Abstract

   Metabolic dysfunction–associated steatohepatitis (MASH) is a
   progressive disease driven by obesity-related hepatic inflammation and
   oxidative stress. Recently, cysteine persulfidation (PSSH), a
   protective post-translational modification by hydrogen sulfide (H[2]S),
   was established to play a role in redox regulation. Despite the role of
   the liver in H[2]S metabolism, the function of PSSH in MASH remains
   underexplored. We demonstrated that H[2]S-producing enzymes are
   downregulated in both human and mouse livers with steatosis and
   fibrosis, resulting in a decline in global PSSH levels. Dimedone-switch
   mass spectrometry in dietary mouse models of distinct
   obesity-associated liver disease stages revealed dysregulated PSSH on
   specific proteins. Surprisingly, increased hepatic PSSH levels of
   protein tyrosine phosphatases and redox regulators were found in
   advanced disease stages, suggesting a targeted adaptive response to
   oxidative stress. Overall, our findings demonstrated that impaired
   H[2]S production disrupts protective PSSH networks in MASH. However,
   selective PSSH preservation on redox-sensitive proteins may represent a
   compensatory mechanism, underscoring the therapeutic potential of
   persulfidation in restoring redox homeostasis during obesity-associated
   chronic liver disease.

1. Introduction

   Metabolic dysfunction–associated steatotic liver disease (MASLD) is one
   of the most prevalent chronic liver conditions globally, closely linked
   to the obesity epidemic [[57]1,[58]2]. Approximately 80 % of MASLD
   cases are associated with high body mass index (BMI) [[59]3]. MASLD
   progresses from steatosis, characterized by excessive fat accumulation
   in hepatocytes, to metabolic dysfunction–associated steatohepatitis
   (MASH) a more severe necroinflammatory state, that can lead to fibrosis
   and potentially cirrhosis or hepatocellular carcinoma (HCC) [[60]4].

   In obesity, excess reactive oxygen species (ROS), such as hydrogen
   peroxide (H[2]O[2]) and superoxide radicals (O[2]^•^−), accumulate in
   the liver and contribute to MASH progression through several mechanisms
   [[61][5], [62][6], [63][7]]. A key process involves the oxidative
   post-translational modification of protein cysteines by H[2]O[2],
   leading to S-sulfenylation (Cys-SOH) and, in severe cases,
   overoxidation to sulfinic and sulfonic acids (Cys-SO[2-3]H) [[64][6],
   [65][7], [66][8]].

   Recent studies have identified hydrogen sulfide (H[2]S), a key gaseous
   signaling molecule that regulates numerous physiological processes
   [[67]9,[68]10], as having protective roles against protein
   overoxidation and oxidative stress [[69]11,[70]12]. H[2]S is primarily
   produced enzymatically via the transsulfuration pathway, through
   cystathionine-β-synthase (CBS) and cystathionine-γ-lyase (CSE/CGL), or
   3-mercaptopyruvate sulfurtransferase (MPST/3-MST) ([71]Fig. S1A)
   [[72]13,[73]14]. H[2]S production is tissue-dependent and can be
   influenced by enzyme expression, physiological state, and co-factors.
   Notably, high-fat diets and obesity impair H[2]S production, while
   dietary restriction has the opposite effect [[74][15], [75][16],
   [76][17]]. Remarkably, mice lacking both CSE and MPST develop
   MASLD-like characteristics even without obesity, a phenotype mitigated
   by H[2]S donor treatment [[77]18]. These findings indicate that H[2]S
   plays a critical role in safeguarding the liver against MASLD onset.
   However, the mechanisms by which H[2]S levels evolve during MASLD
   progression and whether H[2]S supplementation can reverse later stages
   of the disease remain unknown.

   In general, H[2]S exerts many of its biological effects by reacting
   with oxidized protein thiols (either sulfenic acids or a disulfides),
   with sulfenic acids being kinetically favored to form protein
   persulfides (PSSH) [[78]10,[79]19]. Thiols can also be persulfidated
   directly via polysulfides or cysteine, catalyzed by enzymes like MPST
   or cysteinyl-tRNA synthase (CARS) [[80]20,[81]21], the latter even
   acting co-translationally. This modification protects proteins from
   irreversible oxidation: the more nucleophilic (due to the α-effect)
   outer cysteine is preferentially oxidized, allowing reduction to
   restore the original thiol [[82]10,[83]22]. In rodents, 10–25 % of
   liver proteins are estimated to be persulfidated [[84]23,[85]24],
   indicating its widespread regulatory role.

   The liver is a critical organ for regulating H[2]S levels
   [[86]25,[87]26]. Hepatocytes, which are essential for functions such as
   detoxification, iron storage, fatty acid metabolism, and protein
   synthesis, often operate under conditions of elevated ROS. We
   hypothesize that hepatocytes use liver-derived H[2]S as an intrinsic
   defense against oxidative damage, particularly in redox-sensitive
   proteins like protein tyrosine phosphatases (PTPs) [[88]27,[89]28]. A
   decline in H[2]S impairs PSSH-mediated protection against cysteine
   overoxidation and disrupts broader persulfidation-dependent regulatory
   networks, contributing to the onset and progression of MASLD.

   Here, we assessed H[2]S production and PSSH levels across several
   murine models representing different stages of MASLD. Using
   dimedone-switch mass spectrometry, we generated the first comprehensive
   liver persulfidome profile across distinct stages of MASLD progression
   under metabolic stress. This dataset revealed protein-specific patterns
   of PSSH loss or retention, particularly in pathways related to redox
   balance, metabolism, and inflammation. We identified selective PSSH
   remodeling as a hallmark of MASLD progression and offer a valuable
   resource for future research into redox signaling in liver disease.

2. Results

2.1. H[2]S-producing enzymes are downregulated in MASLD

   To determine whether H[2]S metabolism is altered in the context of
   MASLD, we compared the total proteome of human liver biopsies at
   different MASLD stages, from steatosis (fatty liver) to MASH/cirrhosis,
   using liquid chromatography-tandem mass spectrometry (LC-MS/MS)
   ([90]Fig. 1A and [91]Table S1). The proteomic analysis revealed
   significant downregulation of H[2]S-producing enzymes, CBS and MPST in
   MASH patient livers ([92]Fig. 1A). Interestingly, sulfide quinone
   oxidoreductase (SQOR), a key enzyme involved in H[2]S clearance through
   sulfide oxidation and, was also upregulated in MASH ([93]Fig. S1B). In
   addition, given the well-established connection between MASH and
   oxidative stress, we examined the expression levels of several redox
   regulators, and found that several, including peroxiredoxin 1, 2, and 6
   (PRDX1/2/6), and thioredoxin 2 (TXN2), were downregulated in samples
   from MASH patients ([94]Fig. S1B).

Fig. 1.

   [95]Fig. 1
   [96]Open in a new tab

   H[2]S-producing enzymes are downregulated in metabolic
   dysfunction–associated steatohepatitis (MASH) (A) Heatmap of protein
   expression levels of H[2]S-producing enzymes and redox regulators in
   liver biopsies from patients with healthy liver (H, n = 3),
   metabolic-associated steatotic liver (S, n = 4), and
   metabolic-associated steatohepatitis (M, n = 4). The proteins were
   identified and analyzed by mass spectrometry.

   (B) Expression of CTH (CSE), CBS and MPST across different clusters in
   human livers. The size of each circle represents the percentage of
   cells expressing the gene, while the colour of circles indicates the
   average expression level of the gene. Data obtained from human liver
   atlas.

   (C) mRNA expression levels of CTH (CSE), CBS, and MPST in human livers
   from [97]GSE126848 from patients with healthy liver (n = 14), obesity
   (n = 12), metabolic-associated steatotic liver (MASL, n = 15), and
   metabolic-associated steatohepatitis (MASH, n = 16).

   (D) mRNA expression levels of CTH (CSE), CBS, and MPST in human livers
   from [98]GSE164760 representing different stages of MASLD. Samples from
   patients with healthy liver (n = 6), MASH (n = 74), cirrhosis (n = 8),
   peritumor (n = 29), and tumor (n = 53).

   (E) Heatmap of protein expression levels of H[2]S-producing enzymes and
   redox regulators in the livers from mice fed with different diets
   (n = 4–5). Samples were analyzed by mass spectrometry.

   (F) Western blotting to validate protein expression levels of
   H[2]S-producing enzymes in the livers of mice fed different diets. The
   data are mean ± SD (n = 4–5). ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001 and
   ∗∗∗∗P < 0.0001.

   To confirm the downregulation of H[2]S-producing enzymes, we analyzed
   human MASLD transcriptomic datasets ([99]GSE164760 and [100]GSE126848).
   This revealed a consistent decrease in mRNA expression of CBS, CSE
   (gene name CTH), and MPST during the transition from healthy to
   advanced disease stages ([101]Fig. 1C and D), which is consistent with
   reduced protein levels of H[2]S-producing enzymes observed in advanced
   disease stages, including MASH/fibrotic and cirrhotic livers.

   The liver is composed of diverse cell types, each with distinct
   physiological functions. To determine which cell type produces H[2]S,
   we analyzed publicly available human liver single-cell RNA seq-data
   ([102]GSE192742). Our analysis revealed that the main H[2]S-producing
   enzymes, CSE (CTH) and CBS, are predominantly expressed in hepatocytes,
   the dominant cell type making up ∼80 % of liver volume and responsible
   for most core liver functions ([103]Fig. 1B) [[104]29].

   To model human MASLD and validate our findings, we used diet-induced
   mouse models to recapitulate the human MASLD characteristics. C57BL/6N
   mice were fed a control diet, a high-fat diet (HFD),
   high-fat/high-fructose/high-cholesterol diet (HFHFHCD), or
   choline-deficient high-fat diet (CDHFD) for 12 weeks. HFD and HFHFHCD
   diets induced steatosis, whereas CDHFD diet led to steatohepatitis with
   fibrosis and inflammation, closely mimicking human MASH ([105]Fig. S2A
   and S2B). Consistent with our findings in human livers, livers from
   CDHFD-fed mice showed marked downregulation of H[2]S-producing enzymes
   ([106]Fig. 1E and F), while expression of H[2]S/persulfide clearing
   enzymes SQR remained unchanged ([107]Fig. S3A). Notably, mitochondrial
   MPST (lower band) was selectively decreased, while cytosolic MPST was
   unaffected [[108]14,[109]30]. By contrast, MPST and CSE were modestly
   upregulated in HFD-fed mice with simple steatosis, potentially
   reflecting a compensatory response to fat-induced oxidative stress
   ([110]Fig. 1E and F) [[111]31,[112]32]. These findings indicate that
   impaired H[2]S and sulfide metabolism contributes to MASLD progression,
   particularly in steatohepatitis and fibrosis.

2.2. H[2]S production capacity and protein persulfidation are reduced in
steatohepatitis

   We next verified whether the observed downregulation of H[2]S-producing
   enzymes translates to lower sulfide production to a degree that
   correlates with the MASLD stages. Given the challenges of measuring
   absolute H[2]S concentrations in frozen tissues—due to its low
   abundance and transient nature—we focused on functional output,
   evaluating the hepatic H[2]S production capacity rather than measuring
   absolute H[2]S levels. Using the lead acetate assay in CDHFD-fed mouse
   livers [[113]15,[114]33], where enzyme downregulation was most
   pronounced, we observed a significant decrease in H[2]S production
   capacity ([115]Fig. 2A and [116]Fig. S3D).

Fig. 2.

   [117]Fig. 2
   [118]Open in a new tab

   Protein persulfidation is downregulated in mouse liver with
   steatohepatitis

   (A) H[2]S production capacity in liver tissue from control and
   CDHFD-fed mice (100 μg of proteins). H[2]S production was measured
   using lead acetate plate assay with the substrate after 2 h of
   incubation. The value is the mean intensity measured by ImageJ. Each
   dot represents an individual mouse liver sample. The data are mean ± SD
   (n = 5–7).

   (B) PSSH levels in liver tissues of mice fed with control and CDHFD.
   The data are mean ± SD (n = 4)

   (C) PSSH levels and H[2]S-producing enzymes in isolated primary mouse
   hepatocytes treated with TGF- β1 (5 ng/ml) for 24 h. The data are
   mean ± SD (n = 5).

   (D) H[2]S-producing enzymes and albumin expression levels in isolated
   primary mouse hepatocytes cultured for up to 48 h. The data are
   mean ± SD (n = 4).

   (E) H[2]S production capacity in isolated primary mouse hepatocytes
   culture for up to 48 h H[2]S production from lysate (100 μg) was
   measured by lead acetate plate assay with the substrate after 3 h
   incubation. The data are mean ± SD (n = 3)

   (F) PSSH levels in isolated primary mouse hepatocytes cultured for up
   to 48 h. The data are mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01, and
   ∗∗∗P < 0.001.

   Although these findings are significant, it is important to emphasize
   that H[2]S metabolism in cells is highly dynamic and non-linear. The
   process involves not only its production and oxidation to the
   excretable sulfate within the mitochondria, but also its transfer and
   storage, such as in the form of per- and polysulfides
   [[119]34,[120]35]. Because PSSH is one of the main mechanisms by which
   H[2]S exerts its biological effects [[121]19], we used the
   dimedone-switch method to detect total PSSH [[122]11]. We observed a
   marked reduction in total PSSH levels in CDHFD-fed mice ([123]Fig. 2B),
   consistent with reduced H[2]S-producing enzyme expression ([124]Fig. 1E
   and F). By contrast, livers with only steatosis showed no clear
   difference in total PSSH levels compared to control livers, as
   indicated by gel analysis ([125]Fig. S3E), in agreement with the lack
   of changes in H[2]S-producing enzyme expression levels ([126]Fig. 1E).

   We next sought to identify the factors leading to the observed decrease
   in H[2]S generation and, by extension, PSSH levels. MASH is
   characterized by advanced liver injury and fibrosis ([127]Figs. S2B and
   S3F), which led us to hypothesize that fibrogenesis may be one of the
   contributing factors. To mimic this condition, we treated freshly
   isolated primary hepatocytes with TGF-β1, a key fibrosis-related growth
   factor. This resulted in reduced expression of CSE, CBS as well as PSSH
   levels ([128]Fig. 2C). By contrast, other pro-inflammatory cytokines
   and insulin had no effect ([129]Fig. S4A–D). This suggests that liver
   fibrosis contributes to the reduction in H[2]S-producing enzyme levels
   during disease progression ([130]Fig. 1E).

   The reduction induced by TGF-β1 ex vivo was, however, less pronounced
   than that observed in MASH, indicating that additional mechanisms may
   contribute to the downregulation of H[2]S-producing enzymes in vivo. To
   explore other possibilities, we examined the role of hepatocyte
   de-differentiation a hallmark of MASLD pathogenesis. In MASH, liver
   fibrosis and cirrhosis disrupt the hepatic architecture, impair liver
   function, and result in the loss of hepatocyte identity
   [[131]36,[132]37]. This de-differentiation process can be modeled by
   prolonged culture of primary mouse hepatocytes, which gradually lose
   their specialized functions [[133]38,[134]39]. In our experimental
   setup, we observed this transition as the hepatocytes adopted a
   spindle-shaped, mesenchymal/fibroblast-like morphology ([135]Fig. S4E),
   along with a reduction in the expression levels of the master
   hepatocyte transcription factor HNF4α ([136]Fig. S4F). A similar
   decrease was also observed in CDHFD-fed mice ([137]Fig. S4G).

   This morphological transition was accompanied by a decrease in the
   expression of CSE, CBS, and albumin (a marker of hepatocyte function),
   along with a progressive reduction in H[2]S production capacity and
   total PSSH levels ([138]Fig. 2D–F). These changes closely mirror the
   de-differentiation and reduced H[2]S signaling observed in MASH.
   Together, these findings reveal a strong negative correlation between
   MASH progression and H[2]S-mediated PSSH in hepatocytes, possibly
   caused by fibrosis and de-differentiation.

2.3. Selective remodeling of the hepatic persulfidome in MASLD reveals
adaptive redox regulation

   To investigate changes in persulfidation patterns during MASLD
   progression in more detail, we performed liver persulfidome analysis
   using dimedone-switch PSSH in mice fed control, HFD, HFHFHCD, and CDHFD
   diets. To account for the overall protein abundance across dietary
   groups, persulfidome data were normalized to total proteome levels
   ([139]Fig. 3A and B).

Fig. 3.

   [140]Fig. 3
   [141]Open in a new tab

   Mapping the protein persulfidation landscape of MASH

   (A) Volcano plot illustrating the changes in total proteome and
   persulfidome profiles in CDHFD-/control-fed mice.

   (B) Venn Diagram of the proteins identified in the total proteome and
   persulfidome.

   (C) Fold change of persulfidated proteins in CDHFD-fed mice compared
   with control diet-fed mice after normalization. The fold change less
   than 0.76 (marked as blue) and more than 1.3 (marked as red) are
   considered changed.

   (D) KEGG pathway enrichment analysis of significantly decreasing and
   increasing proteins at PSSH levels was performed using DAVID and
   plotted. The size of the bubbles is indicative of the number of
   proteins annotated with that term; bubbles are color-coded according to
   the significance of the enrichment.

   (E) PSSH changes in subcellular compartments. Colors indicate the
   proportion of proteins with increased (red) or decreased (blue) PSSH
   levels in CDHFD-fed mice compared with control-fed mice; grey
   represents the proportion of proteins that did not change. Protein
   subcellular localization was determined using DAVID and Gene ontology
   cellular compartment.

   (F) Box plot of log2 transformed abundancies of PRDX2 (peroxiredoxin
   2), THIO (Thioredoxin), mitochondrial superoxide dismutase (SOD2), and
   TRXR2 (thioredoxin reductase 2) identified in the total proteome (left)
   and persulfidome (right). The data are presented as box-and-whisker
   plots, showing the minimum, maximum, and interquartile range (IQR) for
   n = 4–5.

   (G) ROS levels in liver tissue from control and CDHFD-fed mice were
   measured using DCFDA. The data are mean ± SD (n = 5–7).

   (H) Lipid peroxidation (MDA) levels in liver tissue from control and
   CDHFD-fed mice were measured using the TBRAS assay. The data are
   mean ± SD (n = 5–7).

   (I) Networks of enriched biological processes in CDHFD-Control group.
   The biological processes were selected based on a p-value (<0.05) and
   an FDR (<0.25) cutoff from the GSEA enrichment (red for processes
   enriched in the CDHFD group, while blue for processes enriched in
   control group).

   (J) NADPH levels in liver samples from mice fed a control diet or CDHFD
   were analyzed. The data are mean ± SD (n = 5–7). ∗P < 0.05, ∗∗P < 0.01,
   and ∗∗∗P < 0.001.

   In HFD-fed mice, which model early-stage MASLD and display only mild
   steatosis, global PSSH levels appeared similar to controls based on gel
   analysis ([142]Fig. S3E). However, persulfidome profiling revealed an
   increased number of proteins with elevated PSSH ([143]Fig. S5C). This
   selective enhancement likely reflects a compensatory response, possibly
   driven by the observed upregulation of H[2]S-producing enzymes in
   HFD-fed livers ([144]Fig. 1E and F). The proteins showing increased
   PSSH were enriched in pathways related to amino acid metabolism, the
   pentose phosphate pathway, proteasome function, and
   xenobiotic/detoxification metabolism ([145]Fig. S5D). These results
   suggest that during the early stages of MASLD, PSSH may support
   metabolic adaptation and help maintain cellular homeostasis. Beyond its
   known role in protecting proteins from irreversible overoxidation, PSSH
   may also modulate the activity of key enzymes within these pathways,
   thereby fine-tuning cellular stress responses and metabolic flux during
   early disease progression.

   By contrast, HFHFHCD-fed mice, which developed more severe metabolic
   dysfunction, moderate steatosis and slight decrease in H[2]S-producing
   enzymes ([146]Fig. 1E and F), showed a greater number of proteins with
   decreased PSSH, particularly within pathways related to biosynthesis of
   cofactors, porphyrin metabolism and other metabolic functions
   ([147]Fig. S5E and S5F). Interestingly, proteins associated with
   peroxisomal, and ribosomal functions showed increased PSSH in this
   group, suggesting intracellular compartmentalization of the change.
   Nonetheless, the overall decline in PSSH across key metabolic enzymes
   suggests a shift toward dysregulated PSSH as the disease progresses.

   The most striking changes were observed in CDHFD-fed mice, a model that
   closely mirrors advanced MASH, including liver fibrosis and
   inflammation. In this group, we identified 212 proteins with decreased
   PSSH and 165 proteins with increased PSSH compared with controls
   ([148]Fig. 3C). This overall decrease is consistent with the global
   reduction in PSSH observed by in-gel analysis ([149]Fig. 2B). Pathway
   analysis indicated that proteins with altered PSSH are involved in
   carbon metabolism, apoptosis, diabetic cardiomyopathy, and other
   metabolic pathways, hinting at a potential role for PSSH in regulating
   cellular stress responses in chronic liver disease ([150]Fig. 3D).
   Notably, several pathways associated with neurodegenerative diseases
   were also enriched, reinforcing the emerging link between disrupted
   PSSH signaling, aging, and neurodegeneration [[151]40].

   At the organelle level, the most significant loss of PSSH occurred in
   the endoplasmic reticulum (ER), which accounted for the highest
   proportion of proteins with reduced PSSH—47 out of 105 in the
   normalized dataset—compared to 43 out of 170 in mitochondria ([152]Fig.
   3E). Among the affected ER proteins was protein disulfide isomerase
   (PDI), known for its increased activity when persulfidated at its CXXC
   motif ([153]Fig. S6A) [[154]41]. Additional ER oxidoreductases,
   including thioredoxin domain-containing protein 5 (TXNDC5) and
   endoplasmic reticulum oxidoreductase 1 (ERO1), also showed decreased
   PSSH ([155]Fig. S6B and S6C). Although the exact functional
   implications remain to be fully understood, these alterations may
   influence ER protein-folding capacity, redox signaling, or stress
   adaptation, indicating the potential regulatory impact of PSSH on
   maintaining ER homeostasis during MASH progression.

   Perhaps one of the most intriguing findings from the persulfidome
   analysis was that, despite the overall decline in H[2]S-producing
   enzymes and global PSSH, certain key proteins, particularly the redox
   regulators peroxiredoxin 2 (PRDX2), thioredoxin (THIO), and thioredoxin
   reductase 2 (TRXR2), showed increased PSSH levels in the livers of
   CDHFD-fed mice ([156]Fig. 3F). Additionally, persulfidated
   mitochondrial SOD (MnSOD) was increased, suggesting that this
   modification might protect MnSOD from inhibition by H[2]O[2], thereby
   preserving its superoxide scavenging activity ([157]Fig. 3F) [[158]11].
   A common feature of these proteins is that they all contain
   redox-sensitive cysteines, which can readily oxidize to sulfenic acids,
   making them potential targets for H[2]S to form PSSH. It should be
   noted that the functional implications of PSSH are context-dependent;
   persulfidation can either enhance or suppress protein activity
   [[159]10,[160]19], or in some cases act primarily as a protective
   mechanism against irreversible oxidative damage ([161]Fig. 4A).

Fig. 4.

   [162]Fig. 4
   [163]Open in a new tab

   Persulfidation protects protein tyrosine phosphatases

   (A) Schematic diagram of PTP oxidation and persulfidation. Protein
   persulfidation can serve as a protective mechanism against irreversible
   oxidation. Liver samples were tag-switched with DCP-Bio1, and
   persulfidated proteins were enriched using magnetic streptavidin beads.

   (B) Total PTP oxidation levels in liver samples from control and
   CDHFD-fed mice. The data are mean ± SD (n = 4)

   (C) PTP persulfidation levels in livers. Samples after biotin
   enrichment were analyzed by western blot. 20 μg of tag-switched samples
   were loaded as input. Samples without tag-switch using DCP-Bio1 were
   used as negative controls. The data are mean ± SD (n = 5–7).

   (D) Total PTP oxidation in isolated primary mouse hepatocytes collected
   at different time points up to 48 h. Total PTPN2 in the same blot was
   detected using a fluorescent secondary antibody. The data are mean ± SD
   (n = 3).

   (E) PTPN2 persulfidation levels in isolated primary mouse hepatocytes
   collected at different time points. Samples were tag-switched with
   DCP-Bio1, and persulfidated proteins were enriched using magnetic
   streptavidin beads. 20 μg of tag-switched samples were loaded as input.
   The data are mean ± SD (n = 4).

   (F) Activity of untreated or treated recombinant human PTPN2 with
   Na[2]S[4] (20 μM) or H[2]O[2] (200 μM) for 30 min. After 10 min of
   measurement, the samples were treated with 1 mM DTT. The data are
   mean ± SD (n = 5).

   (G) Activity of recombinant human PTPN2 untreated or treated with
   Na[2]S (200 μM) + H[2]O[2] (200 μM) or H[2]O[2] alone (200 μM) for
   30 min. After 10 min of measurement, the samples were treated with 1 mM
   DTT. The data are mean ± SD (n = 5). ∗P < 0.05 and ∗∗P < 0.01.

   To illustrate that persulfidation serves as a protective mechanism
   against protein overoxidation, we first sought to demonstrate that ROS
   levels increase with MASLD progression. Although increased ROS levels
   and oxidative stress have been well documented in diet-induced obesity
   models—including those leading to steatosis and MASH
   [[164]27,[165]42,[166]43]—we sought to validate this in our CDHFD-fed
   mouse model. As expected, these mice exhibited clear signs of oxidative
   stress, including elevated ROS levels, increased lipid peroxidation (as
   evidenced by the marker malondialdehyde (MDA), a relatively stable
   byproduct of lipid peroxidation), and an accumulation of sulfonylated
   peptides (PSO[3]H) in the liver [167](Fig. 3G and H and [168]S6F).
   Additionally, we also observed a decrease in NADPH levels ([169]Fig.
   3J) along with decreased expression of proteins involved in NADPH
   regeneration ([170]Fig. S3B), indicating a diminished cellular capacity
   to reduce oxidized proteins or scavenge H[2]O[2] via peroxiredoxins or
   glutathione peroxidases. In agreement with these results, global
   proteome analysis of CDHFD-fed mice revealed enrichment of pathways
   associated with inflammation, immune response, and oxidative stress,
   while sulfur compound metabolism was notably downregulated ([171]Fig.
   3I). Moreover, the expression levels of key antioxidant enzymes such as
   catalase (CAT) and superoxide dismutase 1 (SOD1) were significantly
   decreased ([172]Fig. S3C), consistent with previous reports of reduced
   CAT and SOD expression and activity in MASH [[173]44,[174]45]. Despite
   this oxidative burden, a compensatory antioxidant response driven by
   Nrf2 activation was observed, reflected by increased expression of
   NQO1, HO-1, GSR, GSS, GPX4, GSTA1, PRDX3, PRDX4, and TRXR1 ([175]Fig.
   S3C).

   To further document the protective role of PSSH under oxidative stress
   conditions, we analyzed our proteomic data for the presence of oxidized
   and overoxidized persulfide species, specifically, perthiosulfonic
   acids (P-SSO[3]H). Even though our proteomic workflow was not
   specifically optimized to enrich or target these low-abundance and
   labile modifications, we successfully detected P-SSO[3] species in the
   dataset ([176]Fig. S6E). Crucially, all free thiols and sulfenic acids
   (PSOH) were blocked by NBF-Cl during sample preparation, effectively
   preventing artificial oxidation and confirming that the observed
   P-SSO[3]H was not generated during processing. In addition, we used the
   sulfenic acid-specific probe DCP-Bio1, with and without DTT treatment,
   to probe for oxidized thiol species in primary hepatocytes exposed to
   fatty acids. This analysis showed elevated levels of
   per/polythiosulfenic species (P-SS[n]OH) and sulfenic acids under these
   conditions ([177]Fig. S7A–C) [[178]46], further supporting the role for
   PSSH in buffering protein thiols against oxidative damage.

   Taken together, our findings show that hepatocytes respond to elevated
   ROS levels and lipid peroxidation by dynamically reshaping the
   persulfidome and adjusting the expression of key antioxidant proteins.
   This coordinated response appears to selectively protect critical
   redox-sensitive proteins, reflecting a nuanced and tightly regulated
   antioxidant defense strategy tailored to counteract oxidative stress
   during disease progression.

2.4. Persulfidation protects protein tyrosine phosphatases (PTPs)

   While our persulfidome data underscore the complexity of persulfidation
   changes—making it difficult to attribute effects to a single protein or
   protein class—we focused on a group of proteins particularly
   susceptible to redox regulation and oxidative damage: protein tyrosine
   phosphatases (PTPs), which are also well-established targets of PSSH
   [[179][47], [180][48], [181][49]]. Dysregulated PTP activity worsens
   MASLD progression, leading to the transition toward MASH and the
   development of hepatocellular carcinoma [[182]27,[183]28,[184]50].
   Consistent with previous studies, we observed elevated levels of
   oxidized PTPs in the livers of CDHFD-fed mice ([185]Fig. 4B) [[186]27].
   Given their central regulatory roles and redox sensitivity, we
   hypothesized that PSSH may serve a protective function for PTPs,
   similar to its role in safeguarding redox-regulating enzymes.
   Supporting this idea, persulfidome analysis confirmed increased PSSH of
   PTPN1 ([187]Fig. S8A). However, due to the generally low abundance of
   PTPs, many were either undetectable or only partially represented in
   the persulfidome and proteome datasets.

   To overcome these limitations, we applied a targeted dimedone
   switch-based biotin pulldown followed by immunoblotting. This approach
   not only confirmed the increased PSSH of PTPN1 but also revealed
   similar increases in PTPN2, PTPN6, and PTPN12 ([188]Fig. 4C and
   [189]S8B). In contrast, glyceraldehyde-3-phosphate dehydrogenase
   (GAPDH/G3P), a highly abundant and redox-sensitive protein
   [[190]51,[191]52], showed no significant change in PSSH ([192]Fig. 4C
   and [193]S5D). These results highlight the selective nature of PSSH
   remodeling in MASH, suggesting that hepatocytes preferentially target
   specific proteins—such as PTPs—for protection in the face of oxidative
   stress.

   We next validated these findings using the same primary hepatocyte
   de-differentiation model shown in [194]Fig. S4E, which is characterized
   by elevated ROS levels [[195]39]. In this model, we observed a
   time-dependent increase in PTPN2 oxidation ([196]Fig. 4D, black arrow),
   accompanied by a corresponding rise in its PSSH, as detected by biotin
   pulldown ([197]Fig. 4E). These observations indicate that PSSH of
   specific proteins can still be induced in response to oxidative stress,
   functioning as a targeted protective mechanism—even when overall H[2]S
   production and global PSSH levels are diminished.

   To explore the protective mechanism of PSSH on PTPs, we conducted in
   vitro activity assays using recombinant human PTPN2 in HepG2 cells.
   When exposed to H[2]O[2] together with either the H[2]S donor, sodium
   sulfide (Na[2]S) or the direct persulfide donor sodium tetrasulfide
   (Na[2]S[4]), PTPN2 underwent PSSH. This modification temporarily
   inactivated the enzyme but shielded it from irreversible overoxidation
   ([198]Fig. 4A and F). This protective effect was evident from the fact
   that PTPN2 activity could be at least partially restored by treatment
   with the reducing agent dithiothreitol (DTT), whereas the activity of
   PTPN2 treated with H[2]O[2] alone remained irreversibly lost ([199]Fig.
   4F and G and [200]S8C-E).

   Comparable protective effects were observed in other members of the
   PTP-family, including PTPN1 and PTPRK ([201]Fig. S8F), which are
   involved in hepatic metabolic function [[202]53,[203]54]. However,
   PTPN6 did not benefit from H[2]S-mediated protection under the same
   conditions, likely due to its rapid susceptibility to overoxidation,
   which may outpace persulfidation.

2.5. H[2]S donors mitigate inflammatory signaling

   Having shown that H[2]S donors can protect key proteins like PTPs from
   irreversible overoxidation, we sought to investigate whether this
   protective effect could influence their ability to regulate downstream
   signaling, particularly in the context of inflammation. Our data
   suggest that PSSH may serve a dual function: inhibiting and stabilizing
   PTPs while simultaneously dampening pro-inflammatory signaling by STAT1
   and STAT3, two central regulators of the inflammatory response
   ([204]Fig. 5A). First, we confirmed that treatment with Na[2]S or other
   H[2]S-releasing donors increased PSSH levels in both HepG2 cells and
   primary hepatocytes ([205]Fig. S9A–C). Next, we analyzed the activation
   status of STAT1 and STA3 by assessing their phosphorylation status and
   observed that Na[2]S suppresses STAT3 phosphorylation under basal
   conditions in both HepG2 cells ([206]Fig. 5B) and isolated primary
   mouse hepatocytes ([207]Fig. 5C). This indicates that H[2]S-mediated
   PSSH ameliorates basal inflammatory signaling.

Fig. 5.

   [208]Fig. 5
   [209]Open in a new tab

   H[2]S regulates the inflammatory pathway in hepatocytes

   (A) Schematic diagram of PTP and pro-inflammatory signaling pathways.
   PTPN2 negatively regulates JAK/STAT signaling. Enhanced STAT1/3
   signaling drives the progression of MASH/fibrosis and HCC development.
   While H[2]S can both protect PTP activity, its effects on downstream
   signaling cascades remain unclear. The net impact of H₂S on
   inflammatory signaling pathways is yet to be determined.

   (B) HepG2 cells were treated with Na[2]S for 30 min. The
   phosphorylation of STAT3 (Tyr705) was measured by western blotting. The
   data are mean ± SD (n = 3).

   (C) Freshly Isolated primary mouse hepatocytes were treated with Na[2]S
   for 30 min. Phosphorylation of STAT3 (Tyr705) were measured by western
   blotting. The data are mean ± SD (n = 3).

   (D) HepG2 cells were starved in medium without serum for 2 h and were
   non-treated or pre-treated with 500 μM Na[2]S or Na[2]S[4] for 10 min.
   After removing the medium, cells were treated with (C) IL-6 (1000 U/ml)
   and collected at the indicated time points. Phosphorylation of STAT3
   (Tyr705) was measured by western blotting. The data are mean ± SD
   (n = 3). Statistical analyses were performed using two-way ANOVA.

   (E) HepG2 cells were starved in medium without serum for 2 h and were
   non-treated or pre-treated with 500 μM Na2S or Na[2]S[4] for 10 min.
   After removing the medium, cells were treated with IFN-γ (50 U/ml) and
   collected at the indicated time points. Phosphorylation of STAT1
   (Tyr701) was measured by western blotting. The data are mean ± SD
   (n = 3). Statistical analyses were performed using two-way ANOVA.

   (F) Freshly Isolated primary mouse hepatocytes were starved in medium
   without serum for 2 h and were non-treated or pre-treated with 500 μM
   Na[2]S for 10 min. After removing the medium, cells were treated with
   IL-6 (1000 U/ml) and collected at the indicated time points.
   Phosphorylation of STAT3 (Tyr705) was measured by western blotting. The
   data are mean ± SD (n = 3). Statistical analyses were performed using
   multiple t-test.

   (G) Freshly Isolated primary mouse hepatocytes were starved in medium
   without serum for 2 h and were non-treated or pre-treated with 500 μM
   Na[2]S for 10 min. After removing the medium, cells were treated with
   IFN-γ (50 U/ml) and collected at the indicated time points.
   Phosphorylation of STAT3 (Tyr705) was measured by western blotting. The
   data are mean ± SD (n = 3). Statistical analyses were performed using
   the multiple t-test.

   (H) Box plot of log2 transformed abundancies of STAT3 identified in the
   total proteome (upper) and persulfidome (lower). The data are presented
   as box-and-whisker plots showing the minimum, maximum, and
   interquartile range (IQR) with n = 4–5. ∗P < 0.05, ∗∗P < 0.01.

   To explore whether this effect extends to inflammation induced by
   external stimuli, we treated cells with pro-inflammatory cytokines that
   are known to be elevated during hepatic inflammation and fibrosis
   [[210]4]. In both HepG2 cells ([211]Fig. 5D and E) and isolated primary
   mouse hepatocytes ([212]Fig. 5F and G), Na[2]S treatment similarly
   reduced cytokine-induced STAT1 and STAT3 phosphorylation. These
   findings indicate that H[2]S can modulate both basal and
   stimulus-induced inflammatory responses in hepatocytes—potentially
   delaying the progression from steatosis to MASH and hepatocellular
   carcinoma [[213]28].

   We also evaluated the impact of an H[2]S donor on insulin signaling.
   While treatment with Na[2]S did not significantly affect the activation
   of insulin receptor tyrosyl phosphorylation ([214]Fig. S9D), it
   modestly delayed dephosphorylation ([215]Fig. S9D), likely through the
   persulfidation-mediated inhibition of PTPN1/PTPN2 and/or other related
   phosphatases.

   Collectively, our results indicate that H[2]S-mediated PSSH plays a
   dual protective role: it inhibits and shields PTPs from overoxidation,
   while also curving excessive activation of pro-inflammatory signaling
   proteins. The identification of multiple STAT family members within the
   persulfidome raises the possibility that PSSH may directly modulate
   components of inflammatory signaling pathways ([216]Fig. 5H), although
   indirect effects through upstream or downstream regulators cannot be
   excluded. It is important to note that PSSH does not cause permanent
   inhibition of protein functions. Instead, it is a reversible and
   dynamic modification, integrated into broader redox networks that rely
   not only on H[2]S availability but also on NADPH-dependent reductive
   systems to restore PTP activity. This reversible nature allows cells to
   fine-tune signaling in response to oxidative cues. Thus, PSSH operates
   as a flexible regulatory mechanism—buffering PTPs and inflammatory
   pathways against oxidative damage while maintaining the adaptability
   necessary for rapid cellular signaling responses.

2.6. Discussion

   In this study, we present the first comprehensive analysis of the
   hepatic persulfidome across the progression of MASLD using
   dimedone-switch labeling and multiple dietary mouse models. We observed
   stage-dependent remodeling of protein persulfidation (PSSH),
   characterized by a global decrease in the total PSSH levels in advanced
   disease (CDHFD-fed mice), alongside selective increases in
   redox-regulated proteins. These data postulate that PSSH plays a
   protective and regulatory role in metabolic liver disease.

   The reduction in total PSSH levels coincided with a marked
   downregulation of hepatic hydrogen sulfide (H[2]S)-producing enzymes in
   CDHFD-fed mice. This was not observed in HFD-fed mice, which showed
   increased expression of these enzymes. These results implicate impaired
   H[2]S synthesis as a key contributor to PSSH depletion in severe liver
   damage. Because PSSH can protect protein thiols from irreversible
   oxidative damage, its loss may exacerbate the progression to
   steatohepatitis and fibrosis.

   Mechanistically, persulfides typically form through the reaction of
   sulfenic acids (PSOH) with hydrogen sulfide (H[2]S), rather than via
   H[2]S and thiol groups [[217]24]. Consequently, their formation depends
   on local concentrations of oxidants and H[2]S in compartmentalized
   redox niches [[218]55]. Notably, the PSSH landscape can vary across
   membraneless organelles, which concentrate ROS and redox-active
   proteins [[219]56], thereby creating hotspots for selective oxidative
   signaling. At the transcriptional level, the observed reduction in
   H[2]S-producing enzymes may result from hepatocyte de-differentiation
   and TGF-β1 signaling, both known to suppress CBS and CSE expression
   [[220]57]. We also identified downregulation of HNF4α, a master
   transcription factor that controls transsulfuration pathway genes and
   glutathione homeostasis [[221]58,[222]59]. HNF4α deficiency has been
   linked to altered sulfur amino acid metabolism and reduced H[2]S
   biosynthesis [[223]59]. SELENBP1, a recently identified
   H[2]S-generating enzyme and HNF4α target, was also downregulated in
   both CDHFD-fed mice and human MASH livers [[224]60].

   Despite the overall decrease in PSSH, we observed selective enrichment
   in redox-regulated proteins such as Prdx2, Trx1 (THIO), and TrxR2. The
   accumulation of PSSH on Trx/TrxR is not unexpected, as Trx and TrxR are
   known to transiently form persulfide intermediates in their
   “pseudo-catalytic” cycle upon protein depersulfidation. Normally, the
   persulfide formed on the catalytic Cys32 of Trx1 would be resolved by
   Cys35 to release H[2]S. Due this instability of the Cys32 persulfide,
   it has been extremely challenging to assess how Cys32 persulfidation
   would alter the nucleophilic activity of Trx1 [[225][10], [226][61]].
   Nevertheless, the observed PSSH accumulation of both Trx1 and TrxR is
   expected to compromise the depersulfidation, as well as reduction
   capacity of the Trx system [[227]19,[228]61]. Furthermore, Prdxs and
   SOD harbour cysteines that are highly susceptible to H[2]O[2]. With
   increased ROS and reduced Trx/TrxR activity, these proteins accumulate
   oxidized forms like PSOH, which could selectively drive PSSH formation
   through reaction with residual H[2]S. Overall, the functional
   implications of these changes remains unclear, as persulfidation is
   transient and enzyme activity is often regulated by complex redox
   mechanisms. Future work should explore how these redox PTMs influence
   specific enzymatic functions in vivo.

   Of proteins that did show a decrease in PSSH in CDHFD-fed livers a
   class that stands out are proteins involved in lysosomal degradation
   and apoptosis—including cathepsins and caspases ([229]Table S2). Since
   cathepsins rely on redox-sensitive cysteines for enzymatic activity
   [[230]62], their altered persulfidation may impair autophagy and
   apoptosis. As fibrosis is potentially reversible [[231]63], reversible
   PTMs like PSSH may serve as stress-buffering mechanisms that are lost
   when the redox balance breaks down. Indeed, CDHFD-fed livers showed
   elevated oxidative stress and depleted NADPH, likely disrupting
   persulfide turnover and contributing to disease progression.

   We propose that the hepatic persulfidome is a dynamic reservoir that
   adapts to the disease state. Because H[2]S is rapidly metabolized in
   the liver [[232]64], its local availability depends on both enzyme
   activity and the redox environment. As such, in CDHFD-fed mice, reduced
   H[2]S production may lead to alternative H[2]S sources—such as
   depersulfidating proteins or low-molecular-weight persulfides. Albumin,
   for instance, can carry and deliver PSSH in circulation [[233]34].
   Alternative H[2]S release routes—e.g., via autophagy-related pathways
   or LMW persulfides—may also contribute [[234]65], although their role
   in total PSSH remains debated [[235]19,[236]66]. Notably,
   CARS2-mediated translation-linked persulfidation provides a route
   independent of oxidation [[237]20]. Furthermore, extrahepatic sources
   of H[2]S may influence hepatic redox signaling under metabolic stress
   [[238]67].

   While our study focused on the protective role of H[2]S and PSSH on the
   protein level, their broader implications in regulating biological
   processes such as metabolism, inflammation, and antioxidant defense,
   should not be overlooked [[239]68,[240]69]. Xu et al. demonstrated that
   farnesoid X receptor (FXR) persulfidation by H[2]S enhances Zn^2+
   binding, leading to improved metabolic regulation [[241]69]. PSSH of
   Keap1 has also shown to facilitate Nrf2 release, leading to the
   activation of antioxidant response genes and enhanced cellular defense
   against oxidative stress [[242]70].

   Our data support the therapeutic potential of restoring H[2]S signaling
   in MASLD. H[2]S donors such as NaHS, GYY4137, S‐propargyl‐cysteine
   (SPRC), and diallyl disulfide (DADS) have been shown to reduce
   steatosis, correct lipid dysregulation, and attenuate oxidative stress
   and inflammation in experimental models
   [[243]15,[244]68,[245]69,[246][71], [247][72], [248][73], [249][74]].
   For example, DADS, a garlic-derived donor, alleviated lipotoxicity and
   liver inflammation in MASH models [[250]74]. While earlier studies
   proposed that PSSH inhibits protein tyrosine phosphatases (e.g., PTP1B)
   and blunts growth factor signaling [[251]47], our findings suggest a
   broader regulatory role. PSSH may not only stabilize PTPs against
   irreversible oxidation but may regulate other targets in
   pro-inflammatory pathways, helping to maintain balanced signaling.
   Together, these effects suggest that therapeutic H[2]S benefits are at
   least partly mediated by targeted PSSH. However, the clinical
   translation of conventional H[2]S donors remains limited by poor
   specificity, short half-life, and suboptimal tissue targeting.
   Therefore, future efforts should focus on the development of
   next-generation H[2]S or persulfide delivery systems with improved
   stability and organ selectivity, alongside mechanistic studies to
   delineate how PSSH modulates liver function at the molecular level.

   In conclusion, we provide novel insights into hepatic persulfidome
   remodeling during MASLD progression. Rather than a uniform decline,
   PSSH changes are selective and dynamic, reflecting adaptive redox
   signaling. Modulating H[2]S bioavailability and PSSH holds potential
   for therapeutic intervention in metabolic liver disease.

3. Materials and methods

3.1. Animal

   Mice (C57BL/6N) were housed and managed in compliance with the Belgian
   Regulations for Animal Care, and the animal protocols underwent
   approval from the Commission d’Ethicque du Bien-Être Animal (CEBEA),
   Faculté de Médecine, Université libre de Bruxelles (dossier No. 732).
   Animals were housed at 22 °C on a 12:12-h light-dark cycle with ad
   libitum access to food and water. At 8weeks old, mice were fed a
   standard control diet (Control diet, Research Diets, D09100304i), a
   high-fat diet (HFD; 60 % kcal fat, Research Diets, D16042106i),
   high-fat high-fructose high-cholesterol diet (HFHFHCD, Research Diets,
   D09100310i), or a choline-deficient HFD (CDHFD; l-Amino Acid Diet,
   Research Diets, A20012301i) for 12 weeks. Liver samples were
   consistently collected from the center of the left lobe.

3.2. Histological analysis

   Mouse liver tissues intended for histological analysis were collected
   from euthanized mice, dissected, and subsequently rinsed with PBS. The
   obtained tissue specimens were fixed in 4 % buffered formaldehyde (pH
   7.4) and embedded in paraffin blocks. The paraffin blocks were then
   sectioned into slices measuring 5–7 μm using a Leica rotator microtome.
   Hematoxylin and Eosin (H&E) staining was employed for the sections.
   This staining process involved initial deparaffinization, followed by a
   sequence of ethanol-based rehydration steps. The sections underwent
   successive treatment with Harris hematoxylin solution, acid alcohol,
   and ammonia water. Eosin Y solution was used for counterstaining. The
   subsequent steps included dehydration, xylene-based clearing, drying,
   and slide mounting.

3.3. Primary mouse hepatocyte isolation

   Mouse primary hepatocytes were isolated from wild-type C57BL/6N mice
   following overnight ad libitum feeding, using a two-step collagenase
   perfusion method through the vena cava. The process was initiated by
   anaesthetizing the mice through an intraperitoneal injection of a
   ketamine (100 mg/kg) and xylazine (10 mg/kg) mixture, the peritoneum
   was opened, and the intrahepatic segment of the vena cava was
   cannulated for subsequent perfusion. The hepatic portal vein was cut to
   clear blood from liver at the initiation of liver perfusion. In the
   first perfusion step, the liver was exposed to HBSS (no calcium, no
   magnesium, Thermo Fisher Scientific, #14170138) supplemented with 10 mM
   4‐(2‐hydroxyethyl) ‐1‐piperazine ethanesulfonic acid–NaOH (pH 7.4),
   saturated with O[2]/CO[2] (95:5 vol/vol), at 37 °C for 10 min. The
   second step involved adding collagenase type IV (0.3 mg/mL) to
   William's E Medium (Thermo Fisher Scientific, #32551087) and further
   perfusing for 10 min, effectively softening the liver tissue. The
   softened liver was then transferred to a sterile plastic dish, and
   cells were dispersed using a coarse‐toothed comb in cold Williams
   medium, followed by filtration through a 100‐μm cell filter to
   eliminate cell clumps. The resulting clump-free cell suspension was
   pelleted through centrifugation at 50×g for 5 min at 4 °C, and the
   pellet was resuspended in William's E Medium and layered onto Percoll
   solution (Millipore Sigma, # GE17-0891-01) (10 ml Percoll + 1.25 ml PBS
   10X + 1.25 ml H[2]O) and centrifuged for 10 min at 190×g. The pellet
   was washed three times with William's E Medium. Viability assessment
   using the trypan blue exclusion test yielded around 15–20 million cells
   with approximately 85 % viability. The isolated hepatocytes were then
   subjected to cell culture and used for subsequent analyses.

3.4. Cell culture

   HepG2 cell lines were cultured in DMEM-GlutaMax with 10 %
   heat-inactivated FBS and 1 % Pen-Strep (P/S). Mouse primary hepatocytes
   were cultured in Attachment Medium (William's Medium with Glutamax
   supplemented with 10 % FBS, 1 % P/S, 10 mM HEPES). Following a 3–4 h
   attachment period, the plating medium was replaced with a Maintenance
   Medium (William's medium with Glutamax supplemented with 10 % FBS, 1 %
   P/S, 1 % Non-Essential Amino Acids (NEAA), 10 mM HEPES, and 5 μM
   Hydrocortisone). For chronic treatment, isolated cells were treated
   with cytokines (IL-6 1000 U/ml, IFN-γ 50 U/ml, TNF-α 50 U/ml), TGF-β1
   (5 nM), or insulin (10 nM) 3 h after seeding in William's medium with
   GlutaMax supplemented with 1 % FBS, 10 mM HEPES, 1 % P/S for 24 h. For
   fatty acid treatment, palmitate acid (0.4 mM) and oleate acid (0.8 mM)
   were conjugated with 1 % BSA in William's medium with GlutaMax
   supplemented with 1 % FBS, 10 mM HEPES, and 1 % P/S for 30 min prior to
   treatment. Unless explicitly indicated otherwise, treatments were
   conducted using maintenance medium supplemented with a low FBS
   concentration of 1 %. For cytokine and insulin pulse and chase
   treatment, cells were allowed to incubate overnight at 37 °C before
   subsequent experiment procedures and were subjected to serum starvation
   for 3 h before experiments.

3.5. Expression and purification of human PTP

   NEB Turbo competent cells were transformed with an expression plasmid
   (his-sumo-hPTPN2) and cultured at 37 °C, 210 rpm in TB medium
   supplemented with kanamycin. The cells were grown at 37 °C until
   OD[600] = 0.6. Protein expression was induced by adding
   isopropyl-thio-β-d-galactopyranoside (IPTG) at a final concentration of
   1 mM. The proteins were expressed for 4 h, after which the cells were
   harvested and frozen at −20 °C until use. Cells were lysed in lysis
   buffer (50 mM Tris, pH = 8.0, 400 mM NaCl, 1 mM DTT, 1 mg/ml Leupeptin,
   0.1 mg/ml AESBF, 50 mg/ml DNase1, 20 mM MgCl[2]) and homogenized by
   sonication at 70 % amplitude for 3 min with a 30 s on-30 s off cycle on
   ice. The cell lysate was clarified at 48000×g for 45 min and
   subsequently filtered through 0.45 μm membrane. Immobilized metal
   affinity chromatography (IMAC) was performed using the AKTA pure
   system, and all buffers were filtered by 45 μm pore-size membrane.
   His-tagged proteins were incubated with Ni-NTA Agarose (Cytiva,
   #GE17-5268) for 2 h at 4 °C. His-tagged proteins were eluted by a
   gradient of imidazole. The eluted fractions were collected and
   concentrated using Amicon centrifugal filters 10 kDa MWCO (Millipore,
   UFC5010). To remove the His tag from the proteins, pooled fractions
   were dialyzed in dialysis buffer (50 mM Tris-HCl, pH 8.0, 50 mM NaCl,
   2 mM DTT) to remove imidazole to perform the second IMAC. His-Ulp1 was
   added to the elution to cleave the His-sumo tag from the proteins, and
   a second IMAC was performed to collect untagged PTPN2. In brief,
   dialyzed samples were again incubated with Ni-NTA beads for 2 h and
   flow-through was collected. Samples were concentrated and loaded on
   SDS-PAGE and stained with instant-Blue (Abcam, ab119211) to check for
   purity. Buffer exchange to storage buffer (25 mM, Tris-HCl, pH 7.5,
   1 mM DTT, 1 mM EDTA, 20 % glycerol) was performed, and concentration
   was determined by measuring absorbance at 280 nm. The purified proteins
   were snap frozen in liquid nitrogen and stored at −80 °C. For PTPRK,
   the recombinant PTPRK intracellular domain (PTPRK ICD) was purified
   according to Hay Iain et al. [[252]75], with small modifications. Human
   recombinant PTPN1 (ab51277) and PTPN6 (ab51289) were purchased from
   Abcam.

3.6. Protein tyrosine phosphatase activity assay

   Recombinant PTPs were reduced with 10 mM DTT for 30 min at room
   temperature. Desalting was performed using Amicon 10 kDa MWCO
   concentrators four times to remove DTT and exchange to activity assay
   buffer (20 mM HEPES (adjusted to pH 7.4), 100 mM NaCl, 0.1 mM DTPA and
   1 mM sodium azide and 0.05 % BSA). The assay buffer was degassed by
   flushing it with Argon. PTPs were subjected to the indicated treatments
   (H[2]O[2], Na[2]S, Na[2]S[4]) and loaded into a clear 96-well plate. A
   final concentration of 15 mM of p-Nitrophenyl Phosphate (Sigma, #4876)
   was added to the samples before the start of the measurement to
   initiate the reaction. The final enzyme concentration in the reacion
   were 20 nM for PTPN1, PTPN2, PTPN6, and 100 nM for PTPRK. The assay was
   performed using the ID5 spectrometer (Molecular Devices) by measuring
   the formation of p-nitrophenol at 405 nm at 27 °C. To stop the reaction
   with H[2]O[2], 200 U/ml of catalase (Sigma, #C9322) was added to the
   samples.

3.7. Western blotting

   Cell lysis buffer (Cell Signaling, #9803) supplemented with Halt
   protease and phosphatase inhibitor cocktails (Thermo Fisher, cat
   #78442) was used to extract total protein lysates from cell lines and
   primary mouse hepatocytes. The protein concentration was determined
   using the BCA protein assay (Thermo Scientific, #23225). Samples
   (20–30 μg) were mixed with SDS loading buffer containing
   β-mercaptoethanol and boiled at 95 °C for 10 min Samples were separated
   on 10–12 % SDS gel and transferred to 0.22 μM nitrocellulose membrane
   (Bio-rad, Hercules, #1620112). The membrane was blocked by 5 % skim
   milk in 0.3 % v/v TBS-tween 20 (TBST) for 1 h at room temperature. The
   membrane was then incubated with the diluted primary antibody in TBST
   overnight at 4 °C with agitation. The membrane was washed three times
   for 10 min with TBST and incubated with goat anti-rabbit or anti-mouse
   IgG-horseradish peroxidase conjugate secondary antibody (1:5000–1:7000,
   #P0447 & P0448, Dako Agilent, Santa Clara). Alternatively, goat
   anti-rabbit or anti-mouse StarBright Blue 700 fluorescent secondary
   antibodies (Bio-Rad, #12004162) were used.

   The membrane was washed three times for 10 min with TBST and once with
   distilled water. The signal was detected and visualized using ECL
   western blotting substrate (Thermo Fisher Scientific, #34579) on
   ImageQuant 800 (Amersham, Cytiva).

3.8. H[2]S production (lead acetate plate assay)

   The method was performed as previously described with some
   modifications [[253]15,[254]33]. Tissues were mechanically homogenized
   in PBS containing 0.5 % Triton X-100 and protease inhibitor using
   stainless steel disruption beads with a cell disruptor. Supernatants
   were collected by centrifugation at maximum speed. Protein BCA assay
   was performed to determine the protein concentration. Cell lysates or
   tissue homogenates were added to a reaction mixture containing PBS (pH
   7.4) supplemented with 2 mM pyridoxal-5-phosphate, 5 mM L-Cys, and 1 mM
   homocysteine. For MPST-dependent production, the mixtures were
   supplemented with 1 mM 3-mercaptopyruvate (3-MP) and 1 mM DTT.
   Non-enzymatic production (mainly from 3-MP) was subtracted before
   quantification. A lead acetate paper (Whatman indicator papers, Merck)
   was placed to cover the 96-well plate to absorb H[2]S released into the
   air (sealed by stick cover) and incubated at 37 °C in the dark for the
   indicated time (1-2 h) until lead sulfide formation.

3.9. Determination of ROS levels

   The tissue homogenates were prepared as previously described. 50 μl of
   homogenate was mixed with 150 μL of PBS, followed by the addition of
   DCFDA (Sigma, D6883) to a final concentration of 1 mM. The mixture was
   incubated at 37 °C in the dark for 30 min. The fluorescent signal was
   measured at Excitation 488 nm/Emission 535 nm using a microplate
   reader. The relative ROS levels were normalized to the protein
   concentration determined using the BCA assay.

3.10. Thiobarbituric acid reactive substances (TBRAS) assay

   The TBARS assay was performed to assess lipid peroxidation product
   (Malondialdehyde, MDA). Tissue homogenates were prepared as previously
   described. For TBARS analysis, 100 μl of homogenate or cell lysate was
   mixed with 200 μl of ice-cold 10 % trichloroacetic acid (TCA) to
   precipitate proteins. The mixture was incubated on ice for 15 min,
   followed by centrifugation at 5000×g for 15 min at 4 °C. Subsequently,
   200 μl of the supernatant was mixed with an equal volume of 0.67 %
   Thiobarbituric acid (Sigma, T5500) and incubated in a boiling water
   bath for 10 min. After cooling, the absorbance was measured at 532 nm
   using a microplate reader. A standard curve was generated using
   1,1,3,3-tetramethoxypropane (Sigma, 108383) as a TBARS standard. TBARS
   levels were calculated and normalized to protein concentrations
   determined using the BCA assay.

3.11. Protein persulfide detection in cell lysates

   The dimedone-based labelling method was based on a previously described
   protocol [[255]11]. 500 μl of HENS lysis buffer (50 mM HEPES, 1 mM
   EDTA, 0.1 mM neocuproine, 1 % NP-40 and 2 % SDS, adjusted to pH 7.4,
   1 % protease inhibitor) was supplemented with 20 mM
   4-chloro-7-nitrobenzofurazan (NBF-Cl, add before lysis). Cells were
   washed twice with cold PBS and lysis buffer was added into the wells.
   Cells were gently scrapped using tips and the lysates were collected.
   The lysates were homogenized using a needle and syringe or sonicated in
   a water bath for 1 min and incubated at 37 °C in the dark for 1 h.
   Proteins were precipitated by methanol/chloroform precipitation twice,
   washed with cold methanol, and re-dissolved in 50 mM HEPES (adjusted to
   pH 7.4) supplemented with 2 % SDS. DAz-2: Cy5 click mix (50 μM) was
   added to the redissolved samples and incubated at 37 °C for 1 h in the
   dark. The samples were precipitated and re-dissolved as previously
   described. The protein concentration was determined using the BCA
   protein assay, and the samples were prepared for SDS-PAGE. The gels
   were recorded at 520 nm for the Cy2 signal and at 705 nm for the Cy5
   signal on ImageQuant 800 (Amersham, Cytiva).

3.12. Protein persulfide detection in mouse liver tissue lysates

   The tissue was cut into small pieces and quickly snap-frozen in liquid
   nitrogen after scarification. The frozen tissue was mechanically
   homogenized in HEN lysis buffer supplemented with protease inhibitor
   and 20 mM NBF-Cl using stainless steel disruption beads in a cell
   disruptor. After incubation 1 h at 37 °C, samples were precipitated as
   previously described and tag-switched for persulfidation detection
   using Daz-2/Cy5 click mix.

3.13. DAz-2:Cy-5 click mix

   A DAz-2:Cy-5 Click Mix was prepared by sequentially combining 1 mM
   DAz-2 (Cayman Chemical, 13382), 1 mM Cyanine5 alkyne (Lumiprobe), 2 mM
   copper (II)-TBTA complex (Lumiprobe), and 4 mM ascorbic acid in PBS
   containing 30 % (v/v) acetonitrile sequentially [[256]11]. The click
   reaction was incubated at 50 °C for 2 h, followed by overnight
   incubation at room temperature. The reaction was quenched with 18 mM
   EDTA (1 M Tris buffer, pH adjusted to 9.0) for 2 h at room temperature.
   The final click mix was stored at −20 °C protected from light.

3.14. Biotin pulldown

   Samples blocked with NFB-Cl were tag-switched to 100 μM DCP-Bio1 for
   1 h at 37 °C. Excess reagent was removed by precipitating the samples
   as mentioned above. DCP-Bio1-labeled samples were re-dissolved in 50 mM
   HEPES (adjusted to pH 7.4) supplemented with 1 % SDS. 100 μg of lysate
   was incubated with Dynabeads™ MyOne™ Streptavidin C1 (Thermo
   Scientific, 65001) at a final concentration of 0.1 % SDS overnight at
   4 °C with rotation. The beads were collected by placing the tube at the
   magnetic stand (DynaMag™-2 Magnet, Invitrogen, 12321D) and washed three
   times with PBS supplemented with 0.1 % SDS and one time with PBS.
   Biotinylated proteins were eluted from the beads at 95 °C in 2X SDS
   sample loading buffer supplemented with 10 % β-mercaptoethanol for
   10 min. The eluted supernatants were resolved by SDS-PAGE. A sample
   without DCP-Bio1 was used as negative control.

3.15. Persulfidome proteomic analysis of mouse liver tissue

   Persulfidome and total proteome measurements and data analysis in liver
   tissue samples were performed as previously described
   [[257]11,[258]76]. In brief, proteins were extracted from liver tissue,
   processed, and labeled using NBF-Cl and DCP-Bio1, followed by
   methanol/chloroform precipitation. The protein solution was incubated
   with Pierce™ High Capacity NeutrAvidin™ Agarose (Thermo Fisher
   Scientific) at RT for 4 h with agitation. Proteins were eluted using
   ammonium hydroxide and lyophilized. Samples were resuspended in 50 mM
   amonium bicarbonate, 1 mM CaCl[2], and digested using trypsin. The
   digestion products were desalted on Supel^TM-Select HLB SPE columns,
   dried and resuspended in 0.1 % TFA. Peptides were separed on a
   Acclaim^TM PepMap^TM C18 column using a 90 min (for persulfidation) or
   a 120 min (for total proteome) linear gradient (3 to 35 % B; B: 84 %
   acetonitrile, 0.1 % formic acid) and analyzed on Orbitrap Eclipse™
   Tribid™ mass spectrometer at a flow rate of 250 nl min^−1. Data were
   acquired in DDA mode, with survey scans from m/z 300–1500 at a
   resolution of 120,000 using Orbitrap mode. MS2 scans were carried out
   for 3 s using high-energy collision dissociation (HCD) with normalized
   collision energy of 32 % and analyzed with a normal speed IonTrap mode.
   Peptide identification was performed using PEAKSONLINE software with a
   precursor mass tolerance of 15 ppm, fragment mass tolerance of 0.5 Da,
   and up to three missed cleavages. Variable modifications included NBF
   (+163.0012 Da), DCP-Bio1 (+394.1557 Da), N-term acetylation (+42.010565
   Da), and methionine oxidation (+15.994915 Da). For total proteome,
   DCP-Bio1 was omitted. PSM Proteins were filtered at a 1 % false
   discovery rate, and data normalization was performed using the EigenMS
   R package. Statistical analysis and enrichment (GO and KEGG) were
   performed using the Perseus and DAVID tools.

3.16. Protein sulfenylation (PSOH) and per/polythiosulfenylation (PSS[n]OH)

   Analysis of protein sulfenylation and per/polythiosulfenylation was
   performed according to Heppner, David E et al. with small modifications
   [[259]46]. Cells were lysed in cell lysis buffer (Cell Signaling,
   #9803) containing 1 mM DCP-bio1 (Kerafast) and 10 mM N-ethylmaleimide
   and incubated for 1 h on ice. After centrifugation, cell lysates were
   mixed with SDS loading buffer, either in the presence (P-SOH) or
   absence (P-SSOH/P-SS[n]OH) of 10 % β-mercaptoethanol and separated by
   10 % SDS-PAGE for western blotting with anti-Biotin-peroxidase antibody
   (Sigma-Aldrich, #A0185).

3.17. Total PTP oxidation detection

   Livers were processed as previously described [[260]27,[261]28]. In
   brief, liver samples were mechanically homogenized in ice cold oxPTP
   lysis buffer (20 mM HEPES, adjusted to pH 6.5, 150 mM NaCl, 10 %
   glycerol, 1 % NP40, 10 mM N-ethylmaleimide to alkylate free thiol and
   prevent post-lysis oxidation) supplemented with protease inhibitor
   cocktail (Roche, #04693159001) and incubated for 1 h at 4 °C with
   rotation. Cell debris was then removed by centrifugation (20,000×g,
   20 min at 4 °C) and buffer exchanged (NAP-5 columns, Prepacked with
   Sephadex® G-25 DNA Grade, GE Healthcare) into 20 mM HEPES (adjusted pH
   to 7.5) and incubated with 10 mM DTT to reduce oxidized species before
   further buffer exchange into 20 mM HEPES (adjusted to pH 7.5). The
   samples were then treated with freshly prepared 100 μM pervanadate to
   overoxidized reduced PTPs to the sulfonic acid (SO[3]H) state for
   analysis by SDS-PAGE and immunoblotting using oxPTP antibody.

3.18. Proteomic analysis of human and mouse liver tissue

   Hepatocytes were lysed in lysis buffer (8 M Urea, 50 mM ammonium
   bicarbonate) supplemented with PhosSTOP (Roche) and cOmplete™,
   EDTA-free Protease Inhibitor Cocktail (Roche). Lysates were centrifuged
   at 15060 rpm at 16 °C for 1 h and protein concentration was determined
   with Quick Start™ Bradford (Cat NO. 5000205, BioRad). The total amount
   of protein was subjected to reduction and alkylation by incubating the
   samples with 5 mM DTT for 1 h, followed by incubation with 10 mM
   iodoacetamide for 30 min in the dark at room temperature. Proteins were
   digested with Lys-C (1:75) for 4 h and Trypsin (1:100) overnight.
   Peptides were desalted on reverse-phase Sep-Pak C18 cartridges (Cat.
   No. 045132007A, Waters). The eluted peptides were vacuum dried, and
   stored unitl further use.

   Peptides were dissolved in 80 % acetonitrile and 0.1 % TFA. The total
   proteome was analyzed by high-resolution LC-MS/MS using an UltiMate
   3000 UHPLC system (Thermo Fisher Scientific) coupled with an Orbitrap
   Exploris 480 Mass Spectrometer via an EASY-spray (Thermo Fisher
   Scientific). Operated in a Data Dependent Acquisition (DDA) mode.
   Peptide separation was carried out on an Acclaim™ Pep-Map™100C18 column
   (Thermo Fisher Scientific) using a 175min gradient from 10 % to 40 % of
   B (80 % ACN, 0.1 % FA) at a flow rate of 300 nL min^−1. MS1 survey
   scans were acquired from 375 to 1600 m/z at a resolution of 60,000
   using the Orbitrap mode. MS2 scans were carried out with 28 %
   high-energy collision-induced dissociation (HCD). MS raw files were
   searched with MaxQuant (2.1.0.0) against the mouse UniProt database
   (downloaded in March 2021), using integrated Andromeda search engines.
   Cysteine carbamidomethylation was added to fixed modification, while
   N-terminal acetylation and methionine oxidation were added as variable
   modifications. Trypsin/P was set as the enzyme for digestion, and a
   maximum of 2 missed cleavages. Label-free quantification (LFQ) and
   match-between-runs were enabled for protein identification. A false
   discovery rate (FDR) of 1 % was set for both peptide spectrum match
   (PSM) and protein.

   Data filtering was performed using Perseus software (version 1.6.15.0).
   Potential contaminants and reverse peptides were excluded. Intensities
   were log2 transformed. Proteins detected in at least three out of five
   replicates were retained, and missing values were imputed based on
   normal distribution. A two-sided paired Student's t-test was performed,
   with permutation-based FDR (q-values) calculated from 250
   randomizations. Proteins with q-values of 0.05 or less were considered
   significant. Significantly upregulated and downregulated proteins were
   subjected to Gene Set Enrichment Analysis (GSEA, version 4.3.3) using
   the GO Biological Process gene sets from MSigDB (2023). A p-value
   cutoff of 0.05 and an FDR cutoff of 0.25 were applied. Visualization of
   the enriched pathways was performed in Cytoscape (version 3.10.3).

3.19. Bioinformatic analysis

   Publicly available transcriptomic data (RNA-seq) corresponding to
   [262]GSE164760 was downloaded from the NCBI Sequence Read Archive (SRA)
   in fastq format using version 3.0.0 of the SRA Toolkit. Adapter
   sequences were removed using TrimGalore version 0.6.0 with Cutadapt
   version 1.1846. The clean reads were aligned to the reference genome
   using the splice-aware aligner STAR version 02020147, based on the hg38
   genome version. The aligned reads were quantified using HTseq version
   0.11.0. The trimmed mean of M values method was used with EdgeRversion
   3.28.148, R software (version 3.6.3). For [263]Fig. 1D, the data was
   downloaded from the Gene Expression Omnibus (GEO), accession number
   [264]GSE126848
   . Raw gene counts were normalized using transcripts per kilobase
   million (TPM). The gene counts were normalized by gene length and later
   normalized by sequencing depth. To study the expression of
   H[2]S-producing enzymes in different liver cell types ([265]Fig. 1B),
   sa ingle-cell RNA-Seq dataset of human healthy-obese livers was
   obtained from Ref. [[266]77]. The dataset and cell annotations were
   downloaded from the accession number [267]GSE192740. Using the
   information provided by the authors, UMAP and gene expression data were
   plotted using Seurat. The analysis was performed in R (version R
   2024.04.2).

3.20. Statistical analysis

   Sample sizes (n) were calculated based on individual samples, animals,
   cell passage, independent hepatocyte preparations, or separate
   experiments. All results are expressed as mean ± standard deviation
   (SD). For comparisons between two groups, a two-tail student t-test was
   employed, while differences across multiple groups were assessed using
   one-way ANOVA if not specified in the corresponding figure legend.
   Statistical analysis was performed using Prism 10 software (GraphPad).
   Statistical significance was defined as ∗p < 0.05,∗∗p < 0.01,
   ∗∗∗p < 0.001, and ∗∗∗∗P < 0.0001.

CRediT authorship contribution statement

   Tzu Keng Shen: Conceptualization, Data curation, Formal analysis,
   Investigation, Methodology, Writing – original draft, Writing – review
   & editing. Thibaut Vignane: Data curation, Investigation, Methodology,
   Visualization. Eduardo H. Gilglioni: Data curation, Formal analysis,
   Investigation, Writing – review & editing. Leonardo Traini: Formal
   analysis, Investigation, Writing – review & editing. Elisavet
   Kalaitsidou: Data curation, Formal analysis, Writing – review &
   editing. Pierre Conan: Formal analysis, Investigation, Writing – review
   & editing. Ao Li: Formal analysis, Investigation, Writing – review &
   editing. Wadsen St-Pierre-Wijckmans: Formal analysis, Investigation,
   Writing – review & editing. Jose M. Herranz: Formal analysis,
   Investigation. Bernat Elvira: Data curation, Formal analysis, Writing –
   review & editing. Lukas Otero Sanchez: Resources, Writing – review &
   editing. Eric Trépo: Resources, Writing – review & editing. Leo
   Deelman: Resources, Writing – review & editing. Wei Wu: Formal
   analysis, Investigation, Writing – review & editing. Milos R.
   Filipovic: Conceptualization, Formal analysis, Investigation,
   Methodology, Resources, Visualization, Writing – original draft. Joris
   Messens: Conceptualization, Data curation, Investigation, Resources,
   Supervision, Writing – original draft, Funding acquisition. Daria
   Ezeriņa: Conceptualization, Formal analysis, Investigation, Resources,
   Supervision, Writing – original draft. Esteban N. Gurzov:
   Conceptualization, Funding acquisition, Project administration,
   Supervision, Writing – original draft.

Data and materials availability

   The mass spectrometry proteomics, peptidomics and persulfidomics
   datasets used for [268]Fig. 1, [269]Fig. S1, [270]Fig. S3, [271]Fig.
   3A-E, [272]Fig. 3I, [273]Fig. S5, and [274]Fig. S6 have been deposited
   to the ProteomeXchange Consortium via the PRIDE partner repository
   (PXD057932 or PXD066952) and available from the corresponding authors
   upon reasonable request.

Funding

   European Research Council (ERC) Consolidator grant METAPTPs GA817940
   (ENG).

   FNRS-WELBIO grant (35112672), FNRS-Aspirant fellowship (40010598,
   40024372) and ULB Foundation (ENG).

   European Research Council (ERC) under the European Union's Horizon 2020
   research and innovation programme Grant Agreement No. 864921 (MRF).

   VIB grant (JM).

   Singapore Immunology Network (SIgN), Agency for Science, Technology and
   Research (A∗STAR) (WW).

   Biomedical Research Council (BMRC) Core Research Fund for use-inspired
   basic research and IAF-PP project H22J2a0043 (WW).

   Singapore National Medical Research Council (NMRC) project
   MOH-001401-00 (WW).

   FNRS-ASP scholarship and ENG is a Research Associate of the FNRS,
   Belgium (TKS).

Declaration of competing interest

   ENG declares that there are no relationships or activities that might
   bias, or be perceived to bias, the present work.

Acknowledgments