Metabolic control of innate immune activation in TET2-mutant clonal hematopoiesis

Summary

Somatic mutations in TET2 drive hyper-inflammation in clonal hematopoiesis of indeterminate potential (CHIP), but the molecular link between TET2 inactivation and myeloid immune activation remains unclear. We used in vivo genome-wide genetic perturbations enabled by ultra-diverse barcoding in primary wild-type (WT) or Tet2 knockout (KO) Cas9+ hematopoietic stem-progenitor cells (HSPCs) to elucidate the basis of Tet2 KO inflammation. We uncover a metabolic circuit by which Tet2 restrains O-linked N-acetylglucosamine (O-GlcNAc) glycosyltransferase (Ogt), a Tet2 binding partner and metabolic sensor. Tet2 loss disrupts this inhibitory Tet2-Ogt interaction, and dysregulated Ogt facilitates widespread H3K4 trimethylation including lipid-related gene loci and inflammatory lipid droplet formation. We identified that ATP citrate lyase (Acly) is decorated with O-GlcNAc and is a critical node for lipid accumulation and inflammation in Tet2 KO. These findings reveal that TET2 suppresses inflammation by gating nutrient-responsive chromatin remodeling and nominate metabolic interventions to restrain inflammatory disease in TET2-mutant clonal hematopoiesis.

Graphical Abstract

eTOC Blurb

Kim and Hergott et al. uncover a metabolic foundation for the pathogenic inflammation arising from TET2-mutant clonal hematopoiesis (CH). In myeloid cells, TET2 loss disinhibits the nutrient sensor/glycosyltransferase OGT, driving aberrant and inflammatory lipid accumulation via the metabolic enzyme ACLY. Inhibiting ACLY offers a therapeutic strategy for controlling CH-related hyperinflammation.

Introduction

Somatic mutations in hematopoietic stem cells can reprogram the innate immune system and predispose individuals to chronic inflammatory diseases. TET2 loss-of-function mutations are highly prevalent in clonal hematopoiesis of indeterminate potential (CHIP) and myeloid malignancies^1–7 and have particularly powerful associations with inflammatory conditions including atherosclerotic disease, liver disease, chronic obstructive pulmonary disease, and gout^2,5–8. TET2 is an epigenetic enzyme that oxidizes 5-methylcytosine to 5-hydroxymethylcytosine (5hmC) and leads to demethylation of genomic DNA^9–11, but the mechanistic basis of how TET2 inactivation reprograms myeloid cells to a pro-inflammatory state has not been elucidated.

TET2 mutations arise in hematopoietic stem and progenitor cells (HSPCs) and are present within maturing myeloid cells, including neutrophils and monocytes. Murine models of Tet2 inactivation faithfully recapitulate the competitive advantage of hematopoietic stem cells, myeloproliferation¹², and enhanced inflammation observed in humans^2,5–8. Tet2 KO bone-marrow-derived macrophages (BMDMs) display exaggerated responses to microbial products and other inflammatory stimuli^2,8, especially in the setting of nutrient (e.g., lipid) excess^2,5. Prior investigations into the mechanisms by which TET2 mutations augment inflammation have converged primarily on the expression of inflammatory cytokines such as interleukin (IL)-1ß by monocytes and macrophages via activation of the NLRP3 inflammasome^2,5,8, which may be accentuated by signals from the microenvironment^2,13. Others have focused on delayed resolution of inflammation or other cytokines such as IL-6 or MIF, implicating histone deacetylase-mediated transcriptional effects^8,14,15 or alterations of NF-κB transcription factor activity by non-coding RNAs in TET2 KO cells¹⁶. However, the proximal molecular mechanism by which TET2 mutations trigger these inflammatory pathways remains poorly understood.

We employed an unbiased approach to probe the foundations of pathologic inflammation in Tet2 KO myeloid cells. Using genome-wide CRISPR/Cas9 perturbations in primary HSPCs, screening their consequences on myeloid inflammation in vivo, and interrogating our results with metabolomic, epigenetic, and proteomic studies, we uncover metabolic dysregulation as a central driver of TET2-related hyperinflammation, illuminate the key molecular interactions that propel it, and devise therapeutic strategies for metabolic intervention.

Results

Tet2 restrains inflammatory lipid accumulation

To investigate the molecular mechanisms of inflammation driven by Tet2 KO myeloid cells, we performed an in vivo genome-wide CRISPR/Cas9 genetic screen in primary wild-type (WT) or Tet2 KO Lin-Sca1+c-Kit+ (LSK) HSPCs and induced innate immune activation using zymosan-mediated peritonitis following stable engraftment and differentiation (Figure 1a, S1a–d). Zymosan-induced peritonitis is a widely accepted model of acute inflammation with adequate myeloid cellular output in the peritoneal cavity to preserve genome-wide guide RNA (gRNA) representation (Figure S1b)¹⁷. The heterogeneity in cellular expansion that occurs during engraftment and differentiation typically impedes large-scale genetic perturbations¹⁸. To overcome this issue, we barcoded the gRNA library with > 10 million unique-molecular identifier (UMI) sequences, such that each gRNA is linked to a pool of UMIs (Figure S1e), to enhance clone-level sensitivity of genetic screens in vivo on a genome-wide scale (Figure S1f–g). As a proof of principle, upon isolating neutrophils and comparing against other cell types, we identified previously known genetic drivers of human congenital neutropenia (Figure S1h), suggesting that this platform can be leveraged to answer questions related to myeloid biology in vivo at an genome-wide scale and accuracy.

Figure 1 – Lipid metabolism as a target in Tet2 mutant inflammation.

A)Schematic for the UMI-based in vivo CRISPR/Cas9 KO screen in a zymosan-induced peritonitis model using primary WT and Tet2 KO HSPCs (CD45⁺Lin⁻Sca1⁺c-Kit⁺). Migrated neutrophils and monocytes were compared against their counterparts in the bone marrow 12 hours after 1 mg i.p. zymosan exposure within the same genotype. For each gRNA, UMI counts were treated as gRNA abundances in the MaGeCK analysis to control for stochastic ‘jackpotting’ effects. To select for Tet2 KO specific hits, WT effect sizes (β) were compared to that of Tet2 KO following MaGeCK analysis.

B) CRISPR screen results with genes ranked by effect size, displaying top genes selectively required in Tet2 KO for inflammation comapared with WT. Red points indicate significant Tet2 KO-specific hits (FDR < 0.3, Tet2 KO – WT effect size > 2). FDR = false discovery rate.

C) Gene ontology (GO) gene set enrichment analysis based on CRISPR effect sizes (Tet2 KO – WT) to identify Tet2 KO inflammation selective gene sets. NES = normalized enrichment score.

D) Confocal imaging showing neutral lipid staining with BODIPY 493/503 (green) or LipidTOX DeepRed (red) in mouse neutrophils and monocytes after zymosan-induced peritonitis, and in human macrophages derived from CD34⁺ HSPCs. DAPI (blue) marks nuclei. Scale bar = 10 um; inset displays a representative cell.

We investigated genes whose disruption by gRNAs selectively reduced the number of Tet2 KO neutrophils and monocytes to the inflamed peritoneum, compared with WT cells. The metabolic enzymes O-linked N-acetylglucosamine transferase (Ogt) and ATP-citrate lyase (Acly) were among the top-ranked genes selectively required for Tet2 KO inflammation (Figure 1b, Supplemental Table 1). To determine whether these screen hits were related to pathway-specific dependencies in Tet2 KO cells, we performed gene set enrichment analysis. Lipid metabolic pathways emerged as the most enriched gene sets required for Tet2 KO inflammation compared to WT (Figure 1c, S1i). To investigate the phenotype of lipid metabolism in Tet2 KO cells, we performed lipid staining and confocal microscopy. Neutral lipid staining showed that the loss of Tet2 markedly increases lipid droplets in mouse macrophages and neutrophils (Figure 1d, S1j). Similarly, human CD34⁺ progenitor-derived macrophages with Cas9 ribonucleoprotein-mediated KO of TET2 showed increased neutral lipid staining compared to AAVS1-edited control macrophages (Figure 1d, S1k). Tet2 heterozygous mutations also increased neutral lipid accumulation and suggested a dose-dependent relationship between Tet2 protein levels and lipid accumulation (Figure S1l–p).

Tet2 restricts Ogt-mediated glycosylation activity

Our screen highlighted master nutrient sensor Ogt as a critical mediator of inflammation in Tet2 KO cells. Ogt is a glycosyltransferase that integrates several nutrient signals such as glucose to catalyze the addition of O-GlcNAc moieties onto serine and threonine residues of proteins (Figure 2a)^19,20. Having found that Tet2 loss leads to aberrant lipid accumulation, we hypothesized that Tet2 loss could dysregulate Ogt activity and lead to this metabolic abnormality. Although we did not detect a difference in overall glucose uptake between WT and Tet2 KO cells (Figure S2a), using C-13 nutrient tracing studies we found that Tet2 KO macrophages have a strikingly increased incorporation of C-13-labeled glucose into UDP-N-acetylglucosamine (UDP-GlcNAc), the substrate for Ogt (Figure 2a, S2b–c, Supplemental Table 2), as well as its precursors in the pentose phosphate pathway responsible for UDP nucleotide generation (Figure S2d). Collectively, the glucose tracing studies demonstrate that Tet2 KO BMDMs convert more glucose into UDP-GlcNAc, consistent with elevated Ogt activity in the setting of Tet2 KO.

Figure 2 – Increased Ogt-mediated glycosylation in Tet2 KO.

A) Uniformly labeled-C13 [U-C13] glucose tracing study in which BMDMs are treated for 24 hours in defined serum-free media. Central carbon metabolites were compared between WT and Tet2 KO after incorporation of U-C13 glucose, showing a mass shift due to C13 (M+). N = 2. UDP-HexNAc (i.e. GlcNAc, GalNAc) M+13 was among the most increased isotopologues in Tet2 KO compared to WT, and M+6 through M+13 intermediate isotopologues were significantly elevated in Tet2 KO compared to WT. Schematic of the C13 labeling is shown on top of the molecular components of UDP-GlcNAc and their sources. Statistical comparisons made with two-way ANOVA and multiple comparisons were corrected using the two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli (BKY).

B) Immunoblotting for O-GlcNAc (MultiMab; CST) in WT and Tet2 KO RAW264.7 macrophage-like cells. N = 3.

C) IP-MS for O-GlcNAc in WT and Tet2 KO RAW264.7 macrophage-like cells expressing ectopic Myc tagged-Ogt using antibodies specific for O-GlcNAc²¹. Distribution of significantly enriched proteins in WT or Tet2 KO shown as a violin plot (FDR < 0.25, log2 fold change < or > 2 respectively).

Distribution comparisons were made using the Welch’s two sample t-test. N = 4.

We next investigated whether the O-GlcNAc post-translational modification is increased in Tet2 KO macrophages, as would be expected from increased Ogt activity. We examined O-GlcNAc labeling via immunoblotting and immunoprecipitation-mass spectrometry (IP-MS) using antibodies specific for O-GlcNAc²¹. In contrast to previous reports that Tet2 is required for histone O-GlcNAcylation^22–24, both methodologies revealed a striking increase in O-GlcNAc-modified proteins in Tet2 KO compared to WT myeloid cells (Figure 2b–c, S2e–j, Supplemental Table 2).

Tet2 mutations lead to Ogt-dependent lipid accumulation and consequent inflammation

To investigate whether the specific types of lipid species that are altered in the setting of Tet2 inactivation are consistent with increased Ogt activity, we performed mass spectrometry-based lipidomic analyses of mouse neutrophils and BMDMs as well as human CD34⁺ progenitor-derived macrophages. Di- and triacylglycerols (DG, TGs) and cholesterol esters (CEs), which are key components of lipid droplets²⁵, were markedly increased in Tet2 KO cells compared to WT (Figure 3a, Supplemental Table 3).²⁶ Polar metabolite profiling independently revealed that glycerophospholipid metabolism was the top-ranked metabolic pathway enriched in Tet2 KO BMDMs compared to WT (Figure S3a). Short- and long-chain acylcarnitines were also increased in Tet2 KO BMDMs (Figure S3b) and C-13 nutrient tracing of glucose revealed incorporation into palmitate, consistent with a lipid anabolic process that increases lipid droplets (Figure S3c).

Figure 3 – Tet2 KO inflammation depends on increased Ogt activity.

A) Lipidomic analysis of neutrophils during zymosan-induced peritonitis (N = 4), bone-marrow derived macrophages (BMDMs; N=7), and human CD34⁺ derived macrophages (N=5). Neutrophils and macrophages were obtained from independent mice harvests. Human CD34⁺ derived macrophages were obtained from at least two independent cord blood samples.

B). Neutral lipid staining with LipidTOX DeepRed during zymosan-induced peritonitis in WT and Tet2 KO Cas9+ HSPCs transduced with control or Ogt gRNAs. N = 5–7. Statistical comparisons made with one-way ANOVA and multiple comparisons were corrected with BKY.

C) Neutral lipid staining with BODIPY 493/503 in peripheral blood neutrophils, monocytes, or BMDMs treated with OGA inhibitor (Z)-PUGNAc (50 uM) for 4 hours. N = 10. Statistical comparisons made with one-way ANOVA and multiple comparisons were corrected with BKY.

D-F) Single cell RNA-sequencing of Cas9+ WT and Tet2 KO monocytes with control gRNA or Ogt gRNA. D) Expression of Ccr2 is shown. E) Heatmap showing reduction of Tet2 KO gene signature with Ogt gRNA. F) Gene set enrichment analysis quantifying the level of reduction in of Tet2 KO gene signature with Ogt gRNA

G) Relative Tet2 KO vs WT infiltration into the peritoneum following 1 mg zymosan i.p. with ctrl or Ogt gRNA-transduced Tet2 KO vs WT Cas9+ HSPCs. WT HSPCs were transduced with a blue fluorescent protein (BFP) vector and Tet2 KO HSPCs transduced with a red fluorescent protein (RFP) vector. N = 6–8. Comparisons made using Welch’s two sample t-test.

To determine whether the aberrant neutral lipid droplets in Tet2 KO are controlled by Ogt in the setting of physiologic nutrient availability, we assessed neutral lipid staining in vivo after transplantation of WT and Tet2 KO Cas9⁺ HSPCs transduced with control gRNAs or gRNAs targeting Ogt in the setting of zymosan peritonitis. Ogt disruption significantly decreased neutral lipids (Figure 3b), consistent with prior reports that Ogt is required for palmitate levels (Figure S3d)²⁷ and linking selective dependence of Tet2 KO-mediated hyperinflammation on Ogt to intracellular lipid homeostasis. Conversely, inhibiting the deglycosylase Oga chemically using (Z)-PUGNAc led to increased neutral lipids accumulation in WT myeloid cells to levels similar to Tet2 KO cells (Figure 3c, Supplementary Table 4), increased IL1β protein levels (Figure S3e), and increased the expression of genes associated with Tet2 KO (Figure S3f–g).

To confirm whether Ogt is differentially required for Tet2 KO inflammation as indicated by our CRISPR screen, we performed single-cell RNA-sequencing of WT and Tet2 KO Cas9⁺ monocytes with control or Ogt gRNAs in the context of zymosan peritonitis (Figure S3h–i) and examined Tet2 KO inflammatory gene expression (Figure 3d–e). Ogt gRNAs dramatically reversed the expression of genes increased in Tet2 KO monocytes including Il1b (Figure 3e–f, Supplemental Table 5), which was confirmed using chemical inhibition of OGT (Figure S3j). Next, we transplanted WT and Tet2 KO Cas9⁺ HSPCs transduced with control or Ogt gRNAs and compared relative inflammatory accumulation of monocytes and neutrophils in the peritoneum (Figure S3k–l). Whereas Tet2 KO myeloid cells were enriched in the control gRNA group compared to WT, Ogt gRNAs reduced Tet2 KO inflammatory response to levels similar to WT (Figure 3g). Changing the fluorescent proteins used to track gRNA-transduced cells, altering the dose of inflammatory stimuli, or accounting for the baseline chimerism did not affect the results (Figure S3m–o). Notably, Ogt gRNAs did not significantly change the number of WT myeloid cells in this setting (Figure S3p–q). In summary, Ogt glycosylation activity is enhanced in Tet2 KO myeloid cells, associates with increased lipid accumulation, and is required for Tet2 KO hyperinflammation, consistent with results of our CRISPR screen.

Disruption of Ogt-Tet2 interaction exacerbates inflammation

Having found that loss of Tet2, Ogt’s binding partner, increases Ogt’s glycosylation activity, we investigated whether disrupting the Tet2-Ogt interaction is sufficient to alter lipid accumulation and inflammatory responses. While prior studies have found that TET2 and OGT interact, the binding interface has not been defined, and the functional effect of this binding has not been definitively shown^22,23. To identify amino acids on TET2 that are critical for OGT interaction, we developed a split mNeon reporter system tagged to OGT and TET2 that fluoresces when mNeon1-10-OGT and TET-mNeon11 associate (Figure S4a–b), previously used to study protein-protein interactions²⁸. We then performed an adenine base editor screen of the OGT-binding domain on TET2²⁹. We identified a disordered region of TET2 that is required for OGT binding (Figure 4a). In parallel, we employed AlphaFold to model the TET2-OGT interface and found the same disordered region of TET2 [amino acids 1606–1712] to interact with OGT (Figure S4c–f)^23,30,31.

Figure 4 – TET2 restricts OGT activity upon its binding to the TET2 disordered region.

A) Adenine base editor screens in the human U937 cell line demonstrating TET2 amino acid residues critical for TET2-OGT interaction. U937 cells stably expressing the split-mNeon constructs TET2-mNeon11 and mNeon1-10-OGT were subjected to a TET2 base editor screen. Higher score indicates residues important for the TET2-OGT interaction. Red-highlighted residues were selected for further interaction analysis.

B) Time-resolved Förster resonance energy transfer (TR-FRET) assay using 5’-GFP labeled peptides and terbium-labeled OGT in vitro. Mutated residues include 1609 and 1659, the top two residues identified by the base editor screen. N = 6.

C) In vitro O-GlcNAc glycosylation assay using UDP-Glo in the presence of increasing concentrations of recombinant TET2 protein. N = 4.

D) Zymosan-induced peritonitis using WT CD45.1/2+, Tet2 KO CD45.2+ HSPCs transduced with RFP-vector control, GFP-TET2 peptide (amino acids 1606–1712), or GFP-TET2 peptide with Ala substitutions. N = 8–35. Both peptides contain a nuclear export signal. See Figure S4k–n. Statistical comparisons made with one-way ANOVA and multiple comparisons were corrected with BKY.

To validate the TET2-OGT interaction surface in vitro, we used a time-resolved Förster resonance energy transfer (TR-FRET) assay using a GFP-labeled TET2 peptide containing the OGT-binding domain described above [amino acids 1606–1712] and terbium-labeled OGT (Figure 4b). Alanine substitutions in both key TET2 residues identified in the base editor screen abrogated OGT binding (Figure 4b). In contrast, a peptide containing the distal TET2 C-terminal region did not bind OGT (Figure S4c). To investigate the effect of TET2 binding on OGT function, we performed in vitro glycosylation assays, which revealed that purified TET2 potently restricts OGT activity (Figure 4c). Further studies reveal that aberrant lipid accumulation can be reversed with induction of catalytically inactive TET2 (Figure S4g–h).^9,32 Lastly, we did not appreciate significant expression of Tet1 in myeloid cells, though Tet1 has been reported to bind to Ogt in other cell types (Figure S4i–j).³³ Together, these results confirm that TET2 selectively binds and sequesters OGT activity.

Having found that the TET2-OGT interaction restricts OGT activity and having validated a TET2 peptide that binds OGT by TR-FRET, we examined whether forced expression of the OGT-binding TET2 peptide is sufficient to suppress inflammation in Tet2 KO myeloid cells in vivo. Recognizing that OGT shares >98% amino acid sequence conservation between mouse and humans, we introduced the human OGT-binding TET2 peptide in WT or Tet2 KO mouse HSPCs via lentiviral transduction and assessed its effect on inflammatory responses in the zymosan-induced peritonitis model. TET2 peptide expression was sufficient to reduce inflammation resulting from Tet2 KO while the peptide with impaired OGT-binding had minimal effect (Figure 4d, S4k–n). We found that the addition of a nuclear export signal (NES) to the peptide strengthened our phenotype on inflammatory reduction in Tet2 KO (Figure 4d, S4k). As TET2 restricted OGT glycosylation activity (Figure 4c, S4g), we evaluated the effect of TET2 peptide on O-GlcNAcylation in the WT and TET2 KO U937 cell line and found that it inhibited O-GlcNAcylation in TET2 KO cells and reduced neutral lipid accumulation (Figure S4o–p). Taken together, we have identified a region of TET2 that is required for binding to OGT and inhibition of OGT activity; the overexpression of this peptide in Tet2 KO cells is sufficient to inhibit lipid droplet formation and inflammation.

Loss of Tet2 potentiates Ogt activity on chromatin including on lipid gene loci

Having found that Tet2 restricts OGT activity (Figure 4) and the transcriptional profile of Tet2 KO is consistent with increased OGT activity (Figure 3e, S3f–g), and based on the established role of OGT as a nutrient sensor with effects on histones^20,23, we examined whether Tet2 loss impairs a feedback loop in response to nutrient availability. Our metabolomic studies demonstrated increased incorporation of C-13 glucose in Tet2 KO cells, so we examined whether Tet2 KO cells have aberrant Ogt activity across various glucose concentrations. We assessed Ogt activity on chromatin in WT and Tet2 KO BMDMs by chromatin immunoprecipitation sequencing (ChIP-seq) and found increased Ogt binding and increased O-GlcNAcylation at the same loci, in various glucose concentrations but not sucrose or galactose (Figure 5a–b, S5a–h). To determine the consequences of increased Ogt on chromatin, we examined active and repressive chromatin marks via ChIP-seq. We observed increased activating marks including H3K27 acetylation (H3K27ac) and H3K4 trimethylation (H3K4me3) but not repressive marks such as H3K27 trimethylation (H3K27me3), in Tet2 KO compared to WT BMDMs (Figure S5i). We also utilized a vector system in WT and TET2 KO human U937 cells in which doxycycline-inducible OGT is tagged with the TurboID proximity labeling system, allowing capture of transient interactions³⁴. Upon OGT-induction, TET2 KO cells displayed a dose-dependent increase in H3K4me3 marks compared to WT, which correlated with OGT-proximity labeling and suggested SET1-mediated effect (Figure 5c–d, S5j–k)^22,33.

Figure 5 -. Unrestrained Ogt in Tet2 KO contributes to an inflammatory lipid gene signature.

A-B) ChIP-seq analysis in WT or Tet2 KO BMDMs for Ogt or O-GlcNAc in A) 12 mM glucose or B) 55mM glucose for 24 hours in defined serum-free media. All ChIP-seq peak loci were ordered in descending Ogt peak intensity. Genomic window is −/+ 5kb around the peak.

C) ChIP-seq in WT or TET2 KO human U937 cells with a doxycycline-inducible OGT tagged with a 5’ TurboID proximity labeling system. Cells were processed for (B) H3K4me3 ChIP-seq and (C) strep-avidin ChIP-seq after OGT proximity labeling via 300 nM biotin supplementation plus doxycycline induction for 18 hours. Cells were otherwise grown in biotin-free RPMI. OGT binding to chromatin at 0.01ng/ml doxycycline mirrored control data prior to the induction. Genomic window is −/+ 5kb around the peak.

D) Examination of gene loci for H3K4me3, Ogt, O-GlcNAc ChIP-seq peaks corresponding to transcripts significantly upregulated in Tet2 KO compared to WT BMDMs on RNA-seq (FDR < 0.25, log2 fold change > 0.5). TSS = transcriptional start site. TES = transcriptional end site. Genomic window is −/+ 5kb around the peak.

E) Analysis of promoter and gene bodies for H3K4me3, Ogt, and O-GlcNAc ChIP-seq peaks at lipid-associated O-GlcNAc modified gene loci using Gene Ontology associations.

To examine Ogt’s effect on transcription, we performed RNA-sequencing on WT and Tet2 KO BMDMs. We first examined all genes significantly increased in Tet2 KO BMDMs compared to WT, and then lipid metabolism-associated genes. For both sets of genes, Tet2 KO BMDMs showed increased Ogt binding in the promoter region that is also associated with increased O-GlcNAc and H3K4me3 (Figure 5d–e, Supplemental Table 6). Accordingly, these lipid genes with increased transcription in Tet2 KO BMDMs compared to WT were largely confirmed as hits on the Tet2 KO inflammation CRISPR screen (Figure S5l–m). In summary, Tet2 loss increases Ogt chromatin binding and enzymatic activity, enhances Ogt-dependent accumulation of activating chromatin modifications including H3K4me3, and stimulates a transcriptional program important for lipid accumulation.

ACLY dependence in TET2 KO inflammation

Our findings indicate that Tet2 KO cells have Ogt-dependent neutral lipid droplet formation and Ogt-mediated chromatin modifications that lead to expression of lipid metabolism-associated genes. To identify mediators of inflammatory lipid droplet formation in Tet2 KO myeloid cells, we performed a targeted CRISPR/Cas9 screen focused on lipid and metabolic mediators in the setting of zymosan-peritonitis and selected neutrophils and monocytes based on dual staining intensity of BODIPY 493/503 and LipidTOX Deep Red. This screen revealed that the ATP-citrate lyase (Acly) gene scored most highly for Tet2 KO-dependent lipid droplet formation (Figure 6a Supplemental Table 7). Consistent with this finding, Acly was also among the top Tet2 KO-selective hits in our genome-wide CRISPR screen (Figure 1b). Acly is a rate-limiting regulator of lipogenesis that catalyzes the conversion of citrate to acetyl-CoA, a substrate for both fatty acids and cholesterol biosynthesis³⁵. Acly expression was elevated in Tet2 KO cells, and experimental modulation of Ogt expression produced a similar change in Acly levels (Figure S6a–e). Additionally, we found that Acly O-GlcNAcylation is increased in Tet2 KO cells compared to WT (Figure 6b), a modification that has been reported to increase Acly activity dramatically by facilitating binding to CoA³⁶. Tet2 loss would thereby lead to increased Acly activity and consequent lipid biogenesis.

Figure 6 – Acly is a therapeutic target critical for lipid accumulation in Tet2 mutants.

A) Targeted CRISPR screen for genes involved in metabolism or lipid processing in Tet2 KO HSPCs, using dual BODIPY 493/503 and LipidTOX Deep Red staining in neutrophils and monocytes during zymosan-induced peritonitis. Overlap with prior CRISPR screen results on zymosan-induced peritoneal inflammation is shown on the x-axis.

B) IP-MS for O-GlcNAc corresponding to Figure 2C.

C) Inflammatory gout model using monosodium urate (MSU) crystals. Mice were treated with BMS303141 vs vehicle 5mg/kg i.p. for 4 days prior to a 1mg MSU injection with or without 125ug BMS303141 into the paw. Paw edema was assessed 12 hrs post-injection. N = 5–10. Statistical comparisons made with one-way ANOVA and multiple comparisons were corrected with BKY. Transected paw histology (hematoxylin and eosin stain) is shown on right, with epidermal tissue oriented upwards. Tet2 KO mice treated with vehicle exhibit increased dermal/subcutaneous edema and polymorphous inflammatory infiltration compared with WT. Infiltrating inflammatory cells are notably decreased with BMS303141 treatment. 600x magnified images (bottom) highlight myeloid infiltration into underlying musculature. Black scale bars = 100 um.

D) Single cell RNA-sequencing of Cas9+ WT and Tet2 KO monocytes with control gRNA or Acly gRNA. On left, heatmap shows reduction of Tet2 KO gene signature with Acly gRNA. The list of genes is identical to the list of genes displayed with Ogt gRNA in Figure 3E. On right, gene set enrichment analysis quantifies the level of reduction in of Tet2 KO gene signature with Acly gRNA.

E) Relative Tet2 KO vs WT infiltration into the peritoneum with 1 mg zymosan, following transplantation of HSPCs transduce with control or Acly gRNAs in Cas9⁺ HSPCs. N = 15–20. Comparisons made with Welch’s two sample t-test.

F) Quantification of aortic atherosclerosis after transplantation of Cas9+ CD45.2 WT or Tet2 KO HSCs (CD150+CD48-CD135-LSK) transduced with control gRNA or Acly gRNA, competing against WT CD45.1/2 HSCs at a 20:80 ratio. After transplantation into male Ldlr KO recipients, mice were harvested after 10 weeks on high-fat diet. Aortic plaques areas were measured using ImageJ after Oil Red O staining. N = 5–8. Statistical comparisons made with one-way ANOVA and multiple comparisons were corrected with BKY.

To investigate whether ACLY inhibition reduces TET2 KO inflammation in inflammatory models associated with CHIP, we first tested the effect of a specific ACLY inhibitor, BMS303141^37,38. BMS303141 treatment specifically inhibited lipid accumulation and reduced inflammasome activation preferentially in Tet2 KO BMDMs (Figure S6f–h), and reduced inflammatory paw edema and infiltration in a mouse model of monosodium urate (MSU) crystal arthropathy (gout) (Figure 6c)⁶. We next targeted Acly genetically using CRISPR/Cas9 KO in the context of zymosan peritonitis and found that Acly KO decreased Tet2 KO monocyte inflammatory gene expression (Figure 6d, Supplemental Table 5) and decreased inflammatory accumulation of neutrophils and monocytes (Figure 6e). CRISPR/Cas9-mediated KO of other genes involved in lipogenesis downstream of Acly such as Acacb, Fasn, Hmgcr or Ptges2; or chemical inhibition of lipid droplet formation reduced Tet2 KO-specific inflammation (Figure S6i–l)²⁶.

To examine the impact of Acly KO on Tet2 KO-mediated atherosclerosis, we transduced WT and Tet2 KO Cas9+ HSCs with either control or Acly gRNAs. We then conducted competitive transplantation into atherosclerosis-susceptible Ldlr KO recipient mice, using a 20:80 ratio against WT HSCs. Following 10 weeks on a high-fat diet, Tet2 KO transplant mice with control gRNAs exhibited significantly increased aortic atherosclerotic plaque formation compared to WT, an effect that was reversed with Acly gRNAs (Figure 6f, S6m–q). In aggregate, these results demonstrate that altered lipid metabolism is a critical mediator of Tet2 KO hyperinflammation, and highlights the importance of Acly in its regulation.

Discussion

We used a genome-wide, UMI-tagged CRISPR/Cas9 screens in vivo to illuminate disrupted metabolic circuit by which loss of Tet2 promotes hyper-inflammation in myeloid cells, highlighting a nutrient-sensitive checkpoint that governs innate immunity. We found that TET2 restrains the activity of the nutrient sensor OGT. In TET2 KO cells, OGT activity is dysregulated, resulting in increased O-GlcNAcylation, an aberrant chromatin state, altered gene expression, increased lipid biogenesis, and lipid droplet accumulation. Dysregulated OGT activity fuels ACLY-dependent lipogenesis, positioning ACLY as a critical metabolic effector and potential therapeutic target for patients with TET2-related inflammation^35,39.

OGT functions as a metabolic integrator in the hexoamine biosynthesis pathway (HBP), translating nutrient availability via its glycosyltransferase activity into posttranslational signals that ultimately influence transcriptional outputs. During metabolic stress or nutrient deprivation, OGT integrates levels of glucose, amino acids, fatty acid, and nucleotide availability as substrates in the hexosamine biosynthesis pathway and relays those signals into posttranslational modifications^19,20. In Tet2 KO cells, dysregulated OGT leads to maladaptive nutrient sensing causing excessive lipid accumulation and exaggerated activation of inflammatory signaling. Our CRISPR/Cas9 screening approach was designed to capture this complexity by measuring Tet2 KO inflammation in vivo with physiologic nutrient availability. These findings reposition TET2 as not only an epigenetic modifier but also a gatekeeper of nutrient sensing and metabolic homeostasis in the immune system. Previous studies have demonstrated an interaction between TET2 and OGT^23,40,41, but results have been conflicting on whether TET2 promotes OGT activity^22–24, and it is not known whether OGT activity might be linked to inflammation in TET2 mutant cells. Mechanistically, we identified a binding interface on TET2 that interacts with OGT and negatively regulates OGT enzymatic activity. There may be several mechanisms of metabolic feedback to govern TET2 function. Accumulating evidence suggests TET2 responds to metabolic signals, and our findings indicate that the role of TET2 as a metabolic sensor is mediated through altered OGT activity. Glucose-dependent phosphorylation of TET2 by AMPK stabilizes TET2⁴², thus augmenting negative feedback on OGT upon glucose excess. TET2 suppresses amino acid metabolism via mTORC1 signaling⁴³, which may reduce the glutamine levels necessary for amplified HBP activity. While the metabolic feedback on TET2 function is multifactorial, our results demonstrate that loss of the TET2-OGT interaction causes lipid accumulation and hyperinflammation in the setting of TET2 deficiency.

Our unbiased screens demonstrate that these metabolic abnormalities are central drivers of aberrant inflammation in Tet2 KO cells. Our study indicates that abundant nutrient availability accentuates TET2-mutant inflammation via OGT-driven neutral lipid accumulation. These results are consistent with a prior report showing that hyperglycemia cooperates with Tet2 loss to induce leukemia driven by proinflammatory cytokines⁴⁴. We and others have observed that adding palmitate or low-density lipoprotein rich in cholesterol esters potentiates TET2-mutant inflammation in response to lipopolysaccharide and other stimuli^2,5. Lipid droplets propel inflammation through multiple routes, including liberation of eicosanoids such as prostaglandins²⁶, facilitating the emergence of inflammatory cytokines²⁶, and fueling acetylation-driven inflammatory transcription by supplying acetyl-CoA⁴⁵. Similarly, the macrophage “foam cells” essential for pathology within atherosclerotic plaques are laden with lipid droplets⁴⁶. Given the established association between TET2-mutant clonal hematopoiesis and increased risk of atherosclerosis^2,8, our work suggests that TET2 mutations and intracellular lipid excess are not just coincident, but rather inextricably linked from metabolic rewiring of myeloid cells. Recent reports also implicate metabolic reprogramming and enhanced mitochondrial respiration as key drivers of DNMT3A-mutated clonal expansion^47–49, although inflammation was not the primary focus of those studies. These findings support a broader model in which nutrient excess synergizes with CHIP mutations to reshape myeloid inflammatory tone, with implications for CHIP-associated diseases and trained immunity.

Our findings support a model in which TET2 binds OGT in the nucleus and suppresses both its glycosylation activity and its contributions to other activating modifications on chromatin, most strikingly H3K4me3. We report that OGT binding to the TET2’s disordered region inhibits inflammation via suppression of OGT activity. A recent publication described this disordered region as important for biomolecular condensation and TET2’s interaction with other epigenetic regulators, suggesting a role for phase separation in steady-state TET2-OGT association⁵⁰. As OGT is known to associate with the Set1/COMPASS family of H3K4me3 methylases^22,33, the enhanced H3K4me3 we observe in the absence of TET2 may result from a direct effect of OGT on Set1/COMPASS activity or indirect influence via H2B glycosylation²³ or RNA polymerase II regulation⁵¹. In either model, dysregulated OGT chromatin activity increases transcripts associated with lipid metabolism and leads to overt accumulation of cytoplasmic lipid droplets, consistent with previous studies^27,52. Prior studies using embryonic stem cells to examine the epigenetic consequences of OGT-TET family interactions have yielded opposing results (including effects on H3K4me3)^22–24,33. While this may reflect cell-type, nutrient or inflammatory context specific differences, the dose-dependent relationship between OGT and H3K4me3 abundance in our inducible OGT system supports a causal role for OGT in augmenting H3K4me3 deposition and, thus, lipid accumulation in TET2 KO myeloid cells.

Identifying Acly as a key mediator of lipid accumulation and inflammation in Tet2 KO cells connects Ogt-mediated transcriptional changes with inflammatory lipid droplets in the cytoplasm. Acly itself is O-GlcNAcylated by Ogt and O-GlcNAcylation of Acly is increased in the context of Tet2 loss. This modification dramatically increases Acly activity³⁶, and Ogt and Acly may act in a complex on chromatin, thus explaining why the genetic disruption of Acly impairs Tet2 KO myeloid cells as much or more than Ogt disruption. This observation indicates that increased nutrient flow through this critical node plays a major role in Tet2 KO inflammation, and that glucose or lipid management may have therapeutic implications for TET2-CHIP. A liver-restricted ACLY inhibitor, bempedoic acid, is currently a component of an FDA-approved medication for cardiovascular disease⁵³, providing proof-of-concept for therapeutic tractability. Multiple highly specific, widely biodistributed ACLY inhibitors are currently in varying stages of development³⁹. The ability of one such compound, BMS-303141, to nearly ablate Tet2 KO hyperinflammation in our murine gout experiments suggests that our elucidation of these mechanistic pathways may yield promising options for patients with TET2 mutations.

In summary, our findings reveal dysregulated metabolic circuit as an underlying driver of hyperinflammation in TET2 mutant myeloid cells, driven by OGT and ACLY hyperactivity and resulting in intracellular lipid overload. This biology implicates multiple approaches for amelioration, including sequestration of OGT, inhibition of ACLY, and modulation of the nutrient microenvironment.

Limitations of the Study

This study focuses on the mechanisms underlying hyper-inflammation in TET2-mutated clonal hematopoiesis, but the same mechanisms may not apply to clonal expansion of TET2-mutated HSPCs. While this study cannot account for all the mutations found in TET2-mutated CHIP and hematological malignancies, nearly two thirds of the mutations are truncating and are predicted to influence OGT biology, although this has not yet been systematically studied. Missense mutations may exert additional effects (e.g. influencing post-translational modifications, protein stability or degradation). Finally, while our studies focus on the de-repression of OGT upon TET2 loss, we cannot exclude other potential mechanisms by which TET2 mutants increase OGT activity.

Resource Availability:

Lead Contact:

Further requests for reagents and resources should be directed to and will be fulfilled by the lead contact, Benjamin L. Ebert (E-mail: Benjamin_Ebert@dfci.harvard.edu)

Materials Availability:

All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Data and Code Availability:

• Sequencing data was deposited in GEO/SRA. Mass spectrometry results have been deposited in PRIDE or UCSD Metabolomics Workbench. Accession numbers are listed in the key resources table. All deposited data will be publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper will be made upon request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-OGT	Cell Signaling Technology	Cat#24083S
O-GlcNAc MultiMab™ antibody	Cell Signaling Technology	Cat#82332
Mouse monoclonal anti-O-GlcNAc (RL2) (unconjugated)	Thermo Fisher Scientific	Cat#MA1-072
Mouse monoclonal anti-O-GlcNAc (RL2) Alexa Fluor 647 conjugate	Novus Biologicals	Cat#NB300-524AF647
Rabbit polyclonal anti-H3K4me3	Abcam	Cat#ab8580
Rabbit polyclonal anti-H3K27ac	Abcam	Cat#ab4729
Rabbit polyclonal anti-H3K4me1	Abcam	Cat#ab8895
Rabbit polyclonal anti-H3K27me3	Abcam	Cat#ab6002
Rabbit polyclonal anti-ACSS2	Abcam	Cat#ab133664
Rabbit monoclonal anti-Myc-tag	Cell Signaling Technology	Cat#2276S
Mouse monoclonal anti-β-actin	Thermo Fisher Scientific	Cat#MA5-15739
Rabbit monoclonal anti-GAPDH	Cell Signaling Technology	Cat#2118
Mouse monoclonal anti-α-tubulin	Sigma-Aldrich	Cat#T9026
Rabbit monoclonal anti-HDAC1	Cell Signaling Technology	Cat#5356S
Rabbit polyclonal anti-IL-1β (human)	GeneTex	Cat#GTX130021
Rabbit polyclonal anti-IL-1β (mouse)	GeneTex	Cat#GTX74034
Mouse monoclonal anti-Vinculin	Santa Cruz Biotechnology	Cat#sc-73614
Mouse monoclonal anti-Tet1	Thermo Fisher Scientific	Cat#MA5-16312
Bacterial and virus strains
One Shot™ Stbl3™ Chemically Competent E. coli	Thermo Scientific	Cat#C737303
Chemicals, peptides, and recombinant proteins
Zymosan	Sigma-Aldrich	Cat#Z4250
Palmitate	Sigma-Aldrich	Cat#P0500
OSMI-1 (OGT inhibitor)	Cayman Chemical	Cat#21894
(Z)-PUGNAc (OGA inhibitor)	Cayman Chemical	Cat#17151
BMS-303141 (ACLY inhibitor)	Cayman Chemical	Cat#16239
16,16-dimethyl PGE2	Cayman Chemical	Cat#14750
Doxycycline	Selleck Chemicals	Cat#S4163
Biotin	Sigma-Aldrich	Cat#B4501-5G
RetroNectin	Takara	Cat#T100B
FuGENE 6 transfection reagent	Promega	Cat#E2312
UDP-GlcNAc	Promega	Cat#V7071
OGT peptide substrate	AnaSpec	Cat#63726
Recombinant OGT protein	Antibodies-Online	Cat#ABIN1313360
Recombinant TET2 protein	Active Motif	Cat#31418
D-(+)-Glucose-13C6	Cayman Chemical; Sigma-Aldrich	Cat#26707, Cat# 389374
D-(+)-Glucose	Sigma-Aldrich	Cat#G7021
Sodium acetate-13C2	Sigma-Aldrich	Cat#282014
Sodium acetate	Sigma-Aldrich	Cat#S2889
Palmitic acid-13C16	Sigma-Aldrich	Cat#605573
L-Glutamine	Sigma-Aldrich	Cat#49419
L-Glutamine-13C5	Sigma-Aldrich	Cat#605166
Critical commercial assays
PTMScan® O-GlcNAc [GlcNAc-S/T] Motif Kit	Cell Signaling Technology	Cat#95220
UDP-Glo™ Glycosyltransferase Assay	Promega	Cat#V6961
Glucose Uptake-Glo™ Assay	Promega	Cat#J1342
RNeasy Plus Mini Kit	Qiagen	Cat#74106
NEBNext® Ultra™ RNA Library Prep Kit for Illumina	New England Biolabs	Cat#E7645S
ThruPLEX® DNA-Seq Kit	Takara Bio	Cat#R400674
DNA Clean & Concentrator	Zymo Research	Cat#D4014
AAVpro® Purification Kit Maxi	Takara Bio	Cat#6675
AAVpro® Titration Kit	Takara Bio	Cat#6233
CD34 MicroBeads	Miltenyi Biotec	Cat#130-046-702
Deposited data
Bulk and single-cell RNA-seq; ChIP-seq data	GEO	GSE298064; GSE309775
O-GlcNAc IP–MS data	PRIDE	PXD072791
Lipidomics data	Metabolomics Workbench	ST004537
Biological samples
Human umbilical cord blood-derived CD34+ cells	Dana-Farber Cancer Institute	N/A
Experimental models: Cell lines
HEK293T	Broad Institute Genetic Perturbation Platform	N/A
U937	Broad Institute Genetic Perturbation Platform	N/A
U937 TET2 KO	This paper
RAW264.7	ATCC	Cat#TIB-71
RAW264.7 Tet2 KO	Kim et al.4
Experimental models: Organisms/strains
Mouse: Tet2^fl/fl	The Jackson Laboratory	JAX:017573
Mouse: Vav1-Cre	Georgeaides et al.54
Mouse: Rosa26-Cas9 knock-in	The Jackson Laboratory	JAX:026179
Mouse: Ldlr knockout	The Jackson Laboratory	JAX:002077
Mouse: B6.SJL (CD45.1)	The Jackson Laboratory	JAX:002014
Mouse: C57BL/6J	Charles River	Cat#027
Oligonucleotides
Mouse Tet2 crRNA Sequence: GAACAAGCTCTACATCCCGT	IDT	N/A
Human TET2 crRNA Sequence: GATTCCGCTTGGTGAAAACG	IDT	N/A
Human CD34+ TET2 gRNA #1 Sequence: ACGGCACGCTCACCAATCGC	Loke and Kim et al.18 Waarts et al.55	N/A
Human CD34+ TET2 gRNA #2 Sequence: TGTAGCCAGAGGTTCTGTC	Loke and Kim et al.18 Thotova et al.56 Waarts et al.55	N/A
AAVS1 gRNA Sequence: GGGGCCACTAGGGACAGGAT	Loke and Kim et al.18 Waarts et al.55	N/A
CRISPR library sequencing primer (Read 1) Sequence: GTTTCAGACCCACCTCCCAAC	Loke and Kim et al.18	N/A
CRISPR library sequencing primer (Read 2) Sequence: GAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCC	Loke and Kim et al.18	N/A
Recombinant DNA
pMD2.G	Addgene	Plasmid #12259
psPAX2	Addgene	Plasmid #12260
pInducer21	Addgene; Meerbrey et al.57	Plasmid #46948
SFFV gRNA expression vectors	Addgene	Plasmids #169941 and #169940
pAAV-MCS	Cell Biolabs	Cat#VPK-411
pAAV-AAVS1 targeting vector for HDR	Waarts et al.55 Loke and Kim et al.18
pAAV-TET2 targeting vector for HDR	Waarts et al.55 Loke and Kim et al.18
CROPseq-guide-puro	Datlinger et al.58	Plasmid #86708
Software and algorithms
MAGeCK v0.5.9.2	Li et al.59	https://sourceforge.net/projects/mageck/
R v3.6.1	R Project	https://cran.r-project.org/
Prism v7.0	GraphPad	https://www.graphpad.com/
STAR aligner v2.7.2b	Dobin et al.60	https://github.com/alexdobin/STAR
RSEM v1.3.1	Li and Dewey61	https://github.com/deweylab/RSEM/releases
DESeq2 v1.50.2	Love et al.62	https://bioconductor.posit.co/packages/3.19/bioc/html/DESeq2.html
Cell Ranger v9.0.1	10x Genomics	https://www.10xgenomics.com/
Seurat v3	Stuart et al.63	https://cran.r-project.org/package=Seurat
Bowtie2 v2.3.5	Langmead and Salzberg64	https://github.com/BenLangmead/bowtie2
Picard v2.19.0	Broad Institute	https://broadinstitute.github.io/picard/
MACS2 v2.1.1	Zhang et al.65	https://pypi.org/project/MACS2/
deepTools v3.4.3	Ramírez et al.66	https://deeptools.readthedocs.io/
CellProfiler v4.2.8	Broad Institute	https://cellprofiler.org
MaxQuant v1.6.2.6	Cox and Mann67	https://maxquant.org/
TraceFinder	Thermo Fisher Scientific	https://www.thermofisher.com/us/en/home/industrial/mass-spectrometry/liquid-chromatography-mass-spectrometry-lc-ms/lc-ms-software/lc-ms-data-acquisition-software/tracefinder-software.html
Progenesis QI	Nonlinear Dynamics	https://www.nonlinear.com/progenesis/qi/download/
MZmine v3.10	Schmid et al.68	https://github.com/CSi-Studio/mzmine3
MetaboAnalyst v6.0	Pang et al.69	https://www.metaboanalyst.ca/
Other
High-fat, high-cholesterol diet	Teklad	TD.96121
Sony MA900 cell sorter	Sony Biotechnology	Model MA900
Covaris E220 sonicator	Covaris	E220
Illumina NextSeq 550	Illumina	NextSeq 550
Illumina NovaSeq (SP)	Illumina	NovaSeq SP
Streptavidin magnetic beads	Thermo Fisher Scientific	Cat#88817

STAR Methods

Experimental model and study participant details

Animals -

To generate Tet2 KO, Tet2-floxed line (JAX 017573)¹² were crossed to Vav1-Cre⁵⁴. To generate Cas9-expressing mice, Tet2 KO mice (vavCre+Tet2fl/−)¹² or WT mice (vavCre+) were crossed to Cas9 knock-in mice (JAX 026179)⁷⁰. To generate hematopoietic-specific KOs, B6.SJL and C57BL/6J mice at 8 weeks were exposed to 10 Gy split dose irradiation and transplanted through retro-orbital injection HSPCs collected from sex-matched Tet2 KO or WT donor mice aged between 8 and 12 weeks. For zymosan-induced peritonitis, mice were transplanted with Lin-Sca1+c-Kit+ HSPCs, allowed to engraft 4-weeks post-transplantation. Mice were then injected with 1mg intraperitoneal (i.p.) zymosan 12-hours prior to harvesting peritoneal and bone marrow cells. For atherosclerosis experiments, Ldlr KO (JAX 002077) mice were used as recipients after backcrossing to B6.SJL (JAX 002014) mice to generate mice homozygous for CD45.1. Mice were then started on a high-fat, high-cholesterol diet (TD.96121; 21% milk fat, 1.25% cholesterol diet; Teklad) 4 weeks after transplantation. Mice were on diet for 8–11 weeks prior to harvest. Mouse aortic root were subjected to histological analysis. Mouse aorta was fixed in formalin and stained in Oil Red for quantification of total atherosclerosis. For gout experiments, mice were treated with BMS303141 vs vehicle 5mg/kg i.p. for 4 days prior to a 1mg MSU injection with or without 125ug BMS303141 into the paw. Paw edema was assessed 12 hrs post-injection. Mice were genotyped using Transnetyx. Experimental animals were housed in an accredited vivarium under approved animal protocols, with standard husbandry (controlled temperature, humidity, light-dark cycle), routine health monitoring, and ad libitum access to food and water. All experiments were approved by the Institutional Animal Care and Use Committee.

Primary cell cultures -

For CD34+ cells, deidentified human cord blood was obtained from full-term deliveries from healthy donors (Brigham and Woman’s Hospital) or from the Pasquerello tissue bank (Dana-Farber Cancer Institute) after IRB approval. Mononuclear cells were obtained by Ficoll centrifugation and were enriched for CD34+ cells using the CD34 MicroBeads (Miltenyi). CD34+ cells were incubated in StemSpan SFEM II (StemCell Technologies) supplemented with 50 ng/mL SCF and FLT3L, 7.5 ng/mL TPO, 10ng/ml IL3 and IL6, PVA-F (0.1%, Sigma), 1uM UM729, 1uM SR1 (StemCell Technologies) and Pimocin (InvivoGen) as described⁵⁵. Complexed Cas9 RNPs and gRNAs were delivered into CD34+ using electroporation enhancer and Lonza 4D-Nucleofection system (CA-137). Media was changed and AAV containing the AAVS1 or TET2 targeting construct with mNeonGreen or mScarlett was added 15 minutes later at a MOI of 50000. We isolated transfected cells within one week of nucleofection using cell sorting using MA-900 using the GFP+ or BFP+ gate. gRNAs were picked by the CRISPick algorithm (Broad institute). For CD34+ cells, TET2 gRNAs included 5-ACGGCACGCTCACCAATCGC-3 and 5-TGTAGCCAGAGGTTCTGTC-3⁵⁶. AAVS1 gRNA sequence is 5-GGGGCCACTAGGGACAGGAT-3. Validation of decreased TET2 expression was performed previously.⁵⁵ For U937 cells, TET2 gRNA sequence included 5-GATTCCGCTTGGTGAAAACG-3 was used with the Cas9 RNP delivery system. Single cell clones were picked and frameshift mutations were validated using primers 5-TTTGGTAGCAGTGGAGAGC-3 and 5- GGGTTGATACTGAAGAATTGATGG-3.

Cell lines -

KO cells were generated via from RAW 264.7 cells or U937 cells (Broad Institute) both of male origin using recombinant Cas9 protein, synthetic locus-specific CRISPR RNAs (crRNA), negative control crRNA, and transactivating crRNAs (tracrRNA) (Integrated DNA Technologies) as described previously using isogenic single cell clones.⁴ Mouse Tet2 specific crRNA is 5-GAACAAGCTCTACATCCCGT-3 and results were verified using the primer set 5-AGGCTTCTGCGCTTGTT-3 and 5’-TCCTTATTCGAAGATCTTAACACCA-3’ previously. Human TET2 specific crRNA is 5- GATTCCGCTTGGTGAAAACG-3 and results were verified using the primer set 5- TTTGGTAGCAGTGGAGAGC-3 and 5’- GGGTTGATACTGAAGAATTGATGG-3’. Both were generated and validated from the CRISPick algorithm (Broad Institute). 2xMyc-tagged TurboID-tagged OGT or 2xMyc-tagged OGT was synthesized into ENTR vectors from Twist Biotechnologies and cloned into pInducer21 (Addgene #46948)⁵⁷ using LR Gateway reactions. BFP-T2A-Ogt or BFP vectors were synthesized into Lenti SFFV vector from Twist Biotechnologies. For CRISPR KO in Cas9+ cells, gRNA pools targeting a single gene or non-targeting controls from Brie were cloned into SFFV promoter expression vectors (Addgene #169941, 169940) using BsmBI-digestion (NEB) and Gibson Assembly (NEB). Cell line identity was confirmed by the Genetic Perturbation Platform (Broad Institute). Cultures were routinely tested for mycoplasma prior to experiments.

Cell Culture.

The human HEK293T (female) or U937 (male) cells were provided by the Genetic Perturbation Platform (Broad Institute). RAW264.7 (male) cells were obtained from ATCC. HEK293Ts and RAW264.7 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM [Gibco]), and U937 cells were cultured in RPMI (Gibco) with 10% fetal bovine serum (FBS; Invitrogen) and 5% glutamine and penicillin-streptomycin (PSG; Invitrogen) at 37°C and 5% CO2. Mouse BMDMs were generated as described previously² using 20ng/ml M-CSF and 10% FBS and 5% PSG in DMEM. Human macrophage differentiation was induced using GMCSF (5ng/ml) and MCSF (20ng/ml) in RMPI and 10% FCS for 8–12 days. For ChIPseq, BMDMs were incubated in a defined media containing DMEM without glucose or glutamine (Thermo A1443001), supplemented with ITS-X (Thermo), 20ng/ml M-CSF, HEPES (10mM), supplemented with 0.65mM glutamine (Sigma 49419), 5.5mM glucose, 0.5mM acetate (Sigma S2889), and 100uM palmitate (Sigma P0500) for 24 hours, washed with PBS twice, then incubated in 5.5mM glucose or 55mM glucose (Sigma G8270), 0.5mM acetate (Sigma S2889), and 100uM palmitate (Sigma P0500) 24 hours prior to harvest. For sucrose or galactose experiments, 55mM glucose was replaced with equimolar sucrose (Sigma 84097) or galactose (Sigma G5388) but also supplemented with 5.5mM glucose. For isotopologue tracing, BMDMs were incubated in a defined media containing DMEM without glucose or glutamine ((Thermo A1443001), supplemented with ITS-X (Thermo), 20ng/ml M-CSF, HEPES (10mM), supplemented with 0.65mM glutamine (Sigma 49419), 5.5mM glucose (Sigma G8270), 0.5mM acetate (Sigma S2889), and 100uM palmitate (Sigma P0500) for 24 hours, then washed with PBS twice, and then replaced with the same media but each nutrient was replaced with a uniformly labeled C13 version at the same concentration for 24 hours (Sigma 605573, 605166, 282014, 660663). For lipidomics or metabolomics experiments, mouse neutrophils were obtained from transplanted WT or Tet2 KO mice 4 weeks after transplantation, isolated 12 hours after zymosan 1mg i.p. injection using the neutrophil isolation kit (Miltenyi, 130-097-658). BMDMs were obtained from the bone marrow using differentiation method described above². Samples were cell number normalized prior to extraction. For immunofluorescence experiments, BMDMs were incubated in defined media containing DMEM without glucose or glutamine (Thermo A1443001), supplemented with ITS-X (Thermo), 20ng/ml M-CSF, HEPES (10mM), supplemented with 0.65mM glutamine (Sigma 49419), 5.5mM glucose, 0.5mM acetate (Sigma S2889), and 100uM palmitate (Sigma P0500) for 24 hours, washed with PBS twice, then various concentrations of glucose was added. For drug treatments, the following drugs were used OSMI1 (Cayman Chemical 21894), 16,16-dimethylPGE2 (Cayman Chemical 14750), (Z)-PUGNAc (Cayman Chemical 17151), doxycycline (Selleck Chemicals S4163), biotin (Sigma B4501–5G), BMS303141 (Cayman Chemical 16239).

Method Details

CRISPR/Cas9 Screens.

The LentiGuide-Puro vector was converted into the CROPseq-guide-puro using a ligase cycling reaction as previously described⁵⁸. Barcoding UMIs were constructed based on LARRY design^71,72 (Integrated DNA Technologies) and contained a 20bp flanking sequence for cloning into the KflI site of the CROPseq-guide-puro vector. The UMI fragment was annealed and incubated with 1U Klenow DNA polymerase and 50nmoles dNTPs as described previously⁷². After cutting the CROPseq-guide-puro with KflI restriction enzyme, the UMIs were cloned into the vector using Gibson Assesmbly (NEB). The mouse Brie library was then cloned into the barcoded vector⁷³. Sequencing of gRNAs and UMIs were performed using 200bp paired end sequencing on NovaSeq SP using 5’-GTTTCAGACCCACCTCCCAAC-3’ and 5’-GAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCC-3’ primers. Diversity was estimated by error-correction for hamming distance of 3⁷⁴ and modeled using a logistic regression for sequencing saturation. For in vivo genomewide screens, Lin-Sca1+c-Kit+ were sorted using Sony Sorter MA900 from Rosa26-Cas9 knockin mice (Jax 026179) bred to the vavCre+Tet2fl/− or control vavCre+ background. Sorted LSK cells were plated on RetroNectin (Takara) using a chemically defined HSC media⁷⁵. Brie virus was generated using HEK293T cells 2–3 days prior to transduction, then concentrated via ultracentrifugation for 2 hours at 25000rpm prior to LSK transduction. For the zymosan-peritonitis screen, at least 3 recipient C57BL/6J mice per genotype were transplanted with 1e6 – 5e6 Cas9+ WT or Tet2 KO LSKs one-day after transduction of genome-wide Brie library carrying UMI/barcodes. After 4 weeks of hematopoietic reconstitution, mice were challenged with zymosan 1mg i.p. 12-hours before harvesting bone marrow and peritoneal infiltrates. Library composition and representation in vivo was characterized in Figure S1. Screen analysis is discussed separately in the Statistical Analysis section. For subsequent validation experiments, gRNAs targeting non-targeting controls or specific genes were obtained from the Brie library and cloned into lentiviral vectors (Addgene #169941, 169940).

Confocal imaging.

Neutrophils and monocytes obtained via peritoneal lavage or human CD34+ derived macrophages were spun using Thermo Scientific Cytospin 4 Centrifuge at 300rpm for 5 min. Cells were stained with BODIPY 493/503 (Thermo D3922) or LipidTOX DeepRed (H34477) and DAPI according to manufacturer’s instructions. Images were acquired on a Nikon Ti inverted microscope.

LC-MS.

a. C8-pos:

Reversed-phase C8 chromatography/positive ion mode MS detection to measure lipids (Broad Institute). Analyses of polar and non-polar plasma lipids were conducted using an LC-MS system comprised of a Shimadzu Nexera X2 U-HPLC (Shimadzu Corp.) coupled to an Exactive Plus orbitrap mass spectrometer (Thermo Fisher Scientific). Plasma samples (10 μL) were extracted for lipid analyses using 190 μL of isopropanol containing 1,2-didodecanoyl-sn-glycero-3-phosphocholine (Avanti Polar Lipids) as an internal standard. After centrifugation, supernatants were injected directly onto a 100 × 2.1 mm, 1.7 μm ACQUITY BEH C8 column (Waters). The column was eluted isocratically with 80% mobile phase A (95:5:0.1 vol/vol/vol 10mM ammonium acetate/methanol/formic acid) for 1 minute followed by a linear gradient to 80% mobile-phase B (99.9:0.1 vol/vol methanol/formic acid) over 2 minutes, a linear gradient to 100% mobile phase B over 7 minutes, then 3 minutes at 100% mobile-phase B. MS analyses were carried out using electrospray ionization in the positive ion mode using full scan analysis over 200–1100 m/z at 70,000 resolution and 3 Hz data acquisition rate. Other MS settings were: sheath gas 50, in source CID 5 eV, sweep gas 5, spray voltage 3 kV, capillary temperature 300°C, S-lens RF 60, heater temperature 300°C, microscans 1, automatic gain control target 1e6, and maximum ion time 100 ms. Raw data were processed using TraceFinder software (Thermo Fisher Scientific) for targeted peak integration and manual review of a subset of identified lipids and using Progenesis QI (Nonlinear Dynamics) for peak detection and integration of both lipids of known identify and unknowns. Lipid identities were determined based on comparison to reference plasma extracts and are denoted by total number of carbons in the lipid acyl chain(s) and total number of double bonds in the lipid acyl chain(s).

b. HILIC-pos:

Hydrophilic interaction liquid chromatography/positive ion mode MS detection to measure polar metabolites (Broad Institute). HILIC analyses of water soluble metabolites in the positive ionization mode were conducted using an LC-MS system comprised of a Shimadzu Nexera X2 U-HPLC (Shimadzu Corp.) coupled to a Q Exactive Plus hybrid quadrupole orbitrap mass spectrometer (Thermo Fisher Scientific). Plasma samples (10 μL) were prepared via protein precipitation with the addition of nine volumes of 74.9:24.9:0.2 v/v/v acetonitrile/methanol/formic acid containing stable isotope-labeled internal standards (valine-d8, Sigma-Aldrich; St. Louis, MO; and phenylalanine-d8, Cambridge Isotope Laboratories; Andover, MA). The samples were centrifuged (10 min, 9,000 × g, 4°C), and the supernatants were injected directly onto a 150 × 2 mm, 3 μm Atlantis HILIC column (Waters). The column was eluted isocratically at a flow rate of 250 μL/min with 5% mobile phase A (10 mM ammonium formate and 0.1% formic acid in water) for 0.5 minute followed by a linear gradient to 40% mobile phase B (acetonitrile with 0.1% formic acid) over 10 minutes. MS analyses were carried out using electrospray ionization in the positive ion mode using full scan analysis over 70–800 m/z at 70,000 resolution and 3 Hz data acquisition rate. Other MS settings were: sheath gas 40, sweep gas 2, spray voltage 3.5 kV, capillary temperature 350°C, S-lens RF 40, heater temperature 300°C, microscans 1, automatic gain control target 1e6, and maximum ion time 250 ms. Raw data were processed using TraceFinder software (Thermo Fisher Scientific) for targeted peak integration and manual review of a subset of identified metabolites and using Progenesis QI (Nonlinear Dynamics) for peak detection and integration of both metabolites of known identify and unknowns. Metabolite identities were confirmed using authentic reference standards.

c. C18-neg:

Reversed-phase C18 chromatography/negative ion mode MS detection to measure metabolites of intermediate polarity (Broad Institute). Analyses of metabolites of intermediate polarity, including free fatty acids and bile acids, were conducted using an LC-MS system comprised of a Shimadzu Nexera X2 U-HPLC (Shimadzu Corp.) coupled to a Q-Exactive orbitrap mass spectrometer (Thermo Fisher Scientific). Plasma samples (30 μL) were extracted for analyses using 90 μL of methanol containing 50ng/mL 15-methyl PGE1, 15-methyl PGA2, 15-methyl PGE2 (Cayman Chemical Co.) as internal standards. The samples were centrifuged (10 min, 15000 × g, 4°C). After centrifugation, supernatants (2 μL) were injected directly onto a 150 × 2.1 mm, 1.8 μm ACQUITY HSS T3 C18 column (Waters). The column was eluted isocratically with 80% mobile phase A (0.01% formic acid in water) for 3 minutes followed by a linear gradient to 100% mobile phase B (0.01% acetic acid in acetonitrile) over 12 minutes. MS analyses were carried out using electrospray ionization in the positive ion mode using full scan analysis over 70–850 m/z at 70,000 resolution and 3 Hz data acquisition rate. Other MS settings were: sheath gas 45, in source CID 5 eV, sweep gas 10, spray voltage −3.5 kV, capillary temperature 320°C, S-lens RF 60, probe heater temperature 300°C, microscans 1, automatic gain control target 1e6, and maximum ion time 250 ms. Raw data were processed using TraceFinder software (Thermo Fisher Scientific) for targeted peak integration and manual review of a subset of identified metabolites and using Progenesis QI (Nonlinear Dynamics) for peak detection and integration of both metabolites of known identify and unknowns.

d. O-GlcNAc IP-MS and Proteomics.

Cell lysates were processed according to manufacturer’s instructions using the PTMScan^® O-GlcNAc [GlcNAc-S/T] Motif Kit (CST #95220). Samples were processed at Creative Proteomics (Shirly, NY) using the Ultimate 3000 nano UHPLC system (Thermo Scientific, Waltham, MA) using a trapping column (PepMap C18, 100Å, 100 μm × 2 cm, 5 μm) and an analytical column (PepMap C18, 100Å, 75 μm × 50 cm, 2 μm), and 1ug of sample. Mobile phases included A: 0.1% formic acid in water; B: 0.1% formic acid in acetonitrile and total flow rate: 250 nL/min. LC linear gradient: from 2 to 8% buffer B in 5 min, from 8% to 20% buffer B in 60 min, from 20% to 40% buffer B in 33 min, then from 40% to 90% buffer B in 4 min. The full scan was performed between 300–1,650 m/z at the resolution 60,000 at 200 m/z, the automatic gain control target for the full scan was set to 3e6. The MS/MS scan was operated in Top 20 mode using the following settings: resolution 15,000 at 200 m/z; automatic gain control target 1e5; maximum injection time 19ms; normalized collision energy at 28%; isolation window of 1.4 Th; charge sate exclusion: unassigned, 1, > 6; dynamic exclusion 30 s. Raw MS files were analyzed and searched against Mus musculus protein database based on the species of the samples using Maxquant (1.6.2.6). The O-GlcNAc IP-MS results are deposited in Supplementary Table 1 and the ProteomeXchange Consortium via the PRIDE repository (PXD072791).

e. Isotope Labeling with BMDMs.

150 μL of 80% methanol was added to each pellet. All the samples were homogenized on a MM 400 mixer mill with the aid of two metal beads at 30 Hz for 3 min, followed by centrifugation at 21,000 g for 10 min. To measure 13C-isotopomers of the TCA cycle related carboxylic acids and glucose, 20 μL of the clear supernatant of each sample was mixed with 20 μL of 3-nitrophenylhydrazine solution and 20 μL of an EDC solution and 20 μL of 5% pyridine. The mixture was incubated at 40 °C for 30 min. 10 μL aliquots of the resultant solutions were injected to run UPLC-MRM/MS on an Agilent 1290 UHPLC system coupled to an Agilent 6495C QQQ mass spectrometer operated with negative-ion detection. A C18 (2.1*150 mm, 1.8 μm) column was used for chromatographic separation, with the use of 0.01% formic acid in water and in acetonitrile as the mobile phase for binary solvent gradient elution under an optimized condition. To measure 13C-isotopomers of other metabolites, 20 μL the clear supernatant of each sample was mixed with 80 μL of water. 10 μL aliquots of the resultant sample solutions were injected to run LC-MRM/MS on a Waters Acquity UPLC system coupled to a Sciex QTRAP 6500 Plus mass spectrometer with negative-ion detection. A C18 (2.1*100 mm. 1.9 μm) column was used for chromatographic separation, with the use of a tributylamine buffer solution and acetonitrile as the mobile phase for binary-solvent gradient elution under an optimized condition. These results are deposited in Supplementary Table 1.

e. Untargeted LC/MS(/MS) lipidomics analysis with human macrophages.

Lipids were extracted from cells using chloroform/dichloromethane/KCl in a 100:100:0.88% ratio in glass vials. Splash lipid standards were incorporated from Avanti Research. Lipid extracts were concentrated 10x using a centrifugal vacuum concentrator and reconstituted in 1:1 acetonitrile:IPA. The mass spectrometer (QExactive HF-X) was equipped with HESI II probe and coupled to a Vanquish binary UPLC system (Thermo Fisher Scientific, San Jose, CA). For chromatographic separation prior to mass analysis, 5 μL of the lipid extract was injected into a Poroshell EC-C18 column (Agilent). Mobile phase A is 5 mM ammonium formate and 0.1% formic acid in 50% water and 50% acetonitrile, and mobile phase B is 5 mM ammonium formate and 0.1% formic acid in 88% isopropanol, 10% acetonitrile, and 2% water, the column oven is held at 45°C and autosampler at 15°C. The chromatographic gradient from 0% B to 100% B is carried out at a flow rate of 0.26 ml/min over the course of 28 minutes. The mass spectrometer was operated in full-scan positive mode, with the spray voltage set to 3.5 kV, the capillary temperature to 320 °C, and the HESI probe to 300 °C. The sheath gas flow is set to 30 units (50 for free fatty acids), the auxiliary gas flow is set to 8 units (10 for free fatty acids), and the sweep gas flow is set to 1 unit. Mass acquisition is performed in a range of m/z = 250–1,200 (100–1,000 for free fatty acids), with the resolution set to 120,0000. Top-15 data dependent acquisition was carried out for a pooled study sample to create a library of untargeted MS2 spectra. Raw data were converted to .mzML for processing. Lipids were annotated in pooled study sample using “Spectral library search” to match MS2 spectra to Pacific Northwest National Laboratory (PNNL) lipid library as well as LipidBlast in silico library in mzmine 3.^68,76 Some lipid species were identified de novo based on expected MS2 fragments corresponding to theoretical head and/or tail group m/z values. Fatty acids were identified using authentic standards when available; in the absence of standards, retention time was predicted by related species’ retention times, and m/z match. Retention times were manually defined for 11 lipids in the internal standard mix, and each lipid was normalized to the internal standard carrying the same headgroup when available; otherwise, the lipid was normalized to the mean of all internal standards. Lipids were quantified using emzed (Kiefer et al Bioinformatics 2013). Raw peak areas are produced by integrating an RT window determined by m/z, fragmentation patterns, authentic standards, and isotopologue patterns.

Glycosylation Assay.

The OGT Glycosylation assay was performed using UDP-Glo^™ Glycosyltransferase Assay (Promega) according to manufacturer’s instructions. The assay utilized UDP-GlcNAc (V7071, Promega), OGT substrate (Anaspec Cat.# 63726), OGT protein (ABIN1313360; Antibodies-Online), and TET2 protein (Active Motif, 31418).

Glucose Uptake Assay.

WT and Tet2 KO BMDMs were obtained between day 7–10 of differentiation and glucose uptake was assessed using Glucose Uptake-Glo Assay (Promega J1342) according to manufacturer’s instructions.

ChIP-seq and analysis.

Cells were processed as described previously⁷⁷. Briefly, cells were double cross-linked with disuccinimidyl glutarate (Thermo) and formaldehyde for 30 minutes at room temperature and quenched with 2.5M glycine. Cross-linked cells were treated with LB1 buffer (50mM HEPES-KOH, pH7.5, 140mM NaCl, 1mM EDTA, 10% Glycerol, 0.5% NP-40, 0.25% Triton X-100 in distilled water) × 10 minutes at 4C, LB2 buffer (10mM Tris-HCl, pH 8.0, 200mM NaCl, 1mM EDTA, 0.5mM EGTA in distilled water) × 10 minutes at 4C, then sonicated in LB3 buffer (10mM Tris-HCL, pH 8.0, 100mM NaCl, 1mM EDTA, 0.5mM EGTA, 0.1% Na-deoxycholate, 0.5% Na-Lauroylsarcosine, 1% Triton X-100, in distilled water) using Covaris, E220. ChIP antibodies include H3K4me3 (abcam ab8580), H3K27ac (abcam ab4729), OGT (Cell Signaling 24083), O-GlcNAc MultiMab (Cell Signaling 82332), H3K4me1 (abcam 8895), H3K4me3 (abcam 8580), H3K27me3 (abcam 6002), ACSS2 (abcam ab133664). For anti-streptavidin antibody incubations, overnight incubation was omitted. Following antibody incubation overnight, immunoprecipitation (IP) was performed using Protein-A or Protein-G Dynabeads (Thermo) or Strepavidin magnetic beads (Thermo 88817) for 1 hour. Beads were collected via a magnet and washed six times with RIPA buffer (50mM Hepes-KOH, pH 7.5, 500mM LiCl, 1mM EDTA, 0.7% Na-deoxycholate, 1% NP-40 in distilled water), twice with buffer 500 (0.5g deoxycholic acid, 1mM EDTA, 5mM Tris-HCl, pH 8.1, 1% Triton X-100, and 0.02% NaN3 in distilled water), and twice with LiCl buffer (2.5 g deoxycholic acid, 1 mM EDTA, 250 mM LiCl, 0.5% NP-40, 10 mM Tris-HCl pH 8.1, 0.02% NaN3 in distilled water). Beads were washed briefly with TE buffer and DNA was eluted in ChIP elution buffer (50mM Tris-HCl, pH 8.0, 10mM EDTA, 1% SDS in distilled water) and de-crosslinked at 65C for 2 hours. DNA was purified by column purification (DNA Clean and Concentration, Zymo Research). Library preparation was done according to manufacturer’s instructions (ThruPlex DNA-Seq Kit, Takara or NEBNext^® Ultra, NEB). Fragment length and concentrations were assessed using TapeStation (Agilent) and Qubit (ThermoFisher), and libraries were sequenced on an Illumina NextSeq 550 sequencer using paired end reads of 42bp. Trimmed reads were aligned to the mm10 or hu19 genome via bowtie2.3.5⁶⁴ using --very-sensitive -X700 --no-mixed --no-discordant. Duplicates were marked using MarkDuplicates in Picard/2.19.0 (Broad Institute) or MACS2⁶⁵. Peaks were called using Macs2/2.1.1 using -f BAMPE -B --nomodel --broad --keep-dup 1 --extsize 200 -p 1e-2. For transcription factors, --broad parameter was omitted. ENCODE blacklist regions were excluded from analysis⁷⁸. BigWig files were generated via bedGraphToBigWig⁷⁹. Peaks were analyzed further using computeMatrix or plotHeatmap functions from deepTools 3.4.3⁶⁶. Genomic window for computeMatrix is 5kb around the peak. Sequencing data is deposited under GSE298064 and GSE309775.

RNA-seq and analysis.

RNA was isolated using the RNeasy Plus Mini Kit (Qiagen) at 4C then mRNA library preparation was performed poly A enrichment using the NEBNext^® Ultra^™ RNA Library Prep Kit for Illumina or obtained using Novogene (Sacramento, CA). For internal libraries, NextSeq 550 sequencing (Illumina) was used to generate single-end reads of 75 bp. Sequencing data were aligned with STAR v2.7.2b to mm10 genome database,⁶⁰ and counts were obtained using RSEM 1.3.1.⁶¹ Differential expression was determined using DESeq2⁶² using R/3.6.1. Single cell RNA sequencing was performed using 5’ GEX kit from 10x Genomics according to manufacturer’s instructions. CD45+CD11b+ cells were sorted from BL6 recipients transplanted with WT or Tet2 KO Cas9+ HSPCs transduced with control, Ogt, or Acly gRNAs labeled by BFP or RFP. Cells were collected from bone marrow 8 hrs after 1mg zymosan i.p. injection. Monocytes were analyzed only significant egress of neutrophils from the bone marrow after zymosan injection. Libraries were sequenced on the Illumina Novaseq SP platform (Broad Institute). Reads were aligned to the mm10 genome using CellRanger 9.0.1 (10x Genomics), and further analyzed using Seurat v3⁶³. Cells were filtered for percent mitochondrial count < 10%. Sequencing data is deposited under GSE298064 and GSE309775.

Viral transduction.

Lentiviral particles were generated via transfection of HEK293T cells using FuGENE 6 (Promega). Viral constructs were cotransfected with pMD2.G (plasmid 12259; Addgene) and psPAX2 (plasmid 12260; Addgene), and lentiviral particles were harvested on day 2–3 of transfection. For adeno-associated virus (AAV) production, mNeonGreen or mScarlet fluorescent cassettes and homology arms were cloned into the pAAV-MCS vector (VPK-411, CellBio Labs) using Gibson Assembly (NEB). For virus production, HEK293T cells were transfected with AAV vector and pDGM6 (AddGene). Virus was collected 48-hour posttransfection using AAVpro Purification Kit Maxi (Takara) and titered by qPCR using the AAVpro Titration Kit (Takara).

Immunoblotting.

Cells were washed with PBS and lysed in Pierce IP Lysis Buffer (ThermoFisher, 87787), RIPA Lysis Bufffer (Thermo 89900) or 8M Urea supplemented with Halt Cocktail protease inhibitor (ThermoFisher, 87786). After lysis, cells were centrifuged for 15 minutes at maximum speed to clear the lysate. Protein concentration was quantified via the Bradford assay and equal amounts of lysates were run on SDS-PAGE 3–8% Tris-Acetate or 4–12% Bis-Tris Protein Gels (NuPAGE, ThermoFisher), and then transferred to a PVDF membrane using the semi-dry Trans-Blot Turbo Transfer System (Biorad) or wet transfer using the Tris-Glycine SDS transfer buffer with 10% methanol. After blocking for at least 30 minutes using the Odyssey Blocking Buffer/PBS (LI-COR Biosciences), the membrane was incubated with primary antibodies at 4 °C overnight. Membranes were washed thrice using Tris-buffered saline with Tween 20 (TBS-T) and incubated for 1 hour at room temperature with secondary IRDye-conjugated antibodies (LI-COR Biosciences), then washed again in TBS-T × 3 times prior to imaging (Odyssey Imaging System, LI-COR Biosciences). Antibodies include OGT (CST 24083S), Myc-tag (CST 2276S), Bactin (MA5–15739, Thermo), O-GlcNAc (CST 82332, Thermo MA1–072, Novus Biologicals NB300–524), Gapdh (CST 2118), alpha-tubulin (Sigma T9026), Hdac1 (CST 5356S), IL1B (GeneTex GTX130021 for human cells and GTX74034 for mouse cells), Vinculin (scbt sc-73614), and Tet1 (Thermo MA5–16312). Antibodies were used at 1:100 unless otherwise specified.

Immunofluorescence.

For cytospins, cells were spun down onto a glass slide using 600rcf for 5 min, fixed with 4% paraformaldehyde for 15 min, and washed with PBS for 5 min for 3 washes. Cells were then permeabilized using 0.01% Tween and blocked with 5% bovine serum albumin (BSA) for 1 hour. Permeabilized cells were stained with Ogt (1:50, CST 24083S) and Hoechst 33342 (1:2000, Thermo 62249) or OGT (1:50, CST 24083S), Hoechst 33342 (1:2000, Thermo 62249) and anti-HA antibody (1:50, Roche Roche 11867423001). overnight at 4C in 0.01% Tween with 5% BSA. After staining, cells were washed thrice with PBS, the following secondary antibodies were added at 1:2000 for 1 hour at room temperature: anti-Rat AF647 (Thermo A48272), anti-Rabbit (Thermo A27040), or anti-Rabbit AF488 (Thermo A11008). After washing with PBS three times, the slides were imaged on Leica inverted microscope. For paraffin-embedded sections of formalin-fixed samples, paraffin was dissolved with xylene for 10 min and then rehydrated with 100% ethanol (w/v in H2O), followed by 95% ethanol, 90% ethanol, 85% ethanol, 50% ethanol, then in H2O for 5 minutes each. Antigen retrieval was performed using citrate-based buffer in 0.1% Tween, pH6 at boiling temperatures for 5 min. Samples were then stained with 660-DEVD-FMK (abcam ab270785) according to manufacturer’s instructions along with Hoechst 33342 (Thermo 62249).

TR-FRET.

2xMyc-tagged OGT protein or GFP-3xFLAG peptides were purified from transfected HEK293T cells using 50 mM HEPES pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 5% (v/v) glycerol, 1 mM TCEP supplemented with Halt protease inhibitors (Thermo). After clearing the lysate × 15 minutes at 21000g, lysates were incubated with Myc-Trap Nanobody Magnetic Agarose (Chromotek) or anti-FLAG magnetic beads (M8823, Sigma) × 1 hour at 4C. Protein of interest-bound magnetic beads were washed with 50 mM HEPES pH 7.4, 200 mM NaCl, 1 mM EDTA, 1% NP-40, 5% (v/v) glycerol, 1 mM TCEP three times. OGT biotinylation was performed in the presence of BirA enzyme, 50 mM HEPES pH 7.4, 200 mM NaCl, 10 mM MgCl2, 200 uM biotin, 20 mM ATP, 1 mM TCEP for 1 hour at room temperature and subsequently washed 50 mM HEPES pH 7.4, 300 mM NaCl, 1 mM EDTA, 1% NP-40, 5% (v/v) glycerol, 1 mM TCEP three times. Proteins were eluted using 3xMyc or 3x FLAG peptide at 65C × 5 min. Proteins were fractionated via ion exchange spin columns (Pierce) and concentrated using 7K or 40K or MWCO Zeba Spin Desalting Columns. Biotinylated OGT was bound to Strepavidin-Terbium (Tb) at 1nM. Protein concentrations were obtained using the standard Bradford assay. Assays were performed in 384-well microplates (Corning, 4514) with a 15 μL final assay volume in 25 mM HEPES, 100 mM NaCl, 0.01% Triton X-100, 5% (v/v) glycerol, 1 mM TCEP with increasing concentrations of the peptide using the Tecan aD300e Digital Dispenser. After excitation at 337 nm, emission at 490 nm (OGT-Tb) and 520 nm (GFP) was read using CLARIOstar Plus (BMG Labtech). TR-FRET signal of each data point was calculated using the 520/490 nm ratio.

Flow cytometry.

Peripheral blood was collected retro-orbitally into EDTA tubes. Bone marrow cells were obtained from the femur. After red cell lysis, cells were analyzed on the Sony MA900 cell sorter, BD FACSymphony or BD FACSCanto. The following antibodies and stains were used: DAPI (Sigma), Zombie Aqua Viability Dye (Biolegend, 423101) or LIVE/DEAD^™ Fixable Near IR (780) Viability Kit (Thermo, L34994), PerCP/Cyanine5.5 anti-mouse Ly-6G (Biolegend, 1A8), PE/Cyanine7 anti-mouse Ly-6C (Biolegend HK1.4), APC anti-mouse CD45.1 (Biolegend, A20), APC/Cyanine7 anti-mouse CD45.2 (Biolegend, 104), Brilliant Violet 711^™ anti-mouse/human CD11b (Biolegend M1/70), PE anti-mouse CD45.1 (Biolegend, A20), PE/Cyanine7 anti-mouse Ly-6G (Biolegend A18), APC/Cyanine7 anti-mouse Ly-6C (Biolegend HK1.4), BUV395 Mouse Anti-Mouse CD45.2 (BD, 104), Brilliant Violet 605^™ anti-mouse CD45.2 (Biolegend, 104), PE/Cyanine7 anti-mouse CD3 (Biolegend, 17A2), PE/Cyanine7 anti-mouse CD19 (Biolegend 6D5), Alexa Fluor^™ 700 CD101 Monoclonal Antibody (eBioscience, Moushi101), FITC anti-mouse CD45.1 (Biolegend, A20), Pacific Blue^™ anti-mouse CD45.1 (Biolegend, A20). For intracellular antibody staining, the following antibodies were used after fixation/permeabilization (BD, 555028): Alexa Fluor647 O-GlcNAc Antibody (Novus Biologicals RL2, NB300–524AF647). For neutral lipid staining, BODIPY 493/503 (Thermo D3922) or LipidTOX Deep Red (Thermo H34477) were used according to manufacturer’s instructions. For sorting LSK HSPCs, the following antibodies and dyes were used: DAPI (Sigma), V450 Mouse Lineage Cocktail (BD 561301), PE anti-mouse CD45.2 (Biolegend, 104), APC anti-mouse CD117 (Biolegend 2B8), PE/Cyanine7 anti-mouse Ly-6A/E (Biolegend, E13–161.7), anti-mouse CD117 magnetic beads (Miltenyi Biotec, 130-122-937) and cells were magnetically separated using Miltenyi LS Columns prior to sorting on the MA900 cell sorter.

Histology.

Mouse paws were fixed in 10% formalin. Paraffin-embedded tissue blocks were sectioned and stained using haematoxylin and eosin. Mouse aortas were stained with Oil Red stain. Atherosclerotic area was quantified using ImageJ using image thresholding.

Quantification and Statistical Analysis

Analyses were performed in R 3.6.1 statistical environment or Prism 7.0. A two-sided unpaired Welch’s t test was used to assess significant differences between the two samples. Mann-Whitney test was used if samples did not pass the Shapiro-Wilk normality test. For multiple comparisons against one sample, one-way ANOVA was used and p-values were adjusted for multiple comparisons using the two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli (BKY) unless otherwise noted. For grouped comparisons, significance was assessed by two-way ANOVA adjusted for multiple comparisons with the two-stage linear step-up procedure of BY. CRISPR sequencing results were analyzed using MAGeCK/MAGeCK-MLE using UMI counts as read counts⁵⁹. Zymosan peritonitis screen data were analyzed to estimate genotype-specific effect sizes (ß) for peritoneal recruitment relative to the bone marrow within the genotypes, followed by a between genotype comparison (Δß = ßKO – ßWT) to identify Tet2 KO-specific dependencies. The error bars represent standard deviation (SD) and biological replicates were used unless specified. P-values are indicated with * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Supplementary Material

2

S1 – Results of the in vivo genomewide CRISPR/Cas9 screen in WT and Tet2 KO myeloid cells, related to Figure 1

3

S2 – Mass spectrometry results. C13 glucose uptake and incorporation into central carbon metabolites. Results of the O-GlcNAc Immunoprecipitation-Mass Spectrometry, related to Figure 2.

4

S3 – Results of the lipidomics study, related to Figure 3.

5

S4 – Results of bulk RNA-seq in WT and Tet2 KO BMDMs treated with Z-PUGNAc, related to Figure 3.

6

S5 – Results of single cell RNAseq in Cas9+ WT and Tet2 KO monocytes, transduced with control or guides against Ogt or Acly, related to Figure 3.

7

S6- Results of bulk RNA-seq in WT and Tet2 KO BMDMs, treated with glucose. Related to Figure 5.

8

S7 – Results of targeted lipid droplet CRISPR/Cas9 screen in Cas9+ Tet2 KO neutrophils and monocytes, related to Figure 6.

Significance:

Clonal hematopoiesis of indeterminate potential (CHIP) arises from expansion of blood stem cells with somatic mutations, most commonly in epigenetic regulators such as TET2. CHIP occurs with increasing frequency with aging and it is strongly associated with several chronic inflammatory diseases, including atherosclerotic cardiovascular disease. While TET2-mutant myeloid cells are known to exhibit heightened inflammatory behavior, how this epigenetic lesion connects to immune activation remained unclear. We uncover a nutrient-responsive mechanism in which TET2 restricts O-linked N-acetylglucosamine transferase (OGT), the enzyme that adds a sugar modification (O-GlcNAc) to proteins. In the absence of TET2, excess O-GlcNAcylation reshapes gene regulation towards lipid building programs, driving lipid droplet accumulation and an exaggerated inflammatory state. We further identify ATP citrate lyase (ACLY) as a key metabolic dependency and show that disrupting OGT or ACLY suppresses the inflammatory lipid phenotype. By connecting a common CHIP mutation to a specific metabolic circuit, this work provides a mechanistic framework for how mutant blood cells can amplify inflammation through dysregulated glycosylation and aberrant lipid accumulation.

Highlights.

• In vivo genome-wide CRISPR screen uncovers regulators of Tet2 mutant inflammation

• Dysregulated Ogt drives inflammatory lipid accumulation upon Tet2 loss

• Acly is a critical node for lipid accumulation and inflammation in Tet2 KO