Macrophage metabolic exhaustion and PANoptotic cell death drive chronic tissue inflammation in rheumatoid arthritis

SUMMARY

In the autoimmune disease rheumatoid arthritis, the inflammatory response evolves from protective to pathogenic, causing tissue destruction. Rheumatoid synovitis persists despite the presence of pro-repair SELENOP^hiMerTK⁺CD206⁺ macrophages, suggesting that these cells acquire pro-arthritogenic functions. Patient-derived synovial SELENOP^hiMerTK⁺CD206⁺ macrophages produced high concentrations of the complement component C1q and concurrently expressed its receptor, C1QBP. Stimulation of these macrophages with C1q induced metabolic exhaustion, characterized by diminished ATP production, cleavage of mitochondrial nicotinamide adenine dinucleotide (NAD), and accumulation of cyclic ADP ribose (cADPR). This metabolic crisis was driven by the mitochondrial enzyme Sterile alpha and Toll/interleukin-1 receptor (TIR) motif containing 1 (SARM1), which catalyzed the conversion of NAD into cADPR, triggering PANoptotic and pro-inflammatory macrophage death. In vivo experiments demonstrated that C1q treatment exacer-bated synovitis, whereas SARM1 inhibition conferred therapeutic benefit. These findings identify the NAD⁺ hydrolase SARM1 as a marker of metabolically stressed macrophages and an executor of pro-inflammatory macrophage death during autoimmune tissue inflammation.

In brief

Rheumatoid joints contain macrophages with reparative potential, yet inflammation persists. Huang et al. demonstrate that these homeostatic macrophages lose their protective role by undergoing PANoptotic death. Mechanistically, C1q activates the mitochondrial enzyme SARM1, driving NAD⁺ cleavage into cADPR, which precipitates metabolic collapse and PANoptosis. The resulting macrophage death amplifies autoimmune tissue inflammation through the release of pro-inflammatory mediators.

Graphical abstract

INTRODUCTION

Host protection against infection and malignancy is fundamentally dependent on tissue inflammation. However, in autoimmune disease, inflammation becomes chronic and leads to irreversible damage. Rheumatoid arthritis (RA), a prototypic autoimmune disease, is characterized by long-lasting and destructive inflammation in both small and large joints. This condition often manifests in individuals with a pre-aged immune system, with major genetic risk factors associated with the human leukocyte antigen (HLA) class II region.^1,2 Pathogenic RA T cells exhibit inadequate repair of nuclear and mitochondrial DNA (mtDNA), resulting in excessive telomeric loss and bioenergetic failure.^3,4 Additionally, the cytosolic leakage of mtDNA can provoke lytic cell death.⁵ While the functions of pathogenic T cells are linked to a hypometabolic state, macrophages exhibit a hypermetabolic profile, similar to macrophages found in cardiovascular disease.^6,7 Synovial macrophages maintain their hypermetabolic state by shifting to glutaminolysis,⁸ allowing them to survive and fulfill their antigen-presenting roles in a glucose-scarce environment.

Abundant in inflammatory RA synovitis, macrophages secrete inflammatory mediators, clear debris, and present antigens, thereby creating high bioenergetic demand. These macrophages require access to local nutrients to keep their metabolic machinery functioning effectively. Inflammatory lesions contain a heterogeneous population of macrophages, with the rheumatoid pannus in inflamed joints serving as a prime example.^9,10 In addition to driving tissue inflammation, macrophages are crucial for maintaining tissue homeostasis, facilitating repair, and ensuring tissue turnover.^11,12 Persistent inflammation may arise from ongoing stimulatory signals or failing protective mechanisms. A key paradox in RA pathogenesis is enduring destructive synovitis despite the presence of macrophage subsets, such as those expressing Mer tyrosine kinase (MerTK), which are thought to be pro-resolving. This indicates that their protective functions may be compromised in the inflamed joint’s hostile microenvironment.

To fulfill their protective roles, tissue-resident macrophages produce antibacterial mediators, including the complement component C1q. This 460-kDa hexamer acts as a soluble pattern-recognition receptor, forming complexes with binding partners in solution and on cell membranes. In addition to its role in the classical complement pathway, C1q enhances phagocytosis to clear complement-opsonized targets and apoptotic cells.¹³ Unlike other complement factors synthesized primarily in the liver, C1q is produced by macrophages, dendritic cells, and fibroblasts in peripheral tissues.¹⁴ C1q also exhibits anti-inflammatory functions, clearing immune complexes and cellular debris while inducing anti-inflammatory mediators.¹⁵ It binds to various receptors, including CR1, two immunoglobulin (Ig) superfamily members (RAGE and CD305), two scavenger receptors (SREC1 and SR-F3), an integrin (α2β1), and two receptors devoid of cytoplasmic tails (calreticulin and C1q-binding protein [C1QBP]).¹⁶ Besides its intracellular presence (mitochondria, nucleus), C1QBP is also found on the cell surface^17,18 and targets mitochondria via an N-terminal signal sequence, playing roles in mitochondrial protein synthesis and oxidative phosphorylation.^19,20

Effective tissue inflammation requires massive recruitment and expansion of immune cells, which eventually necessitates purging. Immune cell elimination through apoptosis, a non-lytic and non-inflammatory mode of cell death, helps maintain immune tolerance. By contrast, lytic and inflammatory death modes (e.g., PANoptosis, pyroptosis, and necroptosis) cause plasma membrane disintegration and release of cellular content into the tissue environment.^21–25 Cellular detection of danger signals triggers the assembly of PANoptosome complexes (e.g., ZBP1-, AIM2-, RIPK1-, and NLRP12-PANoptosomes), activating caspases and receptor-interacting protein kinases (RIPKs)^26–29 and cleaving pore-forming gasdermin D (GSDMD).³⁰

Here, we have identified the mitochondrial matrix enzyme Sterile alpha and Toll/interleukin-1 receptor (TIR) motif containing 1 (SARM1) as an inducer of immunogenic cell death in synovial macrophages. SARM1, comprising an armadillo repeat motif (ARM), two sterile alpha motifs (SAMs), and a TIR domain, functions in the Toll-like receptor (TLR) signaling pathway.³¹ Localized to the mitochondria, the TIR domain acts as a potent NAD⁺ hydrolase,^32,33 regulating mitochondrial nicotinamide adenine dinucleotide (NAD) concentrations and overall fitness. Additionally, SARM1 catalyzes reactions that produce cyclic ADP ribose (cADPR) and nicotinic acid adenine dinucleotide phosphate (NAADP) from nicotinamide adenine dinucleotide phosphate (NADP) and nicotinic acid. Both cADPR and NAADP serve as second messengers, mobilizing Ca²⁺ from endoplasmic and endo-lysosomal stores.³⁴ We provide a molecular definition of macrophage metabolic exhaustion, proposing that low metabolic resilience underlies chronic tissue inflammation. In the rheumatoid joint, SARM1^hi macrophages activated the enzyme via an autocrine and paracrine mechanism involving C1q release. Mitochondrial translocation of C1q-C1QBP complexes led to metabolic crisis, marked by NAD consumption, electron transport chain (ETC) failure, and a surge of cADPR. cADPR triggered the formation of NLRC5- and NLRP12-containing PANoptosomes, activated MLKL and caspases 1/8, cleaved GSDMD, and caused plasma membrane lysis. Metabolically exhausted macrophages released cytoplasmic and nuclear content, created a robust inflammatory nidus, and activated T cells, macrophages, and stromal cells. In vivo, inhibiting SARM1 protected tissues from inflammation. These data classify SARM1 as a pro-destructive molecule that culls metabolically exhausted macrophages and identify cADPR-induced immunogenic cell death as a contributing disease mechanism in RA.

RESULTS

SELENOP^hi FOLR2^hi MerTK⁺ CD206⁺ macrophages release C1q, respond to C1q, and exhibit a pro-inflammatory signature

All subtypes of rheumatoid synovitis contain a heterogeneous population of macrophages.³⁵ Some macrophages drive inflammation through antigen presentation and cytokine production,⁸ while others suppress inflammation and restore tissue homeostasis.^36,37 To define molecular mechanisms regulating pro-resolving macrophages, we analyzed a single-cell RNA sequencing (scRNA-seq) dataset of 71,073 synovial tissue mononuclear cells from 17 untreated RA patients.³⁸ Unsupervised graph-based clustering identified macrophages, T cells, B cells, fibroblasts, dendritic cells, and five minor cell types (Figures S1A and S1B). Macrophages were further stratified into six subclusters based on the expression patterns of SELENOP, SPP1, FCN1, IFI27, MALAT1, and CXCL9 (Figures 1A and S1C). This clustering aligns well with other synovial cell atlases,^35,39 which have also identified MerTK⁺selenoprotein P (SELENOP)^hi subsets as major subpopulations in rheumatoid synovitis (Figure S1D). Only the small CXCL macrophage subset exhibited pro-inflammatory features (e.g., expression of IL6, IL1b, TNFa, IL12b, and IL18) (Figure 1B). Most macrophages were designated as M2-like, displaying gene expression patterns suggestive of homeostatic functions (Figure S1E). Classification of the six subclusters by known markers for tissue residency and pro-resolution (MerTK, CD206, LYVE1, and CD163) predicted SELENOP^hi macrophages as the major tissue-resident homeostatic population (Figure 1B).

Figure 1. Synovial tissue SELENOP^hiCD206⁺MerTK⁺ macrophages produce C1q and express C1q receptors.

Synovial tissue samples from 17 untreated patients with RA were analyzed by scRNA-seq.^8,38

(A) Uniform manifold approximation and projection (UMAP) visualization of tissue macrophage clusters. Relative percentages are indicated.

(B) Expression of gene transcripts relevant for tissue residency and pro-inflammatory functions in the macrophage subsets.

(C–F and K) Synovial tissues collected from RA patients with active synovitis were digested and analyzed by flow cytometry. Tissues were categorized as MerTK^hiCD206^hi macrophage rich or poor based on frequency analysis.

(C) Secretome analysis of synovial tissue macrophages. Volcano plot comparing proteins secreted by macrophage populations rich (>75%) versus poor (<25%) in MerTK^hiCD206^hi macrophages. Positive values indicate protein enrichment in MerTK⁺CD206⁺ macrophages.

(D) Single-cell suspensions from eight inflamed tissues were analyzed by flow cytometry. t-SNE visualization mapped C1q production to CD45⁺CD68⁺MerTK^hiCD206^hi macrophages. See also Figures S2A–S2C.

(E and F) Correlation analysis of C1q secretion with (E) the proportion of MerTK^hiCD206^hi macrophages and (F) the tissue inflammatory score in individual synovial tissues.

(G and H) Monocyte-derived macrophages (MDMs) generated from circulating CD14⁺ precursor cells of RA patients were polarized with an array of stimuli (LPS, IL-4, IL-10, and IL-10 plus LPS).

(G) Gene expression patterns from in vitro stimulated macrophages were integrated with gene panels of tissue macrophage subclusters in scRNA-seq by CCA.

(H) MDM from healthy individuals and RA patients were differentiated with M-CSF (M0) and then polarized with multiple stimuli for 24 h. Macrophage subsets are identified by the leading gene transcript as SPP1^hi, CXCL^hi, or SELENOP^hi. Collected supernatants were assessed for C1q concentration (n = 5–8).

(I) Ligand-receptor pairing between macrophage-derived complement and complement receptors was evaluated in the scRNA-seq dataset for the ten indicated synovial cell types by utilizing CellChat analysis.

(J) Heatmap showing expression of C1q receptors in RA synovial cells identified by scRNA-seq.

(K) Flow cytometric analysis of C1QBP in RA synovial cells. Dot plot and summary of C1QBP mean fluorescence intensities expressed by MerTK^hiCD206^hi and MerTK⁺CD206⁺ synovial macrophages (n = 12).

Data are shown as violin plots or as bar graphs with means. Comparisons by two-tailed paired Student’s t test (K) and one-way ANOVA (H). Correlation was analyzed by linear regression (E and F). Canonical correlation coefficients were calculated by CCA (G).

See also Figures S1 and S2.

Gene set enrichment analysis (GSEA) of SELENOP^hi macrophages yielded enrichment in complement activation pathways, antigen processing, and C-type lectin receptor and TLR signaling (Figure S1F). Downregulated gene sets indicated metabolic reprogramming, with low expression of mitochondrial pathways (Figures S1F and S1G). Notably, SELENOP^hi macrophages expressed a pro-inflammatory gene signature (Figure S1G).

To identify functionally relevant molecules in SELENOP^hi macrophages, we evaluated synovial tissues from patients with refractory RA, focusing on CD206⁺MerTK⁺CD68⁺ cells for phenotyping and secretome analysis. We assigned intensity of synovial inflammation by histologic examination (see STAR Methods). We classified synovial tissues as either poor or rich in CD206⁺MerTK⁺ macrophages based on whether <25% or >75% of their respective CD68⁺ macrophages expressed the tissue-resident, pro-resolving phenotypes of CD206 and MerTK (Figures S2A and S2B). Among 500 secreted proteins, 88 proteins were significantly enriched in SELENOP^hi macrophage supernatants (Figures 1C and S2C; Table S1), including cysteine proteinase inhibitors (Cystatin D, S, SN) and matrix proteins (collagen IV, endorepellin), implicating this macrophage subset in tissue repair. Azurocidin and C1q linked SELENOP^hi macrophages to host defense. We confirmed enrichment of C1QA, C1QB, and C1QC transcripts in rheumatoid synovium (Figure S2D). C1q concentrations in the serum of RA patients were higher than in matched controls (70 μg/mL versus 53 μg/mL, p = 0.0081), indicating that RA is a condition characterized by C1q overproduction (Figure S2E).

Employing multi-parametric flow cytometry, we mapped C1q production to CD45⁺ CD68⁺ MerTK⁺ CD206⁺ cells (Figures 1D, S2F, and S2G). We confirmed that synovial tissues with higher C1q production were enriched in CD206^hiMerTK^hiSELENOP^hi macrophages (Figure 1E). In 20 patient samples, high C1q release correlated with high inflammation scores (Figure 1F), suggesting that SELENOP^hi macrophages might contribute to inflammation.

To facilitate mechanistic studies of SELENOP^hi macrophages, we established priming conditions to generate these macrophages in vitro. CD14⁺ monocytes were differentiated with macrophage colony stimulating factor (M-CSF) and polarized using various stimuli. Canonical correlation analysis (CCA) aligned gene expression profiles of in vitro differentiated monocyte-derived macrophages (MDMs) with the defined scRNA-seq clusters (Figure 1G). Gene expression profiles revealed that lipopolysaccharide (LPS)-stimulated macrophages closely resembled tissue CXCL⁺ macrophages. SELENOP^hi macrophages were identified as interleukin (IL)-10 primed MDM (Figures 1G, S2H, and S2I). IL-10 priming followed by LPS stimulation best mirrored the transcriptional characteristics of SELENOP^hi macrophages in rheumatoid synovitis (Figure 1G). For further experiments, the IL-10 plus LPS combination was utilized to generate SELENOP^hi macrophages in vitro. Macrophage subtypes varied in C1q production capacity, but all patient-derived macrophages outperformed their healthy counterparts, with patient-derived SELENOP^hi macrophages excelling in C1q release (Figure 1H).

Given the abundance of C1q in synovitis, we investigated which cells respond to C1q stimulation in RA. We utilized CellChat to predict cell communication among synovial cell clusters. Ligand-receptor pairs for complement signaling showed macrophages interacting with macrophages, while T cells had limited involvement (Figures 1I and S2J). Expression profiling of nine C1q receptors revealed C1QBP as the most prominent receptor on macrophages and CD93 on endothelial cells (Figure 1J). Flow cytometric analysis identified C1QBP in CD68⁺MerTK^hi CD206^hi macrophages (Figure 1K).

In summary, SELENOP^hiMerTK^hiCD206^hi macrophages, categorized as pro-resolving and reparative, are prevalent in inflamed joints. Secretome analysis and immunophenotyping indicate they are the primary source of the pattern-recognition receptor C1q and express the C1q receptor C1QBP, facilitating autocrine and paracrine complement signaling.

C1q induces PANoptotic cell death of SELENOP^hi macrophages

We noticed that tissue-derived SELENOP^hiCD206^hiMerTK^hi macrophages had a short in vitro survival time, with most cells dying within 48 h.⁸ Live-cell imaging demonstrated cellular swelling, cell membrane disruption, and DNA release into the extracellular space (Figure 2A), indicating non-apoptotic cell death. We quantified the frequency of synovial tissue dead/dying SELENOP^hiCD206^hiMerTK^hi macrophages, relying on nucleic acid staining with SYTOX dyes impermeant to live cells. SYTOX⁺ cells were rare in non-macrophage populations but reached 50% in CD68⁺ cells (Figure 2B). SYTOX positivity ranged from 15% to 70% in freshly isolated CD206^hiMerTK^hi macrophages (Figure 2C), and tissue inflammation was most intense in samples with high proportions of dead/dying macrophages (Figure 2D).

Figure 2. C1q induces PANoptotic cell death of SELENOP^hi tissue macrophages.

Synovial tissue samples were collected from RA patients with active synovitis. Single-cell suspensions generated from digested tissues were analyzed by flow cytometry. CD14⁺ precursor cells were isolated from the peripheral blood of RA patients and age- and sex-matched healthy donors.

(A) Synovial macrophages were stained for CD68 (green), CD206 (white), SYTOX (red), and DAPI (blue). Representative images from four experiments.

(B) Flow cytometric analysis of synovial cells. Macrophages were marked with anti-CD68, and membrane leakage was detected with SYTOX-Deep Red (n = 15 synovial tissues).

(C) Frequencies of dead/dying MerTK⁺CD206⁺ and MerTK⁻CD206⁻ synovial macrophages as determined by SYTOX-Deep Red staining (n = 10).

(D) Frequencies of dead/dying synovial macrophages correlated with the tissue inflammation score in individual tissues (n = 15).

(E and F) IL-10/LPS-polarized MDM from healthy individuals and RA patients were stimulated with increasing doses of C1q. Cell survival was quantified by SYTOX live-cell imaging (E) and LDH release (F).

(G) RA macrophages were transfected with control, C1qA, or C1QBP siRNA. Cell death was quantified by SYTOX staining 96 h after transfection (n = 8).

(H) RA macrophages were transfected with control or C1QBP siRNA prior to rhC1q treatment. Cell death detected with SYTOX staining. Data from three experiments.

(I) RA macrophages were treated with C1q or vehicle and analyzed by immunoblotting. Pro- and activated caspase-1 (P45/P20), caspase-3 (P35/P17), caspase-7 (P35/P20), caspase-8 (P55/P18), GSDMD (P53/P30), gasdermin E (P53/P34), NLRC5, NLRP12, and MLKL (phosphorylated/total). β-actin served as a loading control.

(J) RA macrophages were treated with rhC1q or vehicle. The interaction between ASC and NLRC5/NLRP12 was examined by immunoprecipitation experiments. Whole-cell lysates were subjected to immunoprecipitation (IP) using anti-ASC, followed by immunoblotting of the associated NLRC5/NLRP12.

(K) RA macrophages were transfected with control or NLRC5/NLRP12 siRNA before C1q treatment. Cell survival was quantified by SYTOX live-cell imaging.

Data are shown as violin plots or as bar graphs with mean ± SEM. Comparisons by two-tailed unpaired Student’s t test (G), two-tailed paired Student’s t test (B and C), and one-way ANOVA (E–H and K).

See also Figure S3.

To examine the susceptibility of SELENOP^hi macrophages to lytic cell death in patients versus controls and C1q’s role during macrophage death, we stimulated in vitro differentiated SELENOP^hi macrophages with rhC1q. Both healthy and RA macrophages were protected from C1q-induced membrane rupture and DNA leakage for the first 12 h, followed by markedly higher death rates in RA macrophages (Figures 2E, S3A, and S3B). Monitoring of lactate dehydrogenase (LDH) release confirmed resistance to C1q-induced death in healthy but susceptibility in RA macrophages (Figure 2E). Extended pretreatment with LPS did not exaggerate cell death in RA macrophages (Figure S3C). The differential susceptibility of RA macrophages to C1q-induced death suggested a pre-existing vulnerability, primed by chronic stimulation in the metabolic environment of the synovium.

C1QA small interfering RNA (siRNA) silencing in RA macrophages reduced the rate of spontaneous cell death, confirming the need for intrinsic C1q release to trigger C1q-induced death and providing direct evidence for a cell-autonomous, autocrine suicide loop (Figures 2G and S3D). Also, C1q-induced cell death was dependent on the expression of C1QBP. C1QBP siRNA silencing effectively protected RA macrophages from dying (Figures 2G and 2H). Loss of the receptor molecule was more protective than loss of C1q, identifying C1QBP expression as the gatekeeper of C1q-induced cell death.

To arrive at a molecular definition of the underlying cell death mode, we probed the dependence of cell membrane lysis on caspase activation. Treatment of macrophages with the pan-caspase inhibitor Q-VD-Oph prevented SYTOX positivity (Figure S3E). By contrast, cell death rates were unaffected by diphenyleneiodonium (DPI), which blocks the generation of reactive oxygen species (ROS) (Figure S3F). Transcriptomic analysis of the rheumatoid synovium scRNA dataset for lytic cell death adaptors, sensors, and effectors yielded abundant expression of AIM2, NLRP12, and ZBP1 transcripts in tissue macrophages (Figure S3G). Similarly, CASP1, CASP3, CASP7, and CASP8 transcripts were highly expressed in the SELENOP^hi subset, pointing to PANoptosis as a death pathway. PANoptosis integrates components from several other death pathways and involves the formation of highly complex PANoptosomes, e.g., the NLRP12-PANoptosome,²⁶ which displays its catalytic effects by activating caspase 1 or 8 to cleave GSDMD^40–42 and triggers the formation of large membrane pores.

C1q stimulation of RA macrophages led to the upregulation of NLRC5 and NLRP12; activation of caspases 1, 3, 7, and 8; and cleavage of GSDMD and GSDME (Figure 2I), supporting the activation of PANoptotic machinery. We analyzed the multiprotein PANoptosome complex, confirming the interaction of ASC, NLRC5, NLRP12, and caspase-8 post C1q treatment (Figure 2J), while NLRP3 was absent (Figure S3H). siRNA silencing provided unequivocal evidence for NLRP12 and NLRC5’s role in the death of SELENOP^hi macrophages (Figure 2K). NLRC5 siRNA markedly reduced cell death rates, and NLRP12 siRNA silencing was even more effective, with the combination yielding no additive effect (Figures 2K and S3I). Analysis of synovial biopsies showed cleaved caspase 7 and 8 positivity in sublining CD68⁺CD206⁺ macrophages (Figure S3J).

These data indicate that C1q-C1QBP complexes regulate the life-death decisions of SELENOP^hi macrophages by triggering PANoptotosis.

C1q stimulation causes a mitochondrial energy production crisis

To understand C1QBP’s involvement in macrophage PANoptosis, we explored how C1q stimulation affected receptor availability and localization. C1q stimulation caused intracellular enrichment of C1QBP (Figure 3A, top), while C1QBP mRNA concentrations were unchanged (Figures S4A–S4C). Importantly, C1q stimulation altered the receptor’s subcellular positioning (Figure S4D). Analysis of isolated subcellular compartments revealed that C1q stimulation induced translocation of C1QBP from the plasma membrane to mitochondria (Figures 3A, bottom, and S4E).

Figure 3. C1q causes mitochondrial NAD cleavage.

MDMs generated from RA patients and healthy controls were treated with rhC1q (70 μg/mL). SYTOX DNA dye was used to detect lytic cell death.

(A) RA macrophages were stimulated with rhC1q for 48 h. Protein expression of C1QBP was quantified by western blotting (representative blots and data from eight experiments, upper). Plasma membrane, mitochondrial, and cytosolic compartments were isolated for immunoblot analysis (representative plots and data from six experiments, lower).

(B) Macrophages were treated with rhC1q or vehicle for 48 h and assessed for intracellular ATP (n = 6 patients, n = 6 controls).

(C) RA MDMs were treated with rhC1q for 48 h. Cell death was quantified by SYTOX staining. Mitochondrial membrane potential (TMRM) and mitochondrial ROS (MitoSOX) were assessed by flow cytometry (n = 8 patients, n = 8 controls).

(D) Quantification of intracellular NAD⁺ in rhC1q-stimulated macrophages (n = 8 patients, n = 6 controls).

(E) THP-1 cells expressing nuclear, cytoplasmic, or mitochondrial NAD⁺ sensors were differentiated into macrophages and treated with rhC1q for 24 h. NAD⁺-dependent fluorescence was measured by flow cytometry. Data from three experiments.

(F) Schematic of NAD⁺ consumption pathway.

(G) RA macrophages were treated with rhC1q or vehicle for 24 h (n = 6). The NAD⁺ intracellular cleavage products NAD⁺, NAM, NADH, NR, cADPR, and ADPR were quantified by liquid chromatography-tandem mass spectrometry (LC-MS/MS).

(H) RA macrophages were treated with increasing doses of cADPR-AM for 24 h. Frequencies of dead/dying cells were determined with flow cytometry (n = 5).

(I) RA macrophages were pre-treated with 8-Br-cADPR and then stimulated with rhC1q for 24 h. SYTOX staining was measured by flow cytometry (n = 4).

All data are mean ± SEM. Comparisons by two-tailed unpaired Student’s t test (G), two-tailed paired Student’s t test (A and C), and one-way ANOVA (B, D, H, and I). p values indicated in each panel.

See also Figure S4.

Given the mitochondrial localization of C1QBP, we examined how C1q signaling affects metabolic performance. Healthy macrophages produced 1–2 μM ATP/100,000 cells, and their ATP production capacity was independent of C1q stimulation. By contrast, RA macrophages generated less ATP at baseline, and C1q stimulation further diminished ATP output, connecting C1q signaling to bioenergetic failure (Figure 3B).

Next, we exposed RA macrophages to rhC1q and compared mitochondrial membrane potential in the subpopulations with and without plasma membrane damage. SYTOX⁺ macrophages had distinctly low membrane potential, confirming that C1q caused mitochondrial stress and minimized ETC function (Figure 3C). Macrophages with C1q-induced membrane leakage produced markedly higher ROS amounts, typically seen in ETC dysfunction (Figure 3C).

The profound disruption of the mitochondrial membrane potential raised the question of whether C1q activation interferes with complex I function and NAD metabolism. NAD transfers hydrogen in oxidation-reduction reactions and generates the second messengers cADPR and NAADP.⁴³ Flux balance analysis of scRNA-seq data predicted that tissue macrophages imported NAD and nicotinate, suggesting high demand for the metabolite (Figure S5A).⁴⁴ C1q stimulation lowered NAD⁺ levels in RA macrophages (Figure 3D), drawing attention to a possible role of C1q in NAD⁺ metabolism. Consequently, we monitored NAD⁺ dynamics in living C1q-stimulated macrophages with fluorescent sensors detecting mitochondrial, cytoplasmic, and nuclear NAD⁺ pools (Figure S5B).⁴⁵ In C1q-stimulated cells, mitochondrial NAD declined within the first 2 h to continuously fall over the subsequent 24 h (Figure 3E). C1q only slightly affected cytoplasmic and nuclear NAD⁺ compartments (Figure 3E).⁴⁶ These data indicated that C1q stimulation regulated NAD availability, specifically in patient-derived cells.

NAD⁺ consumption typically results in nicotinamide (NAM) and ADP-ribose (ADPR) production (Figure 3F). We therefore measured the following metabolites in vehicle- and C1q-activated RA macrophages by mass spectrometry: NAD⁺, NADH, NaAD, NR, NAM, ADPR, and cADPR. C1q stimulation of macrophages from RA patients caused a rapid decline of NAD⁺ and NADH. Instead, the macrophages accrued cADPR, yielding 6–7-fold enrichment over vehicle-treated cells (Figure 3G). Together, these data suggested that C1q induced NAD⁺ hydrolysis to form linear ADPR but mostly drove NAD⁺ cyclization to generate cADPR.

We then investigated whether the NAD cleavage product cADPR initiated DNA leakage and lytic cell death in RA macrophages (Figure 3H). One hundred μM cADPR-AM was highly efficient in causing membrane leakage, turning more than half of the cells SYTOX positive (Figure 3H). In support of the concept that the second messenger cADPR controlled macrophage membrane disruption and cell lysis, employing the selective and competitive cADPR antagonist 8-bromo-cADPR successfully protected RA macrophages from membrane rupture (Figure 3I).

Together, these data linked the death-inducing signal in RA macrophages to the induction of a metabolic crisis, characterized by the loss of mitochondrial fitness, NAD degradation, and ETC collapse.

C1q signaling degrades mitochondrial NAD by activating the glycohydrolase SARM1

SELENOP^hi macrophages undergoing chronic stimulation and C1q release entered metabolic exhaustion, selectively killing cells that did not tolerate NAD degradation and ETC disruption. To define the molecular steps connecting C1q signaling to mitochondrial NAD loss and cADPR enrichment, we evaluated NAD cleavage products induced by C1q stimulation. NAM and cADPR accrual combined with NAD and NADH decline suggested NAD cleavage by an NAD-consuming enzyme. We analyzed the transcript expression pattern for 13 NAD-consuming enzymes in all synovial macrophage subpopulations (Figures 4A and S5C). Transcript expression for the different enzymes was subset specific. CD38 transcripts were highly enriched in CXCL⁺ macrophages. By contrast, SELENOP^hi macrophages expressed SIRT2, PARP1, and SARM1 (Figure 4A). SARM1 was almost exclusively expressed in the SELENOP^hi macrophages subcluster and was absent in non-macrophage populations in the rheumatoid joint (Figure S5C). Further support for SARM1 being a SELENOP^hi macrophage marker gene came from the comparison of healthy and RA macrophages (Figures 4B and 4C). C1q treatment further upregulated SARM1 expression, identifying the enzyme as a complement target gene (Figure S5D).

Figure 4. C1QBP targets the NAD-consuming enzyme SARM1.

(A) Gene expression of NAD⁺-consuming enzymes in synovial macrophages.

(B) Expression of NAD⁺-consuming enzymes in control and RA IL-10/LPS-polarized MDM (RT-qPCR). Color scale, fold change.

(C) Immunoblotting of SARM1 protein. Loading control, β-actin (n = 8 patients, n = 8 controls).

(D and E) RA macrophages transfected with control siRNA or siRNA to SARM1, SIRT1, PARP1, and CD38 were treated with rhC1q. Quantification of intracellular NAD⁺ (D). Flow cytometry for the cell membrane portion (E) (n = 6).

(F) RA macrophages were pre-treated with or without nicotinamide (NAM) for 2 h before rhC1q stimulation. Frequencies of dead/dying cells were determined by flow cytometry (n = 5).

(G and H) RA macrophages were treated with vehicle or the SARM1 inhibitor DSRM-3716. Quantification of cellular NAD⁺ (G, n = 7) and cell death in response to C1q stimulation (H, n = 4).

(I) The interaction between SARM1 and C1QBP was analyzed in 293 T cells. IP of whole-cell lysates using anti-GFP, followed by IB analysis of C1QBP.

(J) Correlation of C1QBP and SARM1 gene expression in RA macrophages (n = 33).

Data are mean ± SEM. Comparisons by one-way ANOVA (B–D and F–H) and two-tailed paired Student’s t test (E). p values shown in each panel, or ****p < 0.0001.

See also Figure S5.

To connect enzyme expressions to C1q-induced NAD loss, we knocked down the four NAD⁺-consuming enzymes elevated in RA macrophages and quantified C1q-induced NAD⁺ concentrations (Figures 4D and S5E). C1q stimulation consistently reduced NAD⁺ concentrations by 20%–30% (Figure 4D). SIRT1 or PARP1 siRNA silencing failed to prevent the C1q-induced NAD⁺ reduction. However, SARM1 siRNA silencing abrogated C1q-induced NAD⁺ loss, pinpointing SARM1 as a critical NAD⁺ regulator in C1q-stimulated macrophages (Figure 4D).

We relied on a two-pronged approach to examine whether SARM1 is required for C1q-dependent macrophage death. First, we knocked down SARM1 and compared DNA leakage in control versus RA macrophages (Figure 4E). Reducing the SARM1 content was sufficient to minimize DNA leakage in C1q-stressed RA macrophages, implicating SARM1 in the process of lytic macrophage death. To confirm that SARM1’s NAD hydrolase activity was responsible for cell death, we restored intracellular NAD⁺ in C1q-treated macrophages with NAM, suppressing NAD hydrolase activity by a self-inhibition mechanism.⁴⁷ NAM supplementation partially prevented PANoptosome assembly and rescued macrophages from C1q-induced death, confirming the critical role of NAD⁺ depletion in the process (Figures 4F and S5F). In parallel, we applied 5-Iodoisoquinoline (DSRM-3716), a pharmacologic inhibitor with specificity for SARM1 enzyme activity.⁴⁸ Blocking SARM1 activity restored cellular NAD⁺ concentrations (Figure 4G) and alleviated C1q-induced plasma membrane disruption (Figure 4H).

SARM1 possesses an N-terminal mitochondrial localization signal (MLS) domain, placing the NAD hydrolase into the mitochondria.^49,50 Similarly, C1QBP has a canonical N-terminal MLS,⁵¹ raising the question of whether C1q signaling imposes NAD⁺ breakdown by C1QBP-SARM1 interaction in the mitochondria. Co-immunoprecipitation demonstrated that C1QBP physically associated with SARM1 in co-transfected HEK293 T cells (Figure 4I). Correlative studies in tissue-derived RA macrophages confirmed that C1QBP and SARM1 expression were highly correlated in 33 patient samples (Figure 4J).

Together, these data identify SARM1 as a sensor of metabolic stress and an executioner of cell death of metabolically exhausted macrophages.

C1q is an effective driver of synovial inflammation

We examined whether the C1q-C1QBP-SARM1-cell death axis is relevant in vivo. We employed a chimeric mouse model in which human synovium is engrafted into NSG mice, and synovitis is induced by adoptive transfer of RA peripheral blood mononuclear cell (PBMC) (Figure S6A).^5,8,52–54 To investigate whether C1q exacerbates synovitis, mice received injections of vehicle or 150 μg rhC1q on alternate days. The following outcome parameters were measured in explanted synovial grafts: tissue inflammation score; density of synovial macrophages and fibroblast-like synoviocytes (FLSs); and activation state of synovial T cells, macrophages, and FLSs. Upon C1q treatment, the inflammation score doubled (Figure 5A). Flow cytometry of tissue-derived cells showed expansion of macrophages and T cells in C1q-treated tissues (Figures 5B and 5C). Since adoptively transferred macrophages were CFSE-labeled, we were able to separate tissue-resident and invading macrophages (Figure 5B). Increased macrophage recruitment into synovial lesions accounted for the expansion of the macrophage population, suggesting that C1q enhanced the recruitment of inflammatory cells into synovial space. C1q did not affect the expansion of FLS (Figure S6B). To analyze C1q-induced synovitis on a cellular level, we analyzed the production of IL-1β, IL-6, and tumor necrosis factor (TNF) in macrophages, T cells, B cells, and FLS by flow cytometry (Figures 5D and S6C). C1q-induced synovial inflammation depended primarily on cytokine-producing synovial macrophages, T cells, and FLS. B cell function was unaffected (Figure S6D). The frequency of IL-1β⁺ macrophages increased from 25% in controls to 50% in C1q-treated grafts (Figure 5D). FLS responded to C1q with robust IL-6 induction. TNF-producing T cells expanded markedly in response to C1q (Figure 5D). Similarly, C1q-treated tissues contained higher frequencies of CD68⁺ TNF⁺ macrophages than the controls (Figure 5D). And C1q treatment enhanced the inflammatory status and cytokine production of FLS (Figure 5E).

Figure 5. C1q drives synovial tissue inflammation.

RA PBMC supplemented with autologous CFSE-labeled macrophages were adoptively transferred into NSG mice that were engrafted with human synovium. Chimeras were treated with vehicle or rhC1q (intraperitoneally [i.p.]) on alternate days for 2 weeks. Cells from explanted synovial grafts were analyzed by flow cytometry. n = 9, vehicle; n = 10, rhC1q.

(A) H&E staining of synovial infiltrates. Tissues were assigned inflammation scores in a blinded fashion.

(B and C) Frequency of CD68⁺ macrophages (B) and CD3⁺ T cells (C) as determined by flow cytometry.

(D) Frequency of IL-1β-, IL-6-, and TNF-producing macrophages, synovial fibroblasts (FLSs), and T cells. See also Figure S6D–S6F.

(E) FLS isolated from the tissue grafts were incubated with the secretion inhibitor BFA for 5 h and assessed for IL-6 production by intracellular flow cytometry.

(F) Transcripts for the indicated genes were quantified by RT-qPCR from tissue extracts of vehicle- and C1q-treated synovial grafts.

Data are shown as violin plots or as bar graphs with mean ± SEM with individual values shown. Comparison by two-tailed unpaired Student’s t test (A–C) and two-tailed paired Student’s t test (E and F). p values are provided in panels.

See also Figure S6.

Transcriptomic analysis in tissue grafts showed robust upregulation of C1q-induced cytokine mRNAs (IL23A, TNF, IL1β, and IL6) (Figure 5F). C1q-treated tissue expressed significantly more HLA-DRB1 mRNA, while the anti-inflammatory markers NUPR1 and IL10 were unchanged (Figure 5F).

In these studies, C1q has emerged as a potent intensifier of synovitis, driving activation of synovial macrophages, T cells, and fibroblasts.

Blocking SARM1 activity suppresses synovial inflammation

The above data implicated C1q signaling in the immunogenic death of SELENOP^hi macrophages and localized its action to the mitochondria. To evaluate whether SARM1 regulates macrophage fate in vivo, we treated humanized NSG mice with the SARM1 inhibitor DSRM-3716, a potent selective inhibitor lacking activity against CD38.⁴⁸ Synovitis was induced in human synovium-engrafted NSG mice as in Figure 5, and adoptively transferred macrophages were tagged with CFSE to distinguish them from the tissue-resident population. Treatment of spontaneous synovitis with DSRM-3716 suppressed inflammation scores, protected tissue-resident macrophages from death, and reduced frequencies of SYTOX⁺MerTK⁺CD206⁺ macrophages, all supporting a role of SARM1 in macrophage death-induced synovitis (Figures 6A–6C and S6E–S6J). DSRM-3716 did not suppress the subset of IL-1β- and IL-6-producing macrophages (Figure 6D), consistent with the selective expression of SARM1 in SELENOP^hi macrophages (Figure 4A) and the almost exclusive allocation of cytokines to CXCL^posSARM1^neg macrophages (Figure 1B). The SARM1 inhibitor diminished the density of CD3⁺ T cells (Figure 6E) and minimized IL-6 production in synovial fibroblasts (Figure 6F).

Figure 6. Blocking SARM1 NAD hydrolase activity alleviates synovial tissue inflammation.

Human synovium-NSG chimeras were generated as in Figure 5. Chimeras were treated with vehicle or the SARM1 inhibitor DSRM-3716 (A–F) or with C1q plus/minus DSRM-3716 (G–L). Single-cell suspensions from explanted synovial grafts were analyzed by flow cytometry (n = 13, vehicle; n = 13, DSRM-3716; n = 12, C1q minus DSRM-3716; n = 12, C1q plus DSRM-3716).

(A and G) Representative images of tissue sections from the synovial grafts (H&E) showing density of inflammatory infiltrates. Tissues were assigned inflammation scores (see STAR Methods) in a blinded fashion.

(B–F and H–L) Frequencies of CD68⁺ macrophages (B and H), dying/dead MerTK⁺CD206⁺ macrophages (C and I), macrophages producing IL-1β or IL-6 (D and J), CD3⁺ T cells (E and K), and IL-6/IL-1β-producing synovial fibroblasts (F and L).

(M) RNA-seq analysis of C1q-loaded synovial tissues treated without and with DSRM-3716. Treatment-induced transcriptome changes were presented as a volcano plot (n = 5 experiments).

(N) Gene Ontology (GO) pathway enrichment of significantly upregulated/downregulated genes (p.adjust < 0.05) after DSRM-3716 treatment.

Data are shown as violin plots or as bar graphs with mean ± SEM. Comparison by one-way ANOVA (B–D and H–J) or two-tailed unpaired Student’s t test (A, E–G, K, and L). p values are provided.

See also Figure S6.

To mimic conditions of high C1q concentrations in the synovial tissue, we co-administered rhC1q and the SARM1 inhibitor DSRM-3716 (Figures 6G–6L and S6K). Under conditions of C1q-exacerbated synovitis, SARM1 inhibition was a highly effective anti-inflammatory therapy, efficiently reducing inflammation scores (Figure 6G). In DSRM-treated tissues, CSFE^neg macrophages recovered, and the death rate of MerTK⁺CD206⁺ macrophages fell by 50% (Figures 6H, 6I, and S6L). In C1q-exacerbated synovitis, IL-1β-producing macrophages expanded (Figure 6J) but were successfully suppressed when SARM1 activity was blocked. The SARM1 inhibitor provided effective tissue protection, profoundly reducing tissue-infiltrating T cells and pro-inflammatory FLS (Figures 6K, 6L, and S6M–S6R).

To provide an overview of transcriptional changes following SARM1 inhibition, we performed bulk RNA sequencing on RNA extracts from synovial grafts of vehicle- and DSRM-3716-treated mice. SARM1 inhibition led to broad downregulation of key pro-inflammatory mediators, including chemokines, cytokines, and matrix metalloproteinases (MMPs), alongside significant upregulation of gene sets related to mitochondrial fitness and oxidative phosphorylation (Figures 6M and 6N; Table S2).

These proof-of-principle studies validate SARM1 as a key driver of synovitis and demonstrate that this pathway of tissue inflammation is treatable.

DISCUSSION

Despite a broad range of therapies, RA remains incurable, and joint inflammation often persists for decades. Restoring immune tolerance requires an understanding of how tissue inflammation is perpetuated. The macrophage-rich RA synovium contains diverse macrophage subsets, with only one exhibiting pro-inflammatory features and several classified as homeostatic. Transcriptomic profiling identified SELENOP^hiMerTK⁺CD206⁺ macrophages as resident tissue macrophages with predicted protective and reparative roles. However, functional analysis revealed that this subset acquires tissue-destructive capability by expressing the C1q receptor C1QBP and responding to C1q with metabolic collapse and lytic cell death. We identified the NAD⁺-consuming enzyme SARM1 as a target of mitochondrially translocated C1q-C1QBP complexes. SARM1-mediated NAD⁺ hydrolysis produced the Ca²⁺-mobilizing messenger cADPR and promoted assembly of NLRC5-NLRP12-caspase8 PANoptosomes. PANoptotic macrophages drove severe tissue inflammation, attenuated by pharmacologic SARM1 inhibition. These findings classify SELENOP^hi macrophages as tissue-destructive effector cells and implicate metabolic exhaustion in chronic autoimmune inflammation. A SARM1-dependent metabolic checkpoint upstream of this “self-sacrifice” response prunes chronically stressed macrophages while selecting metabolically resilient ones. Although this process enhances anti-pathogen defense, it also amplifies autoimmune tissue injury, sustaining disease chronicity.

Like in other chronic inflammatory lesions, synovial macrophages are heterogeneous, with each subtype potentially supporting a highly specialized functional domain.^10,55 The origin, longevity, and functional attributes of synovial macrophages remain unclear, particularly the tissue-resident macrophages that do not easily fit into either pro- or anti-inflammatory paradigms.⁵⁶ Pro-inflammatory actions have been assigned to IL-1β⁺HBEGF⁺ macrophages.⁵⁷ Conversely, MerTK⁺ macrophages are considered anti-inflammatory and remission-sustaining, possibly through inflammation-resolving lipid mediators.³⁹ However, a recent study finds synovial MerTK⁺CD206⁺ macrophages to be CD40^pos and express multiple pro-inflammatory mediators.¹⁰ In our analysis, focusing on synovial tissues with high Krenn inflammation scores, MerTK^hiTREM2^hiLYVE1^hiSELENOP^hi macrophages were the most abundant macrophage subset and were positively correlated with inflammation intensity, raising the question of whether they are host-protective or damaging. SELENOP (encoding for selenoprotein P) high expression suggested a potential role in redox stress protection, as selenocysteine is a potent catalytic redox-active residue.⁵⁸ Secretome analysis of MerTK⁺CD206⁺SELENOP^hi macrophages yielded redox protection proteins but showed high abundance of all three C1q components. Physiologically, C1q activates the classical complement pathway and is anti-microbial by modulating adaptive immunity.⁵⁹ Conversely, in the tumor environment, C1q abundance predicts poor outcomes.⁶⁰ Tumor-associated macrophages (TAMs) specialized in C1q release typically express CD206, SELENOP, FOLR2, and APOE but not SPP1 and are potent inducers of T cell exhaustion.⁶⁰ Together, these data associate C1q-producing macrophages with poor outcomes in both tumor immunology and autoimmune tissue inflammation.

In a humanized mouse model, C1q treatment induced aggressive synovitis through lytic macrophage death and T cell, macrophage, and fibroblast activation. Molecular analysis identified a role for NLRP12 and NLRC5, components of a PANoptosome triggering inflammatory cell death.^26,61 Targeted metabolomics indicated low NAD concentrations, consistently and expectedly associated with ETC collapse and ATP loss, in dying macrophages. Notably, metabolically exhausted macrophages sensed a crisis state through cADPR, a metabolite generated by SARM1-dependent NAD hydrolysis. cADPR alone was sufficient to trigger macrophage PANoptosis, and blocking cADPR protected macrophages from lytic death. Thus, for synovial macrophages, the upstream metabolic signal lies in cADPR, a calcium-mobilizing second messenger produced by NAD consumption.

Current data link macrophage survival to the tissue metabolic landscape. Recent studies have defined RA synovium as a glucose-low environment, forcing residents and infiltrating cells toward metabolic adaptations.⁸ Accordingly, synovial macrophages switch to glutaminolysis and utilize glutamine-derived glutamate and α-ketoglutarate as biofuel. This metabolic switch depends on the transcription factor RFX5, which upregulates glutamate dehydrogenase and antigen-presenting machinery.⁸ Thus, the inflamed joint relies on supply from glutamine-producing organs, particularly muscles. Current data confirm the intense metabolic pressure in the inflamed joint and the checkpoint function of metabolite availability (e.g., NAD) in determining macrophage fate. SELENOP^hi macrophages monitor activation by sensing their own C1q production and probing metabolic resilience via SARM1-dependent NAD cleavage. Cells with insufficient NAD reserves accumulate cADPR, triggering lytic death.

Profiling NAD⁺-consuming enzymes identified the NAD⁺ glucohydrolase SARM1 as selectively overexpressed in SELENOP^hiMerTK⁺CD206⁺ macrophages. SARM1 and CD38 both cleave mitochondrial NAD⁺ but differ in their ability to generate ADPR and cADPR.^62,63 CD38 silencing did not affect NAD⁺ loss or macrophage PANoptosis, implicating SARM1 as the sole mediator of C1q-induced NAD⁺ depletion. Most cell types, except neurons, express only low amounts of this potentially destructive NAD⁺ hydrolase.⁶⁴ SARM1-induced NAD⁺ deficiency activates an intrinsic self-destruction program that eliminates damaged axons, and SARM1 is considered a therapeutic target to prevent neuron loss in neurodegenerative disease.⁶⁵

Beyond its role as a cell executioner, SARM1 has complex immunoregulatory functions,⁶⁶ acting as an inhibitory checkpoint through its TIR domain to restrain cytokine production and TLR signaling.⁶⁷ Its inhibition of inflammasome activation is NAD⁺ hydrolase-independent, while suppression of IL-1β secretion involves both hydrolase-dependent and -independent mechanisms.^33,68 Klebsiella pneumoniae exploits SARM1 in a type I interferon (IFN)-dependent manner to evade host immunity.⁶⁹ In our studies, SARM1-marked macrophages engaged in resolution and tissue repair, yet their activation induced severe inflammation through PANoptotic death. RA macrophages showed higher SARM1 levels, consistent with a failed negative feedback response. Elevated SARM1 rendered these cells vulnerable to NAD⁺ depletion and metabolic exhaustion, converting a protective mechanism into one driving immunogenic death and chronic inflammation.

Upstream signals involve an autocrine loop, where chronically stressed cells sense their secretome and probe their metabolic resilience. In chronically inflamed tissue, PANoptotic macrophages may release C1q from cytoplasmic stores, sustaining a feed-forward loop of inflammatory cell death. Overall, SARM1 occupies a strategic position to integrate information from its environment and metabolic conditions to determine macrophage fate. This pathway adds metabolic crises in tissue macrophages to the list of immunometabolic deficiencies in RA⁷⁰ and implicates NAD-dependent signaling in sustaining the chronicity of autoimmunity.

The shift from host protection to aggression poses considerable challenges for developing new therapies, especially for the chronic phase of disease. Immunogenic cell death serves as a potent inflammatory stimulus^71,72 and potential target for enhancing anti-tumor immunity.⁷³ The intricate chain of events identified here, involving the C1q-C1QBP-SARM1-cADPR-NLRP12-NLRC5-caspase 8 axis, presents multiple points for intervention. In a proof-of-principle therapeutic trial, we reduced synovial inflammation by inhibiting SARM1’s enzymatic activity. While C1q-producing macrophages are typically seen as immunosuppressive, they lose protective function in synovitis and instead drive inflammation. Protective C1q-dependent responses necessitate metabolic fitness, which is compromised in the harsh metabolic environment of a chronically inflamed joint.⁸ Therefore, enhancing the metabolic conditions within tissue lesions could help mitigate the self-destructive behavior of SELENOP^hi macrophages. Besides addressing NAD deficiency, strategies to interrupt the macrophage destruction program could involve halting mitochondrial transition of C1q-C1QBP complexes, suppressing the primed state of SARM1, blocking cADPR action, or counteracting the mechanisms leading to lytic cell death.

Limitations of the study

Several limitations warrant consideration. Patients donating tissue for the secretome analysis were biased toward treatment-refractory synovitis. Future studies should include patients with mild synovitis. Furthermore, while our study describes reduced metabolic flexibility and increased sensitivity to PANoptotic death in synovial tissue macrophages, signals leading to this death-primed state are not fully understood and require further investigation.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and reagents should be directed to and will be fulfilled by the lead contact, Cornelia M. Weyand (cweyand@stanford.edu).

Materials availability

This study did not generate any new, unique reagents.

Data and code availability

All data supporting the findings of this study are available within the paper and its supplemental information.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Human samples

Patients enrolled into the study fulfilled the 2010 ACR diagnostic criteria for RA, were positive for rheumatoid factor and/or anti-CCP antibodies and were selected for having active disease. About 15% of RA patients have persistent inflammatory refractory disease, do not benefit from multiple disease-modifying therapies, and have signs of ongoing joint inflammation over years.^74,75 Such patients are now recognized as difficult-to-treat RA (D2T-RA) and serve as a model system to study active synovitis. Individuals with a current or previous diagnosis of cancer, uncontrolled medical disease, or chronic inflammatory syndrome were excluded from enrollment.

To have sufficient tissue for cell isolation, synovial tissue samples were collected from synovectomies or total joint replacement surgeries. All tissue samples derived from the knee joints of patients with active disease. Active disease was defined as the presence of ≥6 swollen or tender joints and concentrations of at least moderate on the physician’s and patient’s assessments of disease activity. Patients on prednisone higher than 10 mg/day were excluded. Steroid injections into the knee joint in the last 90 days were considered an exclusion criteria for tissue collection. Tissues were assessed for synovial hyperplasia, vascularization, and density of the cellular infiltrate before they were selected for further experiments. Samples from patients with mechanical joint injury served as noninflammatory control tissues.

Demographic and clinical characteristics of the patient population are summarized in Table S3. Age-matched healthy individuals without a personal history of cancer or autoimmune disease were recruited from the Blood Bank of Mayo Clinic Rochester. All patients and controls provided informed consent and did not receive compensation. All studies were approved by the Institutional Review Boards (IRB) of Stanford University and the Mayo Clinic.

Synovial tissue histopathology

To classify synovial tissues, serial tissue sections (5 μm) were prepared spanning a total of 500 μm³ of tissue and stained with hematoxylin and eosin (H&E). Two individuals trained in the histomorphology assessment of synovial tissue scored 2–5 sections (100 μm apart) to generate a single grade for each histopathological feature. Synovitis scoring followed the criteria set by Krenn and colleagues⁷⁶ and Najm et al. with slight modifications.⁷⁷ Four major parameters were assessed to semi-quantitatively evaluate the degree of synovial inflammation: 1) synovial lining thickness, 2) density of sub-synovial lymphocytic and plasma cell inflammatory cell infiltrate, 3) activation of stromal elements including multinucleated giant cells, and 4) sub-synovial neovascularization. Scores for each histopathological feature were averaged across the number of sections per sample. To generate a single composite inflammation score, median grades for each component were summed up and each patient was assigned to the following categories: category 1, none/normal, average grade <2.0; category 2, mild, average grade 2.0 to 5.0; category 3, moderate/severe, average grade >5.0.

Cells and tissues

PBMCs were isolated by density gradient centrifugation using Lymphoprep (STEMCELL Technologies). Monocytes were isolated from peripheral blood mononuclear cells as previously described^8,52 and were differentiated into macrophages in RPMI 1640 medium with 20 ng/ml of M-CSF (BioLegend) for a duration of 5 days in 10% fetal bovine serum. To polarize macrophages, we applied three procedures: (1) 100 ng/ml LPS (Sigma-Aldrich) and 100 U/ml IFN-γ (Sino Biological) for 24 hours; (2) 10 units/mL IL-4; (3) 10 ng/mL IL-10 followed by LPS (10 ng/mL). Cells were dissociated from culture plates using Accutase Cell Detachment Solution (Inn Cell Technologies) for 20 minutes at 37 C. HEK293T cells (CRL-11268), THP-1 cells (TIB-202) were obtained from ATCC (Manassas, VA, USA).

Synovial tissues were obtained from individuals diagnosed with inflammatory polyarthritis undergoing synovectomy or total joint replacement and were processed promptly after removal. Fresh synovial tissues were weighed and sliced into thin sections. Single-cell suspensions were prepared by treating with 1 mg/mL collagenase type IV (Worthington, LS004196) and DNase I (100 μg/mL, ZYMO RESEARCH, E1011-A) for 1 hour at 37 °C. Tissue debris was removed using sterile cell strainers (70 μm followed by 40 μm). Macrophages were isolated using Easysep Human Monocyte Cell Isolation kits (STEMCELL Technologies, 19359).

In vivo studies in chimeric NSG mice

All in vivo experiments followed previously applied procedures.^52,53,78,79 NSG mice were obtained from Jackson Laboratory and maintained in specific pathogen-free conditions at 20–22 C and on a 12-hour light/dark cycle. All mice had free access to water and food. The synovial tissues were allocated to experiments based on their arrival order in the laboratory. As described, 8–10-week-old NSG mice were implanted with human synovial tissue into a subcutaneous pocket on the midback, which provides full engraftment and graft vascularization after one week. Seven days later, chimeras were reconstituted with PBMC (10 million/mouse) and in some experiments also received autologous CFSE-labeled monocyte-derived macrophages (1 million/mouse). To test the role of C1q, mice implanted with synovial tissue from the same donor were randomly divided into vehicle control and rhC1q groups and were injected intraperitoneally with vehicle or 150 ug rhC1q every 2nd day for 2 weeks. In treatment trials exploring the role of SARM1, SARM1 activity was inhibited by treating the tissue-engrafted mice with DSRM-3716 (2mg/kg i.p given every second day for 6 treatments). In some experiments, C1q and DSRM-3716 injections were combined. Synovial grafts were collected on day 21 and prepared for RNA extraction or embedded into OCT for immunofluorescence analysis. Tissues were digested as previously described,⁵ and isolated cells were incubated with PMA, ionomycin, and monensin for 5h. All protocols were approved by the Institutional Animal Care and Use Committee of the Mayo Clinic.

METHOD DETAILS

Secretome array

RA synovial tissue-derived macrophages were isolated by monocyte enrichment kit (STEMCELL, 19058) and kept in RPMI 1640 complete medium, and supernatants were collected after 48 hours. Secreted proteins were analyzed via RayBio^® L-Series Human Antibody Array 3 Membrane Kits (Cat: AAH-BLM-3–4). In brief, samples were purified to remove low molecular weight amine derivatives or unwanted buffer by spin column purification. Purified supernatants were then labeled with biotin. After complete blocking, labeled supernatants were added to the array slide champers, which were pre-printed with capture antibodies, for overnight incubation at 4°C. Following washing, Streptavidin-conjugated fluorescent dye (Cy3-equivalent) was then applied to the array. Finally, the glass slides were dried, and laser fluorescence scanning was used to visualize the signals. MFI of each array protein was quantified by imageJ and normalized to initial cell numbers. Fold change and adjusted p-values were calculated comparing macrophage preparations rich (>75%) and poor (25%) in MerTK^hiCD206^hi macrophages (Table S1). C1q produced by monocyte-derived macrophages was measured in the supernatants 48 hours after stimulation using human C1q ELISA arrays (LSBIO, LS-F23976–1).

Plasmids and reagents

Plasmids for the cytoplasmic NAD⁺ biosensor (Cat.186787), the nuclear NAD⁺ biosensor (Cat.186789), the mitochondrial NAD⁺ biosensor (Cat.186791), the cytoplasmic cpVenus control (Cat.186788), the nuclear cpVenus control (Cat.186790) and the mito cpVenus control (Cat.186792) were purchased from Addgene. The plasmid for hSARM1 (Cat. HG20854-UT) was purchased from SinoBiological. C1QBP with a myc-GFP tag was cloned into the pLVX-IRES-Puro Vector from cDNA. Lipofectamine 2000 (11668019) was obtained from Invitrogen. The SHP-1 inhibitor (TPI-1, HY-100463) and SARM1 inhibitor (DSRM-3716, HY-W021879) were purchased from MedChemExpress. cADPR (C7344) was obtained from Sigma-Aldrich. The cADPR antagonist (8-Bromo-cADP-Ribose, sc-201514A) was purchased from Santa Cruze.

RNA extraction, RT–PCR and RNA-seq

Total RNA was isolated using TRIzol reagent (Molecular Research Center Inc.) and Direct-zol RNA MiniPrep kits (Genesee Scientific). cDNA was synthesized from total RNA using ribonuclease H reverse transcriptase (Invitrogen) and oligo(dT) primers. RT-PCR was performed in triplicate using an iCycler sequence detection system (Bio-Rad) and iQTM SYBR Green Supermix (Bio-Rad). Expression of individual genes was calculated from a standard curve and normalized to the expression of β-actin. PCR primers used in this study are listed in Table S4. Tissue total RNA was isolated and subjected to RNA-seq analysis. Differential expression analysis was performed on the count data using the R package DESeq2. Significant genes were defined using a p value at a cutoff of 0.05 and fold-change of 1.0.

Flow cytometry

Cell surface staining was performed as previously described.^8,54,80 Data were collected using a CYTEK NL-3000 flow cytometer and analyzed by FlowJo 10.0 (Tree Star). The following antibodies and dyes were used for staining: SYTOX Deep Red (Invitrogen, S11380), Zombie UV Fixable Viability Kit (BioLegend, 423107), AF594 anti-human C1Q antibody (Abcam, ab4223), PerCP anti-human CD45 antibody (BioLegend, 304026), PE anti-human CD68 antibody (BioLegend, 333808), Brilliant Violet 510 anti-human CD3 antibody (BioLegend, 344828), Brilliant Violet 711 anti-human CD4 antibody (BioLegend, 317440), BUV805 anti-human CD8 antibody (BD, 612942), PerCP/Cyanine5.5 anti-human CD11c antibody (BioLegend, 337210), Brilliant Violet 785 anti-human CD19 antibody (BioLegend, 302240), PE/Cyanine7 anti-human MerTK antibody (BioLegend, 367609), Brilliant Violet 510 anti-human CD206 (MMR) antibody (BioLegend, 321138), Brilliant Violet 510 anti-human CD90 (Thy1) antibody (BioLegend, 328125), Brilliant Violet 421 anti-human podoplanin antibody (BD, 566456), FITC anti-human C1QBP antibody (Santa Cruze, sc-23885 FITC), Pacific Blue anti-human IL-1β antibody (BioLegend, 511709), PE/Cyanine7 anti-human IL-6 antibody (Biolegend, 501120), APC anti-human TNF-α antibody (Biolegend, 502912), APC/Cyanine7 anti-human CD31 antibody (Biolegend, 303120). For intracellular cytokine staining, single-cell suspensions were stimulated for 5 h with PMA plus ionomycin in the presence of monensin and then subjected to intracellular staining followed by flow cytometry analysis. The antibodies used in flow cytometry experiments are listed in key resources table.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-human C1Q, FITC	Abcam	Cat# ab4223; RRID: AB_304387
Goat anti-Rabbit IgG, HRP	Cell Signaling TECHNOLOGY	Cat# 7074S; RRID: AB_2099233
Goat anti-Mouse IgG, HRP	Cell Signaling TECHNOLOGY	Cat# ab6789; RRID: AB_955439
Human TruStain FcX	BioLegend	Cat# 304026; RRID: AB_2818986
anti-human CD45, PerCP	BioLegend	Cat# 422302; RRID: AB_893337
anti-human CD68, PE	BioLegend	Cat# 333808; RRID: AB_1089056
anti-human CD3, BV510	BioLegend	Cat# 344828; RRID: AB_2563704
anti-human CD4, BV711	BioLegend	Cat# 317440; RRID: AB_2562912
anti-human CD8, BUV805	BD Biosciences	Cat# 612942; RRID: AB_2870223
anti-human CD11c, PerCP/Cyanine5.5	BioLegend	Cat# 337210; RRID: AB_1279069
anti-human CD19, BV785	BioLegend	Cat# 302240; RRID: AB_2563442
anti-human MerTK, PE/Cyanine7	BioLegend	Cat# 367609; RRID: AB_2687286
anti-human CD206, BV510	BioLegend	Cat# 321138; RRID: AB_2721530
anti-human CD90 (Thy1), BV510	BioLegend	Cat# 328125; RRID: AB_2562199
anti-human podoplanin, BV421	BD Biosciences	Cat# 566456; RRID: AB_2739735
anti-human C1QBP, FITC	Santa Cruze	Cat# sc-23885; RRID: AB_673544
anti-human IL-1β, Pacific Blue	BioLegend	Cat# 511709; RRID: AB_2124351
anti-human IL-6, PE/Cyanine7	BioLegend	Cat# 501120; RRID: AB_2572042
anti-human TNF-α, APC	BioLegend	Cat# 502912; RRID: AB_315264
anti-human CD31, APC/Cyanine7	BioLegend	Cat# 303120; RRID: AB_10640734
SARM1 (D2M5I) Rabbit mAb	Cell Signaling Technology	Cat# 13022; RRID: AB_2798090
CD68 Monoclonal Antibody (KP1)	Thermo Fisher Scientific	Cat# 14-0688-82; RRID: AB_11151139
Donkey anti-Sheep IgG (H+L) Cross-Adsorbed Secondary Antibody, AF647	Thermo Fisher Scientific	Cat# A-21448; RRID: AB_2535865
Human MMR/CD206	R&D systems	Cat# MAB2534; RRID: AB_2227843
NLRC5 (E1E9Y) Rabbit mAb	Cell Signaling Technology	Cat# 72379S
NLRP12 Antibody (N-term)	Abcepta	Cat# AP14014a; RRID: AB_10893749
NLRP3 Polyclonal Antibody	Invitrogen	Cat# PA5-79740
Caspase-1 (D7F10) Rabbit mAb	Cell Signaling Technology	Cat# 3866; RRID: AB_2069051
Caspase-3 Antibody	Cell Signaling Technology	Cat# 9662; RRID: AB_331439
anti-cleaved caspase-3 (Asp175)	Cell Signaling Technology	Cat# 9661; RRID: AB_2341188
Caspase-7 Antibody	Cell Signaling Technology	Cat# 9492; RRID: AB_2228313
anti-cleaved caspase-7 (Asp198)	Cell Signaling Technology	Cat# 9491; RRID: AB_2068144
Caspase-8 (D35G2) Rabbit mAb	Cell Signaling Technology	Cat# 4790; RRID: AB_10545768
Gasdermin D (E8G3F) Rabbit mAb	Cell Signaling Technology	Cat# 97558; RRID: AB_2864253
Gasdermin E Antibody	Cell Signaling Technology	Cat# 84005
MLKL (D2I6N) Rabbit mAb	Cell Signaling Technology	Cat# 14993; RRID: AB_2721822
Phospho-MLKL (Ser358) (D6H3V) Rabbit mAb	Cell Signaling Technology	Cat# 9661; RRID: AB_2732034
RIP (D94C12) XP® Rabbit mAb	Cell Signaling Technology	Cat# 3493; RRID: AB_2305314
ASC/TMS1 Polyclonal antibody	proteintech	Cat# 10500-1-AP; RRID: AB_2174862
Bacterial and virus strains
E. coli DH5a	AngYuBio	G6016
Biological samples
Human synovium tissue	Mayo Clinic, Rochester	N/A
Chemicals, peptides, and recombinant proteins
collagenase type IV	Worthington	Cat# LS004196
Dnase I	Sigma-Aldrich	Cat# 4716728001
RPMI 1640, no glucose	ThermoFisher	Cat# 11879020
Recombinant Human IFN-gamma	Sino Biological	Cat# 11725-HNAS-100
Recombinant Human IL-4	GenScript	Cat# Z02925-50
Recombinant Human IL-10	R&D systems	Cat# 217-IL-010
Recombinant Human M-CSF	Biolegend	Cat# 574806
Zombie UV Fixable viability	Biolegend	Cat# 423107; RRID: AB_955439
SYTOX Deep Red	Invitrogen	Cat# S11380
Lipofectamine 3000	Invitrogen	Cat# L3000008
SHP-1 inhibitor (TPI-1)	MedChemExpress	Cat# HY-100463
SARM1 inhibitor (DSRM-3716)	MedChemExpress	Cat# HY-W021879
8-Bromo-cADP-Ribose	Santa Cruze	Cat# sc-201514A
Cyclic ADP-ribose ammonium	MedChemExpress	Cat# HY-N7395A
LPS	Sigma-Aldrich	Cat# L2630-25MG
Native human C1q protein	Abcam	Cat# ab282858
CD38 siRNA	Santa Cruze	Cat# sc-29996
NLRC5 siRNA	Santa Cruze	Cat# sc-93466
Cell Activation Cocktail (without Brefeldin A)	BioLegend	Cat# 423301
BD GolgiStop™ Protein Transport Inhibitor (Containing Monensin)	BD Bioscience	Cat# 554724
Q-VD-Oph	MedChemExpress	Cat# HY-12305
PARP1 siRNA (h)	Santa Cruze	Cat# sc-29437
SIRT2 siRNA (h)	Santa Cruze	Cat# sc-40988
SARM siRNA (h)	Santa Cruze	Cat# sc-62976
C1q-A siRNA (h)	Santa Cruze	Cat# sc-43651
Critical commercial assays
Easysep Human Monocyte Cell Isolation kits	STEMCELL Technologies	Cat# 19359
Mitochondria Isolation Kit	Thermo Fisher Scientific	Cat# 89874
Mem-PER™ Plus Membrane Protein Extraction Kit	Thermo Fisher Scientificsd	Cat# 89842
RNeasy Mini Kit	QIAGEN	Cat# 74134
Human L3 Array, Glass Slide	raybiotech	Cat# AAH-BLG-3-8
Human Complement C1q (Sandwich ELISA) ELISA Kit	Lsbbio	Cat# LS-F23976-1
NAD+/NADH Assay Kit	millipore	Cat# MAK460
Deposited data
Raw and analyzed RNA-seq data	This paper	N/A
Experimental models: Cell lines
HEK293T cell line	ATCC	CRL-3216
THP1 cell line	ATCC	TIB-202
Experimental models: Organisms/strains
NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ	The Jackson Laboratory	Strain #:005557; RRID:IMSR_JAX:005557
Oligonucleotides
Primer for qPCR, see Table S2	This paper	N/A
Recombinant DNA
cytoplasmic NAD+ biosensor	Addgene	Cat# 186787; RRID: Addgene_186787
nuclear NAD+ biosensor	Addgene	Cat# 186789; RRID: Addgene_186789
mitochondrial NAD+ biosensor	Addgene	Cat# 186791; RRID: Addgene_186791
cytoplasmic cpVenus control	Addgene	Cat# 186788; RRID: Addgene_186788
nuclear cpVenus control	Addgene	Cat# 186790; RRID: Addgene_186790
mito cpVenus control	Addgene	Cat# 186792; RRID: Addgene_186792
pCDNA3.1 C1QBP-GFP	This paper	N/A
SARM1 cDNA ORF Clone, Human, untagged	SinoBiological	HG20854-UT
Software and algorithms
ImageJ		N/A
FlowJo	FlowJo	N/A
GrapdPad Prism	GraphPad Software, Inc	N/A
R Studio	https://www.r-project.org/	Version 4.5.2
Seurat (CCA analysis)	https://satijalab.org/seurat/	Version 5.0.0

Quantification of metabolic intermediates

NR, NAM, NAAD, NAD, NADH, ADPR, and cADPR were measured using LC-MS/MS on a ThermoFisher Scientific TSQ Quantum Ultra by the Mayo Clinic Metabolomics Core. Standard curves for each compound were generated by analyzing NADomes reconstituted in 5mM ammonium formate. Intracellular NAD concentrations were measured using NAD/NADH assay kits (Millipore, MAK460). Intracellular ATP (adenosine triphosphate) concentrations were quantified by Luminescent ATP Detection Assay Kits (Abcam, ab113849). Lyophilized samples were reconstituted with 5 mM ammonium formate (15 ml), centrifuged (13,000 ×g, 10 min, 4°C), and 10 ml clear supernatant was analyzed.

Immunofluorescence and confocal microscopy

Methods for immunofluorescence analysis of cultured cells and tissue sections have been published.⁸¹ Macrophages were seeded in Nunc Lab-Tek II Chamber Slide (ThermoScientific, 154534PK) and treated with vehicle or C1q. Cells were washed with cold PBS and stained using SYTOX Deep Red DNA dye (Invitrogen, S11380) for 30 min at 37°C. 4% paraformaldehyde solution (Affymetrix) was used to fix the cells; primary antibodies were incubated at 4°C overnight, followed by a 1h incubation at room temperature with fluorescence-conjugated secondary antibodies. For tissue sections, 5 μm thick sections of synovial biopsies were rinsed, preincubated with 5% blocking serum in 0.1% Triton X-100 for 1 hour, and then incubated overnight with primary antibodies at 4°C. Fluorescence-conjugated secondary antibodies were incubated at room temperature for 2h. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (Roche). All the images were accessed on an LSM 780 or LSM 980 (Carl Zeiss) confocal microscope. Data was processed using Carl Zeiss software ZEN (blue) and analyzed by imageJ software. The following antibodies were used: CD163 monoclonal antibody (10D6) (ThermoFisher, MA5–11458), anti-human MMR/CD206 antibody (R&D systems, MAB2534), CD68 monoclonal antibody (KP1) (ThermoFisher Scientific, 14–0688-82), cleaved caspase-7 (CST, 9491S), cleaved caspase-8 (CST, 9496S), goat anti-rat IgG (H+L) Cyanine3 (Invitrogen, A10522), donkey anti-sheep IgG (H+L) Alexa Fluor 647 (ThermoFisher Scientific, A-21448), goat anti-mouse IgG (H+L) Alexa Fluor 488 (ThermoFisher Scientific, A-11001), donkey anti-rabbit IgG (H+L) Alexa Fluor 594 (ThermoFisher Scientific, A-21207).

Live cell imaging

Fluo-4AM (Life Technologies) and SYTOX were prepared fresh and added to macrophages at a final concentration of 2.5 μM for 30 min before the experiment. After treatment, cells were imaged with the Keyence BZX800E system (Keyence). Images were collectively analyzed for Fluo-4AM intensity with CellProfiler software.

Western blotting and immunoprecipitation

Immunoblotting methods followed procedures previously published.⁸ Whole-cell lysates or subcellular extracts were prepared by lysing the cells in lysis buffer supplemented with protease inhibitors and immediately subjected to IB analysis and IP. For in vitro overexpression experiments, an expression vector encoding GFP-tagged C1QBP was transfected into THP-1 cells and then immunoprecipitated by anti-GFP antibody. The samples were resolved by 8.25% SDS–polyacrylamide gel electrophoresis (PAGE). After electrophoresis, the separated proteins were transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore). For immunoblotting, the PVDF membrane was blocked with 5% nonfat milk. After incubation with a specific primary antibody, HRP-conjugated secondary antibody was applied. Immune reactive signals were detected by ECL (Amersham Biosciences).

Analysis of scRNA-seq data

scRNA-seq data from RA synovial tissues generated by Wu et al. were downloaded from the Genome Sequence Archive in BIG Data Center, Beijing Institute of Genomics (BIG), Chinese Academy of Sciences, under the accession code HRA000155. Raw data processing and alignment were followed by the code scripts used in the original publication (https://github.com/yeee904/RA_scRNA). Raw sequencing data was processed by CellRanger software v 7.0 and the 10× human transcriptome GRCh38–1.2.0 was used as the reference. Single cell read counts from all synovial tissue samples were converted to Seurat objects individually by Seurat package (v 4.4.0) in RStudio (v 4.2.2). Seurat objects were filtered by UMI/gene numbers. Genes expressed in at least 5 cells and cells with 500–3500 detected genes together with less than 15% of sequences coming from mitochondrial genes were retained. The Harmony (v 1.1.0) package was used to integrate the individual samples into one Seurat object. The object was then normalized by NormalizaData function and high variable genes were detected by the FindVariableFeatures function in Seurat. After principal component analysis (PCA), clusters were identified using Uniform Manifold Approximation and Projection (UMAP). Differential gene expression (DEG) testing was performed using the FindMarkers function in Seurat with Wilcoxon test and p values adjusted using Bonferroni correction. DEGs were filtered using a minimum log2(fold change) of 0.5 and a maximum adjusted p value of 0.05. Gene Ontology (GO) and Gene Set Enrichment Analysis (GSEA) were presented by the ClusterProfiler package (v 4.6.2) on DEGs of each cluster. M1/M2 expression pattens were calculated by using AddModuleScore function based on genelists in Table S5. Flux balance analysis was performed by using METAFlux package.⁴⁴ Canonical correlation analysis (CCA) was performed by comparing the expression matrix of marker gene panels from scRNA clusters or in vitro priming conditions using the vegan package (v 2.6–8) as previously described.⁹ Cell-cell interaction analysis was performed by using CellChat (v2.1.2) among major synovial cell types.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was performed using Prism software. Two-tailed unpaired Student’s t tests were performed between two groups. One- or two-way ANOVA was used to calculate differences among multiple groups. P values <0.05 were considered significant, and the level of significance is indicated as *, P < 0.05 and **, P < 0.01. In animal studies, a minimum of four mice was required for each group based on the calculated number necessary to achieve a 2.3-fold change (effect size) in a two-tailed t test with 90% power and a significance level of 5%. All statistical tests have been justified as appropriate, and the data have met the assumptions of the tests.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.immuni.2026.01.019.

Highlights.

• Rheumatoid arthritis persists despite pro-repair MerTK⁺CD206⁺ macrophages

• MerTK⁺CD206⁺ macrophages in RA tissue undergo PANoptotic death

• PANoptosis stems from SARM1-driven NAD⁺ depletion and metabolic collapse

• SARM1 is activated via autocrine C1q-C1QBP signaling