 μg of lysate protein was mixed with 50 μM of xanthine and 500 μM NAD⁺ in assay buffer to assess total XOR activity; and 20 μg of protein was mixed with 50 μM of xanthine only to assess xanthine oxidase activity. For both, the final reaction volume was 100 μL. The reaction was performed at 25°C with low vortexing. At each time point, 20 μL of the reaction mix was quenched with 80 μL of 50:50 MeOH:ACN. The quenched sample was spun at max speed, and the supernatant was collected and analyzed by LC-MS for urate production.

ATIC purification and activity assay

To generate FLAG-tagged ATIC expressing HEK293T cells (ATCC^® CRL-3216^™), lentiviral particles were produced by transfecting HEK293T cells with Fugene 6 transfection reagent (Promega: E2691) and OPTI-MEM (Thermo Fisher Scientific: 31985070) containing psPAX2 packaging plasmid (Addgene #12260), pMD2.G envelope plasmid (Addgene #12259) and pLenti-Puro-FLAG-ATIC plasmid (Vector Builder, Vector ID: VB240807–1702zty). Viral particles were added to cells in full DMEM with 8 μg/mL polybrene (Sigma Aldrich: TR-1003) for 24 hours. Cells were selected with 5 μg/mL puromycin (Sigma Aldrich: P4512). Confirmation of stable expression was performed by western blot.

To purify ATIC, cells were washed and lysed in lysis buffer (20 mM Tris-HCl, pH 7.4, 1% Triton, 100 mM NaCl, 5 mM MgCl₂, 1 mM DTT (Fisher Scientific: AAJ1539706), containing protease and phosphatase inhibitors). Total protein was quantified using reducing agent compatible BCA assay (Thermo Fisher Scientific: 23252). 2 mg of protein from lysate was added to washed anti-FLAG M2 resin (Sigma Aldrich: A2220) and rotated end-over-end for 3 hours at 4°C. The resin was washed four times with wash buffer and spun at 2,000 × g for 1 minute at 4°C. The resin was washed in 1 mL final wash buffer (20 mM Tris-HCl, pH 7.4, 100 mM NaCl, and 5 mM MgCl₂) and spun at 2,000 × g for 1 minute at 4°C. 7 μg FLAG peptide (Sigma Aldrich: F4799) in final wash buffer was added to the resin and shaken at 1,200 rpm for 30 minutes at 25°C. The total volume was spun at 1,000 × g for 4 minutes at 4°C in a chromatography column (BioRad: 7326204). The flowthrough was transferred to a 30 kDa MWCO filter (Sigma Aldrich: UFC503024). 300 μL of storage buffer (40 mM Tris-HCl pH 7.5, 100 mM NaCl, 2 mM DTT) was added then spun at 10,000 × g for 5 minutes at 4°C. Buffer addition and spins were repeated 10 times total. Glycerol was added to a final concentration of 15% (v/v) in the final remaining volume. Molarity of purified protein was determined by Nanodrop with Extinction Coefficient of 54,320 M⁻¹cm⁻¹ and molecular weight of 63,342 Daltons. The purified ATIC was checked by SDS-PAGE (Supplementary Fig. 4).

Activity assay was adapted from a previously described protocol.⁶⁸ Assay buffer contained 40 mM Tris-HCl pH 7.4, 5 mM Na₂HPO₄, 5 mM MgCl₂, 2 mM DTT, and 100 mM NaCl. AICAR (Millipore Sigma: 123041) was made at 500 μM in assay buffer and stored for later use. Purified protein (50 nM) in assay buffer containing 50 μM AICAR was incubated with or without 500 μM DPTA-NONOate (Cayman Chemical: 82110) for 3h at room temperature. During this time, 10-formyl THF was synthesized from 5,10 Me⁺-THF (Schricks laboratories: 16.230) according to manufacturer’s instructions. Briefly, 5,10 Me⁺-THF was diluted to a final concentration of 4 mM in 100 mM Tris HCl pH 8.5 containing 14.3 mM of 2-mercaptoethanol (Millipore Sigma: M6250). The tube was vigorously vortexed at 37°C for 1 minute before sitting for 1h at 25°C. After the 3h incubation, 10-formyl-THF was added to the reaction at a final concentration of 100 μM, in a final volume of 100 μL. At each time point, 20μL of the reaction was combined with 80 μL of 50:50 MeOH:ACN to quench. The quenched reaction was spun at max speed, and the supernatant was analyzed by LC-MS for IMP production.

In vivo stimulation and isolation of peritoneal macrophages

Wildtype and Nos2^−/− female mice were intraperitoneally injected with 500 μL of 4% Brewer’s thioglycollate broth to recruit macrophages. 72h later, mice were intraperitoneally injected with both 15 μg of IFNγ and 1.5 μg/g of LPS or an equivalent volume of PBS. The mice were then fasted, and mice of the same genotype were housed together. 24h later, the peritoneal cavity was lavaged with 6 mL of cold RPMI containing 2% dialyzed fetal bovine serum and 2 mM of EDTA using a 23G needle (BD: 305145). All subsequent steps were done on ice or at 4C. The lavage fluid was strained through a 70 μM cell strainer (Fisher Scientific: 08-771-2) and ACK Lysing Buffer (Thermo Fisher Scientific: A1049201) was added to lyse red blood cells. The remaining cell suspension was treated with anti-mouse CD16/32 antibody (Fc block, BioLegend: 156604) and stained with CD11b (BioLegend: 101212), F4/80 (BioLegend: 111603), and DAPI (Fisher Scientific: D3571). CD11b, F4/80 double positive cells⁶⁹ were collected by FACS (BD FACSAria) using a 130 μM nozzle at a pressure of 14 psi to minimize cell perturbations from sorting. Collected cells were washed with cold DPBS, then metabolites were extracted with −80°C LC-MS grade 80:20 MeOH:H₂O (v:v). Metabolites were analyzed by LC-MS as described above. Metabolite level in stimulated conditions were normalized to that in PBS injected controls in the same trial, and statistical analysis between genotypes was performed by paired two-tailed student’s t-test. For gating strategy (analysis performed using FlowJo v.10.4 (TreeStar)) see Supplementary Fig. 2.

ELISA

For cytokine release measurements, 24 hour spent media from cells was collected and spun at 500xg for 5 minutes. Then IL-10 concentration in the supernatant was measured using DuoSet ELISA kits (R&D Systems:DY417–05) following manufacturer’s instruction. Released IL-10 levels were normalized to the total cellular protein content in the culture plate (as a measure of cell density).

Cell migration

Cell migration across transwells was performed using an Incucyte live cell imager. For migration assays, Matrigel (Fisher Scientific: CB-40234) was diluted to 50 μg/mL in DPBS to coat the top and bottom sides of the inserts of an Incucyte Clearview 96-well plate (Sartorius: 4582). After 1h, residual Matrigel solution was aspirated and RAW 264.7 cells (wildtype and Hprt KOs) were plated in RPMI containing 2 mM glutamine and 0.5% dialyzed fetal bovine serum (assay medium) at a seeding density of 9,000 cells per well on the top chamber. The cells were allowed to settle and adhere to the top membrane for 1h. Afterwards, the chemoattractants LPS or C5a (R&D Systems: CB-40234) were added to assay medium in the bottom reservoir at concentrations of 1000 ng/mL or 200 ng/mL, respectively. Cell migration was imaged continually for 48h. Data analysis was performed using the Incucyte Chemotaxis module that quantified the number of cells present on the bottom side of the transwell at each timepoint imaged.

Phagocytosis

Phagocytosis was measured by imaging analysis using inactivated Salmonella enterica conjugated to a pH-sensitive fluorescent dye pHrodo (Essen BioScience: 4649). Macrophages were seeded in 96-well plates for imaging at the density of 50,000 per well for RAW 264.7 cells or 40,000 per well for BMDMs, in 100 μL of RPMI 1640 with 10% dialyzed fetal bovine serum, 25 mM HEPES, 2 mM glutamine, 1% penicillin/streptomycin. To prepare pHrodo labeled S. enterica, 20 μg (100μL) inactivated Salmonella enterica (Sigma Aldrich: MBD0010) was diluted 1:10 with labeling buffer containing 1:200 pHrodo and remained at 37°C in the dark for 1h. Afterwards, the conjugated S. enterica was diluted 1:10 in labeling buffer, and 200 ng (100μL) of the diluted pHrodo labeled S. enterica was added to either WT or Hprt KO RAW 264.7 cells or BMDMs pre-treated for 24h with 6-mercaptopurine. Cells were imaged using an Incucyte S3 Live-Cell Analysis System and total integrated red intensity was quantified using the analysis software.

Analysis of Toxoplasma gondii growth

T. gondii tachyzoites were cultured in human foreskin fibroblasts in DMEM (Gibco: 11960–051) supplemented with 10% fetal bovine serum, 2 mM glutamine, and 1% antibiotics at 37°C with 5% CO₂. T. gondii growth in macrophages was measured by two imaging-based methods. To determine overall T. gondii burden in live culture over time, 5×10⁴ viable RAW 264.7 macrophages (wildtype or Hprt KO) were seeded in a 48-well plate and infected with 5×10⁴ mCherry ME49 (type II) tachyzoites. The plate was incubated at 37°C and 5% CO₂ and imaged using an Incucyte Live-Cell imager. The red channel used an acquisition time of 400 milliseconds for all replicates and samples. The parasites (red) were measured at a threshold of 0.90 Red Calculated Units with surface-fit segmentation. To image T. gondii proliferation at higher resolution and analyze parasite number in vacuoles by immunofluorescence, a total of 1×10⁵ viable macrophages were seeded on glass coverslips and infected with 1×10⁵ mCherry ME49 (type II) tachyzoites. At 24h post infection, cells were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100 (Fisher Scientific: BP151–500) and blocked with 5% BSA. Cells were incubated with primary antibody against T. gondii surface antigen 1 (SAG1, rabbit, 1:300; Genetex GTX38936) overnight at 4°C and then thoroughly washed with DPBS. Secondary antibody against Alexa Fluor 594 (goat anti-rabbit, 1:1000; Thermo Scientific: A11012) was incubated for 2h at room temperature. Cells were counterstained with DAPI (Sigma D9564) and mounted using VECTASHIELD antifade Mounting Medium (Vector Laboratories: H-1000). Cells were imaged using an epifluorescence microscope (Imager.M2, Carl Zeiss, DEU). Vacuoles, parasites per vacuole, infected macrophages, and non-infected macrophages were counted in 10 random fields per coverslip. Each timepoint was performed on triplicate coverslips per condition. All coverslips were blinded before counting.

Statistics and reproducibility

For quantitative measurements, the experimental sample sizes and error bars are noted in corresponding figure legends. Statistical analysis was performed on GraphPad Prism (v.9) or Microsoft Excel using student’s t-test for comparisons between two groups and one-way ANOVA with multiple comparisons for assessment of more than two groups. Comparisons among specific groups were made using post-hoc tests as indicated in the figure legends. Statistical analysis methods are also noted in specific figure legends. The immunoblots shown are representative of at least two independent experiments. The investigators were not blinded for most experiments as one person was typically responsible for carrying out experiments from start to finish. However, coverslips were blinded before quantifying vacuoles in the T. gondii infection experiment. No successfully collected data were excluded from experiments. Data distribution was assumed to be normal but this was not formally tested. No statistical method was used to predetermine sample size, but our sample sizes are similar to those reported in previous publications.^12,30 Cells and animals were randomly assigned to experimental conditions.

Extended Data

Extended Data Figure 1: Reprogramming of nucleotide metabolism is independent of changes in cell proliferation or viability.

a) Relative cell number of BMDMs over a timecourse with or without continual stimulation. Mean ± SD (n=4 independent samples). b) Cell viability in a population of BMDMs that are unstimulated and cultured for 72h (n=4 independent samples), continually stimulated for 72h (n=4), or acutely stimulated for 96h (n=3). Mean ± SD. c) Metabolomic changes in pentose phosphates, nucleotides, nucleosides, and nitrogenous bases in RAW 264.7 cells over a timecourse of continual stimulation. d) Relative cell number of RAW 246.7 cells over a timecourse with or without continual stimulation. Mean ± SD (n=3 independent samples). e) The top 10 genes enriched along principal component 1 of the transcriptomic changes in BMDMs over a timecourse of acute stimulation. c, e) Relative (c) metabolite abundances or (e) gene expression are compared to unstimulated cells (0h) and displayed on a heatmap, with saturating color representing 10-fold change (or 3.32 on log2 scale). Each box represents the mean of n=3 independent samples. a, d) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated to stimulated conditions at 72h. b) Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells. ns indicates not significant (p>0.01).

Extended Data Figure 2: Kinetic glutamine tracing in nucleotide de novo synthesis pathway.

a) Labeling kinetics of UDP and UTP in unstimulated or stimulated BMDMs. b) Relative intracellular abundance of UMP in unstimulated or stimulated BMDMs. Mean ± SD (n=3 independent samples). c) Labeling kinetics of dCTP in unstimulated or stimulated BMDMs. d–e) Labeling kinetics of (d) glutamine, carbamoyl aspartate, orotate, UMP and (e) dTMP, CTP in unstimulated or stimulated RAW 264.7 macrophages. f) Labeling kinetics of ADP and ATP in unstimulated or stimulated BMDMs. g) Labeling kinetics of AICAR, IMP, AMP, and XMP in unstimulated or stimulated RAW 264.7 macrophages. a, c–g) Cells were labeled with γ−¹⁵N-glutamine for various time points as indicated on x-axis. Mean ± SD (n=3 independent samples). a–c, f) Stimulated BMDM cells are continually stimulated for 48h. d–e, g) Stimulated RAW 264.7 cells are continually stimulated for 24h. a–g) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated to stimulated cells at the final time point. ns indicates not significant (p>0.01).

Extended Data Figure 3: LPS+IFNγ stimulation-induced changes in nucleotide degradation and salvage.

a–b) Labeling kinetics of (a) IMP and (b) inosine in unstimulated or 48h continually stimulated BMDMs after labeling with U-¹³C-glucose for 0–4 hours. c) Extracellular metabolites released into the media by BMDMs over a timecourse of continual stimulation. Cells were incubated in the media for 22h before analysis. Data are presented relative to the level in spent media incubated with 24h stimulated BMDMs on a log scale. N.d. indicates not detected. Mean + SD (n=3 independent samples). d) Relative intracellular abundance of inosine in BMDMs over a timecourse of continual stimulation, with or without treatment of 10 μM forodesine (FD) throughout the timecourse. e) Fraction of M+1 labeled GMP from 24h γ−¹⁵N-glutamine labeling in BMDMs over a timecourse of continual stimulation, with or without treatment of 10 μM forodesine throughout the timecourse. f) Relative intracellular PRPP abundance in BMDMs over a timecourse of continual stimulation, with or without treatment of 10 μM forodesine throughout the timecourse. a–b, d–f) Mean ± SD (n=3 independent samples). a–b) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated to stimulated cells at each time point. d–f) Statistical analysis was performed with unpaired t-test comparing untreated to FD treated cells at each time point. ns indicates not significant (p>0.01).

Extended Data Figure 4: One-carbon metabolism supplements do not rescue the stimulation-induced inhibition of nucleotide de novo synthesis.

a) Transcriptional changes of detected genes in one carbon metabolism and serine de novo synthesis in BMDMs over a timecourse of acute stimulation. Relative expression of each gene is presented as relative to unstimulated cells (0h) on a log2 scale as a heatmap, with saturating color representing 10-fold change (or 3.32 on log2 scale). Each box represents the mean of n=3 independent samples. b–e) Labeled fraction of (b) M+1 labeled AICAR, (c) M+2 labeled IMP, (d) M+2 labeled ATP, and (e) M+2 labeled GTP in unstimulated or stimulated BMDMs supplemented with 1mM ¹³C-formate for 24h. f) Relative total intracellular dTTP abundance in unstimulated or stimulated BMDMs supplemented with 1mM ¹³C-formate for 24h. Data are presented relative to abundance in unstimulated cells. g–i) Labeled fraction of (g) IMP (M+2) and (h) dTTP (M+1), and (i) relative intracellular dTTP abundance in RAW 264.7 cells that are unstimulated, stimulated, or unstimulated and treated with 200nM methotrexate (MTX), labeled with γ−¹⁵N-glutamine for 24h. Cells were additionally supplemented with or without 1 μM folinate throughout the timecourse. b–i) Stimulated cells are continually stimulated for 48h. Mean ± SD (n=3 independent samples). b–f) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated to stimulated cells. g–i) Statistical analysis was performed using one-way ANOVA followed by post hoc Tukey’s test. Bars with different lower-case letters (a, b, c, or d) indicate a statistically significant difference with p<0.05.

Extended Data Figure 5: Rescuing altered redox ratio and aspartate level is not sufficient to rescue the stimulation-induced inhibition of nucleotide de novo synthesis.

a) Basal oxygen consumption rate of unstimulated or stimulated BMDMs. Mean ± SD (n=10 independent samples). b) Intracellular NAD to NADH ratio in unstimulated or stimulated BMDMs with or without supplementation of 2.5 mM pyruvate. c) Relative intracellular abundance of aspartate in BMDMs over a timecourse of continual stimulation. d) Relative intracellular abundance of aspartate in unstimulated or stimulated BMDMs, with or without supplementation of 2.5 mM pyruvate or 5 mM aspartate throughout the timecourse. e–i) Labeled fraction of (e) M+2 labeled IMP, (f) M+2 labeled AMP, (g) M+1 and M+3 labeled GMP, (h) M+1 and M+2 labeled CDP, and (i) M+1 labeled UMP after 24h γ−¹⁵N-glutamine labeling in unstimulated or stimulated BMDMs, with or without supplementation of 2.5 mM pyruvate or 5 mM aspartate throughout the timecourse. N.d. indicates not detected. a–b, d–i) Stimulated cells are continually stimulated for 48h. b–i) Mean ± SD (n=3 independent samples). a) Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated to stimulated cells. c) Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells (0h). b, d–i) Statistical analysis was performed using one-way ANOVA followed by post hoc Tukey’s test. Bars with different lower-case letters (a, b, c, d, or e) indicate a statistically significant difference with p<0.05.

Extended Data Figure 6: Dynamic changes in NO production in relation to dynamic changes in nucleotide metabolism upon LPS+IFNγ stimulation.

a) Relative intracellular abundance of citrulline in BMDMs over a timecourse of continual LPS+IFNγ stimulation. Data are presented as log2 fold change compared to abundance in unstimulated cells (0h). b) Transcriptional changes of Nos2 and Tyms in BMDMs over a timecourse of acute LPS+IFNγ stimulation. Relative expression is displayed as log2 fold change compared to unstimulated cells (0h). c) Relative intracellular abundance of citrulline in wildtype or Nos2^−/− BMDMs over a timecourse of continual stimulation. N.d. indicates not detected. d) Fraction of M+1 labeled glutamine from 24h γ−¹⁵N-glutamine labeling in unstimulated or 48h continually stimulated wildtype or Nos2^−/− BMDMs, with or without treatment of 200 μM NO donor DETA-NONOate throughout the timecourse. e) Relative intracellular AICAR abundance in BMDMs over a timecourse of continual stimulation. a–e) Mean ± SD (n=3 independent samples). e) Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells (0h). ns indicates not significant (p>0.01).

Extended Data Figure 7: NO is a key regulator mediating the rewiring in nucleotide metabolism that is specific to classical activation.

a) Arginine uptake, as measured by the changes in arginine level in spent media compared to fresh media after a 22h incubation period, by BMDMs over a timecourse of continual and acute stimulation. Statistical analysis was performed using one-way ANOVA followed by post hoc Dunnett’s test comparing all groups to unstimulated cells (0h). ns indicates not significant (p>0.05). b) p-mTOR, mTOR, p-p70S6K, p70S6K, p-4E-BP1, and 4E-BP1 protein levels in unstimulated Nos2^−/− BMDMs over a timecourse of 200μM DETA-NONOate treatment. c) p-p70S6K, p70S6K, p-4E-BP1, and 4E-BP1 protein levels in wildtype or Nos2^−/− BMDMs that are unstimulated or continually stimulated with LPS+IFNγ, with or without treatment of 200 μM DETA-NONOate for 48h. d) Schematic of differential arginine metabolism by classically activated and alternatively activated macrophages. e) Citrulline and ornithine abundance in RAW 264.7 cells over a timecourse of continual LPS+IFNγ (red) or IL-4+IL-13 (blue) treatment. Data are presented relative to abundance in unstimulated cells (0h). Statistical analysis was performed with unpaired two-tailed student’s t-test comparing stimulated groups at each time point. ns indicates not significant (p>0.01). a, e) Mean ± SD (n=3 independent samples). f) Schematic of the in vivo stimulation experiment. g–h) The fold changes of (g) dTTP and (h) IMP in CD11b⁺F4/80⁺ cells harvested from wildtype or Nos2^−/− mice treated with LPS+IFNγ. Data are presented relative to abundance in CD11b⁺F4/80⁺ cells harvested from PBS injection controls from each trial. Mean ± SEM (n=9 independent mice for PBS injection controls. n=5 for LPS+IFNγ samples, with each replicate consisting of cells pooled together from two stimulated mice). g–h) Statistical analysis was performed with paired two-tailed student’s t-test comparing stimulated cells. i–j) The fold changes of (i) dTTP and (j) IMP in wildtype or Nos2^−/− peritoneal macrophages after 48h continual stimulation with LPS+IFNγ after isolation (ex vivo), compared to unstimulated control. Mean ± SD (n=3 independent mice). Statistical analysis was performed with unpaired two-tailed student’s t-test comparing stimulated cells.

Extended Data Figure 8: HPRT knockout effectively modulates nucleotide metabolism and significantly impacts macrophage gene expression.

a) Cell viability in wildtype and Hprt KO RAW 264.7 cells, either unstimulated or continually stimulated for 48h. b) M+1 labeled fraction of GTP in wildtype or Hprt KO (2 clones) RAW 264.7 cells, either unstimulated or continually stimulated for 48h. Cells were labeled with γ−¹⁵N-glutamine for 24h before analysis. c) Expression levels of Hprt and Aprt (encoding adenine phosphoribosyltransferase) in unstimulated wildtype or Hprt KO RAW 264.7 cells. Data are presented as fold change relative to wildtype unstimulated. d) Metabolomic changes in pentose phosphates, nucleotides, nucleosides, and nitrogenous bases in 48h stimulated wildtype or Hprt KO (2 clones) RAW 264.7 cells. Metabolite abundances are relative to wildtype cells and displayed on a log2 scale as a heatmap. Each box represents the mean of n=3 independent samples. e) Labeling pattern of M+1 GTP in BMDMs over a timecourse of continual stimulation treated with or without 6-mercaptopurine. Cells were labeled with γ−¹⁵N-glutamine for 24h before analysis. a–c, e) Mean ± SD (n=3 independent samples). a, c, e) Statistical analysis was performed using unpaired two-tailed student’s t-test comparing WT to Hprt KO or untreated to 6-MP treated cells at each time point. ns indicates not significant (p>0.01). b) Statistical analysis was performed using one-way ANOVA followed by post hoc Tukey’s test. Bars with different lower-case letters (a, b, c, or d) indicate a statistically significant difference with p<0.05. f–g) Volcano plot showing differentially expressed genes (up-regulated in red and down-regulated in green) in Hprt KO RAW 264.7 cells compared to wildtype cells in f) unstimulated state, and g) 48h stimulated state.

Extended Data Figure 9: HPRT knockout suppresses macrophage migration.

a) Gene ontology enrichment of the pathways that are most downregulated in Hprt KO macrophages compared to wildtype. b) Schematic of a macrophage migration assay. c) Migration of wildtype and Hprt KO (2 clones) RAW 264.7 cells across a Matrigel-coated transwell towards C5a. Mean ± SD (n=5 independent samples). d) Labeled fraction of left: 3PG (M+3) and right: serine (M+3) in wildtype or Hprt KO (2 clones) RAW 264.7 cells over a timecourse of continual LPS+IFNγ stimulation. Cells were labeled with U-¹³C-glucose for 24h before analysis. e) Relative intracellular abundance of 3-phosphoserine in wildtype or Hprt KO (2 clones) RAW 264.7 cells over a timecourse of continual stimulation. f–g) Transcript levels of (f) Phgdh, Psat1, Psph, and (g) Shmt1, and Shmt2 in unstimulated or 48h continually stimulated wildtype or Hprt KO RAW 264.7 cells. d–g) Mean ± SD (n=3 independent samples). c, f–g) Statistical analysis was performed using unpaired two-tailed student’s t-test comparing wildtype to Hprt KO. ns indicates not significant (p>0.05). d–e) Statistical analysis was performed using one-way ANOVA followed by post hoc Tukey’s test. Bars with different lower-case letters (a, b, or c) indicate a statistically significant difference with p<0.01.

Extended Data Figure 10: Various fates of nucleotides in unstimulated or LPS+IFNγ stimulated macrophages, and in macrophages infected with T. gondii.

a) Schematic showing T. gondii replication in a macrophage and its dependence on salvaging host purine. Specifically identified is the parasitophorous vacuole that separates host cell cytosol from T. gondii. b–c) Labeled fractions of (b) guanosine and adenosine units from digested RNA, compared to (c) free GTP and ATP, after 24h U-¹³C-glucose labeling in unstimulated or 24h continually stimulated wildtype BMDMs. d–e) Labeling kinetics of intracellular (d) NAD+ (M+4) and (e) NADP+ (M+4) compared to their precursor ATP (M+4) in unstimulated or 48h stimulated BMDMs supplemented with 50 μM U-¹⁵N-inosine for 0–4 hours as indicated on x-axis. Mean ± SD (n=3 independent samples). Statistical analysis was performed with unpaired two-tailed student’s t-test comparing unstimulated and stimulated cells at the final timepoint.

Supplementary Material

1

Supplementary Figure 1: Transcriptional changes of thymidine kinase Tk1 and Tk2 in BMDMs over a timecourse of acute LPS+IFNγ stimulation. Relative expression is compared to unstimulated cells (0h) and displayed on a log2 scale.

Supplementary Figure 2: Gating strategy used to isolate CD11b⁺F4/80⁺ cells.

Supplementary Figure 3: Wildtype and Hprt knockout clones used in the manuscript were sequenced for validation after PCR amplification of the target region using the primers specified in the method.

Supplementary Figure 4: Purified ATIC protein. After purification, 3xFlag-ATIC was heat denatured, separated on a 4–12% Bolt Bis-Tris gel, and stained with Coomassie.

Data availability

All data and uncropped scans for all western blots are included in the Source Data files. The RNA-seq data reported in this article have been deposited in NCBI’s Gene Expression Omnibus (GEO) and are accessible through GEO Series accession numbers GSE267544 and GSE267546. Unique reagents generated in this study are available upon reasonable request from the corresponding author. Source data are provided with this paper.