Abstract

   Metabolic enzymes, critical for cellular homeostasis, are frequently
   co-opted in a disease-specific manner to drive cancer progression.
   Here, we identify aldo-keto reductase family 1 member B10 (AKR1B10),
   down-regulated in gastrointestinal cancers, as a pivotal metastasis
   suppressor correlating with improved colorectal cancer (CRC) prognosis.
   Mechanistically, AKR1B10 activates protein phosphatase 2A (PP2A) by
   preventing redox-regulated nitration of its B56α subunit, preserving
   holoenzyme assembly and enabling c-Myc dephosphorylation at serine-62.
   Loss of AKR1B10 disrupts this pathway, stabilizing c-Myc, which drives
   integrin signaling and metastatic dissemination in CRC. We further
   demonstrate that lysine-125 of AKR1B10 is essential for its interaction
   with PP2A-Cα and B56α nitration, thereby attenuating CRC metastatic
   aggressiveness. Pharmacological restoration of PP2A activity
   effectively mitigates metastasis associated with AKR1B10 loss. In
   addition, c-Myc transcriptionally represses AKR1B10, establishing a
   feedback loop that sustains its down-regulation and enhances metastatic
   progression. This study uncovers an antimetastatic mechanism involving
   AKR1B10-mediated PP2A activation and highlights its potential as a
   biomarker and therapeutic target.
     __________________________________________________________________

   Loss of AKR1B10 in colorectal cancer drives metastasis by disrupting
   PP2A-mediated c-Myc inactivation.

INTRODUCTION

   Colorectal cancer (CRC) ranks as the third most prevalent malignancy
   worldwide, with high rates of relapse and mortality ([44]1). Despite
   notable advancements in CRC therapeutics, metastasis and recurrence
   remain major challenges to improving patient outcomes ([45]2).
   Approximately 60 to 70% of CRC cases are diagnosed at advanced stages,
   with only 30 to 35% of patients surviving for more than 3 years and
   fewer than 20% achieving 5-year survival ([46]3, [47]4). The elevated
   abnormal gene expression, high mutational burden, extensive molecular
   heterogeneity, and tumor cell plasticity of CRC collectively pose
   tremendous obstacles in treating metastatic disease ([48]2).
   Elucidating the molecular mechanisms driving metastasis is critical for
   developing biomarker-driven, personalized therapeutic strategies.

   Protein phosphatase 2A (PP2A) is a ubiquitously expressed, multifaceted
   serine threonine phosphatase comprising a catalytic (C), scaffold (A),
   and a highly diverse B regulatory subunit. PP2A’s dephosphorylation
   activity is essential for embryogenesis and key cellular processes,
   including signal transduction, cell proliferation, apoptosis, and
   metabolic regulation ([49]5–[50]7). Widely recognized as a celebrated
   tumor suppressor, PP2A is frequently altered or inactivated in various
   malignancies, leading to aberrant hyperphosphorylation of oncogenic
   signaling molecules, such as constitutive RAS activity ([51]8–[52]10).
   This dysfunction synergizes with other oncogenic events, contributing
   to cancer progression, metastasis, and therapy resistance
   ([53]11–[54]16). The structural diversity of PP2A’s regulatory B
   subunits determines substrate specificity, enabling the phosphatase to
   target key tumorigenic regulators such as extracellular
   signal–regulated kinase (ERK), AKT, p53, Bcl-2, and c-Myc
   ([55]17–[56]20). These B subunits are classified into four distinct
   families: PPP2R5/B56, PPP2R2/B55, PPP2R3/PR72/PR130, and Striatin
   ([57]21), and their accessibility to bind the core AC dimer, along with
   posttranslational modifications across PP2A subunits, dynamically
   fine-tunes PP2A’s functional versatility ([58]22). For example,
   methylation at L309 on the C subunit facilitates selective binding of
   B55/B56 family members to the AC dimer ([59]23–[60]25). Nitration of B
   subunits suppresses holoenzyme assembly ([61]18, [62]26). Thus, PP2A
   dysregulation profoundly reprograms the cancer proteome, positioning
   PP2A as an appealing therapeutic target.

   Aldo-keto reductase family 1 member B10 (AKR1B10) is a reduced form of
   nicotinamide adenine dinucleotide phosphate (NADPH)–dependent monomeric
   enzyme that belongs to the AKR 1B subfamily ([63]27, [64]28). It
   catalyzes the reduction of carbonyl substrates, including aldehydes,
   lipid peroxidation products, and xenobiotics, thereby protecting cells
   from carbonyl-induced toxicity ([65]29–[66]31). In addition, AKR1B10
   attenuates retinoic acid signaling through retinaldehyde reduction
   ([67]32, [68]33) and recycles farnesal or geranylgeraniol,
   intermediates in cholesterol synthesis ([69]34). While AKR1B10 is
   highly up-regulated in malignancies such as hepatocellular carcinoma
   (HCC), pancreatic carcinoma, cholangiocarcinoma, lung cancer, and
   breast cancer, where it promotes oncogenesis and drug resistance
   ([70]28, [71]35–[72]38), it is paradoxically down-regulated in
   gastrointestinal cancers ([73]39). Despite being recognized as a
   prognostic marker and therapeutic target, particularly in HCC ([74]40),
   the implications of AKR1B10 down-regulation in gastrointestinal cancer
   progression remain poorly understood.

   In this study, we demonstrate that AKR1B10 suppresses gastrointestinal
   cancer metastasis through redox-dependent posttranslational
   modification of the PP2A–c-Myc axis. Loss of AKR1B10 disrupts PP2A
   holoenzyme assembly, sustaining phosphorylation of c-Myc at serine-62
   (S62), which promotes epithelial-mesenchymal transition (EMT) and
   enhances the metastatic potential of CRC. Furthermore, activated c-Myc
   transcriptionally represses AKR1B10, creating a self-sustaining
   feedback loop. These findings reveal AKR1B10 as a critical regulator of
   PP2A enzymatic activity and underscore the therapeutic potential of
   restoring PP2A activity to counteract AKR1B10 loss–driven CRC
   metastasis.

RESULTS

Down-regulation of AKR1B10 correlates with CRC progression and poor prognosis

   To assess the clinical relevance of AKR1B10 in CRC, we analyzed its
   mRNA and protein expression levels through The Cancer Genome Atlas
   (TCGA) and cProSite databases. Unlike its up-regulation in HCC, breast
   cancer, and lung cancer, AKR1B10 was down-regulated in CRC and gastric
   cancer (GC) tissues ([75]Fig. 1, A and B, and fig. S1, A and B). This
   finding was validated in paired CRC and GC tissue samples from the
   First Affiliated Hospital of Sun Yat-sen University (SYSU-FAH)
   ([76]Fig. 1, C and D, and fig. S1C) and further corroborated by
   immunohistochemical (IHC) analysis of tissue microarrays (TMAs)
   comprising 217 paired CRC tissue samples ([77]Fig. 1, E and F, and
   table S1). Reduced AKR1B10 expression was strongly associated with
   advanced stages and higher incidence of lymph nodes and distant organ
   metastases ([78]Fig. 1, G to J, and fig. S1, D and E). Survival
   analyses from TCGA, Gene Expression Omnibus (GEO), and SYSU-FAH
   datasets consistently demonstrated that low AKR1B10 expression
   predicted poor outcomes in patients with CRC and GC ([79]Fig. 1, K and
   L, and fig. S1, F to H). Furthermore, multivariate analysis identified
   AKR1B10 as an independent prognostic factor for CRC (fig. S1I). These
   data establish a clear inverse relationship between AKR1B10 expression
   and CRC aggressiveness and prognosis.

Fig. 1. AKR1B10 down-regulated in CRC correlates with poor prognosis.

   [80]Fig. 1.
   [81]Open in a new tab

   (A) Paired analysis of AKR1B10 mRNA expression in adjacent normal
   tissues versus primary tumor samples from TCGA colon adenocarcinoma
   (COAD; n = 41), rectum adenocarcinoma (READ; n = 9), and liver HCC
   (LIHC; n = 49) database. FPKM, fragments per kilobase of exon model per
   million mapped fragments. (B) Protein expression of AKR1B10 in COAD
   tissues compared to paired normal tissues based on cProSite database
   (n = 96). (C) Quantitative polymerase chain reaction (qPCR) analysis of
   AKR1B10 mRNA expression in paired CRC (n = 30)/GC (n = 21) tissues and
   adjacent normal tissues from SYSU-FAH. (D) Western blot analysis of
   AKR1B10 protein expression in 14 paired adjacent normal tissues (N) and
   CRC tissues (T) from SYSU-FAH. Rel, relative. (E and F) Representative
   IHC staining (E) and quantification (F) of AKR1B10 expression in CRC
   tumor microarrays (TMAs). Scale bars, 50 μm. (G to J) Analysis of
   AKR1B10 protein expression in CRC TMAs stratified by T stages (G),
   National Comprehensive Cancer Network stages (H), lymph node metastases
   (I), and distant metastases (J). n.s., not significant. (K and L)
   Kaplan-Meier overall survival curves of patients with low versus high
   AKR1B10 expression, derived from TCGA-COAD (n = 430)/READ (n = 154)
   databases (K), and IHC analysis of patients with CRC from SYSU-FAH (L)
   (n = 93). Data are analyzed with unpaired Student’s t test (F and H to
   J), paired Student’s t test (A to D), and log-rank test (K and L).

AKR1B10 deficiency promotes CRC metastasis

   To explore the functional role of AKR1B10 in CRC progression, we
   ectopically overexpressed or silenced AKR1B10 in CRC cell lines,
   including SW1116, HCT116, and LoVo ([82]Fig. 2A). Of them, SW1116 is
   microsatellite stable, HCT116 is microsatellite instability high
   (MSI-H), and LoVo displays a unique hybrid profile with MSI-low
   features. All three cell lines constitutively express the oncoprotein
   c-Myc. Knockdown of AKR1B10 marginally affected CRC cell propagation
   both in vitro and in vivo, whereas overexpression of AKR1B10 slightly
   enhanced CRC cell proliferation ([83]Fig. 2, B to E, and fig. S2A).
   Moreover, AKR1B10 depletion did not alter cell cycle progression but
   led to a modest reduction in apoptosis in CRC cells (fig. S2, B to E).
   AKR1B10 deficiency markedly increased cell migration, colony formation,
   and wound healing abilities compared to control groups ([84]Fig. 2, F
   and H, and fig. S2, F and G). Conversely, AKR1B10 overexpression
   suppressed these aggressive traits ([85]Fig. 2, G and I, and fig. S2H).
   We also observed similar results in GC cell lines (MKN45 and AGS) (fig.
   S2, I to O). To confirm these observations in vivo, we used an
   intravenous injection model of pulmonary metastasis and an intrasplenic
   injection model for liver metastasis. Mice injected with
   AKR1B10-overexpressing LoVo cells exhibited fewer lung metastatic foci
   ([86]Fig. 2J), while those receiving AKR1B10-deficient HCT116 cells
   showed a marked increase in metastatic nodules in both lung and liver
   tissues, as evidenced by ex vivo quantification and hematoxylin and
   eosin (H&E) staining ([87]Fig. 2, K to N). The AKR1B10 expression was
   markedly down-regulated in metastatic lesions compared to primary
   tumors, as evidenced by analyses of the colon adenocarcinoma (COAD)–GEO
   dataset and our independent cohort ([88]Fig. 2, O to Q). These findings
   collectively demonstrate that AKR1B10 acts as a suppressor of CRC
   metastasis irrespective of microsatellite instability status.

Fig. 2. AKR1B10 curbs CRC metastasis in vitro and in vivo.

   [89]Fig. 2.
   [90]Open in a new tab

   (A) Western blot analysis of AKR1B10 protein levels in human CRC (hCRC)
   cell lines. (B) Validation of AKR1B10 knockdown efficiency in HCT116
   and SW1116 cells via Western blot and assessment of cell proliferation
   with Cell Counting Kit-8 (CCK-8) assays (n = 4). (C) Verification of
   AKR1B10 overexpression in LoVo cells by Western blot, with
   corresponding CCK-8 proliferation analysis (n = 4). (D and E)
   Representative images (D) and statistical analysis (E) of xenograft
   tumor volumes in nude mice implanted with HCT116 (shNC&shAKR1B10) cells
   (n = 5 mice per group). (F and G) Representative transwell assay images
   (left) and quantification (right) in HCT116/SW1116 cells (F) or LoVo
   cells after 48 hours (G). (H and I) Representative clonogenic assay
   images and quantification in HCT116/SW1116 cells (H) or LoVo cells (I).
   (J) H&E staining of lung metastatic nodules following intravenous
   injection of LoVo (Ctrl&AKR1B10-OE) cells. Ctrl, control; OE,
   overexpression. Scale bars, 250 μm. (K and L) Representative images (K)
   and quantification of lung tumor nodule area per section (L) in nude
   mice intravenously injected with HCT116 (shNC&shAKR1B10) cells (n = 5
   mice per group). (M and N) Representative images and quantification of
   liver metastatic nodules in nude mice after splenic injection of HCT116
   (shNC&shAKR1B10) cells (M), with corresponding H&E staining of liver
   metastatic lesions (N) (n = 5 mice per group). (O) mRNA expression
   levels of AKR1B10 in primary tumors (n = 56) and metastases (n = 27;
   liver = 23, peritoneum = 3, and lung = 1) based on [91]GSE28702. (P and
   Q) Representative IHC images (P) and quantification (Q) of AKR1B10 in
   paired primary CRC tissues and liver metastatic lesions from 20
   patients. Scale bars, 50 μm. (F to I) n = 3. Data are presented as
   mean ± SD, with two-way analysis of variance (ANOVA) test in (C) and
   (E) or unpaired Student’s t test (F to I, L, M, O, and Q). *P < 0.05,
   **P < 0.01, and ***P < 0.001.

AKR1B10 curbs CRC metastasis by repressing integrin transcription

   While AKR1B10 has been implicated in reducing glycolytic capacity in
   breast cancer cells, supporting their survival under low glucose
   conditions and facilitating metastatic colonization in the lung
   ([92]41), our Seahorse XF Glycolysis Stress tests showed that AKR1B10
   depletion did not modulate compensatory glycolysis but attenuated basal
   glycolytic activity in CRC/GC cell lines, suggesting cancer
   type–specific metabolic adaptation mechanisms (fig. S3A). Similarly,
   down-regulation of Akr1b8, the murine ortholog of human AKR1B10, did
   not affect MC38 cell proliferation in vitro ([93]Fig. 3, A and B).
   Notably, tumors derived from MC38-shAkr1b8 cells exhibited a notable
   increase in liver metastasis in both nonobese diabetic (NOD)–severe
   combined immunodeficient (SCID) and C57BL/6 mouse models ([94]Fig. 3, C
   and D), suggesting that AKR1B10 and Akr1b8 suppress CRC metastasis via
   cell-intrinsic mechanisms.

Fig. 3. AKR1B10 suppresses EMT by down-regulating integrin signaling.

   [95]Fig. 3.
   [96]Open in a new tab

   (A) Western blot analysis of Akr1b8 knockdown efficiency in MC38 cells.
   (B) Cell proliferation rates of MC38 (shNC and shAkr1b8) cells assessed
   using the CCK-8 assay (n = 4). (C) Representative images of liver
   metastatic nodules in NOD-SCID mice on day 15 following splenic
   injection of MC38 shNC and shAkr1b8 cells. Liver weight and the number
   of metastases were quantified (n = 6 mice per group). (D)
   Representative images of liver metastatic nodules in C57BL/6 mice on
   day 18 following splenic injection of MC38 (shNC and shAkr1b8) cells.
   Liver weight and the number of metastases were quantified (n = 6 mice
   per group). (E) Western blot analysis of EMT markers and EMT-related
   TFs in HCT116 and SW1116 cells. GAPDH, glyceraldehyde-3-phosphate
   dehydrogenase. (F) Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analysis of the top 20 up-regulated biological
   processes in shAKR1B10 cells based on RNA sequencing (RNA-seq). PI3K,
   phosphatidylinositol 3-kinase; MAPK, mitogen-activated protein kinase;
   TGF-β, transforming growth factor–β; AGE-RAGE, Advanced Glycation End
   products–Receptor for Advanced Glycation End products; FoxO, Forkhead
   box O. (G) Heatmap depicting mRNA levels of genes from selected KEGG
   pathways. (H) Western blot analysis of ITGB8 protein and its downstream
   factors in HCT116 and SW1116 cells. p-ERK, phosphorylated ERK. (I)
   Western blot analysis of ITGB8 protein and its downstream EMT markers
   in HCT116 and SW1116 cells after ITGB8 silencing. (J and K)
   Representative images (J) and quantification (K) of transwell assays in
   HCT116 and SW1116 cells after ITGB8 silencing. (C, D, and K) n = 3.
   Data are presented as mean ± SD, with two-way ANOVA test (B) or
   unpaired Student’s t test (C, D, and K). ***P < 0.001.

   The EMT is a crucial process conducive to tumor dissemination ([97]42).
   Upon AKR1B10 depletion, CRC cells underwent a morphological shift from
   cuboidal epithelial shapes to elongated mesenchymal forms (fig. S3B),
   accompanied by reduced E-cadherin levels and increased expression of
   mesenchymal markers, including N-cadherin, vimentin, and fibronectin,
   as well as EMT drivers such as Twist, ZEB1, and Snail ([98]Fig. 3E).
   RNA sequencing (RNA-seq) transcriptome profiling further revealed
   up-regulation of cell adhesion–related pathways, including
   extracellular matrix (ECM)–receptor interactions and focal adhesion, in
   shAKR1B10 cells ([99]Fig. 3, F and G). Among these pathways, integrins
   emerged as prominently up-regulated, with TCGA database analysis
   corroborating an inverse correlation between AKR1B10 expression and
   integrin gene expression in CRC and GC tissues (fig. S3C). Downstream
   integrin pathways, such as AKT and ERK signaling, were activated in
   AKR1B10-deficient cells, along with a notable up-regulation of ITGB8
   expression ([100]Fig. 3H and fig. S3D). Knocking down ITGB8 in CRC
   cells reduced mesenchymal marker expression (e.g., N-cadherin and
   vimentin) at both mRNA and protein levels, effectively reversing the
   migratory and colonization advantages conferred by AKR1B10 abrogation
   ([101]Fig. 3, I to K, and fig. S3, E and F). Similar results were
   observed in GC cells (fig. S4). Together, these findings suggest that
   AKR1B10 loss enhances CRC cell migratory and invasive potential through
   integrin transcriptional induction.

AKR1B10 transcriptionally regulates integrins by modulating c-Myc stability

   To identify transcription factors (TFs) mediating AKR1B10’s regulation
   of integrin expression, we analyzed proteins that bind integrin gene
   promoters in HCT116 cells and the ITGB8 promoter across various tumor
   cell lines using the chromatin immunoprecipitation (ChIP)–Atlas
   database. This analysis identified three candidates: c-Myc, RUVBL2, and
   MAX ([102]Fig. 4A). Notably, AKR1B10 depletion increased c-Myc protein
   levels in CRC and GC cells, while RUVBL2 and MAX levels remained
   unchanged ([103]Fig. 4B and fig. S5, A and B). Furthermore, c-Myc
   knockdown substantially reduced mRNA levels of integrins, particularly
   ITGB8, while silencing RUVBL2 had minimal effects. Knocking down MAX,
   an obligate dimerization partner of c-Myc, displayed a similar
   suppressive effect on integrin expression, supporting the role of the
   c-Myc–MAX heterodimer as the transcriptional driver of integrin genes
   in gastrointestinal tumor cells ([104]Fig. 4C and fig. S5, C to E).
   Depleting c-Myc in shAKR1B10 CRC/GC cells also suppressed mesenchymal
   marker expression and impaired their migratory capacity and
   anchorage-independent and anchorage-dependent growth ([105]Fig. 4, D to
   H, and fig. S5, F to I). c-Myc has a short half-life that is rigorously
   controlled by phosphorylation and proteasomal degradation ([106]43).
   Treatment with the proteasome inhibitor MG132 elevated c-Myc levels in
   control cells but had less effect in AKR1B10-deficient cells,
   suggesting that AKR1B10 loss hinders the proteasomal degradation of
   c-Myc ([107]Fig. 4I and fig. S5J). Cycloheximide (CHX) chase assays
   further revealed a reduced c-Myc protein decay rate in
   AKR1B10-deficient cells, supporting its role in c-Myc stabilization
   ([108]Fig. 4J and fig. S5K). Together, these data demonstrate that
   AKR1B10 facilitates c-Myc proteasomal degradation, thereby repressing
   integrin transcription and suppressing metastatic traits in CRC.

Fig. 4. AKR1B10 destabilizes c-Myc to repress integrin transcription.

   [109]Fig. 4.
   [110]Open in a new tab

   (A) The Venn diagram showing three TFs capable of binding to integrin
   gene promoter regions in HCT116 cells and to the ITGB8 promoter region
   in various cancers, based on the ChIP-Atlas database. (B) Western blot
   analysis of c-Myc, RUVBL2, and MAX protein expression in HCT116 and
   SW1116 (shNC&shAKR1B10) cells. (C) qPCR analysis of integrin mRNA
   expression in HCT116 and SW1116 cells after c-Myc silencing. (D)
   Western blot analysis of ITGB8 protein and its downstream EMT markers
   in HCT116 and SW1116 cells after c-Myc silencing. (E and F)
   Representative images (E) and quantification (F) of transwell assay in
   HCT116 and SW1116 cells after c-Myc silencing. (G and H) Representative
   images (G) and quantification (H) of spheroid formation assays in
   HCT116 and SW1116 cells after c-Myc silencing. Scale bars, 50 μm. (I)
   Western blot analysis of c-Myc expression in HCT116 and SW1116
   (shNC&shAKR1B10) cells after treatment with MG132 (10 μM for 8 hours).
   c-Myc protein levels were quantified and normalized to loading control
   (vinculin). (J) Detection of c-Myc protein turnover in HCT116 and
   SW1116 cells by Western blot after CHX (100 μg/ml) treatment and
   quantified by ImageJ software. h, hours. (C, F, H, and J) n = 3. Data
   are presented as mean ± SD, with unpaired Student’s t test (C, F, and
   H) or with two-way ANOVA test (J). **P < 0.01 and ***P < 0.001.

AKR1B10 destabilizes c-Myc by facilitating PP2A holoenzyme assembly

   The stability of c-Myc is meticulously regulated by phosphorylation at
   S62 and threonine-58 (T58). Phosphorylation at S62 (pS62) stabilizes
   c-Myc, while its dephosphorylation by PP2A enables proteasomal
   degradation of phospho-T58 (pT58) Myc ([111]44, [112]45). In
   AKR1B10-depleted CRC/GC cells, we observed elevated levels of pS62
   c-Myc and total c-Myc, while pT58 c-Myc remained unchanged ([113]Fig.
   5A and fig. S6A). To explore the mechanisms, immunoprecipitation–mass
   spectrometry (IP-MS) identified the catalytic subunit of PP2A (PPP2CA)
   as a top AKR1B10 interactor, consistent with Kyoto Encyclopedia of
   Genes and Genomes (KEGG) analysis linking AKR1B10 to posttranslational
   modification and signal transduction pathways ([114]Fig. 5B and fig.
   S6B). Coimmunoprecipitation (Co-IP) confirmed interactions between
   Flag-tagged AKR1B10 and hemagglutinin (HA)–tagged PPP2CA, as well as
   endogenous AKR1B10 and PPP2CA in CRC/GC cells, with colocalization
   observed primarily in the cytosol ([115]Fig. 5, C to E, and fig. S6, C
   and D). AKR1B10 loss did not affect the protein expression or
   proteasomal degradation of PPP2CA and B56α (PPP2R5A), the regulatory
   subunit known to mediate PP2A-driven c-Myc dephosphorylation ([116]Fig.
   5F and fig. S6, E to G) ([117]19, [118]26). However, AKR1B10 depletion
   substantially reduced PP2A activity ([119]Fig. 5G and fig. S6H).
   Overexpression of AKR1B10 in shAKR1B10 CRC cells restored
   PPP2CA–B56α–c-Myc interactions and promoted c-Myc degradation in a
   dose-dependent manner ([120]Fig. 5H and fig. S6I).

Fig. 5. Targeting AKR1B10 enhances c-Myc stability by disrupting PP2A
assembly.

   [121]Fig. 5.
   [122]Open in a new tab

   (A) Western blot analysis of c-Myc protein expression, and its
   phosphorylation status in HCT116 and SW1116 (shNC&shAKR1B10) cells. (B)
   IP-MS analysis to identify AKR1B10-interacting kinases and
   phosphatases. m/z, mass/charge ratio. (C) Co-IP analysis showing the
   physical interaction of exogenous AKR1B10 with PPP2CA in HCT116 and
   SW1116 cells. (D) Physical interaction of endogenous AKR1B10 with
   PPP2CA was detected by Co-IP with anti-AKR1B10 antibody in HCT116 and
   SW1116 cells. (E) Immunofluorescent staining showing the colocalization
   of endogenous AKR1B10 with PPP2CA in HCT116 and SW1116 cells. Scale
   bars, 20 μm. DAPI, 4′,6-diamidino-2-phenylindole. (F) Western blot
   analysis of PPP2CA and PPP2R5A protein expression in HCT116 and SW1116
   (shNC&shAKR1B10) cells. (G) PP2A activity assay in HCT116 and SW1116
   (shNC&shAKR1B10) cells (n = 3). (H) Co-IP analysis of HA-PPP2CA in
   HCT116 cells coexpressing Flag-AKR1B10 with varying plasmid amounts, as
   indicated. (I) Co-IP analysis of PPP2R5A indicating increased
   3-nitrotyrosine (3-NT) levels and reduced PPP2CA interaction in
   HCT116-shAKR1B10 cells. Pan-Ace, Pan Acetylation. (J and K) Co-IP
   analysis of PPP2R5A showing reduced 3-NT levels and increased
   interaction with PPP2CA in the presence of DPI [10 μM for 20 hours (J)]
   or FeTPPS [10 μM for 3 hours (K)] in HCT116 (shNC&shAKR1B10) cells. (L
   and M) Representative fluorescence-activated cell sorting (FACS) images
   (L) and quantification of intracellular reactive oxygen species (ROS)
   levels, and relative PP2A activities [(M); n = 3] in HCT116
   (shNC&shAKR1B10) cells treated with DPI (10 μM for 20 hours). DMSO,
   dimethyl sulfoxide. (N and O) His-PPP2R5A-WT or His-PPP2R5A-Y238F
   plasmid was transfected into HCT116-shAKR1B10 cells with or without
   Flag-AKR1B10. Co-IP analysis showing reduced 3-NT levels of
   His-PPP2R5A-Y238F (N). Western blot demonstrating decreased total
   c-Myc, pS62 c-Myc, and reduced ITGB8 expression in the presence of
   His-PPP2R5A-Y238F (O). Data in (G) and (M) are presented as mean ± SD,
   with unpaired Student’s t test. **P < 0.01 and ***P < 0.001.

   The nitration of B56α by peroxynitrite (ONOO^−), a product of
   superoxide (O2^●–) and nitric oxide (NO), inhibits PP2A holoenzyme
   assembly required for c-Myc dephosphorylation ([123]26).
   Immunoprecipitated B56α lysates from CRC cells showed increased
   3-nitrotyrosine (3-NT) levels and reduced B56α-PPP2CA interactions in
   AKR1B10-depleted cells. This was accompanied by elevated pS62 c-Myc and
   total c-Myc, with no changes in B56α phosphorylation (serine/threonine)
   or acetylation ([124]Fig. 5I). These findings indicate that AKR1B10
   deficiency fosters a prooxidant microenvironment promoting nitrative
   inhibition of B56α, impairing PP2A activity and c-Myc destabilization.
   Treatment with the NADPH oxidase inhibitor diphenyleneiodonium chloride
   (DPI) reduced O2^●– levels, mitigating B56α nitration and restoring
   B56α-PPP2CA interaction. DPI treatment accelerated c-Myc
   dephosphorylation and degradation in AKR1B10-depleted cells ([125]Fig.
   5J and fig. S6J). Similarly, Fluorescent-tagged Thiol-Protected
   Proteolytic Substrate (FeTPPS), a ONOO^− scavenger, reversed nitrative
   inhibition of B56α and restored PP2A activity by reducing NO production
   ([126]Fig. 5K and fig. S6M). DPI and FeTPPS treatment profoundly
   enhanced PP2A activity, with DPI reducing intracellular reactive oxygen
   species (ROS) levels, countering the effects of AKR1B10 loss ([127]Fig.
   5, L and M, and fig. S6, K, L, and N to Q). Critically, the
   nonnitratable B56α Y238F mutant ([128]26) effectively reversed AKR1B10
   loss–induced B56α nitration. This mutant synergized with AKR1B10
   reconstitution, amplifying the reduction in 3-NT levels ([129]Fig. 5N
   and fig. S6R). Unlike wild-type (WT) B56α, the Y238F mutant failed to
   stabilize phosphorylated c-Myc or activate c-Myc signaling ([130]Fig.
   5O and fig. S6S). Collectively, these results demonstrate that AKR1B10
   facilitates PP2A holoenzyme assembly, destabilizing c-Myc by promoting
   its dephosphorylation at S62.

L125 of AKR1B10 is essential for PP2A assembly and CRC metastasis suppression

   While DPI and FeTPPS partially restored B56α-PPP2CA interactions in
   shAKR1B10 CRC cells, the levels remained below those of control cells
   ([131]Fig. 5, J and K). This indicates that AKR1B10 contributes to PP2A
   holoenzyme assembly not solely by reducing nitrative inhibition but
   also through direct interaction with PPP2CA. AKR1B10 was found to
   interact with PPP2CA independent of its N-terminal residues (1 to 100)
   ([132]Fig. 6A). Protein-protein docking simulation identified potential
   interaction sites between AKR1B10 and the PP2A A/Cα core enzyme, rather
   than B56α. Specifically, there are three residues in AKR1B10 that may
   form interactions with corresponding amino acids in PPP2CA ([133]Fig.
   6B and fig. S7A). Among them, AKR1B10^K125L markedly weakened the
   interaction with PPP2CA, reducing its association with B56α and c-Myc,
   supporting our hypothesis that AKR1B10 facilitates c-Myc–bound PP2A
   holoenzyme assembly through direct PPP2CA binding ([134]Fig. 6C).
   Furthermore, glutathione S-transferase (GST) pull-down assays with
   recombinant proteins confirmed that AKR1B10 selectively interacts with
   PPP2CA via the K125 residue while showing no direct binding to the PP2A
   Aα (PPP2R1A) ([135]Fig. 6, D and E, and fig. S7, B and C).

Fig. 6. AKR1B10^K125L impedes PP2A assembly to drive CRC metastasis.

   [136]Fig. 6.
   [137]Open in a new tab

   (A) Schematic representation of AKR1B10 truncations and Co-IP analysis
   of Flag-tagged AKR1B10 truncations in HCT116 cells, as indicated. FL,
   full length. (B) Computational docking model predicting the interaction
   between AKR1B10 (red) and PP2A (purple) using the Global RAnge
   Molecular Matching (GRAMM) algorithm. (C) Co-IP analysis showing the
   interaction between PPP2CA with AKR1B10^WT, AKR1B10^K125L,
   AKR1B10^P219S, and AKR1B10^K263R in HCT116 cells, using an anti-HA
   antibody. (D and E) Recombinant GST-tagged AKR1B10 (WT or K125L) was
   incubated with His-PPP2CA (D) or His-PPP2R1A (E). Bound proteins were
   analyzed by Western blot. Data are representative of three independent
   experiments. (F) Representative FACS images and quantification of
   intracellular ROS levels in HCT116 shNC and shAKR1B10 cells expressing
   mock, AKR1B10^WT, AKR1B10^K125L, and AKR1B10^C299S. (G) Alignment of
   K125 and C299 amino acid residues in AKR1B10 across multiple species.
   H. sapiens, Homo sapiens; M. musculus, Mus musculus; R. norvegicus,
   Rattus norvegicus; P. troglodytes, Pan troglodytes; M. mulatta, Macaca
   mulatta. (H) Co-IP analysis of PPP2R5A demonstrating increased 3-NT and
   reduced PPP2CA interaction in AKR1B10^K125L and AKR1B10^C299S
   overexpressing HCT116-shAKR1B10 cells. (I) Western blot analysis of
   c-Myc, phospho-Myc (p-Myc), and ITGB8 protein expression in HCT116 shNC
   and shAKR1B10 cells expressing mock, AKR1B10^WT, AKR1B10^K125L, and
   AKR1B10^C299S. (J and K) Representative images (left) and
   quantification (right) of transwell assay (J), and relative PP2A
   activity (K) in HCT116 shNC and shAKR1B10 cells expressing mock,
   AKR1B10^WT, AKR1B10^K125L, and AKR1B10^C299S. (L to O) Representative
   images (L) and quantification (M) of liver metastatic nodules in nude
   mice following splenic injection of HCT116 shNC and shAKR1B10 cells
   expressing mock, AKR1B10^WT, and AKR1B10^K125L. Representative H&E
   staining and IHC images of AKR1B10, ITGB8, and c-Myc staining in liver
   sections (N), with quantified histoscores (O) (n = 5 mice per group).
   (F, J, and K) n = 3. Data in (F), (J), (K), (M), and (O) are presented
   as mean ± SD, with unpaired Student’s t test. *P < 0.05, **P < 0.01,
   and ***P < 0.001.

   AKR1B10 exhibits superior catalytic efficiency for retinaldehyde and
   detoxifies lipid-derived reactive carbonyls ([138]46, [139]47). To
   assess the importance of AKR1B10’s catalytic activity in PP2A assembly,
   we reconstituted shAKR1B10 CRC cells with AKR1B10 WT or catalytically
   impaired mutants: K125L, V301L (reducing catalytic activity for
   all-trans-retinaldehyde) ([140]32), and C299S (targeting catalytic
   activity for 4-hydroxynonenal) ([141]47). While AKR1B10^WT and
   AKR1B10^V301L normalized intracellular ROS levels, AKR1B10^K125L and
   AKR1B10^C299S failed to reduce ROS. K125 and C299, unlike V301, are
   evolutionarily conserved, underscoring their critical roles in
   AKR1B10’s reductase function ([142]Fig. 6, F and G, and fig. S7, D to
   F). AKR1B10^WT reconstitution also restored B56α-PPP2CA interactions,
   which were impaired in cells expressing AKR1B10^K125L and AKR1B10^C299S
   ([143]Fig. 6H and fig. S7G). Functionally, AKR1B10^WT reconstitution
   suppressed c-Myc and ITGB8 expressions, reduced shAKR1B10 cells’
   migratory capability, and restored PP2A activity, eliminating
   off-target effects of short hairpin RNA (shRNA). However, AKR1B10^K125L
   and AKR1B10^C299S mutants did not achieve comparable outcomes
   ([144]Fig. 6, I to K, and fig. S7, H to L). Notably, AKR1B10^K125L
   showed a stronger inhibitory effect on PP2A assembly and downstream
   signaling than AKR1B10^C299S, which did not disrupt PPP2CA binding,
   highlighting the dual requirement of AKR1B10’s catalytic activity and
   direct interaction with PPP2CA for PP2A functionality (fig. S7M). In
   vivo, AKR1B10^WT, but not AKR1B10^K125L, attenuated liver metastasis of
   AKR1B10-deficient CRC cells in the intrasplenic injection mouse model
   ([145]Fig. 6, L and M), nor did AKR1B10^K125L reduce c-Myc and ITGB8
   expression within liver metastatic lesions ([146]Fig. 6, N and O).
   These findings demonstrate that K125 is indispensable for
   AKR1B10-mediated PP2A assembly and metastasis suppression in CRC.

Pharmacological activation of PP2A suppresses AKR1B10 depletion-driven
metastasis

   To evaluate the therapeutic potential of PP2A activation, we used
   DT-061 [a small-molecule activator of PP2A (SMAP)], which stabilizes
   the B56α-PP2A holoenzyme in its active state to dephosphorylate c-Myc
   ([147]22). DT-061 treatment reduced c-Myc and ITGB8 protein levels and
   suppressed the migratory and colonization ability of AKR1B10-deficient
   CRC/GC cells. We observed similar effects with FTY-720, another PP2A
   activator that targets the endogenous PP2A inhibitor SET ([148]Fig. 7,
   A to C, and fig. S8, A to H). Conversely, pharmacological inhibition of
   PP2A with LB-100 at subinhibitory concentrations robustly reversed
   c-Myc signaling in control groups, with less pronounced effects in
   shAKR1B10 CRC cells, underscoring PP2A’s role in regulating
   c-Myc–driven oncogenic plasticity, particularly under AKR1B10-deficient
   conditions (fig. S8, I to L). In an HCT116 intrasplenic liver
   metastasis model, DT-061 treatment markedly suppressed liver metastasis
   induced by AKR1B10 depletion ([149]Fig. 7, D and E). In an orthotopic
   CRC model, mice implanted with HCT116-shAKR1B10 cells in the caecum
   exhibited more and larger hepatic metastases, compared to the controls,
   and DT-061 treatment reduced both the number and size of metastases
   ([150]Fig. 7, F to J, and fig. S9A). However, DT-061 did not
   significantly affect the size or weight of orthotopic primary tumors,
   although it modestly reduced Ki67 expression ([151]Fig. 7, H and J, and
   fig. S9B), suggesting limited antitumor efficacy in primary lesions. In
   addition, knocking down Akr1b8 promoted hepatic metastases in the CT26
   syngeneic orthotopic CRC model. Conversely, DT-061 treatment
   effectively reduced the incidence of liver metastasis in this system
   ([152]Fig. 7, K to N, and fig. S9, C to E). These data underscore the
   critical role of PP2A signaling in AKR1B10-mediated metastasis and
   highlight the potential of a pharmacological PP2A activator as a
   therapeutic strategy against AKR1B10-deficient CRC metastasis.

Fig. 7. PP2A activators counteract AKR1B10 loss–driven metastasis.

   [153]Fig. 7.
   [154]Open in a new tab

   (A and B) Representative images (A) and quantification (B) of transwell
   assay in HCT116 cells treated with DT-061 or FTY-720. (C) Western blot
   analysis of c-Myc, p-Myc, and ITGB8 protein expression in HCT116 cells
   treated with DT-061 or FTY-720. (D and E) Intrasplenic injection mouse
   models treated with DT-061 (SMAP) via gavage at 5 mg/kg on alternate
   days (n = 5 mice per group). Representative images and quantification
   of liver metastatic lesions 30 days following HCT116 injection (D).
   Representative H&E staining of metastatic nodules in the liver (E). (F
   to J) HCT116 orthotopic implantation mouse models treated with SMAP via
   gavage at 5 mg/kg on alternate days (n = 8 mice per group).
   Representative images of primary tumors and liver metastases (F).
   Quantitative analysis of liver metastasis incidence (G), primary tumor
   volume and weight (H), number of metastatic nodules (I), and Ki67 IHC
   staining in primary tumors (J). (K to N) CT26 orthotopic implantation
   models treated with SMAP via gavage at 5 mg/kg on alternate days (n = 8
   to 10 mice per group). Representative images of liver metastases (K).
   Quantitative analysis of liver metastasis incidence (L), primary tumor
   volume (M), and number of metastatic nodules (N). Data in (B), (D),
   (H), (I), (J), (M), and (N) are presented as mean ± SD, with unpaired
   Student’s t test. *P < 0.05, **P < 0.01, and ***P < 0.001.

c-Myc represses AKR1B10 transcription in gastrointestinal cancer cells

   To investigate the mechanism underlying the reduced AKR1B10 mRNA
   expression in CRC, especially its opposing expression pattern in
   gastrointestinal cancer versus HCC, we performed TF prediction analyses
   using PROMO, hTFtarget, and Gene Transcription Regulation Database
   (GTRD) databases. Among the identified candidates, CEBPB, c-Myc, and
   signal transducer and activator of transcription 1 (STAT1) exhibited an
   inverse coexpression pattern with AKR1B10 in gastrointestinal cancers
   compared to HCC ([155]Fig. 8A). Functional experiments revealed that
   c-Myc knockdown increased AKR1B10 expression in CRC and GC cells,
   whereas silencing CEBPB or STAT1 resulted in minor or negligible
   effects ([156]Fig. 8, B and C, and fig. S10, A to C). Overexpression of
   exogenous c-Myc reduced AKR1B10 protein levels in gastrointestinal
   cancer cells ([157]Fig. 8D). Analysis of TCGA datasets consistently
   revealed a robust negative correlation between c-Myc and AKR1B10
   expression in colon [correlation coefficient (r) = −0.35], rectal
   (r = −0.2), and stomach adenocarcinoma (r = −0.11) samples ([158]Fig.
   8E). By comparison, a modest positive correlation is evident in liver
   HCC (r = 0.11), with stronger positive correlations in lung
   adenocarcinoma (r = 0.18) and lung squamous cell carcinoma (r = 0.36)
   (fig. S10D). Bioinformatics revealed three putative c-Myc binding sites
   within the AKR1B10 promoter region ([159]Fig. 8F). ChIP confirmed c-Myc
   enrichment at these regions with different binding affinity ([160]Fig.
   8G). Mutating sites −850 ~ −861 and −1422 ~ −1433 together abolished
   c-Myc–mediated suppression of AKR1B10 promoter activity, as shown in
   luciferase reporter assays ([161]Fig. 8H). c-Myc knockdown in HCC cell
   lines (MHCC97-H and Huh7) exerted negligible effects on AKR1B10
   expression at either protein or transcriptional level, implying a
   cancer type–specific regulatory mechanism (fig. S10, E and F). These
   results identified AKR1B10 as a direct transcriptional target of c-Myc
   and established a c-Myc–dependent positive feedback loop, contributing
   to AKR1B10 down-regulation in CRC.

Fig. 8. c-Myc suppresses AKR1B10 transcription in gastrointestinal cancers.

   [162]Fig. 8.
   [163]Open in a new tab

   (A) Bioinformatics analysis identifying three putative TFs that
   regulate AKR1B10 expression and show an inverse coexpression pattern
   with AKR1B10 between gastrointestinal cancer and HCC, based on TCGA
   database. coeff., coefficient. (B) Quantitative reverse transcription
   (qRT)–PCR analysis of AKR1B10 mRNA expression in HCT116, SW1116, and
   AGS cells transfected with small interfering RNAs (siRNAs) targeting
   c-Myc, CEBPB, or STAT1. (C and D) Western blot analysis of AKR1B10
   protein expression in HCT116, SW1116, and AGS cells after c-Myc
   knockdown (C) or overexpression (D). (E) Correlation between AKR1B10
   and c-Myc mRNA levels in human COAD (n = 494), READ (n = 173), and
   stomach adenocarcinoma (STAD; n = 405) tissues, based on TCGA database.
   (F) Identification of putative c-Myc binding sites within the AKR1B10
   gene promoter region using the JASPAR database. (G) ChIP-qPCR analysis
   depicting the binding sites of c-Myc within the AKR1B10 promoter region
   in HCT116 cells. (H) Dual-luciferase reporter assay showing that
   mutations at both site 1 (−850 ~ −861) and site 2 (−1422 ~ −1433)
   abrogated c-Myc–mediated repression of AKR1B10 promoter activity. Luc,
   luciferase. Data in (B), (G), and (H) are presented as mean ± SD
   (n = 3), with unpaired Student’s t test. ***P < 0.001.

AKR1B10 is down-regulated in metastatic hepatic lesions and correlates with
CRC aggressiveness

   To assess the clinical significance of AKR1B10 in CRC metastasis, we
   performed IHC analyses to compare expression levels in primary CRC
   tissues and matched liver metastatic lesions. The AKR1B10 expression
   was markedly reduced in metastatic lesions, while c-Myc and ITGB8
   levels were elevated compared to primary tumors ([164]Fig. 9, A and B).
   Consistently, in an orthotopic CRC mouse model, AKR1B10 depletion
   markedly increased c-Myc and ITGB8 expression in both primary tumors
   and liver metastatic foci compared to controls ([165]Fig. 9, C and D).
   Correspondingly, TMA analysis of human CRC (hCRC) samples revealed a
   strong negative correlation between AKR1B10 and both c-Myc and ITGB8
   expressions ([166]Fig. 9E). Low AKR1B10 expression, especially in
   combination with high c-Myc or ITGB8 expression, strongly predicted
   poor prognosis in patients with CRC based on TMA analysis. Notably,
   concurrent high expressions of both c-Myc and ITGB8 correlated with
   unfavorable clinical outcomes. Patients with c-Myc^lowITGB8^high tumors
   exhibited worse prognosis compared to those with c-Myc^lowITGB8^low
   tumors, underscoring the critical oncogenic roles of c-Myc and ITGB8 in
   CRC progression ([167]Fig. 9, F and G). Together, these findings
   underscore the clinical relevance and prognostic value of AKR1B10, with
   key mechanistic insights into the AKR1B10–c-Myc–integrin axis in CRC
   metastasis.

Fig. 9. Dysregulation of the AKR1B10–c-Myc–ITGB8 axis drives CRC progression.

   [168]Fig. 9.
   [169]Open in a new tab

   (A and B) Representative H&E and IHC images (A) and quantification (B)
   of AKR1B10, ITGB8, and c-Myc staining in paired primary CRC tissues and
   liver metastatic lesions from 20 patients. (C and D) Representative IHC
   images (C) and histoscore (D) of AKR1B10, ITGB8, and c-Myc staining in
   primary tumors and liver sections from the HCT116 orthotopic
   implantation mouse model (n = 6). (E) Correlation analysis of IHC
   scores between AKR1B10 and ITGB8 or c-Myc in CRC TMAs. (F and G)
   Kaplan-Meier survival curves of patients with CRC stratified by high or
   low c-Myc/ITGB8 expression levels (F) and AKR1B10/ITGB8 or
   AKR1B10/c-Myc expression levels (G) from CRC TMAs. (H) Schematic
   diagram illustrating how down-regulated AKR1B10 accelerates CRC
   metastasis. Ub, ubiquitin. Data are presented as mean ± SD, with paired
   (B) or unpaired (D) Student’s t test or log-rank test (G).

DISCUSSION

   This study was motivated by the observed aberrant down-regulation of
   AKR1B10 in gastrointestinal cancers and its enigmatic role in disease
   progression. We demonstrate that AKR1B10 acts as a crucial suppressor
   of CRC metastasis by functioning as a bona fide endogenous activator of
   the PP2A holoenzyme through its redox enzymatic activity and direct
   interaction with PP2Ac. Loss of AKR1B10 stabilizes c-Myc via sustained
   S62 phosphorylation, promoting integrin transcription and EMT. Notably,
   the phospho-stabilized c-Myc transcriptionally represses AKR1B10,
   creating a positive feedback loop that amplifies its reduction in CRC.
   L125 of AKR1B10 is essential for PP2Ac binding and redox-regulated B56α
   nitration ([170]Fig. 9H). Pharmacologically restoring PP2A activity may
   offer a promising therapeutic avenue to counter AKR1B10
   down-regulation–driven metastasis. This study highlights AKR1B10 as a
   critical nexus between the redox milieu and metastasis, advancing
   understanding of redox-regulated posttranslational modification in
   fine-tuning the PP2A–c-Myc axis and offering insights into
   antimetastatic strategies.

   The contrasting expression patterns of AKR1B10 between gastrointestinal
   cancers and nondigestive tract malignancies highlight its dichotomic
   role in tumorigenesis. While AKR1B10 promotes tumorigenesis in HCC,
   breast cancer, lung cancer, and pancreatic cancer ([171]41,
   [172]48–[173]50), we demonstrate that its expression predicts better
   prognosis in gastrointestinal cancers, based on analyses of public
   databases and patient samples from FAH-SYSU. Our findings link AKR1B10
   down-regulation to c-Myc–ITGB8 activation, which drives EMT and
   facilitates secondary organ colonization. This conclusion is further
   supported by their inverse correlation in TMAs and metastatic samples
   derived from patients with CRC, underscoring AKR1B10–c-Myc–integrin
   signaling in governing inherent plasticity of CRC.

   AKR1B10 is known to exert oncogenic functions in nondigestive solid
   tumors by promoting fatty acid synthesis, activating KRAS via protein
   prenylation, and enhancing metabolic adaptability ([174]35, [175]41,
   [176]51). Our findings reveal a contrasting role for AKR1B10 as a tumor
   suppressor in CRC metastasis, mediated through posttranslational
   activation of PP2A. While AKR1B10 depletion exerted marginal effects on
   the glycolytic capacity of CRC cells, it suppressed gastrointestinal
   cancer metastasis through its reductive enzymatic activity targeting
   retinal and reactive aldehydes. However, its influence on other
   metabolic pathways, such as fatty acid synthesis and lipid metabolism,
   and their potential impact on CRC progression need to be clarified. In
   addition, as a secretory protein pivotal in managing inflammatory
   signaling pathways ([177]39), AKR1B10’s paracrine effects within the
   tumor microenvironment and its immunomodulatory roles in tumor
   progression warrant further elaboration.

   Deregulated redox homeostasis is a hallmark of cancer cells, driving
   tumor growth, promoting angiogenesis, and enhancing invasiveness and
   metastatic potential ([178]52, [179]53). As an NADPH-dependent
   oxidoreductase, AKR1B10 plays a crucial role in redox regulation
   through its enzymatic activity. It efficiently reduces reactive
   aldehydes like 4-hydroxynonenal, a byproduct of lipid peroxidation
   ([180]54). AKR1B10 knockdown elevates cellular carbonyl levels,
   exacerbating ROS accumulation and oxidative stress in CRC and HCC cells
   ([181]48, [182]55). In this study, we observed that AKR1B10 depletion
   increased ROS production, which serves as a second messenger through
   oxidative modification on PP2A heterocomplex. Specifically,
   ONOO^−-mediated nitration of PP2A subunits, such as B56α, inhibited
   holoenzyme assembly, sustaining S62 phosphorylation of c-Myc ([183]18,
   [184]26, [185]56). We uncovered that AKR1B10 depletion in CRC cells
   enhanced redox-dependent c-Myc stabilization via PP2A-B56α nitration,
   pinpointing its central role in redox-modulated c-Myc activation.
   Notably, AKR1B10 and PP2A synergistically regulate c-Myc
   signaling—enhancement of either suppresses it, while inhibition
   amplifies it ([186]Fig. 5, N and O, and fig. S8, I to L). These
   findings suggest that AKR1B10 also modulates the c-Myc–ITGB8 axis
   through a minor PP2A-independent mechanism. Further exploration is
   required to determine whether AKR1B10 affects redox modification of
   other PP2A subunits and their downstream substrates. Expanding this
   research to proteins sensitive to tyrosine nitration, such as p53,
   procaspase-3, and Bcl-2 ([187]18, [188]57, [189]58), may uncover
   broader mechanisms by which AKR1B10 governs redox-dependent signaling
   in cancer.

   AKR1B10 also displays noncanonical functions independent of its
   enzymatic activity. It has been shown to interact with HSP90α, which
   facilitates its lysosomal secretion as a molecular chaperone ([190]59).
   In addition, AKR1B10 binds to acetyl–coenzyme A carboxylase α, a
   rate-limiting enzyme in de novo fatty acid synthesis, preventing its
   ubiquitination and degradation ([191]51). In this study, we demonstrate
   that AKR1B10 promotes PP2A holoenzyme assembly by directly interacting
   with PP2Ac, serving as an endogenous activator of PP2A through both its
   reductase activity and moonlighting function. Beyond regulating retinal
   reductase activity, we find the critical role of K125 in mediating B56α
   nitration and facilitating AKR1B10-PP2Ac interaction. Notably, the
   AKR1B10^K125L mutant amplified c-Myc’s oncogenic effects and promoted
   CRC metastasis by suppressing PP2A activity. Our study emphasizes the
   clinical relevance of AKR1B10 as both a diagnostic biomarker and a
   molecular target for tailored therapeutic strategies aimed at
   alleviating c-Myc–driven metastasis.

   PP2A is a well-established tumor suppressor, frequently subject to
   genetic alterations or functional inactivation, contributing to the
   development of various human malignancies ([192]60–[193]64). Its
   enzymatic activity is finely regulated by posttranslational
   modifications, including phosphorylation, nitration, and methylation
   ([194]18, [195]24–[196]26). In this study, we present the first
   evidence that AKR1B10 acts as a precise regulator of PP2A activity.
   Moreover, we showed that pharmacological restoration of PP2A activity
   efficiently mitigates the metastatic potential induced by AKR1B10 loss
   during CRC progression. Our findings highlight the feasibility of
   repurposing PP2A activators to counteract AKR1B10–PP2A–c-Myc
   axis–mediated CRC metastasis.

   In summary, this study uncovers the tumor-suppressive attributes of the
   aldo-keto reductase AKR1B10 in CRC (phenocopies in GC), where it
   unleashes PP2A activity through both its redox regulatory capacity and
   moonlighting, binding-dependent function. We propose pharmacological
   restoration of PP2A activity as a promising strategy to counteract the
   prometastatic effects mediated by AKR1B10 depletion. Given AKR1B10’s
   contrasting expression patterns across various malignancies, these
   findings provide valuable insights into the intricate mechanisms of
   AKR1B10 dysregulation in tumorigenesis, paving the way for promising
   therapeutic strategies.

MATERIALS AND METHODS

Cell culture

   hCRC cells [HCT116 (Dukes’ stage D, mesenchymal subtype), SW1116
   (Dukes’ stage A, canonical epithelial subtype), LoVo (Dukes’ stage C,
   immune-activated subtype), and HT29 and SW620], human GC (hGC) cells
   (MKN45, AGS, HGC27, and MKN74), human gastric epithelial cells (GES-1),
   mouse CRC cells (MC38 and CT26), human HCC cells (MHCC97-H and Huh7),
   and human embryonic kidney (HEK) 293T cells were obtained from the
   American Type Culture Collection (Manassas, VA, USA) or maintained in
   our laboratory. Cells were cultured at 37°C with 5% CO[2] in a
   humidified incubator with Dulbecco’s modified Eagle’s medium (DMEM) or
   RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS),
   penicillin/streptomycin (100 U/ml), and 1 mM sodium pyruvate. All cell
   lines were routinely checked using the short tandem repeat method and
   tested negative for mycoplasma.

Clinical specimens

   Human tissue specimens, including TMAs containing paired CRC and
   adjacent normal tissues (between December 2019 and December 2021), 93
   CRC tissues for IHC staining, RNA and protein samples from matched
   CRC/GC tissues and adjacent normal tissues, and 20 paired CRC primary
   tumor samples and metastatic liver samples were obtained from the
   SYSU-FAH with the consent of the patients. Patients with CRC were
   stratified into four groups based on the expression combinations of
   AKR1B10-ITGB8, AKR1B10–c-Myc, and c-Myc–ITGB8, categorized as low
   (H-score < 40) or high (H-score ≥ 40). Kaplan-Meier curves were
   generated to evaluate overall survival among the four groups. This
   study was approved by the Medical Ethics Committee, SYSU-FAH
   ([2024]476), with the consent of the patients. The clinicopathological
   characteristics of the patients with CRC whose samples were included in
   TMAs are summarized in table S1.

Plasmid construction and transfection

   pCMV-HA-PPP2CA, pCMV-3×Flag-AKR1B10, and pCMV-His-PPP2R5A were obtained
   from MiaolingBio (Wuhan, China). Mutant constructs of AKR1B10 (K125L,
   V301L, C299S, P219S, and K263R) and PPP2R5A (Y238F) were generated by
   site-directed mutagenesis and cloned into the pCMV-3×Flag and
   pcDNA4/myc-His A vectors, respectively. The primers used in the
   site-directed mutagenesis are listed in table S2. To generate
   reconstituted cell lines in hCRC-shAKR1B10 or hGC-shAKR1B10, cells were
   transfected with pCMV-3×Flag-AKR1B10, pCMV-3×Flag-AKR1B10^K125L,
   pCMV-3×Flag-AKR1B10^V301L, or pCMV-3×Flag-AKR1B10^C299S using
   Lipofectamine 3000 (Invitrogen, L3000008). The stable cell lines were
   selected with G418 (Sigma-Aldrich, A1720) and confirmed by quantitative
   polymerase chain reaction (qPCR) and Western blot. For small
   interfering RNA (siRNA)–mediated ITGB8, c-Myc, MAX, RUVBL2, STAT1, and
   CEBPB silencing, cells were transfected with siRNA duplex and
   Lipofectamine RNAiMAX (Invitrogen, 13778075) according to the
   manufacturer’s instructions. A nonspecific siRNA oligo (Sigma-Aldrich,
   SIC002) was used as a negative control. To establish stable
   AKR1B10/Akr1b8 knockdown cell lines, shRNAs targeting the genomic
   sequences of AKR1B10 and Akr1b8 were inserted into pLKO.1-puro
   (Addgene) empty vector. Lentiviral particles were produced by
   transfecting HEK293T cells with pMD2.G, psPAX2, and pLKO.1-shRNA
   plasmids at a 1:2:2 ratio using Lipo8000 reagent (Beyotime, C0533) for
   48 hours. Parental cells were then infected with shRNA-containing
   lentivirus for 24 hours and selected with puromycin (Selleck, S7417)
   for 1 to 2 weeks. The sequences of shRNA and siRNA are listed in table
   S3.

RNA extraction and quantitative PCR

   Total RNA was extracted using a FastPure Cell/Tissue Total RNA
   Isolation Kit (Vazyme, RC112) according to the manufacturer’s
   instructions and reversely transcribed into cDNA using the Evo M-MLV RT
   Kit (Accurate Biology, AG11706). Real-time PCR was performed on a
   QuantStudio 5 (Applied Biosystems) using the Premix Pro Taq HS qPCR Kit
   II (Accurate Biology, AG11702). The primers used for quantitative
   reverse transcription (qRT)–PCR were synthesized by TSINGKE
   Biotechnology, and their sequences are detailed in table S4.

RNA-seq and analysis

   Total RNA was extracted from HCT116 cells [negative control shRNA
   (shNC) versus shAKR1B10] using TRIzol reagent (Invitrogen, 15596026).
   RNA quality was evaluated using the Agilent 2100 Bioanalyzer. Only
   samples that exhibited an RNA integrity number of 7 or higher were
   deemed acceptable for further procedures. Subsequent RNA-seq was
   performed on the Illumina NovaSeq 6000 sequencing machine (BGI
   Genomics, China). Differential genes (DEGs) were selected using the
   DESeq2 R package with the following requirements: fold change > 1 and
   adjusted P value < 0.05. The KEGG pathway analysis was performed to
   annotate the biological function of the DEGs using the clusterProfiler
   R package. The heatmap of differential gene clusters was generated
   using the pheatmap function on the differential gene set.

ChIP assay

   ChIP assays were performed using the EZ-ChIP kit (Merck Millipore,
   17-375) according to the manufacturer’s instructions. Briefly, HCT116
   cells were harvested after crosslinking in 1% formaldehyde at room
   temperature for 15 min, followed by quenching with 0.125 M glycine
   (Sigma-Aldrich, V900144) for 5 min. Cells were then washed twice with
   ice-cold phosphate-buffered saline (PBS), scraped, and pelleted by
   centrifugation. The cell pellet was resuspended in lysis buffer and
   subjected to freeze-thaw cycles via liquid nitrogen. Cell lysate was
   centrifuged to isolate nuclei, which were subsequently resuspended in a
   diluted enzymatic cocktail to enzymatically shear genomic DNA into
   fragments ranging from 100 to 200 base pairs (bp). Cleaved chromatin
   was incubated with 5 μg of anti-Myc antibody (Abcam, ab32072) or only
   Protein A Agarose/Salmon Sperm DNA. Following washes and
   decrosslinking, qPCR and agarose gel electrophoresis assays were
   performed to detect the enriched regions of c-Myc within the AKR1B10
   promoter. Primers used for ChIP are listed in table S5.

Cell viability and colony formation assay

   Cells (1 × 10^4) were seeded in a 96-well plate for cell proliferation
   assay. Cell viability was measured at 24, 48, 96, and 120 hours using
   the Cell Counting Kit-8 [MedChemExpress (MCE), HY-K0301]. For the
   colony-forming assay, the indicated number of cells was seeded in a
   six-well plate for 12 to 14 days. The colonies were fixed using 4%
   paraformaldehyde (Biosharp, BL539A-1), stained with crystal violet
   staining solution (Beyotime, C0121), and counted with the ImageJ v1.53c
   software. The experiment was performed in three biological replicates.

Cell migration assays

   The transwell assay was performed using polycarbonate membrane chambers
   with an 8-μm pore (Corning, 3422). Cells (1 to 1.5 × 10^5) were seeded
   into the upper transwell chamber with 300 μl of serum-free medium for a
   1- to 2-day migration period. The lower chamber was filled with a
   500-μl medium containing 20% FBS as a chemoattractant. After 24 to
   72 hours of migration, cells that migrated to the lower side were fixed
   by 4% paraformaldehyde and stained with crystal violet, photographed,
   and counted. The results of the experiment were determined by three
   replications.

   For the wound healing assay, 3 × 10^5 cells were inoculated in a
   six-well plate, and the cell layer was scraped with a sterilized 10-μl
   pipette tip when the cells reached 90% confluence. Subsequently, the
   cells were cultured in serum-free medium, and images were captured by
   an inverted microscope at 0, 12, and 24 hours. ImageJ v1.53c software
   was used to calculate the wound healing areas.

Cell cycle and cell apoptosis assay

   Cell cycle analysis was performed using the Propidium Iodide Cell Cycle
   Detection Kit (4A Biotech, FXP0211) according to the manufacturer’s
   instructions. Briefly, cells were collected and fixed with 70% ethanol
   at 4°C overnight. Subsequently, the cells were incubated with propidium
   iodide (PI) staining solution at 37°C in the dark for 30 min. Following
   three successive washes with PBS, the cells were analyzed using a flow
   cytometer CytoFLEX (Beckman Coulter). For cell apoptosis analysis,
   cultured cells were washed, dissociated, and suspended in 1× binding
   buffer. The cells were then stained with annexin V and PI (4A Biotech,
   FXP023) at room temperature for 15 min in the dark, followed by washing
   and further flow cytometry analysis.

H&E and immunohistochemistry

   IHC assays were conducted as previously described ([197]65).
   Paraffin-embedded tumor tissue sections were deparaffinized in xylene
   and rehydrated with graded ethanol. Postrehydration, the sections were
   subjected to H&E staining. For IHC staining, the sections underwent
   rehydration, after which endogenous peroxidase activity was quenched,
   and heat-induced epitope retrieval was performed. After blocking with
   normal bovine serum albumin (BSA) antigen for 1 hour at 37°C, the
   tissue sections were incubated with anti-AKR1B10 (Signalway Antibody,
   36071), anti-ITGB8 (Proteintech, 29775-1-AP), or anti–c-Myc (Abcam,
   ab32072) antibodies at 4°C overnight. Following incubation with
   horseradish peroxidase (HRP)–conjugated secondary antibodies, the
   sections were visualized with DAB Detection Kit (ZSGB-BIO, ZLI-9017)
   and then dehydrated and stabilized with mounting medium, and the images
   were acquired with a KF-PRO-020 scanner (Konfoong Tech). The
   specificity of antibodies used for IHC was validated using standard
   approaches, as demonstrated in fig. S11. IHC analysis was conducted by
   two independent observers who were blinded to the results of the other
   markers and clinical outcomes. Staining intensity was categorized as 0
   (no staining), 1+ (weak staining), 2+ (moderate staining), or 3+
   (intense staining). Multiplying the staining intensity by the reaction
   extension value (range 0 to 300) yielded the H-score for the tumor
   tissue.

Immunofluorescence

   Cells seeded on 15-mm glass-bottomed cell culture dishes (Nest, 801002)
   were fixed with 4% paraformaldehyde for 10 min and permeabilized with
   0.5% Triton-X 100 (in PBS) for 10 min at room temperature, followed by
   blocking, overnight incubation with primary antibody [AKR1B10 mouse
   monoclonal antibody (1:300; Santa Cruz Biotechnology, sc-365689) and
   PPP2CA rabbit polyclonal antibody (1:500; ABclonal A6702)], and
   incubation with secondary antibodies [Alexa Fluor 555–labeled Donkey
   anti-Rabbit IgG (immunoglobulin G) (1:500, Beyotime, A0453) and Alexa
   Fluor 488–labeled Goat anti-Mouse IgG (1:500, Beyotime, A0428)] for 1
   hour in the dark. Confocal microscopy images were captured using an
   Olympus FV3000 Laser Scanning Confocal Microscope and evaluated with
   FV31S-SW software.

Western blot and immunoprecipitation

   Cells were harvested and lysed with radioimmunoprecipitation assay
   buffer (Beyotime, P0013B) or immunoprecipitation (IP) lysis buffer
   (Beyotime, P0013). The cell lysates underwent sonication for 10 cycles
   and were centrifuged at ×12,000 rpm for 10 min. Supernatants were
   collected, and protein concentrations were determined by bicinchoninic
   acid (BCA) protein assay kit (Beyotime, P0011). Protein lysates were
   separated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and
   transferred to polyvinylidene difluoride membranes (Merck Millipore,
   IPVH00010). Membranes were blocked with 5% nonfat milk in 1 × PBS–Tween
   20 for 1 hour at room temperature and probed overnight at 4°C with
   specific antibodies. The immunoblots were visualized using Immobilon
   Western HRP substrate (Merck Millipore, WBKLS0500) and scanned using an
   Amersham ImageQuant 800 system (Cytiva). To determine the half-life of
   c-Myc, shNC and shAKR1B10 CRC/GC cells were exposed to CHX
   (Sigma-Aldrich, C7698) for varying durations (0, 2, 4, 8, 16, and
   24 hours) and subsequently analyzed through Western blot.

   After BCA quantification, equal amounts of protein lysates for Co-IP
   were subjected to a preclearing step using 25 μl of Protein A/G
   PLUS-Agarose (Santa Cruz Biotechnology, sc-2002) for 1 hour on ice and
   then incubated with the primary antibodies and Protein A/G PLUS-Agarose
   on a rocker platform at 4°C overnight. Immunoprecipitates were
   collected by centrifugation at 2000 rpm for 5 min at 4°C, washed thrice
   with cold lysis buffer, and eluted by boiling in 1 × SDS-PAGE loading
   buffer for 6 min and then subjected to Western blot analysis.
   Antibodies used for Western blot and IP are listed in table S6. The
   specificity of antibodies targeting c-Myc and pT58 c-Myc was validated
   in fig. S11.

Purified protein pull-down assay

   The pull-down protein-protein interaction assay was performed as
   previously described ([198]14). Glutathione-Sepharose slurry (Cytiva,
   17075601) beads were equilibrated by incubating for 3 × 10 min in PBS
   buffer at 4°C. Following equilibration, 3 μg of recombinant GST-AKR1B10
   protein (CusaBio, CSB-EP001540HU) or GST-AKR1B10 (K125L) protein
   [CusaBio, CSB-EP001540HU(M)] was incubated with the beads at room
   temperature for 1 hour. Then, 3 μg of recombinant His-PPP2CA (CusaBio,
   CSB-EP018559HUc7) or His-PPP2R1A (CusaBio, CSB-EP018562HUa0) was added
   to the buffer containing protein-bound glutathione beads overnight at
   4°C. After three washes with PBS, the bead-bound complex was
   resuspended in SDS loading dye and boiled at 95°C. Eluted proteins were
   subjected to SDS-PAGE and immunoblotted with indicated antibodies.

Flow cytometry analysis

   The DCFH-DA Assay Kit was used for the detection of cellular ROS. Cells
   were plated in six-well culture dishes at a density of 1 × 10^5 cells
   per well and treated with DPI (Selleck, S8774) or FeTPPS (MCE,
   HY-131697) for indicated durations. Following treatment, cells were
   trypsinized, centrifuged, and resuspended in PBS. They were then
   incubated in the dark at 37°C for 30 min with 5 μM DCFH-DA probe (MCE,
   HY-D0940). Following three successive washes with PBS, the abundance of
   ROS was detected using a flow cytometer CytoFLEX (Beckman Coulter), and
   the data were processed via FlowJo 10.8.1.

Luciferase reporter assay

   The hAKR1B10 promoter regions, spanning from −2500 to +1 of exon 1,
   were cloned into the pGL3-Basic vector. Luciferase activities were
   evaluated using the Dual-Luciferase Reporter Assay System (Promega,
   E1910) according to the manufacturer’s instructions. HCT116 cells were
   seeded into a 12-well plate and transiently transfected with WT or
   mutant luciferase reporter plasmids plus Renilla DNA (pRL-TK) at a
   ratio of 10:1, after transfection with control or HA-MYC plasmid. HCC
   cells (MHCC97-H and Huh7) were seeded into a 12-well plate and
   transiently transfected with WT luciferase reporter plasmids plus
   Renilla DNA (pRL-TK) at a ratio of 10:1, after transfection with
   negative control siRNA or si–c-Myc RNAs. At 24 hours posttransfection,
   cells were washed with PBS, lysed with 1× passive lysis buffer, and
   transferred into a white 96-well plate for bioluminescence detection.
   Firefly luciferase activities were normalized by Renilla luciferase
   signal.

PP2A activity assay

   PP2A activity was measured by the Serine/Threonine Phosphatase Assay
   Kit (Promega, V2460). Briefly, cells were collected and lysed by
   phosphatase storage buffer [20 mM imidazole-HCl, 2 mM EDTA, 2 mM EGTA
   (pH 7.0) with 10 μg/ml each of aprotinin, leupeptin, antipain, soybean
   trypsin inhibitor, 1 mM benzamidine, and 1 mM phenylmethylsulfonyl
   fluoride]. Lysates were then sonicated for 3 × 1.5 s, centrifuged at
   13,000 rpm at 4°C for 30 min, and added to Sephadex G-25 spin columns
   to remove endogenous free phosphate. Following protein quantification
   by BCA Protein Assay Kit, protein lysates (500 μg) were incubated with
   5 μg of antibody anti-PPP2R5A (12675-2-AP) ([199]66) and 50 μl of
   Protein A/G PLUS-Agarose at 4°C overnight with constant rocking. A
   total of 5 μl of phosphopeptide RRA(pT)VA (1 mM) and 10 μl of PP2A
   reaction buffer [250 mM imidazole (pH 7.2), 1 mM EGTA, 0.1%
   β-mercaptoethanol, and BSA (0.5 mg/ml)] were premixed and added to each
   96-well plate for preincubation at 37°C. The beads were washed three
   times with tris-buffered saline and resuspended in 150 μl of
   phosphatase storage buffer. The reaction was started by adding the
   phosphatase samples (35 μl) to the wells containing the premixed
   reaction solution, and the reactions were incubated for 30 min. The
   reaction was stopped by adding 50 μl of molybdate dye/additive mixture.
   Following 30-min incubation, absorbance was measured at 630 nm in a
   96-well microplate reader (Varioskan LUX).

Tumor spheroid culture

   HCT116 or SW1116 cells were cultivated in DMEM or RPMI 1640 medium,
   respectively, supplemented with 10% (v/v) FBS and 2.5% (v/v) Matrigel
   (Corning, 356234). A total of 4000 cells per well were seeded into
   ultralow-attachment 96-well plates (Corning, 7007). Following
   centrifugation at 1000g for 10 min, the cells were cultured for 7 days
   and then photographed. Images were analyzed using ImageJ.

Animal experiments

   All mice, including BALB/c nude mice, BALB/c mice, C57BL/6 mice, and
   NOD-SCID mice, were purchased from Guangdong GemPharmatech. All animal
   experiments were conducted in accordance with the National Institute of
   Health Guide for the Care and Use of Laboratory Animals. The animals
   were maintained under specific pathogen–free conditions. The
   experimental procedures were approved by the Institutional Animal Care
   and Use Committee of SYSU (approval numbers: SYSU-IACUC-2022003193 and
   SYSU-IACUC-2023355).

   For the subcutaneous xenograft model, 4- to 6-week-old female BALB/c
   nude mice were randomly allocated to two groups. HCT116 cells
   (3 × 10^6; shNC&shAKR1B10) were subcutaneously injected into the dorsal
   flank of BALB/c nude mice. Tumor volume was measured every 3 days since
   day 7 of tumor implantation and calculated using the formula
   V = 0.5 × D × W^2 (V for volume, D for diameter, and W for width). The
   mice were euthanized on day 25, and the tumors were photographed and
   weighed.

   For the metastasis model, 4- to 6-week-old female BALB/c nude mice were
   randomly allocated to two groups. HCT116 cells (1 × 10^6) were
   administered through intravenous injection to establish a pulmonary
   metastasis model or intrasplenic injection to establish a liver
   metastasis model. Three to four weeks after the injection, the mice
   were euthanized, and the lungs or livers were photographed and paraffin
   embedded for H&E staining and IHC. For mice receiving drug treatment,
   after 7 days of intrasplenic injection, mice were orally gavaged with
   O-carboxymethylcellulose (CMC)–Na (negative control) or SMAP (Selleck,
   S8774) at 5 mg/kg on alternate days for 3 weeks.

   MC38 (5 × 10^5) cells were administered via injection into the spleen
   of NOD-SCID mice or C57BL/6 mice (4- to 6-week-old; female). After
   inoculation for 2 to 5 weeks, mice were euthanized, and the metastatic
   liver specimens were removed, photographed, weighed, paraffin embedded,
   and lastly validated by H&E staining and IHC.

   To establish the orthotopic mouse model, 4- to 6-week-old female BALB/c
   nude mice or BALB/c mice were randomly allocated to several groups.
   Mice were anesthetized by intraperitoneal injection, and a small
   incision was made to exteriorize the caecum. HCT116 cells (2 × 10^6) or
   CT26 cells (0.8 × 10^6) in 50 μl of PBS were injected into the caecum
   termini of the mouse, followed by the suturing of the incision. After 7
   days of injection, mice were orally gavaged with CMC-Na or SMAP at 5
   mg/kg on alternate days. Upon the onset of moribundity in one or more
   mice, all animals were humanely euthanized. Subsequently, the
   intestinal and liver tissues were excised, photographed, and paraffin
   embedded.

MS analysis

   HCT116-shAKR1B10 cells stably expressing Flag-AKR1B10 or Flag-vector
   were lysed in IP lysis buffer (Beyotime, P0013) and immunoprecipitated
   with anti-FLAG antibody (Sigma-Aldrich, F1804) and Protein A/G
   PLUS-Agarose (Santa Cruz Biotechnology, sc-2003). The
   immunoprecipitated complexes were eluted by boiling in 1× loading
   buffer and then subjected to SDS-PAGE. After staining with Coomassie
   Brilliant Blue (Beyotime, P0017), gel fragments containing proteins
   were excised, destained, reduced, and subjected to in-gel trypsin
   digestion overnight. The extracted peptides were redissolved in 0.1%
   formic acid mobile phase and centrifuged at 20,000g for 10 min, and the
   supernatants were divided by EASY-nLC 1200 System (Thermo Fisher
   Scientific) and subjected to liquid chromatography–tandem MS (LC-MS/MS)
   on the Orbitrap Exploris 480 (Thermo Fisher Scientific) by BGI
   Genomics. For protein identification, spectra were processed using
   Mascot v2.3 and searched against the Swiss-Prot database (European
   Bioinformatics Institute). The output was preprocessed and prescored
   using Percolator for quality control. Peptides (false discovery
   rate < 0.01) were selected, and the proteins that contain high-quality
   and unique peptides were considered qualified. The extracted ion
   current of each peptide was used for the calculation of peak areas. The
   final intensity-based absolute quantification (iBAQ) value was based on
   the iBAQ algorithm by dividing the total peak area by the number of
   theoretically observable peptides.

Computational molecular docking

   The three-dimensional structure data for AKR1B10 (AF-[200]O60218-F1)
   were obtained from the AlphaFold Protein Structure Database
   ([201]https://alphafold.ebi.ac.uk/). The structure of PP2A protein
   [Protein Data Bank (PDB) ID: 6NTS] was retrieved from PDB
   ([202]www.rcsb.org). The protein-protein docking model was performed
   using GRAMM (Global RAnge Molecular Matching)
   ([203]https://gramm.compbio.ku.edu/) algorithm ([204]67). From the top
   10 ranked models generated, the pose exhibiting the highest docking
   score (indicating the most favorable binding energy) was selected to
   predict the interaction interface between AKR1B10 and PPP2CA.
   Prediction of hydrogen bonding between amino acid residues and
   structural illustrations was prepared with UCSF Chimera
   ([205]www.cgl.ucsf.edu/chimera/).

Bioinformatic analysis

   The data on AKR1B10 mRNA expression were extracted from the TCGA
   database. Pearson correlations between AKR1B10 and integrin family
   genes or MYC are based on the TCGA data. The survival data were
   extracted from TCGA database and GEO datasets ([206]GSE39582,
   [207]GSE17536, and [208]GSE14210). AKR1B10 protein expression was
   collected from the Proteomic Data Commons and analyzed by cProSite.
   AKR1B10 expression in different stages from TCGA data was collected
   from the University of Alabama at Birmingham Cancer data analysis
   portal.

Statistical analysis

   All in vitro experiments were independently replicated at least three
   times, and in vivo animal studies were repeated at least twice. The
   numbers of biological replicates (n) are provided in the figure
   legends. All values shown in graphs are presented as individual data
   points and the mean ± SD, unless stated otherwise. Statistical
   significance was performed using GraphPad Prism 9, with Student’s t
   test for two independent groups and one- or two-way analysis of
   variance (ANOVA) test for multiple groups (≥3). Survival analyses were
   conducted using the log-rank test, and correlation analyses were
   performed using Pearson’s correlation coefficient (P < 0.05 was
   considered statistically significant). Researchers blinded to group
   assignments performed data collection and analysis for animal
   experiments and IHC staining.

Acknowledgments