An iron-regulated methionine redox axis governs adipose browning and cancer cachexia

Main

Individuals with cancer who develop cachexia, a metabolic disorder characterized by morbid wasting of adipose tissue and skeletal muscle, suffer involuntary weight loss, poor responses to therapy and elevated mortality rates¹. While cachexia affects nearly half of the population with cancer, it is especially prevalent among those with pancreatic ductal adenocarcinoma (PDA), the most lethal pancreatic malignancy². Although the biological factors that induce cachexia remain poorly defined, recent studies indicate that adipose tissue wasting precedes muscle loss in individuals with cancer^3–5 and that muscle mass can be preserved in tumor-bearing animals by experimental blockage of adipose depletion^6,7. Thus, elucidating the factors that initiate adipose loss should reveal underlying mechanisms of cancer cachexia and provide new clinical strategies to forestall its development.

Mammals have both white adipose tissue (WAT), which stores energy as triglycerides, and brown adipose tissue (BAT), which can release energy as heat. WAT, the predominant type of fat tissue in adults, is composed mostly of white adipocytes that have a unilocular morphology typically marked by a single massive lipid droplet. With relatively few mitochondria, these cells have limited capacity for oxidative phosphorylation. In contrast, the brown adipocytes of BAT, which possess many small (multilocular) lipid droplets and abundant mitochondria, can robustly support oxidative phosphorylation. Under certain physiological conditions, such as cold exposure, white adipocytes can be transformed into ‘beige’ adipocytes that have BAT-like features, including a multilocular morphology and high mitochondrial content⁸. As a result of this ‘browning’ process, beige adipocytes gain a thermogenic potential that is readily induced by sympathetic tone. In particular, the norepinephrine released by parenchymal neurites can activate β3 adrenergic signaling in beige adipocytes, mobilizing their lipid stores for thermogenesis^9–12.

As the fat depots of adult humans are predominantly comprised of white adipocytes, WAT browning is often associated with metabolic health and the mitigation of obesity¹³. However, when dysregulated, sustained thermogenic remodeling can drive pathological energy expenditure and systemic wasting¹⁴. Indeed, WAT browning has been shown to be an early hallmark of cancer cachexia that accompanies adipose atrophy and precedes skeletal muscle wasting in both cancer mouse models^7,15–17 and patients^5,18,19. Here, we describe an adrenergic iron-dependent signaling axis (β3 adrenergic receptor (β3AR)–iron–methionine sulfoxide reductase A (MSRA)) that promotes WAT browning by preserving the proper redox state of select methionine residues in the adipocyte proteome. This axis is required for browning in both pathological (for example, cancer cachexia) and physiological (cold-induced thermogenesis) settings and its abrogation can alleviate cachexia and prolong the survival of tumor-bearing animals.

Methionine oxidation is altered during WAT browning in PDA-induced cachexia

Previous studies have shown that PDA-bearing mice develop a cachectic phenotype marked by profound loss of adipose and muscle tissue^20,21. Thus, to identify factors that induce adipose browning during cancer cachexia, we used the autochthonous KPC (Kras^G12D;p53^R172H;PdxCre)²² mouse model of PDA. As expected^20,21, PDA-bearing KPC mice (mPDA) with tumors of 5–7 mm diameter lost significant weight, with depletion of both adipose and lean mass, relative to age-matched nontumor wild-type (WT, mNT) controls (Fig. 1a). Adipose atrophy occurred to the same degree in male and female mice upon adjusting for body mass differences (Extended Data Fig. 1a).

Interestingly, WAT of mPDA mice expresses elevated levels of tyrosine hydroxylase (TH), a marker of adipose sympathetic innervation²³ (Fig. 1b,c). Global proteomic analysis of subcutaneous WAT (sWAT) (Supplementary Table 1) further revealed a marked upregulation of proteins involved in oxidative phosphorylation in both male (Fig. 1d) and female (Extended Data Fig. 1b) mPDA mice relative to mNT controls. A similar induction of oxidative phosphorylation occurred in nontumor animals treated with the β3AR agonist CL-316243, which mimics physiological induction of WAT browning (Extended Data Fig. 1b). Likewise, activated p38 kinase, a central regulator of thermogenesis²⁴, was induced in the sWAT of both PDA-bearing animals and CL-316243-treated nontumor animals (Fig. 1c). In addition, white adipocytes from mPDA mice contain smaller lipid droplets, exhibit a BAT-like multilocular morphology and show elevated expression of UCP1, a mitochondrial uncoupling protein that mediates thermogenesis⁹ (Fig. 1e,f and Extended Data Fig. 1c,d). In normal (that is, nontumor) mice, UCP1-mediated energy expenditure occurs in response to specific physiological stimuli, such as cold exposure⁸. Interestingly, mPDA mice maintained at ambient temperature generated greater diurnal heat within their sWAT depots relative to age-matched mNT controls (Fig. 1g). These mice also displayed an elevated diurnal energy expenditure (Extended Data Fig. 1e) and a disproportionate increase in energy expenditure per unit of body mass (Fig. 1h). This hypermetabolic state was not associated with changes in core body temperature (Fig. 1g), activity (Extended Data Fig. 1f), water intake (Extended Data Fig. 1g), food intake (Fig. 1i and Extended Data Fig. 1h), food grinding (Fig. 1j) or malabsorption (Extended Data Fig. 1i). Together, these findings indicate that early-stage PDA-bearing mice undergo hypermetabolic adipose remodeling associated with increased heat dissipation and negative energy balance.

As mitochondrial respiration generates reactive oxygen species (ROS)²⁵, we expected that the intracellular oxidation state of WAT would be increased in PDA-bearing mice relative to controls and adipose atrophy might reflect aberrant induction of ferroptosis. However, PDA WAT did not display characteristic indicators of lipid oxidation^26,27, such as specific depletion of phospholipids or triglycerides with polyunsaturated fatty acid (PUFA) tails (Extended Data Fig. 1j,k) or increased lipid peroxide adduction (that is, 4-hydroxynonenal (4-HNE) abundance) (Extended Data Fig. 1l,m). Thus, ferroptosis is an unlikely outcome of cachectic WAT browning and atrophy. We next surveyed the sWAT proteome for major oxidative protein modifications, namely, cysteine oxidation (loss of biotinylated iodoacetamide (BIAM) reactivity; Fig. 1k,l) and methionine oxidation (loss of oxaziridine-alkyne (Ox-Alk) reactivity; Fig. 1k,m and Extended Data Fig. 1n–p). Of these, only methionine reactivity was significantly perturbed in the tumor setting (Fig. 1m), suggesting that oxidation of proteinaceous methionines may contribute to adipose wasting during cancer cachexia. Interestingly, the sWAT proteome also displayed decreased Ox-Alk but not BIAM reactivity in normal mice subjected to acute cold exposure (Fig. 1n,o). This observation implies that global methionine oxidation is a common feature of adipose browning during both cancer cachexia and physiological thermogenesis.

MSRA induction occurs during both physiological and cachectic adipose browning

The sulfur atom of some proteinaceous methionines can act as a functional redox switch that toggles between its reduced and oxidized states^28,29. This toggling occurs because ROS can convert methionine residues to their oxidized form, methionine sulfoxide, which in turn can be enzymatically reduced by MSRA³⁰. Interestingly, this oxidoreductase was strongly upregulated in both the sWAT (Fig. 1p) and the epididymal WAT (eWAT) (Extended Data Fig. 1q) of tumor-bearing KPC mice. MSRA expression was also elevated, albeit more modestly, in the BAT compared to WAT depots of healthy mice (Extended Data Fig. 1r). In addition, MSRA could be robustly induced in the WAT of healthy mice by treatment with the β3AR-selective agonist CL-316243 (Fig. 1q and Extended Data Fig. 1s), suggesting that MSRA expression in adipocytes is also responsive to sympathetic stimulation.

To ascertain whether similar metabolic changes occur during human cachexia, we examined a panel of visceral adipose tissues procured from (1) nontumor control patients undergoing laparoscopic cholecystectomy (hNT) and (2) PDA patients with cachexia (>5% weight loss, hPDA) (Supplementary Table 2). The WAT adipocytes of PDA patients displayed smaller diameters (Extended Data Fig. 1t), a multilocular morphology (Fig. 1r), increased mitochondria, regionally increased UCP1 levels (Fig. 1r–t) and induction of mitochondrial TNAP (Fig. 1s,t), a key regulator of UCP1-independent thermogenesis³¹ in humans^32,33. Increased MSRA expression (Fig. 1u–x and Extended Data Fig. 1u) and activating phosphorylation of the lipolytic enzyme hormone-sensitive lipase (HSL) (Extended Data Fig. 1u) were also observed in WAT specimens from cachectic PDA patients. Thus, MSRA expression is upregulated in the WAT of PDA patients and PDA-bearing mice, along with a metabolic shift toward mitochondrial respiration that occurs during WAT browning.

Iron uptake stabilizes MSRA protein and induces adipose browning during cancer cachexia

The characteristic brown color of thermogenic adipocytes is derived from the high-density iron–porphyrin complexes of their mitochondrial electron transport chains³⁴. To meet the increased demand for oxidative metabolism in brown and beige adipocytes, β3AR signaling promotes iron uptake for synthesis of iron-bound mitochondrial enzymes³⁵. As this physiological influx of iron is facilitated by upregulation of the transferrin receptor (TfR1), genetic disruption of TfR1 impairs mitochondrial biogenesis and thermogenesis in brown and beige adipocytes³⁵. Accordingly, β3AR stimulation of cultured WAT cells from WT neonates led to an increase in intracellular labile iron, as visualized using the ferrous iron-sensitive fluorescent probe FerroOrange (Extended Data Fig. 2a). We also observed markedly increased TfR1 levels in the WAT of both mPDA mice (Extended Data Fig. 2b) and cachectic PDA patients (hPDA) (Extended Data Fig. 2c). Moreover, inductively coupled plasma mass spectrometry (ICP–MS) revealed elevated levels of total iron content in the adipose depots of tumor-bearing mice (Extended Data Fig. 2d) and correspondingly reduced levels in their sera (Extended Data Fig. 2e). These observations suggest that iron influx not only is essential for physiological thermogenesis³⁵ but also facilitates adipose browning during cancer cachexia.

To determine whether iron is required for MSRA induction during physiological browning, we coadministered healthy mice with CL-316243 and the iron chelator deferiprone (DFP). As expected³⁵, iron chelation mitigated the browning effects of CL-316243 at the histological level (Extended Data Fig. 2f). In addition, the robust induction of MSRA expression that occurred in these cells upon β3AR activation and coincided with both TfR1 and UCP1 upregulation was reversed by DFP treatment (Fig. 2a). Conversely, expansion of the labile iron pool by CRISPR-mediated depletion of the iron storage protein ferritin heavy chain (FTH1) allowed post-transcriptional MSRA induction to occur in WAT adipocytes even in the absence of adrenergic stimulation (Extended Data Fig. 2g–i). Similarly, iron–dextran infusion in WT mice effectively induced MSRA expression in WAT (Extended Data Fig. 2j–l). Of note, iron administration alone also led to a modest but significant reduction in the diameter of eWAT and sWAT lipid droplets (Extended Data Fig. 2m). These results suggest that iron can directly elicit MSRA upregulation and regulate the properties of white adipocytes.

In mammalian cells, iron homeostasis is partly maintained by the iron regulatory proteins IRP1 and IRP2, which coordinate expression of various iron metabolism genes post-transcriptionally by recognizing iron-responsive elements (IREs) within the 5′ and 3′ untranslated regions (UTRs) of their mRNA transcripts. Although iron-dependent induction of MSRA also occurs post-transcriptionally (Extended Data Fig. 2g–i), an in silico search³⁶ failed to identify candidate IREs in the UTR sequences of mouse or human MSRA. Therefore, we asked whether iron could upregulate post-translational MSRA expression by enhancing its protein stability. Indeed, pulse–chase experiments revealed that the half-life of MSRA is markedly prolonged when adipocyte cultures (Fig. 2b) and 293T cells (Extended Data Fig. 2n) were supplemented with iron in the form of ferric ammonium citrate (FAC). Iron-induced stabilization of MSRA is independent of ferroptotic responses, as it persists even when cells are cotreated with the ferroptosis inhibitor liproxstatin 1 (Extended Data Fig. 2o).

Iron promotes the formation of an active MSRA oligomer

To determine whether the stabilization of MSRA involves a direct physical interaction with iron, we first analyzed its metal-binding properties. Of note, the absorbance spectrum of purified recombinant mouse (Fig. 2c) and human (Extended Data Fig. 2p) MSRA was markedly shifted by exposure to iron but not to other metals, including Zn²⁺, Co²⁺, Mn²⁺ and Cu²⁺ (Extended Data Fig. 2q–t). Of note, the absorbance of iron-bound MSRA on the ultraviolet (UV)–visible spectrum is consistent with a μ-oxo-bridged diiron (Fe–O–Fe)³⁷ (Fig. 2c). Furthermore, using ICP–MS, we confirmed that iron content was substantially enriched in MSRA immunoprecipitants relative to other metals of equal or lower cellular abundance (Fig. 2d) and this enrichment was further increased in cells supplemented with FAC (Extended Data Fig. 2u). Together, these data suggested that MSRA is a metalloprotein with selective iron-binding capacity.

To ascertain how MSRA interacts with iron, we examined its amino acid sequence using the metal-ion-binding site prediction algorithm MIB2 (ref. ³⁸). This analysis suggested that residues E203 and H206, which are highly conserved in MSRA orthologs across the evolutionary spectrum (Fig. 2e), have iron-binding potential (Extended Data Fig. 3a). Indeed, the absorbance spectrum of MSRA-AXXA, a mutant lacking both conserved residues (EDYH→ADYA), is insensitive to iron, unlike that of WT MSRA (Extended Data Fig. 3b). Moreover, the protein expression levels of MSRA-AXXA but not MSRA mutants lacking the myristoylation site (G22A) or active-site cysteine (C72S), were significantly suppressed (Fig. 2f). The EXXH motif of MSRA bears some resemblance to the Fe-binding centers of several nonheme proteins that coordinate iron through the carboxylate and imidazole groups of their conserved glutamic acid and histidine residues, respectively³⁹. However, unlike MSRA, most members of the ferritin-like protein superfamily, including the β subunit of class I ribonucleotide reductases (RNRs), possess two conserved EXXH motifs that cooperate to form a diiron center containing two oxygen-bridged iron atoms^40,41. For example, the two EXXH motifs of RNR, albeit widely separated within its primary amino acid sequence (Extended Data Fig. 3a), converge in space to coordinate its diiron center (Fig. 2g)^40,41. Interestingly, the electrostatic environment surrounding the predicted iron-binding EXXH motif of MSRA bears a striking resemblance to a half-site of the RNR diiron center (Fig. 2g). This observation suggested that MSRA dimerization might allow for the assembly of a full diiron center, which in turn might enhance the stability and even the enzymatic activity of MSRA.

To determine whether MSRA can oligomerize in vivo, we initially used proximity-dependent biotinylation to evaluate interactions between MSRA protomers under normal (that is, iron-replete) cellular conditions. Therefore, an MSRA construct fused to an engineered biotin ligase (TurboID)⁴² was expressed ectopically and the streptavidin-bound fraction of cell lysates was subjected to MS. If MSRA exists primarily as a monomer, then the ratio of peptides derived from MSRA versus TurboID sequences should approximate 1:1 (Extended Data Fig. 3c). Instead, MSRA peptides were robustly enriched in the biotinylated (that is, spatially proximal) proteome relative to the global proteome, indicating that ectopic MSRA–TurboID polypeptides readily interact with endogenous MSRA in vivo (Extended Data Fig. 3d,e). Moreover, size-exclusion chromatography coupled to multiangle light scattering (SEC–MALS) showed that iron supplementation directly promotes in vitro formation of a higher-molecular-weight species consistent with the size of an MSRA dimer (Fig. 2h). Indeed, native PAGE analysis of cell lysates revealed two distinct in vivo MSRA species: a lower-molecular-weight form (monomer) and a higher-molecular-weight form (dimer). Of these, the latter is more labile but is stabilized by exogenous iron supplementation (Fig. 2i,j and Extended Data Fig. 3f,g) and persists following liproxstatin 1 cotreatment (Fig. 2k). Increasing iron concentration promoted dimer formation, whereas iron chelation with DFP decreased the abundance of both forms, with a more pronounced decrease in the dimer (Extended Data Fig. 3h,i). Importantly, both in vivo dimer formation in adipocytes (Fig. 2l) and iron-induced dimerization of recombinant MSRA in vitro (Fig. 2m) were abolished by individual (E203A, H206A) and combined (AXXA) substitutions in the EXXH motif.

To investigate how MSRA dimerizes at the molecular level, we first performed in silico modeling of the mouse apo MSRA crystal structure using AlphaFold3 (ref. ⁴³) and HADDOCK⁴⁴. Both algorithms generated a structurally similar dimeric interaction between two MSRA protomers (Extended Data Fig. 3j,k), with low-confidence scores that likely reflect the absence of iron. Consistent with our experimental evidence that iron directly promotes MSRA dimerization (Fig. 2h), the EXXH motifs from each protomer are spatially juxtaposed in the predicted dimer configuration in a manner that would coordinately bridge iron and potentially stabilize the MSRA dimer (Extended Data Fig. 3l). Strikingly, PropKa analysis⁴⁵ predicted that iron-mediated dimerization would also decrease the pKa of the catalytic residue C72 relative to an apo form MSRA monomer (Fig. 2n), thereby favoring its deprotonated (thiolate) state. As such, dimerization through EXXH–mediated iron coordination should enhance the nucleophilicity and catalytic efficiency of the active-site cysteine (Fig. 2n). Consistent with this prediction, iron supplementation increased the activity of recombinant WT MSRA in a dose-dependent manner (Fig. 2o). In contrast, the MSRA-E203A mutant showed a diminished response to iron (Fig. 2p), while MSRA-AXXA was entirely unresponsive (Extended Data Fig. 3m). Furthermore, dimeric MSRA exhibited significantly higher reductase activity than the monomeric form (Fig. 2q and Extended Data Fig. 3n–p). Together, these findings indicate that iron enhances the enzymatic activity of MSRA by promoting dimerization through its juxtaposed EXXH motifs.

Chemoproteomic analysis identifies MSRA substrates during cancer-induced browning

To identify methionine residues that serve as MSRA enzymatic substrates in adipose tissue, we surveyed the WAT proteome using Ox-Alk bioconjugation followed by MS to distinguish between reduced and oxidized methionines^29,46 (Fig. 3a). From this analysis, we identified and quantified 4,655 peptides containing reactive (that is, reduced) methionine, derived from 1,204 distinct proteins across all samples (Supplementary Table 3). Principal component analysis revealed a clear separation between the reactive methionine proteomes of tumor-bearing (mPDA) and control (mNT) mice (Fig. 3b), with mPDA samples showing a net increase in methionine oxidation—88 peptides significantly more oxidized (log₂FC (fold change) < −1, P < 0.05) versus 47 that were reduced (log₂FC > 1, P < 0.05) (Fig. 3c), consistent with our global oxidation analysis (Fig. 1j). As MSRA is upregulated in PDA WAT, we reasoned that its direct substrates would show decreased methionine oxidation in the tumor setting and increased oxidation in MsrA-null (knockout, KO) WAT. Applying these criteria and adjusting for expression differences (Fig. 3d), we identified 19 candidates with significantly reduced methionine oxidation in mPDA samples (log₂FC > 1, P < 0.05), seven of which are more oxidized in MsrA-null WAT (Fig. 3e). These findings suggest that MSRA can coordinate the pleiotropic program of adipose browning by regulating methionine redox states of multiple select protein substrates.

Among these candidate substrates, the catalytic subunit of protein kinase A (PKAC) stood out for its established role in β3AR-induced thermogenesis⁴⁷. To validate our proteomic screen, the PKAC oxidation status of independent WAT samples was compared between PDA-bearing versus nontumor individuals. In both human (Fig. 3f) and mouse (Fig. 3g) WAT specimens, we observed a consistent decrease in PKAC methionine oxidation in PDA-bearing individuals, while PKAC cysteine oxidation remained unchanged (Fig. 3f,g). In line with the elevated sympathetic tone of WAT in PDA-bearing mice (Fig. 1b,c), we also observed altered phosphorylation patterns of global PKA substrates in cachectic WAT from both tumor-bearing mice (Extended Data Fig. 4a) and humans (Extended Data Fig. 4b,c). Moreover, pharmacological inhibition of β3AR signaling in PDA mice suppressed PKA substrate phosphorylation (Extended Data Fig. 4d), abrogated sWAT-mediated heat dissipation (Extended Data Fig. 4e) without altering core body temperature (Extended Data Fig. 4f) and attenuated PDA-associated weight loss (Extended Data Fig. 4g) and the histological features of browning (Extended Data Fig. 4h). Together, these data identify PKAC as a candidate effector protein of MSRA and support a role for β3AR–PKA signaling in cancer-induced cachexia.

The sole redox-sensitive methionine residue of PKAC (M71) (Fig. 3h), which is highly conserved across evolution (Extended Data Fig. 4i), lies immediately adjacent to the ATP-binding lysine K72 on the β3 strand of the N-terminal lobe and within the PKAC catalytic cleft that coordinates ATP, Mg²⁺ and substrate binding (Extended Data Fig. 4j)⁴⁸. In its inactive (open) conformation—before ligand binding—the catalytic cleft of PKAC is accessible to both solvent and ligand but it adopts a closed (active) conformation upon ligand engagement (Extended Data Fig. 4k). Thus, we hypothesize that oxidation and reduction of M71 likely occur in the open state, before both substrate and ATP binding. Supporting this, solvent-accessible surface area (SASA) analysis⁴⁹ of published crystal structures revealed that M71 is solvent exposed and readily oxidizable in the open-state apo form of PKAC, as well as in its holoenzyme complex with the PKA regulatory subunit (Extended Data Fig. 4l). By contrast, in the closed conformation, M71 is completely buried within a hydrophobic pocket (Extended Data Fig. 4m) and no longer solvent accessible (Fig. 3i and Extended Data Fig. 4l,m), suggesting that M71 must be in a reduced state for PKAC to adopt its active conformation. Consistent with this model, DynaMut2 (ref. ⁵⁰) analysis of protein stability following the M71E substitution—a structural and electrostatic mimic of oxidized methionine—showed decreased protein stability across all available structures, supporting the idea that M71 oxidation destabilizes PKAC (Fig. 3i). Moreover, AlphaFold-based modeling predicts weaker interactions between the P-loop and ATP in mutant PKAC-M71E relative to WT PKAC; in particular, the M71E substitution in the WT structure decreased protein stability by −1.82 kcal mol⁻¹, while reversing the substitution (E71M) increased stability by +2.9 kcal mol⁻¹ (Fig. 3j). To validate these in silico data, we next probed the ATP–protein interaction in vitro using a fluorescent ATP analog (TNP-ATP). While a kinase-dead PKAC mutant (K72H) known to lack ATP-binding capacity⁵¹ displayed no detectable TNP-ATP binding, the PKAC-M71E mutant retained only ~50% of the ATP-binding capacity of WT PKAC (Fig. 3k) and required a fourfold higher ATP concentration to achieve the same level of binding as its WT counterpart (Fig. 3l). Furthermore, cellular expression of either PKAC WT or its redox-insensitive M71L mutant promoted global PKA substrate phosphorylation, while the M71E and K72H mutants failed to do so (Fig. 3m). Collectively, these findings suggest that M71 can serve as a redox-sensitive site to modulate PKAC activity and its reduction by MSRA is essential to maintain full PKA activity in β3AR-stimulated adipocytes.

MSRA is required for β3AR-induced signaling in adipocytes

To assess the functional relationship between MSRA and PKA in adipocytes, we first examined adipose tissues from MsrA^−/− mice, which mature to adulthood with no apparent phenotype in the absence of oxidative stress⁵². The BAT and WAT of MsrA-null animals were histologically indistinguishable from those of WT mice (Extended Data Fig. 5a) and BAT-specific expression of UCP1 was unaffected by MsrA nullizygosity (Extended Data Fig. 5b). These results indicate that MsrA is dispensable for adipose tissue development in vivo. Likewise, ex vivo differentiation of cultured preadipocytes from neonatal adipose tissues⁵³ did not require MsrA (Extended Data Fig. 5c).

As expected, PKAC was more oxidized in MsrA-null adipocyte cultures (Fig. 4a) and several lines of evidence indicate that its oxidation compromises PKA signaling output. First, while the pan-βAR agonist isoproterenol induced PKA substrate phosphorylation and UCP1 expression in WT WAT (Fig. 4b–d) and BAT (Fig. 4e,f and Extended Data Fig. 5d) cultures, these effects were impaired in MsrA-null cells, despite comparable PKA expression (Extended Data Fig. 5e). Second, MsrA loss dampened β3AR–PKA-mediated phosphorylation of HSL (Fig. 4g and Extended Data Fig. 5f) similar to pharmacological PKA inhibition (Extended Data Fig. 5f). Of note, MSRA loss did not impair βAR induction of cAMP levels, indicating that MSRA functions downstream of cAMP production, at the level of PKA catalytic activity, within this adipose signaling pathway (Extended Data Fig. 5g). Third, rescue experiments in MsrA^−/− adipocytes confirmed that β3AR–PKA signaling is dependent on MSRA, as well as its iron-binding capacity (Extended Data Fig. 5h) and catalytic activity (Fig. 4h,i).

Further supporting its physiological role in adipocyte β3AR–PKA signaling, MSRA expression and PKA substrate phosphorylation were both induced in WAT by acute cold exposure (Fig. 4j). PKAC M71 was also more oxidized in the adipose tissue of MsrA-null mice compared to WT mice maintained at ambient temperature (22 °C) and this effect was further augmented upon cold exposure (4 °C) (Fig. 4k). Accordingly, the ability of MsrA^−/− mice to maintain dorsal and core body temperature in response to cold was significantly impaired (Fig. 4l,m and Extended Data Fig. 5i). Indeed, local restoration of MSRA expression in MsrA^−/− mice using adeno-associated virus (AAV) delivered to the inguinal fat pad restored cold-induced sWAT local heat generation (Extended Data Fig. 5j) and PKA activity (Fig. 4n,o). In contrast, catalytically inactive (C72S) and iron domain (E203A) mutants of MSRA did not. Thus, the enzymatic activity of iron-bound MSRA is required for β3AR-mediated induction of PKA enzymatic activity and the thermogenic program in white adipocytes, establishing a causal relationship linking MSRA activity, PKAC redox state, PKA enzymatic function and thermogenesis.

MSRA is essential for PDA-associated adipose atrophy and cachexia

To determine whether the thermogenic functions of MSRA are also required for adipose remodeling in response to tumor development, we interbred KPC and MsrA^−/− mice to generate KPCM compound mutant (Kras^G12D;p53^R172H;PdxCre;MsrA^−/−) animals. No apparent effects of MsrA loss were observed on primary tumor burden (Extended Data Fig. 6a), the macroscopic or histological appearance of other organs (Extended Data Fig. 6b), food intake (Extended Data Fig. 6c,d), sWAT sympathetic tone (Extended Data Fig. 6e) or sWAT β3AR expression (Extended Data Fig. 6f). However, comparing KPC and KPCM mice with similar tumor burdens revealed that body mass attrition was significantly mitigated by the absence of MsrA, both during early (tumor diameter: 5–8 mm; Extended Data Fig. 6g) and late (tumor diameter: 9–10 mm; Fig. 5a and Extended Data Fig. 6h) stages of disease. MsrA deletion also preserved adipose tissue volume (Fig. 5b) and lipid droplet diameter in mice bearing early-stage tumors (Fig. 5c,d). Meanwhile, loss of MSRA in nontumor mice had no impact on body mass (Extended Data Fig. 6g) or adipocyte diameter (Extended Data Fig. 6i). Interestingly, although MSRA expression was not induced in the skeletal muscles of KPC mice (Extended Data Fig. 6j), the absence of MsrA did preserve lean mass (Extended Data Fig. 6k) and skeletal myocyte diameter (Extended Data Fig. 6l) in KPCM mice, consistent with the notion that adipose atrophy precedes and possibly facilitates muscle wasting during cancer cachexia^7,15,16,18.

To evaluate whether MSRA has a cancer-cell-extrinsic role in PDA-associated cachexia²⁹, we tested whether PDA cells induce adipose and muscle atrophy to differing extents when orthotopically transplanted into wild-type (WT-host) or MsrA-null (KO-host) mice. As expected, WT-host tumor recipients progressively lost body mass during tumor progression (Extended Data Fig. 7a), with weight loss correlating inversely with overall survival (Extended Data Fig. 7b). Consistent with prior reports that adipose atrophy is an early event in cancer cachexia^14,54,55, loss of adipose tissue coincided with tumor growth and preceded detectable changes in lean mass (Extended Data Fig. 7c). At 4–5 weeks after transplantation—when adipose depletion is most robust (Extended Data Fig. 7c)—the sWAT transcriptome (Supplementary Table 4) was enriched for oxidative phosphorylation genes (Extended Data Fig. 7d) and thermogenic markers such as Elovl3, Ucp1 and Dio2 (Extended Data Fig. 7e). Using this timeline, we then compared cachexia development in age-matched MsrA^+/+ (WT-host) and MsrA^−/− (KO-host) males following orthotopic PDA cell injection. As in the autochthonous KPC model (Fig. 5a–d), MsrA deletion in the host attenuated both adipose (Fig. 5e) and muscle (Extended Data Fig. 7f) wasting 5 weeks after tumor inoculation, without significantly altering primary tumor burden (Extended Data Fig. 7g) or metastatic incidence (Extended Data Fig. 7h). Histologically, MsrA loss also reduced the multilocular transformation and wasting of WAT adipocytes in tumor-bearing mice (Fig. 5f,g).

At the adipocyte level, PKAC methionine reactivity (Fig. 5h), markers of lipolysis (Fig. 5i), local sWAT temperature (Fig. 5j) and UCP1 expression (Fig. 5k–n) were all decreased in both autochthonous (KPC) and orthotopic transplant PDA models by MsrA loss. Consistent with these thermogenic defects, the hypermetabolic phenotype of KPC mice was markedly attenuated upon MsrA deletion (Fig. 5o). These defects were not accompanied by compensatory upregulation of methionine sulfoxide reductase B family members (MSRB2 and MSRB3) (Extended Data Fig. 7i,j) or by changes in sympathetic innervation (Extended Data Fig. 7k) or iron content (Extended Data Fig. 7l), indicating a specific requirement for MSRA.

Although proinflammatory cytokines⁵⁶ and insulin⁵⁷ have been implicated in cancer cachexia, we did not observe consistent MSRA-dependent changes in circulating cytokines or adipokines levels in either autochthonous or orthotopic PDA models, regardless of tumor status (Extended Data Fig. 8a,b and Supplementary Table 5). Glucose tolerance (Extended Data Fig. 8c–f), fasting glucose (Extended Data Fig. 8g), insulin levels (Extended Data Fig. 8h), total blood cell counts (Extended Data Fig. 8i,j) and tumor-infiltrating and peripheral immune populations (Extended Data Fig. 9a–c) were also comparable in WT and KO mice in either PDA model.

As adipose tissue loss correlates with reduced survival PDA patients^58,59, we next asked whether MSRA-dependent adipose remodeling impacts survival in experimental settings. Notably, whereas KPC mice lost body mass in the days preceding death, this effect was attenuated in their MsrA-null counterparts (Fig. 5p). Moreover, the overall survival of tumor-bearing mice was significantly prolonged by MsrA loss in both autochthonous (Fig. 5q) and orthotopic (Fig. 5r) PDA models. This survival benefit occurred independently of primary tumor burden (Extended Data Fig. 10a) or metastatic incidence (Extended Data Fig. 10b). Furthermore, using the AAV-flex system for adipose-specific rescue, expression of WT MSRA—but not its iron-binding E203A mutant—completely reversed the survival advantage conferred by whole-body MsrA deletion (Fig. 5s), independent of primary tumor burden (Extended Data Fig. 10c). Moreover, tumor-bearing MsrA-null mice expressing AAV-encoded WT MSRA—but not the E203A mutant—showed progressive weight loss leading up to death (Fig. 5t), mirroring the pattern observed in KPC mice. Adipocyte diameters similarly declined only in the presence of WT MSRA (Extended Data Fig. 10d,e). Together, these results identify MSRA as an iron-dependent regulator of adipose remodeling required for the initiation of cachexia during pancreatic tumorigenesis (Extended Data Fig. 10f).

Discussion

Although effective therapies to alleviate the lethal impact of cancer cachexia are still not available¹, multiple studies show that adipose depletion precedes and may be required for muscle wasting^5,7,15,16,18. Here, we identified an iron-dependent process that promotes adipose depletion by governing the redox state of proteinaceous methionine residues. This process entails sympathetic induction of β3AR signaling, labile iron accumulation, iron-dependent activation of MSRA and MSRA-catalyzed reduction of select protein substrates. Moreover, this β3AR–iron–MSRA axis is required not only for cachectic adipose browning but also for physiological browning during cold-induced thermogenesis. Importantly, disruption of the β3AR–iron–MSRA pathway in PDA-bearing mice mitigates the morbid outcomes of cancer cachexia, including adipose depletion, muscle wasting, weight loss and reduced survival.

Physiological adipose browning adipocytes is dependent on expansion of the labile iron pool to support the increased mitochondrial capacity³⁵. However, elevated iron content can promote ROS production through the Fenton reaction, thereby rendering adipose proteins more susceptible to oxidative modifications. Here, we showed that iron levels also rise during cachectic adipose browning and the methionine proteome of WAT is more oxidized globally in PDA-bearing animals. Interestingly, MSRA expression is increased in an iron-dependent manner specifically in the browning WAT cells of tumor-bearing mice and the induction of both iron and MSRA are required for WAT browning.

Our results also showed that iron can directly bind a phylogenetically conserved EXXH motif of MSRA in a manner reminiscent of the ferritin superfamily proteins³⁹. However, unlike the ferritin-like polypeptides, each of which possesses two EXXH motifs that coordinate iron intramolecularly, MSRA exploits its single EXXH motif to allow iron-mediated dimerization of two MSRA protomers. Dimerization in turn enables MSRA to function optimally in the iron-rich environment of thermogenic WAT cells by enhancing both its stability and its full catalytic activity. This model may explain why MSRA is largely dispensable for normal development⁵² but essential within the iron-rich oxidative intracellular environment of browning adipocytes. MSRA had often been viewed as a constitutive housekeeping enzyme that broadly maintains redox homeostasis, with overt phenotypes emerging only when mice are challenged with strongly oxidizing agents such as paraquat⁶⁰. Our results revise this view by showing that MSRA is uniquely required in defined physiological and pathological contexts, such as cold-induced thermogenesis and cancer cachexia.

As MSRA is required for cancer cachexia, its relevant substrates should have redox-sensitive methionines that are preferentially reduced in the adipose tissues of tumor-bearing individuals. Our unbiased chemoproteomic screen identified seven proteins, including PKAC, with these properties. Thus, we propose that MSRA helps to orchestrate adipose remodeling by coordinately modulating the functions of multiple substrates. The biological impact of methionine oxidation on each of these substrates will likely vary depending on its particular structural and biochemical properties. Given the known role of PKA in β3AR-stimulated thermogenesis, we examined how methionine oxidation of PKAC affects its enzymatic activity and its impact on adipose browning. Structurally, cAMP activates PKA signaling by binding the PKAR regulatory subunit, which allows the PKAC subunit to freely equilibrate between its open (inactive) and closed (active) conformations⁴⁸. Interestingly, the sole MSRA-regulated residue of PKAC is a highly conserved methionine (M71) that lies adjacent to an ATP-binding residue (K72) within the PKAC catalytic cleft. Our combined in silico, in vitro and in vivo analyses argue that oxidation of PKA diminishes ATP binding, whereas its reduction by MSRA restores full kinase activity. These data imply that PKA activation is, at least during β3AR-induced adipose browning, a two-step process involving cAMP-mediated release of PKAC followed by MSRA-mediated preservation of its reduced redox state.

Understanding the differences between physiological browning (a reversible process associated with metabolic health^8,14) and cachectic browning (irreversible even in the face of increased caloric intake¹) should allow for the development of nontoxic therapies for cancer cachexia. As adipose atrophy preceded skeletal muscle wasting in both genetically engineered and transplant PDA models, our data support the notion that WAT browning is an early and possibly initiating event in the pathogenesis of cancer cachexia^7,15,16,18. Thus, as an essential mediator of adipose atrophy, the β3AR–iron–MSRA axis is a promising target for therapeutic mitigation of cancer cachexia.

Methods

Supplementary information

Source data

Author contributions

J.S.N. and I.I.C.C. conceptualized the study. J.S.N., S.S.A., M.S.D. and P.A. performed the animal and biochemical experiments and analyzed the data. A.V.-C. performed the immunostaining, imaging and analysis of WAT specimens. Y. Shin performed the in silico structural analysis of MSRA and PKA. J.L., M.S.D., A.C. and S.I. measured and analyzed the myocyte and adipocyte diameters. N.A.Y. and S.D. performed and analyzed the metabolic experiments. F.Z. performed the lipidomic analyses. H.C. and L.B. performed the FPLC analyses. Y. Sun performed the MRI analyses. G.K. performed the immunoblot analysis for MSRA expression upon iron–dextran administration. K.S. performed and analyzed the ICP–MS. J.M.G., S.P. and Y.C. performed the bioinformatic and biostatistical analyses. T.C.C., P.G., M.A.H., K.S. and M.D.K. collected the adipose tissues from participants. K.A.K. recruited and obtained consent from the participants with rapid autopsy. B.J.S. performed the pathological scoring of specimens and provided detailed clinical information. A.W.F., S.D. and L.M.Z. assisted with data interpretation. J.S.N. and I.I.C.C. interpreted the results and wrote the paper. All authors participated in the discussion and finalization of the paper.

Peer review

Funding

This work was supported by the National Institute of Health (NIH; R01-CA240654, 1R01CA267870 and 1R01CA273023 to I.I.C.C., P30ES009089 to K.S. and R50CA210240 to P.G.), Department of Defense (HT9425-25-1-0838 to I.I.C.C.), Pershing Square Sohn Research Alliance (to I.I.C.C.), Irma Hirschl Trust (to I.I.C.C.) and Mark Foundation (to I.I.C.C.). Molecular analyses performed in the L.E.B. laboratory were supported by NIH grant R35GM124633 to L.E.B. and NIH training grant T32GM141882. The UNMC RAP for Pancreas is supported by SPORE in Pancreatic Cancer (P50CA127297), the Pancreatic Cancer Detection Consortium (U01CA210240) and a National Cancer Institute Cancer Center Support Grant (P30CA36727). The Herbert Irving Comprehensive Cancer Center at Columbia University is supported by P30-CA13696 and the Columbia University Flow Cytometry Core is supported by P30CA36727.

Data availability

RNA sequencing data were deposited to the Gene Expression Omnibus under accession code GSE274804. Proteomics data were deposited to the ProteomeXchange Consortium through the PRIDE repository with the dataset identifiers PXD054899, PXD063328 and PXD064378 for expression proteomics, chemoproteomics and proximity ligation (TurboID) proteomics, respectively. All data supporting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.