PMID: 40973819 Diet is widely recognized as one of the most significant environmental factors contributing to the development of CRC1. Among the dietary factors that have gained increasing attention is the consumption of SSBs. SSBs are any liquids sweetened with added sugars, including sucrose and high-fructose corn syrup, both of which contain glucose and fructose in an approximate 1:1 ratio2. Since the 1980s, SSB consumption has risen globally3. In the USA, over half of adults and nearly two-thirds of youth consume SSBs daily4,5. This surge parallels alarming increases in young-onset CRC incidence and mortality6,7. Our prior research using precancerous Apc-mutant mouse models, which mimic the early stages of CRC (adenomas)8, showed that SSBs can directly promote tumorigenesis independent of obesity9. A large epidemiological study later confirmed our findings, showing that SSB intake doubled the risk of young-onset CRC10. However, whether SSB consumption affects the later stages of CRC (carcinomas) remains unclear. Although some cohort studies suggest links between SSB intake and CRC recurrence or mortality, causal and mechanistic insights are lacking11,12. Addressing this question is essential, given that A full list of affiliations appears at the end of the paper. e-mail: jyun@mdanderson.org Nature MetabolismLetter https://doi.org/10.1038/s42255-025-01368-w many patients with CRC are diagnosed at advanced stages, and some continue to consume SSBs even after diagnosis. Moreover, oncologists and nutritionists often recommend energy drinks and concentrated juices, which are essentially SSBs owing to their high glucose and fructose content. Therefore, we investigated whether SSB consumption promotes progression and metastasis of advanced CRC and explored the underlying mechanisms. Glucose and fructose together enhance CRC motility and metastasis We previously showed that administering a glucose–fructose mixture in liquid form to Apc-mutant mice, mimicking SSB intake, exposes intestinal tumour cells to both sugars at millimolar (mM) concentrations, enabling efficient uptake9. In this study, we refer to the glucosefructose mixture treatment as the ‘Glu + Fru’ condition, reflecting the physiological impact of high-fructose corn syrup or sucrose, both of which contain glucose and fructose. To examine the specific effects of this combination on advanced CRC, we used 13 CRC cell lines (carcinomas) with various mutations and conducted a cell growth assay under three equimolar sugar conditions (Fig. 1a): glucose only (20 mM), Glu + Fru (10 mM glucose + 10 mM fructose) and fructose only (20 mM). We found that Glu + Fru did not significantly alter cell growth compared to glucose alone (Fig. 1b). Notably, all cell lines showed poor growth in the fructose-only condition (Fig. 1b), which does not reflect physiological conditions, as cells are never exposed to fructose without glucose. Therefore, we excluded the fructose-only condition from further studies unless needed as a control. Next, we conducted a transwell migration assay using 13 CRC cell lines. Nine cell lines did not migrate under either glucose-only or Glu + Fru conditions (Extended Data Fig. 1a), indicating that they lack intrinsic migratory ability. By contrast, the four lines that did migrate— HCT116, DLD1, RKO and HCT8—showed enhanced migration under Glu + Fru conditions compared to the glucose-only condition (Fig. 1c). Consistent with migration results, Glu + Fru also increased invasion compared to glucose, without affecting colony-forming ability (Fig. 1d and Extended Data Fig. 1b). These findings suggest that migratory CRC lines exposed to glucose and fructose uniquely acquire enhanced migration and invasion, key early steps of metastasis. To investigate whether glucose and fructose together promote CRC metastasis in vivo, we used an orthotopic caecum injection model to assess local invasion and spontaneous liver metastasis (Fig. 1e). HCT116 cells were injected into the caecal submucosa of athymic nude mice, which then received water, glucose (25%) or Glu + Fru (25%) in drinking water for 5 weeks. Glu + Fru significantly increased liver metastasis, measured by liver metastatic foci, compared to water or glucose (Fig. 1f), without affecting primary tumour size (Extended Data Fig. 1c). Liver histology and body weight were unchanged between Glu + Fru and water groups (Extended Data Fig. 1d,e), suggesting that Glu + Fru alters neither the liver microenvironment nor body weight under these conditions. Moreover, daily oral gavage of moderate Glu + Fru significantly increased liver metastasis compared to the water group, with metastatic potential comparable to that of the Glu + Fru drinking water group (Extended Data Fig. 1f). Additionally, we confirmed these findings using an intrasplenic injection model, which showed that Glu + Fru enhanced liver metastasis independent of local invasion (Extended Data Fig. 1g). Collectively, these results suggest glucose and fructose together promote CRC metastasis more than glucose alone. Glucose and fructose together drive the SORD reverse reaction To elucidate how the glucose–fructose combination promotes CRC migration and metastasis, we conducted global liquid chromatography–mass spectrometry (LC–MS) metabolomics on four CRC cell lines under glucose-only, Glu + Fru and fructose-only conditions (Fig. 2a). We aimed to identify metabolites consistently and uniquely altered by Glu + Fru across all four cell lines compared to glucose or fructose alone. The metabolomics platform identified over 700 annotated metabolites per cell line. Surprisingly, sorbitol was the only metabolite consistently and significantly increased more than twofold across all four CRC lines under Glu + Fru conditions, exhibiting threefold to 11-fold increases compared to glucose-only and fructose-only conditions (Fig. 2b and Extended Data Fig. 2a–d). We further confirmed that sorbitol was endogenously produced and then secreted over time following Glu + Fru treatment (Fig. 2c and Extended Data Fig. 2e). Sorbitol is a sugar alcohol produced through the polyol pathway, a two-step process that converts glucose into fructose13 (Fig. 2d). In the first, irreversible step, glucose is reduced to sorbitol by aldose reductase (AR) using NADPH as a cofactor. The second, reversible step oxidizes sorbitol to fructose by sorbitol dehydrogenase (SORD), reducing NAD⁺ to NADH (Fig. 2d). Thus, elevated sorbitol under Glu + Fru could arise from glucose, fructose or both by glucose reduction by AR (NADPH-dependent) or fructose reduction by SORD in the reverse reaction (NADH-dependent). Notably, SORD was expressed in all four CRC lines, whereas AR was detected in HCT116 and HCT8 but absent in DLD1 and RKO (Fig. 2e). To determine which sugar is converted to sorbitol and which enzyme catalyses its formation under Glu + Fru conditions, we conducted stable 13C-isotope LC–MS experiments in HCT116 and DLD1 using three labelling conditions: 13C6-glucose-only, 13C6-glucose + unlabelled fructose and 13C6-fructose + unlabelled glucose. As expected, total sorbitol (labelled + unlabelled) was significantly higher under Glu + Fru than glucose-only (Fig. 2f), corroborating our unlabelled metabolomics (Fig. 2b). Remarkably, in cells incubated with 13C-fructose + unlabelled glucose, over 97% of sorbitol was fully 13C-labelled at all six carbons (M+6) (Fig. 2f). By contrast, with 13C-glucose + unlabelled fructose, sorbitol remained mostly unlabelled (Fig. 2f). No intermediate labelling (M+1, M+2, M+3, M+4 or M+5) was detected from either sugar, confirming direct conversion to sorbitol. These results indicate that elevated sorbitol under Glu + Fru is primarily derived from fructose through SORD, not glucose through AR, even in cell lines with high AR expression, such as HCT116. Given that the SORD reverse reaction was activated under Glu + Fru versus glucose-only conditions, we reasoned that the NAD⁺/NADH ratio would likewise increase with sorbitol levels. Indeed, the NAD⁺/ NADH ratio was significantly higher in all four CRC lines under Glu + Fru than glucose (Fig. 2g and Extended Data Fig. 2f). The fructose-only condition did not increase sorbitol (Fig. 2b), probably because of inefficient glycolytic entry and thus inadequate NADH production for the SORD reverse reaction. By contrast, under Glu + Fru conditions, glucose efficiently enters glycolysis, generating NADH through glyceraldehyde-3-phosphate dehydrogenase (GAPDH), while fructose directly serves as a substrate for SORD, together activating the reverse reaction (Fig. 2h). SORD expression is elevated in human colorectal tumours Supporting our findings, SORD mRNA and protein levels were significantly higher in human colorectal tumours than in adjacent normal epithelium (Fig. 3a–c and Extended Data Fig. 3a–f), consistent with previous proteomics studies14,15. In some cohorts, metastatic tumours exhibited higher SORD mRNA levels than primary tumours (Extended Data Fig. 3b,e,f). Higher-grade tumours also showed elevated SORD protein levels compared to lower-grade tumours (Fig. 3b,c). By contrast, AKR1B1 mRNA, encoding AR in the polyol pathway, was downregulated or unchanged in colorectal tumours versus normal tissue (Extended Data Fig. 3g–k), suggesting that SORD—not AR—may have a more prominent role in CRC development. The heterogeneity of CRC tissue—including immune, stromal and epithelial cells—complicates transcriptional analysis from bulk tumour samples. To address this limitation, we analysed single-cell RNA sequencing (scRNA-seq) profiles from 62 CRC and 36 adjacent normal tissue samples16. SORD expression—unlike AKR1B1—was predominantly elevated in tumour epithelial cells across 20 clusters, with minimal expression Nature MetabolismNature Metabolism Letter https://doi.org/10.1038/s42255-025-01368-w in immune and stromal cells, confirming tumour-specific upregulation (Extended Data Fig. 3l). Importantly, SORD was highly expressed in stem-cell/TA-like epithelial subtypes (Extended Data Fig. 3l). Its expression resembled that of known intestinal stem cell markers (LGR5, BMI1), suggesting a stem-cell-like role for SORD (Extended Data Fig. 3l). We further validated that SORD expression was high in undifferentiated, stem-like human colon enteroids but decreased upon differentiation, mirroring LGR5 downregulation (Extended Data Fig. 3m). Two patient-derived CRC organoids exhibited higher SORD expression than differentiated colonic epithelial cells, consistent with patient mRNA Glu Glu + Fru HCT116 DLD1 RKO HCT8 Water Glu Glu + Fru Caecum injection model liver metastasis Number of Liver met. foci Relative invasion Relative migration Glu Glu + Fru HCT116 DLD1 RKO HCT8 HCT116 Glu Glu + Fru Glu Glu + Fru DLD1 RKO HCT8 NCI-H508 Difi Colo205 GP5D HT55 SW620 VACO432 SW48 HCT15 0 20 40 60 80 100 120 Glu Glu + Fru Fru **** **** **** **** **** **** **** **** *** **** **** **** **** a c d e f b Glucose (20 mM) Glu + Fru (10 mM + 10 mM) Fructose (20 mM) 13 CRC lines ●Growth assay ●Migration assay **** * ** ** * * ** ** *** *** ●Colony formation assay ●Invasion assay HCT116 DLD1 RKO HCT8 0 0.5 1.0 1.5 2.0 2.5 HCT116 DLD1 RKO HCT8 0 1 2 3 Water Glu Glu + Fru 0 2 4 6 8 10 ns Cell growth (%) Fig. 1 | Glucose and fructose together enhance the migration, invasion and metastatic potential of CRC cells in vitro and in vivo. a, Schematic of cell culture medium used for multiple assays exploring the specific effects of glucose + fructose treatment across 13 CRC cell lines. b, Cell growth assay conducted with 20 mM glucose (Glu), 10 mM glucose + 10 mM fructose (Glu + Fru) or 20 mM fructose (Fru) in medium containing 10% dialysed FBS for 72 h. Values are normalized to the Glu condition (n = 4). c, Transwell migration assay of CRC cells conducted under Glu and Glu + Fru conditions. Migration was quantified by stained area, normalized to the Glu condition (HCT116, n = 3; DLD1, RKO, HCT8, n = 4). Representative images of migrated cells are shown in the right panel; scale bars, 500 μm. d, Transwell invasion assay of CRC cells conducted under Glu and Glu + Fru conditions. Invasion was quantified by stained area, normalized to the Glu condition (HCT116, n = 4; DLD1, n = 5; RKO, n = 3; HCT8, n = 3). Representative images of invading cells are shown in the right panel; scale bars, 500 μm. e, Schematic of the caecum injection model. f, Number of macroscopic liver tumour foci in athymic nude mice that received caecum injection of HCT116 cells and were treated with water (n = 32), 25% (w/v) Glu (n = 9) or 25% (w/v) Glu + Fru (45:55 ratio; n = 18) for 5 weeks; met., metastatic. Representative images from each group are shown in the right panel; scale bars, 0.5 cm. Data are presented as means and represent biological replicates; error bars, s.e.m. Statistical significance was determined by two-way ANOVA with Holm-Šídák post hoc test (b); one-way ANOVA with Dunnett’s post hoc test (f) or a two-tailed unpaired Student’s t-test (c and d). ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05; n.s., not significant. Illustrations in a and e were created in BioRender.com.Nature Metabolism Letter https://doi.org/10.1038/s42255-025-01368-w data (Extended Data Fig. 3m). Given the known link between stem-like tumour cells and increased aggressiveness17, these findings suggest a role for SORD in CRC migration and metastasis. SORD mediates glucose–fructose-driven CRC motility and metastasis Given the SORD reverse reaction activation under Glu + Fru and elevated SORD expression in human CRC, we reasoned that SORD is essential for the motility and metastasis promoted by glucose and fructose. To test this idea, we generated two independent CRISPR–Cas9 SORD knockout (KO) clones in four CRC cell lines. SORD KO cells showed reduced migration and invasion compared to parental and Cas9 control cells (SORD wild-type (WT)) (Fig. 3d and Extended Data Fig. 4a), while maintaining comparable growth rates (Fig. 3e). Notably, SORD KO cells reduced migration under both glucose-only and Glu + Fru (Extended Data Fig. 4b,c), suggesting an intrinsic role in motility independent of sugar. Although the mechanism remains unclear, SORD KO cells completely abolished the Glu + Fru-induced increase Sorbitol Sorbitol HCT116 DLD1 0 0.5 1.0 1.5 2.0 Relative NAD+/NADH Glu Glu + Fru Relative sorbitol HCT116 DLD1 M+0 M+6 0 10 20 30 ** 0 10 20 30 **** 42 31 42 31 42 31 1 2 SORD AR β–actin 1 2 1 2 1 2 NADPH NADP+ NAD+ NADH D-Glucose D-Sorbitol D-Fructose AR SORD 10 20 30 40 0 1 × 106 2 × 106 3 × 106 4 × 106 5 × 106 6 × 106 Time (h) Glu Glu + Fru 0 10 20 30 40 0 5 × 106 1 × 107 1.5 × 107 2 × 107 2.5 × 107 Time (h) Sorbitol (peak area) HCT116 DLD1 Glu Glu + Fru Fru 0 2 4 6 8 Glu Glu + Fru Fru 0 5 10 15 Glu Glu + Fru Fru 0 1 2 3 4 5 Glu Glu + Fru Fru 0 1 2 3 4 5 Relative sorbitol HCT116 **** **** **** **** **** **** **** **** ** ** DLD1 RKO HCT8 HCT116 DLD1 RKO HCT8 Glucose Intensity m/z Glu + Fru Fructose Untargeted LC–MS > 700 annotated metabolites per cell line Glu + Fru up/down vs glucose/fructose (FD > 2, P < 0.05) Sorbitol a b c e f g d h HCT116 HCT8 DLD1 RKO Glu 13C Glu Glu + Fru 13C Glu + Fru Glu + 13C Fru Glu 13C Glu Glu + Fru 13C Glu + Fru Glu + 13C Fru Glucose Glucose SORD Fructose Fructose NADPHNADP+ NADH NAD+ AR DLD1 HCT116 RKOHCT8 35 0 0 0 0 0 0 0 0 0 1 1 1 4 6 Fig. 2 | Glucose and fructose together activate the SORD reverse reaction, elevating sorbitol and NAD+/NADH ratios. a, Schematic of the metabolomics workflow in four CRC cell lines cultured for 48 h under three different conditions: 20 mM glucose, 10 mM glucose + 10 mM fructose and 20 mM fructose. The Venn diagram highlights metabolites commonly altered under Glu + Fru conditions (fold difference (FD) of >2; P < 0.05; n = 6). b, Relative intracellular sorbitol levels measured by LC–MS, normalized to the Glu group (n = 6). c, Intracellular sorbitol production after incubation in 20 mM glucose (n = 3) or 10 mM glucose + 10 mM fructose (n = 3). Cells were seeded in DMEM with 10% FBS, pre-treated with Glu or Glu + Fru medium containing 10% dialysed FBS for 1 h and then collected at different incubation time points. Sorbitol levels were measured as peak areas by LC–MS, with the average 0 h value set to zero. d, Schematic of the polyol (sorbitol) metabolic pathway. e, Immunoblotting of SORD and AR enzymes in four CRC cell lines. Cells were incubated with 20 mM glucose (labelled as 1) or 10 mM glucose + 10 mM fructose (labelled as 2) in DMEM with 10% dialysed FBS for 48 h. f, Sorbitol levels measured after 14 h of incubation with 20 mM glucose, 20 mM U-[13C]-glucose (13C Glu), 10 mM glucose + 10 mM fructose, 10 mM U-[13C]glucose + 10 mM fructose (13C Glu + Fru) or 10 mM U-[13C]-fructose + 10 mM glucose (Glu + 13C Fru) in DMEM with 10% dialysed FBS, using LC–MS. M+6 denotes sorbitol fully labelled with six 13C carbons. Data were normalized to the Glu group. Statistical comparison of M+6 sorbitol levels between the 13C Glu + Fru and Glu + 13C Fru groups is shown (n = 3). g, Relative intracellular NAD⁺/ NADH ratio in CRC cells assessed by LC–MS after 48 h incubation under Glu or Glu + Fru conditions, normalized to the Glu group (n = 5). h, Schematic showing how glucose and fructose together activate the SORD reverse reaction: glucose fuels glycolysis, generating NADH, which, together with fructose, enables SORD to convert fructose into sorbitol while oxidizing NADH to NAD⁺. Data are presented as means and represent biological replicates; error bars, s.e.m. Statistical significance was determined by one-way ANOVA with Dunnett’s posttest (b); two-way ANOVA with Holm-Šídák post hoc test (f); or two-tailed unpaired Student’s t-test (g). ****P < 0.0001; **P < 0.01. Illustrations in a and h were created in BioRender.com.Nature Metabolism Letter https://doi.org/10.1038/s42255-025-01368-w WT KO 0 500 1,000 1,500 2,000 HCT116 DLD1 RKO HCT8 0 2 4 6 8 Relative growth WT KO1 KO2 n.s. n.s. n.s. n.s. n.s. HCT116 DLD1 RKO HCT8 0 0.5 1.0 1.5 Relative migration WT KO1 KO2 ** ** **** **** * * **** Normal Grade 1 Grade 2 Grade 3 0 5 10 15 n.s. 0 2,000 4,000 6,000 **** **** * **** **** ** *** *** * ** *** * SORD mRNA (log2 (TPM + 1)) Tumour Normal WT KO 0 5 10 15 Number of liver met. foci Number of liver met. foci WT KO 0 2 4 6 8 Number of liver met. foci d e f h g a b c SORD IHC score Grade 1 Grade 3 Normal HCT116 DLD1 WT KO1 KO2 RKO HCT8 KO WT KO WT KO WT KO WT KO WT Fig. 3 | SORD mediates glucose–fructose-driven CRC motility and metastasis. a, SORD mRNA expression in normal tissues (n = 51) and colorectal tumours (n = 612) in The Cancer Genome Atlas Pan-Cancer dataset. b, SORD protein expression by IHC in normal tissues (n = 59) and in colorectal tumours of different grades (Grade 1, n = 8; Grade 2, n = 72; Grade 3, n = 15), assessed by staining intensity (0–4) and coverage. c, Representative IHC images of SORD in CRC samples. Red dashed rectangles in left panels represent zoomed-in images in right panels; scale bars, 400 μm (left) and 100 μm (right). d, Migration of WT and two independent SORD knockout clones (KO1 and KO2) under Glu + Fru conditions, assessed by transwell assay (HCT116, n = 4; DLD1, n = 4; RKO, n = 3; HCT8, n = 5). Values were normalized to WT. Right, representative images; scale bar, 500 μm. e, Growth of WT, KO1 and KO2 cells (n = 3) under Glu + Fru conditions, assessed by CellTiter-Glo. Cell numbers at 96 h were normalized to values at 24 h. f, Macroscopic liver metastases from caecum tumours generated with HCT116 SORD WT (n = 21) or KO (n = 23) cells in athymic nude mice provided with 25% (w/v) Glu + Fru (45:55) in drinking water for 5 weeks. Top right, representative liver images; scale bar, 5 mm. Bottom right, H&E staining. Black dashed squares in left panels represent zoomed-in images in right panels; scale bars, 1 mm (left), 200 μm (right). g, Liver metastases in NSG mice injected in the spleen with luciferase-labelled SORD WT or KO HCT116 cells (n = 10 per group), followed by 5 weeks of 25% (w/v) Glu + Fru (45:55) in drinking water. Right, bioluminescence imaging on day 17. Representative H&E staining of liver metastases is shown in Extended Data Fig. 4i. h, Macroscopic liver metastases from NSG mice receiving colonic injection of SORD WT or KO HCT116 cells (n = 8 per group), followed by 4 weeks of 25% (w/v) Glu + Fru (45:55) in drinking water. Middle, colonoscopic images before, during and 2 weeks after injection. Right, representative images of colon and liver; red arrows indicate metastases; scale bar, 5 mm. Data are presented as mean and represent biological replicates; error bars, s.e.m. Statistical analysis was performed using two-tailed unpaired Student’s t-test (a and f–h) or one-way ANOVA with Dunnett’s post hoc test (b, d and e). ****P < 0.0001; ***P < 0.001; *P < 0.01; P < 0.05; n.s., not significant.Letter https://doi.org/10.1038/s42255-025-01368-w in migration, underscoring SORD’s essential role in motility enhanced by glucose and fructose. By contrast, AR KO cells in HCT116 and HCT8, which exhibit high AR expression, did not impair migration or invasion (Extended Data Fig. 4d,e). To validate our findings in vivo, we used three complementary xenograft metastasis mouse models (Extended Data Fig. 4f). First, in the caecum injection model, Glu + Fru failed to increase liver metastasis in mice injected with SORD KO cells compared to those with SORD WT cells (Fig. 3f and Extended Data Fig. 4g). By contrast, AR KO cells did not abolish Glu + Fru-induced liver metastasis compared to AR WT cells (Extended Data Fig. 4h). Second, in the intrasplenic model, livers from mice with SORD WT cells had numerous metastases, whereas SORD KO or knockdown cells led to fewer foci and preserved histology (Fig. 3g and Extended Data Fig. 4i–k). Third, we used a colonic injection model, wherein cancer cells are injected into the mucosa of the colon using a specialized needle under colonoscopy guidance18. This model uniquely replicates all stages of the growth–invasion–metastasis cascade to the liver. Mice injected with SORD WT cells had more liver tumours than SORD KO cells, while primary colon tumour size was comparable (Fig. 3h and Extended Data Fig. 4l,m), indicating that SORD primarily affects metastasis, not primary tumour growth. Taken together, these orthogonal mouse experiments provide strong evidence that SORD is necessary for motility and metastasis augmented by glucose and fructose. Elevated NAD+/NADH ratio caused by SORD promotes CRC motility and metastasis Given the activation of the SORD reverse reaction under Glu + Fru conditions and elevated SORD expression in human CRC, we reasoned that SORD is essential for motility and metastasis promoted by glucose and fructose. SORD KO cells from DLD1 and RKO, which lack AR expression, showed markedly lower sorbitol levels than WT cells under Glu + Fru conditions (Fig. 4a and Extended Data Fig. 5a,b). By contrast, SORD KO cells from HCT116 and HCT8, which express AR, showed higher sorbitol levels than WT cells under Glu + Fru conditions (Fig. 4a and Extended Data Fig. 5c,d). 13C-tracing showed that over 50% of sorbitol in SORD KO cells from HCT116 cell lines came from glucose (Extended Data Fig. 5c), suggesting AR activation upon SORD loss. Given the mismatch between sorbitol levels and migration ability in SORD WT and KO CRC cells, we excluded sorbitol as a common mediator of Glu + Fru-induced migration. We next tested whether the elevated NAD+/NADH ratio from the SORD reverse reaction under Glu + Fru contributes to the increased CRC migration in SORD WT versus KO cells. Unlike sorbitol levels, the NAD+/NADH ratio—measured by LC–MS or enzymatic assays—was consistently lower in SORD KO than WT cells across all four CRC lines (Fig. 4b and Extended Data Fig. 5e,f). Although widely used, LC–MS or enzymatic assays require cell lysis and cannot distinguish between cytosolic and mitochondrial NAD+/NADH pools or between free and protein-bound forms. To address this limitation, we used Peredox, a fluorescent NAD+/NADH biosensor that detects the cytosolic redox state in live SORD isogenic cells19. Binding of NADH to the Peredox sensor increases green fluorescence without affecting the red fluorescence. Therefore, a lower free cytosolic NAD+/NADH ratio results in cells displaying increased green fluorescence. Indeed, SORD KO cells exhibited a decreased red-to-green ratio—indicative of a lower cytosolic NAD+/NADH ratio—relative to SORD WT cells across all four CRC cell lines (Fig. 4c and Extended Data Fig. 5g). Together, these results suggest that SORD may promote CRC motility by elevating the NAD+/NADH ratio. To determine whether an elevated NAD⁺/NADH ratio contributes to the increased CRC motility, we modulated this ratio. In SORD KO cells, we raised the NAD⁺/NADH ratio using α-ketobutyrate, an exogenous electron acceptor; nicotinamide, an NAD+ precursor; or Lactobacillus brevis water-forming NADH oxidase (LbNOX)20–22 (Extended Data Fig. 6a,b). The increased redox ratio by each approach rescued CRC migration in SORD KO cells (Fig. 4d,e and Extended Data Fig. 6c,d). Conversely, lowering the NAD⁺/NADH ratio in SORD WT cells using FK866 (NAMPT inhibitor) or metformin reduced their migration (Extended Data Fig. 6e–k). FK866 also decreased liver metastasis in Glu + Fru-fed mice injected with SORD WT cells (Extended Data Fig. 6l). Together, these results support a causal role for elevated NAD⁺/NADH in promoting CRC migration and metastasis under Glu + Fru conditions. Fig. 4 | Elevated NAD+/NADH ratio caused by SORD promotes CRC motility and metastasis. a, Sorbitol levels in CRC cells after 48 h of incubation under Glu + Fru (10 mM each) conditions, measured by LC–MS (DLD1, n = 5; HCT116, n = 6). b, Relative NAD⁺/NADH ratio in CRC cells after 48 h of incubation under Glu + Fru conditions, measured by LC–MS (n = 3). c, Quantification of cytosolic Peredox NADH biosensor signals in WT and SORD KO CRC cells under Glu + Fru conditions after 24 h. The red/green fluorescence ratio reflects the NAD⁺/NADH ratio (DLD1 WT, n = 1205; DLD1 KO, n = 1147; HCT116 WT, n = 335; HCT116 KO, n = 311). Right, representative images; scale bars, 50 μm. d, Migration of SORD KO cells treated with vehicle (Veh) or α-ketobutyrate (α-KB) (DLD1, 1 mM; HCT116, 0.1 mM), assessed by transwell assay (n = 3). Values were normalized to the Veh group. e, Migration of SORD KO cells expressing empty vector (Vec) or LbNOX, assessed by transwell assay (n = 3). Values were normalized to the Vec group. Left: schematic of the chemical reaction catalysed by LbNOX. f, Schematic illustrating how SORD may enhance glycolysis under Glu + Fru conditions in CRC cells. The combination of glucose and fructose promotes regeneration of cytosolic NAD⁺, which is required to sustain high rates of aerobic glycolysis. g, LC–MS quantification of key glycolytic and TCA cycle metabolites in SORD WT and KO DLD1 cells after exposure to 10 mM U-[13C]-glucose and 10 mM fructose in 10% dialysed FBS. Sampling times, 10 min for DHAP, GA3P, PEP and pyruvate; 3 h for acetyl-CoA, citrate and succinate; 12 h for malate (n = 5). M+2 and M+3 indicate incorporation of two or three [13C] carbon atoms, respectively. h, Heatmap of relative metabolite levels in sugar metabolism, glycolysis and the TCA cycle in SORD WT and KO DLD1 cells under Glu + Fru conditions, measured by LC–MS (n = 6). Red indicates an increase; blue indicates a decrease. i, Metabolic pathways enriched in SORD WT versus KO cells (DLD1 and HCT116), based on RNA-seq data (n = 4). Cells were incubated under Glu + Fru conditions for 48 h. Pathway analysis was performed using Qiagen Ingenuity Pathway Analysis on genes with P < 0.05 and |log2 (fold change)| > 0.5. Blue bars indicate mevalonate-related pathways. The P value and fold change were calculated using Qiagen Ingenuity Pathway Analysis with Fisher’s exact test. j, Absolute quantities of key mevalonate pathway metabolites in SORD WT and KO DLD1 cells under Glu + Fru conditions, measured by LC–MS (WT, n = 3; KO, n = 4). k, Migration of SORD WT cells assessed by transwell assay under Glu + Fru conditions with vehicle or fluvastatin (Fluv) (n = 3). Values were normalized to vehicle. DLD1, 1.5 μM; HCT116, 1 μM. l, Migration of SORD KO cells assessed by transwell assay with vehicle (or mevalonolactone (MVA; a cell-permeable form of mevalonate) (n = 3). Values were normalized to vehicle. DLD1, 5 mM; HCT116, 2 mM. m, Liver metastatic tumour foci counts in mice that received vehicle (n = 13) or simvastatin (Simv; n = 12) by daily oral gavage. Following caecal injection of HCT116 cells, athymic nude mice were treated with 100 μl day−1 of 10% DMSO (Veh) or 30 mg kg−1 day−1 simvastatin in 100 μl of 10% DMSO. All mice received 25% Glu + Fru (45:55) in drinking water for 5 weeks before liver metastasis assessment. n, Schematic showing how consumption of sugary drinks containing glucose and fructose may promote CRC migration and metastasis via SORD. All in vitro experiments were conducted under Glu + Fru (10 mM each) conditions unless otherwise specified. Data are presented as means and represent biological replicates; error bars, s.e.m. Statistical significance was determined by two-tailed unpaired Student’s t-test (a–e and j–m) or two-way ANOVA with Holm–Šídák post hoc test (g). ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05. Illustration in n created with BioRender.com. LDH, lactate dehydrogenase; DHAP, dihydroxyacetone phosphate; GA3P, glyceraldehyde-3-phosphate; 3PG, 3-phosphoglycerate; G3P, glycerol-3-phosphate; PEP, phosphoenolpyruvate; OAA, oxaloacetate; HMG-CoA, 3-hydroxy-3-methylglutaryl-CoA; MVA-5P, mevalonate-5-phosphate; MVA-5PP, mevalonate-3,5-bisphosphate; IPPP, isopentenyl pyrophosphate; GPP, geranyl pyrophosphate. Nature MetabolismNature Metabolism Letter https://doi.org/10.1038/s42255-025-01368-w WT KO Sugar Glycolysis TCA Glucose Fructose Sorbitol GA3P DHAP 3PG G3P PEP Lactate Pyruvate Acetyl-CoA Citrate α-KG Succinate Fumarate Malate OAA 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0–0.2–0.4–0.6–0.8–1.0–1.2–1.4–1.6–1.8–0.2 Veh Simv 0 2 4 6 Number of liver met. foci VehMVA 0 1 2 3 VehMVA 0 0.5 1.0 1.5 2.0 2.5 DLD1 HCT116 SORD KO SORD KO VehFluv 0 0.5 1.0 1.5 VehFluv 0 0.5 1.0 1.5 DLD1 HCT116 0 0.05 0.10 0.15 0.20 Acetyl-CoA 0 0.05 0.10 0.15 0.20 0.25 HMG-CoA 0 0.05 0.10 0.15 0.20 MVA-5P 0 0.5 1.0 1.5 2.0 MVA-3,5PP 0 0.05 0.10 0.15 IPPP WT KO WT KO WT KO WT KO WT KO WT KO 0 0.05 0.10 0.15 GPP WTKO 0 0.5 1.0 1.5 DHAP **** WTKO 0 0.5 1.0 1.5 GA3P **** **** **** **** **** **** * * * ** ** WTKO 0 0.5 1.0 1.5 PEP **** **** ** * ** ** ** ** ** ** * * WTKO 0 0.5 1.0 1.5 Pyruvate M+0 M+3 **** WTKO 0 0.5 1.0 1.5 Acetyl-CoA **** WTKO 0 0.5 1.0 1.5 Citrate **** WTKO 0 0.5 1.0 1.5 Succinate *** WTKO 0 0.5 1.0 1.5 Malate M+0 M+2 **** Vec LbNOX 0 2 4 6 8 Vec LbNOX 0 1 2 3 DLD1 HCT116 SORD KO SORD KO Veh α-KB 0 0.5 1.0 1.5 2.0 Relative migration Relative migration Relative migration Relative migration Relative migration DLD1 HCT116 Veh α-KB 0 0.5 1.0 1.5 2.0 SORD KO SORD KO a b c d e f g h i j k l m n HCT116 DLD1 WT KO 0 0.5 1.0 1.5 Relative sorbitol abundance Relative abundance Relative abundance WT KO 0 0.5 1.0 1.5 2.0 2.5 WT KO 0 0.5 1.0 1.5 WT KO 0 0.5 1.0 1.5 Relative NAD+/NADH HCT116 DLD1 WT KO 0 0.2 0.4 0.6 0.8 1.0 1.2 Red/green WT KO 0 0.5 1.0 1.5 2.0 HCT116 DLD1 WT KO Red Green Glu + Fru 0 5 10 NAD signalling Pyrimidine salvage Geranyl-PP biosynthesis Mevalonate Cholesterol biosynthesis–log10 (P value) 0 1 2 3 4 Geranyl-PP biosynthesis Mevalonate Pyrimidine salvage Cholesterol biosynthesis Sirtuin signalling–log10 (P value) SORD WT vs KO (DLD1) SORD WT vs KO (HCT116) Glucose Sorbitol Fructose NAD+ NADH AR SORD Pyruvate Lactate NAD+ NADH LDH Glucose Fructose Acetyl-CoA Citrate Glucose Pyruvate TCA SORD NAD+ NADH Fructose Sorbitol Sugary Drinks Various Isoprenoid Metabolites Mevalonate pathway Migration Invasion Metastasis LbNOX NADH +H+ NAD+ H2 O 1/2 O2 Absolute quantity (µmol per 8 × 106 cells) Relative migrationLetter https://doi.org/10.1038/s42255-025-01368-w SORD-mediated NAD⁺ regeneration activates the mevalonate pathway to promote CRC migration Next, we investigated how the increased NAD+/NADH ratio contributes to CRC motility. Cancer cells rely on aerobic glycolysis, known as the Warburg effect23,24. High glycolytic rates are constrained by the need to replenish cytosolic NAD⁺, as GAPDH consumes NAD⁺ and produces NADH25. Aside from two NADH shuttles that regenerate cytosolic NAD+ by the tricarboxylic acid (TCA) cycle and mitochondrial respiration, lactate dehydrogenase is the primary mechanism known for rapidly recycling the cytosolic NAD+ pool; it directly converts NADH to NAD+ in the cytosol by reducing pyruvate to lactate, which is excreted (Fig. 4f). Recent studies suggest that aerobic glycolysis in proliferating cells, such as cancer cells, may be driven more by NAD⁺ demand than by ATP needs26–29. We reasoned that NAD+ generated by the SORD reverse reaction under Glu + Fru conditions might provide an alternative NAD⁺ source for glycolysis, supplementing lactate dehydrogenase. This may relieve CRC cells of NAD⁺ limitation while preserving glucose carbon in pyruvate and downstream metabolites like acetyl-CoA and TCA intermediates (Fig. 4f). As a result, this could accelerate glycolysis and boost metabolite production in glycolysis and the TCA cycle. To test this idea, we measured fractional labelling of glycolytic and TCA metabolites after exposing SORD WT and KO cells to 13C-glucose and unlabelled fructose. As expected, 13C incorporation into glycolytic and TCA metabolites was higher in SORD WT than KO cells (Fig. 4g). Similarly, the relative abundance of these metabolites was elevated in SORD WT cells compared to SORD KO cells (Fig. 4h and Extended Data Fig. 7a,b). Importantly, under Glu + Fru conditions, the levels of glycolytic and TCA cycle metabolites increased compared to those observed under glucose-only conditions (Extended Data Fig. 7c). Notably, expression of LbNOX in SORD KO cells restored the levels of these metabolites, suggesting that NAD⁺/NADH restoration activates glycolysis and the TCA cycle (Extended Data Fig. 7d,e). These findings collectively demonstrate that SORD activation under Glu + Fru conditions enhances glycolysis and TCA activity through NAD⁺/NADH modulation. To understand how increased glycolysis and TCA cycle activity enhance CRC motility, we performed RNA-seq on isogenic SORD WT and KO cells from DLD1 and HCT116. Pathway analysis showed that SORD WT cells upregulated genes in the mevalonate, cholesterol and geranyl diphosphate biosynthesis pathways—components of the mevalonate pathway—compared to SORD KO cells (Fig. 4i and Extended Data Fig. 8a). The mevalonate pathway, which begins with acetyl-CoA from glycolysis, is an essential and complex metabolic pathway that produces numerous de novo isoprenoids, such as cholesterol, vitamin D and bile acids (sterols), as well as dolichol and ubiquinone (non-sterols) (Extended Data Fig. 8b)30–32. Consistent with our RNA-seq results, LC–MS analysis showed elevated mevalonate pathway metabolites in SORD WT compared to KO cells (Fig. 4j and Extended Data Fig. 8c). As expected, Glu + Fru conditions and LbNOX expression also increased mevalonate pathway metabolites (Extended Data Fig. 8d,e), in line with their roles in enhancing the NAD⁺/ NADH ratio, glycolysis and migration. We next investigated the causal role of the mevalonate pathway in CRC migration by treating SORD WT cells with fluvastatin, an HMGCR inhibitor. At doses that did not impair proliferation, HMGCR inhibition significantly suppressed migration in both SORD WT cells and LbNox-expressing SORD KO cells (Fig. 4k and Extended Data Fig. 9a–d). Conversely, adding cell-permeable mevalonate, a downstream HMGCR product, enhanced migration in both SORD KO and WT cells (Fig. 4l and Extended Data Fig. 9e,f). Oral administration of simvastatin, another statin, significantly reduced liver metastasis in a Glu + Fru-fed caecum injection mouse model (Fig. 4m). These findings indicate that SORD-driven activation of the mevalonate pathway under Glu + Fru conditions promotes CRC migration and metastasis (Fig. 4n). Discussion Despite advances in CRC treatment, metastasis remains the primary cause of CRC-related mortality17. Although diet is a key environmental factor in CRC development, the specific dietary components and underlying mechanisms by which diet contributes to metastasis remain unclear, despite prior studies investigating this relationship33. Our study fills this gap by establishing a causal link between SSBs and CRC metastasis through SORD. This finding is further supported by human epidemiological studies demonstrating a positive correlation between SSB intake and increased CRC recurrence and mortality11,12. Our findings underscore the need to use both glucose and fructose when studying the effects of SSBs in preclinical models. Prior studies using fructose alone do not reflect the actual composition of SSBs, which contain both sugars34,35. Moreover, cells in organisms are never exposed to fructose alone in vivo without glucose, even after consuming 100% fructose solutions—such beverages do not naturally occur. In our study, CRC cells were exposed to a combination of glucose and fructose, closely mimicking the physiological environment resulting from the consumption of SSBs. The Glu + Fru condition enabled the SORD reverse reaction, which was absent under glucose-only or fructose-only conditions. Ultimately, our results revealed that the combination of glucose and fructose uniquely enhances CRC motility and metastasis through SORD-mediated NAD⁺ recycling, supporting enhanced glycolysis and subsequent activation of the mevalonate pathway (Fig. 4n). Although our study demonstrated the direct effects of SSBs on cancer migration and metastasis using cell lines and orthotopic xenografts, further investigation is warranted to elucidate how SSBs influence other components of the tumour microenvironment—such as immune cells and the microbiota—and their collective roles in promoting metastasis. Our results also suggest potential strategies to inhibit CRC metastasis. First, targeting SORD may reduce CRC metastasis. Historically, research on the polyol pathway has focused primarily on AR in diabetic complications36,37, with limited understanding of the role of SORD in cancer38,39. Our study showed that SORD deletion suppressed migration even under glucose-only conditions and that SORD is highly expressed in CRC tumours—especially in stem-cell-like tumour cells—together highlighting its fundamental role in CRC motility, independent of sugar conditions. Second, lowering the NAD+/NADH ratio in CRC cells may also inhibit metastasis. Recent studies have linked elevated NAD+/ NADH levels with cell proliferation and survival in various cancers26–29. Our findings reveal a previously unappreciated role for elevated NAD⁺/ NADH levels in promoting CRC migration and metastasis, independent of their effects on cell proliferation. Third, statins may be repurposed to target CRC metastasis, as our study demonstrates that activation of the mevalonate pathway under Glu + Fru conditions contributes to CRC metastasis. However, the detailed underlying mechanisms—such as how SORD deletion leads to decreased expression of mevalonate pathway genes, and how activation of this pathway promotes metastasis—remain unclear, reflecting the broader complexity of this pathway in cancer metastasis31,32,40. Although preclinical studies suggest that statins may serve as effective anti-cancer agents41, epidemiological results have been mixed regarding their effects on CRC survival42,43. Given variability in SORD expression and dietary sugar intake among patients, statins may be especially beneficial in cases with high SORD expression and high SSB consumption—supporting the rationale for evaluating these links through a prospective randomized cohort. Lastly, our preclinical data highlight the potential of SSB consumption to accelerate CRC metastasis, underscoring the urgent need for dietary guidance focused on reducing SSB intake to lower CRC mortality.