Integrated transcriptomics and metabolomics studies reveal the therapeutic effects of Astragalus polysaccharides on cancer cachexia muscle atrophy

Abstract

Cancer cachexia (CC) is a condition causing significant muscle loss in advanced cancer patients, severely impacting their quality of life and life expectancy. The autophagy–lysosome system is a key pathway in muscle depletion in CC, but targeted therapies are lacking. This study investigates how Astragalus polysaccharides (APS) from the traditional Chinese herb Astragalus membranaceus alleviate muscle wasting in CC mice. Mice were divided into control, model, and APS high-dose groups. Results show high-dose APS significantly improved grip strength and muscle fiber cross-sectional area in CC mice. Transcriptomic analysis revealed differentially expressed genes (DEGs) enriched in autophagy and mitochondrial autophagy pathways. Electron microscopy showed APS reduced autophagic vesicles and protected muscle and mitochondria. Metabolomic analysis indicated APS regulates the expression of nitric oxide synthase (NOS). APS protects against muscle wasting in CC by suppressing excessive autophagy and reducing the expression of NOS, suggesting it as a potential therapeutic agent for mitigating muscle depletion in cancer cachexia.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-11342-x.

Introduction

Cancer cachexia is a chronic wasting syndrome that affects more than half of cancer patients¹. The incidence of cancer cachexia in advanced cancer patients is as high as 86%, and approximately 8.6 million people die each year due to cancer cachexia. Notably, weight loss is an independent predictor of mortality in cancer patients^2,3. The incidence of CC is related to the cancer site and metastasis, with a higher risk reported in lung, liver, and pancreatic cancers, followed by head and neck cancers, as well as cancers of the digestive system⁴. CC is diagnosed based on phenotypical criteria, which include one or more of the following: weight loss greater than 5%, body mass index (BMI) less than 20 kg/m², reduced muscle mass within 6 months, or a combination of reduced food intake and increased catabolism, such as chronic malabsorption or a reduced diet⁵. Currently, the catabolism of muscle and fat, negative energy balance, inflammation, and neurological disorders are widely discussed in the study of cachexia⁶.

The 2011 expert consensus identified the loss of skeletal muscle as a key factor in the dysfunction seen in patients with CC⁷. Skeletal muscle degeneration and alterations in muscle protein degradation rates in CC are attributed to impaired protein regulation and energy imbalance⁷. The reduction in skeletal muscle strength, mass, and function severely impacts the quality of life and life expectancy of cancer patients with cachexia⁸. Also skeletal muscle atrophy and incomplete recovery of muscle protein mass increase the risk of adverse outcomes⁹. Therefore, modulating skeletal muscle protein metabolism and energy balance to mitigate skeletal muscle atrophy is a critical strategy for the treatment of CC.

The word “autophagy” is of Greek origin and means "to eat oneself". There are three types of autophagy: macroautophagy, chaperone-mediated autophagy, and microautophagy. This study primarily discusses macroautophagy (hereafter referred to as autophagy), which encapsulates cargo through double-membrane vesicles (autophagosomes) and transports them to lysosomes for degradation¹⁰. Mitochondrial autophagy, or mitophagy, is a selective type of autophagy that degrades damaged mitochondria¹¹. The regulation of the balance between skeletal muscle protein degradation and synthesis is a complex process governed by autophagy. Both excessive and insufficient autophagy can lead to an imbalance in this process¹². In C26-induced CC, overexpression of autophagy positive regulators exacerbates muscle atrophy¹³. Mitochondria, often referred to as the intracellular “energy factories”, produce Adenosine Triphosphate (ATP) to meet cellular energy demands, and mitochondrial dysfunction is strongly associated with the occurrence and development of many diseases. Skeletal muscle relies on oxidative metabolism for energy, placing significant demands on mitochondrial metabolic homeostasis^14,15. Mitochondrial autophagy regulates both the quantity and function of mitochondria, and its imbalance disrupts reactive oxygen species (ROS) homeostasis, resulting in oxidative stress¹⁶. Excessive autophagy reduces mitochondrial function, as evidenced by decreased respiration and insufficient cellular energy supply, which in turn affects muscle strength¹⁷. Both autophagy and mitochondrial autophagy are crucial for maintaining muscle mass and function.

The Chinese herb Astragalus membranaceus has held a critical position in the history of Traditional Chinese Medicine (TCM) and is commonly used in modern clinical treatments. Astragalus polysaccharide or Astragalus membranaceus polysaccharides, extracted and purified from the dried roots of Astragalus membranaceus (belonging to the Fabaceae family), is a water-soluble heteropolysaccharide. Modern medical research has demonstrated its therapeutic effects on diseases such as diabetes and gastrointestinal cancer, as well as its ability to modulate the immune system^14,18. APS has also been reported to exhibit anti-apoptotic properties, antioxidant, and anti-inflammatory, and it helps maintain mitochondrial homeostasis in hepatocytes¹⁹. Our previous study also demonstrated the therapeutic effect of astragalus polysaccharide on the depletion of cachexia fat²⁰. However, the mechanisms by which APS influence muscle atrophy have not yet been studied. Current treatment options for CC primarily focus on anti-inflammatory therapy, appetite stimulation, and promoting voluntary exercise²¹. But CC is often associated with advanced cancer, and these methods are not suitable for all patients. Additionally, there is a lack of targeted therapies for muscle atrophy and functional loss. In this study, we apply the Chinese medicinal monomer APS to address muscle atrophy in CC for the first time and explore its mechanism of action.

Multi-omics research is a comprehensive analytical approach that integrates data from multiple biological levels to provide a holistic view of biological systems. This approach is particularly advantageous for studying metabolic disorders with complex pathological changes. In sarcopenia, multi-omics has identified DEGs between patients and controls, and alterations in the levels of proteins involved in mitochondrial energy metabolism as well as those associated with neuromuscular junctions. Additionally, metabolomics analyses of urine and blood have revealed alterations in acylcarnitines and branched-chain amino acids. These findings have helped identify potential biomarkers²². However, multi-omics studies focusing on muscle atrophy in the setting of CC have not yet been reported. In this study, we used transcriptomics and metabolomics for the first time to investigate the potential mechanisms of muscle atrophy in CC and the medicinal effects of APS. We identified DEGs and differentially expressed metabolites (DEMs) among the control, model, and medicated groups, followed by enrichment and correlation network analyses of both datasets. This study demonstrated that the autophagy-lysosomal pathway, including mitochondrial autophagy, contributes to skeletal muscle atrophy and dysfunction in CC. APS modulated this process, attenuating muscle atrophy and functional decline. Additionally, we found that APS modulates the expression of nitric oxide synthase. Figure 1 shows a flowchart of this study.

Fig. 1.

Study flowchart. This figure was drawn in the Generic Diagramming Platform.

Materials and methods

Cell culture

The CT26 mouse colorectal cancer cell line (CL-0071) was purchased from Procell Life Science & Technology (Wuhan,China). The cells were cultured in medium supplemented with 10% fetal bovine serum (164,210,Procell,Wuhan,China) and Penicillin–Streptomycin Solution (PB180120,Procell,Wuhan,China). Cell cultures were maintained at 37 °C in a humidified incubator with 5% CO₂ atmosphere.

Animal model

Fifty male BALB/c nude mice (6 weeks old, weighing 20–25 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China) (License No.SCXK (Beijing) 2021–0006). The mice were kept in individually ventilated cages (IVCs) with 24-h constant-pressure ventilation, in an environment maintained at 21–24 °C with constant humidity and a 12-h light/dark cycle. Following a one-week acclimatization period, the mice were randomly assigned to five groups (n = 10): control group, model group (with CC), high-dose APS group, medium-dose APS group, and low-dose APS group. Randomization was performed to avoid differences in body weight. CT-26 colon cancer cells (5*10⁵ cells) were suspended in 100 μL of PBS and injected subcutaneously into the right flank of the mice. An identical amount of PBS was injected into the control group. Take the weight of the mice every four days. On day 8, the tumor was palpable, and on day 12, the tumor length and width were measured using a caliper. The tumor volume calculation was based on the formula: tumor volume = 0.56 × length × width^{2 23,24}. The day after tumor injection, APS (A7970, Solarbio, Beijing, China, Lot. No. 2230818001) were administered by gavage to the high-, medium-, and low-dose groups (800 mg/kg, 400 mg/kg, and 200 mg/kg, respectively) in a volume of 0.2 mL per mouse. The model and control cohorts received the same volume of saline through gavage, utilizing an identical method and dosage²⁵. After 4 weeks of treatment, mice were euthanized under 4% isoflurane anesthesia²⁶. The gastrocnemius muscle was dissected for hematoxylin and eosin (HE) staining and electron microscopy experiments. The entire leg muscle was dissected and homogenized, with three randomly selected samples from each group used for transcriptomics sequencing, eight for metabolomics sequencing, and the remaining samples reserved for subsequent experimental validation. For the sake of brevity, the term “muscle” or “leg muscle” will be used to refer to the samples in the following text. The experimental protocol was approved by the Animal Ethics Committee of the Affiliated Hospital of Shandong University of Traditional Chinese Medicine (No. SDSZYYAWE20241118002). This study was conducted in compliance with ARRIVE guidelines. This study complied with the Regulations on the Management of Laboratory Animals issued by the National Science and Technology Commission and the Implementing Rules for the Management of Medical Laboratory Animals issued by the Health Commission, as well as the relevant statutes of the Experimental Animal Ethics Committee of the Affiliated Hospital of Shandong University of Traditional Chinese Medicine, and the conditions of the staffing and the equipment were in accordance with the requirements.

Grip strength

Whole-limb grip strength was measured using a Grip Strength Meter (XR501, Shanghai Xinruan Information Technology Co., Ltd, China). Mice were placed on the grip device with their limbs grasping the grid plate. The body of each mouse was positioned perpendicular to the front edge of the grid plate. The tail of the mouse was gently held parallel to the grid plate’s surface, and the mouse was slowly and steadily pulled backward until its limbs released from the grid. The maximum force generated before detachment was recorded in grams-force (gf). Each mouse underwent three evaluations, with the results averaged for statistical analysis.

mRNA sequences

Total RNA was extracted using the Tianmo #TR205-200 kit. RNA purity and concentration were measured using a Qubit® 3.0 fluorometer (Life Technologies, USA) and a Nanodrop One spectrophotometer (Thermo Fisher Scientific Inc., USA). RNA integrity was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., USA). Samples with an RNA Integrity Number (RIN) > 6.0 and a 28S/18S ratio ≥ 0.7 were selected for sequencing. Library construction was performed using the Stranded RNA-Seq Library Prep Kit (ABclonal, China) with 1 µg of RNA as input. Poly-A mRNA was purified using Poly-T oligo-ligated magnetic beads, followed by reverse transcription to synthesize first-strand cDNA. Second-strand cDNA was synthesized, and the cDNA was end-repaired, adding an 'A' base and ligating adapters. The final cDNA library was amplified by PCR and purified. The libraries were quantified using a Qubit® 3.0 fluorometer and assessed for quality using an Agilent 2100 Bioanalyzer. The library concentration was adjusted to 10 pM, and sequencing was performed on the Illumina NovaSeq 6000 platform. Library construction and sequencing were carried out by Sinotech Genomics Co., Ltd. (Shanghai, China).

Metabolomics analysis by LC–MS

Non-targeted metabolomics analysis of CC muscle tissue was performed using liquid chromatography-mass spectrometry (LC–MS, Waters UPLC, Thermo Q Exactive). Samples were thawed on ice, and 50 mg of tissue was mixed with 800 μL of 80% methanol, vortexed, ground at 65 Hz for 180 s, and sonicated at 4 °C for 30 min. After centrifugation, 120 μL of supernatant was collected and mixed with 3 μL of internal standard (0.20 mg/mL dichlorophenylalanine) before storage. Analysis was conducted on an ACQUITY UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm) at 40 °C with a flow rate of 0.3 mL/min, using mobile phases A (0.05% formic acid in water) and B (acetonitrile). A 5 μL injection volume was analyzed at an autosampler temperature of 4 °C. The LC–MS system performed a full scan (m/z 70–1050) and data-dependent secondary scans (dd-MS2, TopN = 10) with resolutions of 70,000 and 17,500 for primary and secondary scans, respectively, using high-energy collisional dissociation (HCD).

Data analysis

Transcriptomics data analysis

Sequencing data were stored in FASTQ format, evaluated using FastQC, and filtered with fastp. Sequence alignment was performed using HISAT2 with a spliced mapping algorithm. Gene coverage was quantified based on gene locations relative to the reference genome. Post-alignment, fragment counts were obtained using StringTie and normalized with the TMM method. Gene expression abundance was assessed using FPKM. Differential expression analysis was conducted with the edgeR package, adjusting P-values for multiple testing and applying a Q value threshold (Q < 0.05, |log2(FC)|> 1, and FPKM > 1). DEGs were visualized in a heatmap, and functional pathway analysis was performed using Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment. Pathway-gene networks were constructed in Cytoscape (v. 3.9.0) and analyzed with Centiscape 2.2 to identify key genes. Gene set enrichment analysis (GSEA) was also performed on the transcriptomic data.

Metabolomics data analysis

The QE platform employs electrospray ionization (ESI) with positive (POS) and negative (NEG) ion modes to enhance metabolite coverage and detection. Data from both modes included 3 quality control (QC) samples and 24 experimental samples, yielding 7,896 peaks. Peaks were retained if missing values were below 50% in any group. Missing values were imputed using the half-minimum value method, and data were normalized to total ion current (TIC). After preprocessing, 7,896 peaks remained and were log-transformed and UV-scaled using SIMCA software (V14.1, Sartorius Stedim Data Analytics AB). An orthogonal partial least squares-discriminant analysis (OPLS-DA) model was constructed based on the first principal component (VIP), with model quality evaluated via sevenfold cross-validation and permutation tests. Model performance was assessed using R²Y (explained variance) and Q² (predictive capability). Differential metabolites were identified with a Student’s t-test P-value < 0.05 and a VIP score > 1 in the OPLS-DA model. Metabolite pathway analysis was conducted using the KEGG and MetaboAnalyst 6.0 databases.

Multi-omics joint analysis

The KEGG pathway enrichment results were analyzed to identify common enriched pathways between the two histological groups. Correlation clustering of transcriptomic and metabolomic data was conducted using R. Spearman correlation was used to calculate associations between DEMs and DEGs, selecting pairs with P < 0.05 for heatmap visualization. Most of the maps mentioned above were generated using https://www.bioinformatics.com.cn (accessed on 30 Aug 2024), an online platform for data analysis and visualization.

HE staining

Muscle tissues were extracted from paraformaldehyde, set in paraffin blocks, and sectioned into 4 μm slices using a pathology slicer (RM2016, Shanghai, China). The paraffin was removed, and the slices were rehydrated in water prior to staining with Hematoxylin–Eosin (H&E) using the HD constant dye kit (G1076, Servicebio, Wuhan, China). The staining process involved using the pretreatment solution for 1 min, followed by hematoxylin staining for 3 to 5 min after differentiation and returning to blue. The slices were then dehydrated and stained with eosin staining solution for 15 s, after which they were dehydrated and sealed. The H&E-stained sections were microscopically examined using a Nikon Eclipse E100 optical microscope (Nikon, Tokyo, Japan), and images were captured using a NIKON DS-U3 imaging system (Nikon, Tokyo, Japan). The images were analyzed using ImageJ software (NIH, USA) to calculate the muscle fiber cross-sectional area.

Transmission electron microscopy

Fresh mouse gastrocnemius muscle was quickly dissected into 1 mm³ fragments using a razor blade and subsequently preserved in 2.5% glutaraldehyde (G1102, Servicebio, Wuhan, China). Tissues were subsequently treated with 1% osmium tetroxide in 0.1 M phosphate buffer (pH 7.4) in a light-proof room for 2 h. Following rinsing and dehydration at room temperature, tissues were embedded in resin and sectioned into 1.5 μm semi-thin slices using a Leica UC7 ultrathin microtome (Leica, Wetzlar, Germany). The semi-thin sections were then processed into 60–80 nm ultrathin slices, and the tissues were fished out onto the 150-mesh copper grids with formvar film. The sections were stained with a saturated alcohol solution of 2% uranyl acetate and a 2.6% lead citrate solution under specified conditions. Observations were conducted at 80 kV on a HT7800 transmission electron microscope (HITACHI, Tokyo, Japan).

Western blot

Mouse muscle tissues were cut and added to RIPA lysate (P0013B, Beyotime, Shanghai, China), which contains a protease inhibitor mixture (P1005, Beyotime, Shanghai, China) to prevent protein degradation and de-modification. The protein supernatant was obtained by centrifugation after thorough lysis. Protein concentration was determined using the Bicinchoninic Acid (BCA) kit (CW0014S, Cwbio, Jiangsu, China), and the samples were subsequently diluted to a final concentration of 2 µg/µL. Next, proteins were separated according to their theoretical molecular weights using different concentrations of One-Step PAGE gel rapid preparation kits, specifically 8% PAGE gel (E303-02, Vazyme, Nanjing, China), 10% PAGE gel (E303-03, Vazyme, Nanjing, China), and 12% PAGE gel (E303-02, Vazyme, Nanjing, China). Proteins were blotted onto a PVDF membrane with a pore size of 0.45 μm (WGPVDF45, Millipore, Massachusetts, USA) for subsequent immunoblotting. Membranes were treated with 5% skimmed milk powder for 1 h at 25 °C to prevent non-specific interactions, followed by washing and a subsequent overnight incubation at 4 °C with the primary antibody. The primary antibody was subsequently bound with a species-specific HRP-conjugated secondary antibody. Detection of protein bands was achieved using an enhanced chemiluminescence (ECL) kit (PK10002; Proteintech; Wuhan, China), and gray values were quantified using ImageJ software.

Primary antibodies used for the experiments:

Anti-beta Tubulin (M1305-2, 1:5000, HUABIO, Hangzhou, China);

Anti-Fbx32 (Atriogin-1) (ET7109-25, 1:2000, HUABIO, Hangzhou, China);

Anti-BNIP3 (ET1704-08, 1:500, HUABIO, Hangzhou, China);

Anti-LC3B (ET1701-65, 1:1000, HUABIO, Hangzhou, China);

Anti-Phospho-FOXO3a (S253) (ET1609-49, 1:1000, HUABIO, Hangzhou, China);

Anti-nNOS (ET1609-61, 1:1000, HUABIO, Hangzhou, China);

SQSTM1 (p62) (HA721171, 1:5000, HUABIO, Hangzhou, China);

TRIM63 (Murf1) (55,456–1-AP, 1:1000, Proteintech, Wuhan, China);

GABARAPL1 (11,010–1-AP, 1:1000, Proteintech, Wuhan, China);

eNOS (27,120–1-AP, 1:500, Proteintech, Wuhan, China);

FOXO3A (66,428–1-Ig, 1:2000, Proteintech, Wuhan, China).

Statistical analyses

All data were analyzed using GraphPad Prism software (version 9) and are expressed as mean ± SD. Prior to group comparisons, homogeneity of variances was assessed using the Brown-Forsythe test. For multiple-group comparisons, one-way ANOVA was performed when variances were homogeneous (*P* ≥ 0.05 for Brown-Forsythe test). When ANOVA indicated overall significance, Tukey’s post-hoc test was used for pairwise comparisons. When variances were heterogeneous (Brown-Forsythe *P* < 0.05), Brown-Forsythe ANOVA (robust ANOVA) or Welch ANOVA was employed; if significant, Games-Howell post-hoc test was applied for pairwise comparisons. The significance level was set at *P* < 0.05. Exact *P* values for ANOVA, Brown-Forsythe/Welch ANOVA, and respective post-hoc tests will be presented in the figure legends. If not otherwise specified, ANOVA tests used in the following text assume homogeneity of variances.

Result

High-dose APS ameliorates muscle atrophy in colon cancer mice

In this study, the body weight of mice was measured every 4 days after the injection of colon cancer cells. After 28 days of treatment, the mice were euthanized. Subcutaneous tumors were excised to measure the tumor-free body weight. The tumor-free body weight of the model group was significantly lower than that of the control group (P < 0.05), indicating that CT26 colon cancer tumors led to a significant reduction in body weight. The mean tumor-free body weights of the high-, medium-, and low-dose APS groups were all higher than that of the model group, indicating a trend of increased body weight, although these differences were not statistically significant (P > 0.05) (Fig. 2A). One day before euthanasia, a whole-limb grip strength test was performed. Grip strength decreased to varying degrees in all groups, with the model group showing the greatest reduction (P < 0.05). The high-dose group exhibited the smallest decrease, which was significantly higher than that of the control group (P < 0.05) (Fig. 2E). However, compared to the model group, as shown in Fig. 2B,F, the tumor volume measured on day 28 showed no significant difference in the three treatment groups (P > 0.05). HE-stained sections of skeletal muscle (Fig. 2C,D) showed a significant reduction in muscle fibre area in the model group, while the high-dose APS group showed the greatest recovery in muscle fibre area compared to the other two drug administration groups. These results indicate that subcutaneous transplantation of CT26 colon cancer cells induced cachexia in BALB/c mice. Although high-dose APS may not have had a pronounced effect on body weight gain, it exhibited the most significant efficacy in restoring grip strength and muscle fiber area in mice. Therefore, the control, model, and high-dose APS groups were selected for transcriptomic and metabolomic analyses, which are labeled as “MN”, “MM”, and “MP”, respectively, in the figure below.

Fig. 2.

Body weight, tumour, and muscle atrophy in mice. (A) Changes in body weight on day 29 in each group of mice; day 29 reflects the body weight after tumour removal. Brown-Forsythe test: P < 0.05, Significant differences by Brown-Forsythe ANOVA (P < 0.001) and Welch ANOVA (P < 0.05), Games-Howell post-hoc test was applied for pairwise comparisons: N vs. M P = 0.019, M vs. APS-H P > 0.05, n = 10. (B) Tumour volumes in each group on day 29. Significant differences by one-way ANOVA (P > 0.05). (C) Muscle fibre cross-sectional area sizes calculated from the HE sections. Brown-Forsythe test: P < 0.05, Significant differences by Brown-Forsythe ANOVA (P < 0.001) and Welch ANOVA (P < 0.0001), Games-Howell post-hoc test was applied for pairwise comparisons: N vs. M **P = 0.0084, M vs. APS-H ***P = 0.0001, n = 4. (D) HE-stained sections of skeletal muscle in each group. (E) Whole-limb grip strength was measured on day 28 in each group of mice, and grip strength was normalised to body weight. Significant differences by one-way ANOVA (P < 0.0001) and Tukey’s test: N vs. M ****P < 0.0001, M vs. APS-H ****P < 0.0001, n = 4. (F) Tumours in each group.

Transcriptomic results of mice with cancer cachexia after APS treatment

In order to elucidate the mechanism by which APS treats muscle atrophy associated with CC, we conducted transcriptomic sequencing across the control, model, and high-dose groups. Inter-sample relationships were calculated using Spearman’s correlation after gene quantification. The correlation coefficients of biological replicates were generally larger than those of samples from different biological replicates. Principal Component Analysis (PCA) and sample correlation analysis were conducted to assess inter-sample relationships, revealing differences among samples in the three groups (Fig. 3A,B).

Fig. 3.

Transcriptomics Analysis. (A) Correlation plot of individual samples in each group; darker blue indicates a stronger correlation. (B) PCA of the three groups. The first two principal components are represented on the horizontal and vertical axes. The values next to the principal components represent the percentage of variance explained by the data. (C) Volcano plots of DEGs for the MM vs MN group. (D) Volcano plots of DEGs for the MP vs MM group. (E) Venn diagrams of DEGs between the MM vs MN group and the MP vs MM group, with the overlap representing the intersection of the DEGs. (F) Heatmap of differential gene expression in the intersection of the MM vs MN group and the MP vs MM group. (G) Plots of KEGG-enriched DEGs in the MM vs MN group. (H) The co-enriched KEGG plots in both the MM vs MN and MP vs MM, with the dot size representing the number of enriched genes.

The screening criteria for DEGs followed |log2(FC)| > 1, Q < 0.05, and an average FPKM > 1 for either group. 1,607 DEGs were screened in the model group when compared to the control group (MM vs MN), including 577 down-regulated genes and 1,030 up-regulated genes (Fig. 3C). The analysis revealed 131 DEGs in the APS group as opposed to the model group (MP vs MM), comprising 99 down-regulated and 32 up-regulated genes (Fig. 3D). The Venn diagrams of the transcriptomics results from the two groups indicated that there were 119 common differential genes (Fig. 3E), and the heatmaps showed that they were expressed according to the trends of ‘high-low–high’ or ‘low–high-low’ (Fig. 3F). These genes may serve as potential target genes for APS in the treatment of cancer cachexia.

The DEGs were mapped to KEGG pathways to identify significantly enriched pathways. The KEGG results of the MM vs MN group were sorted by P-value from smallest to largest, revealing that the top 30 pathways were enriched (Fig. 3G). Among them, 15 pathways were co-enriched by DEGs in both the MM vs MN and MP vs MM (Fig. 3H). We have observed that there are multiple pathways associated with autophagy in the KEGG pathway, including 'FOXO signaling pathway’, 'Autophagy-animal’, 'Mitophagy-animal’, 'mTOR signaling pathway’ and 'AMPK signaling pathway’. These results suggest that after APS treatment, the expression of some DEGs in the muscle tissue of cachexia showed a tendency to return to normal levels. For instance, genes such as Gabarapl1, Ggad45b, and Igf1r, which are related to autophagy, shifted from upregulated/downregulated states to normal levels following APS treatment. Additionally, the KEGG pathways shared between the MM vs MN and MP vs MM groups in cachectic muscle were related to autophagy. This suggests that the pharmacological effects of APS on muscle atrophy in cancer cachexia may involve autophagy and mitochondrial autophagy mechanisms.

APS regulates autophagy and mitochondrial autophagy-related gene transcription in muscles with cachexia

Based on previous studies, we found that the pathways co-enriched in the MM vs MN and MP vs MM groups were associated with ‘Autophagy’ and ‘Mitophagy’, which we hypothesized could be potential mechanisms for the treatment of malignant muscular dystrophy by APS. Further analysis revealed that in the autophagy-related pathways (Table 1), the genes that increased in the MM vs MN group and decreased in the MP vs MM group included Igf1r, Gadd45b, Prkag2, Gabarapl1, Hoga1, Odc1, Cd82, Ubc, Smcr8, Atp6v1b2, Slc7a5, and Ppargc1a.

Table 1.

Enriched genes in two groups of autophagy-related pathways.

ID	PATHWAY	gene_UP_list MMvsMN	gene_DOWN_list MPvsMM
mmu04068	FOXO signaling pathway	Gabarap,Fbxo32,Igf1r,Pik3r1,Foxo1,Rbl2, Pik3ca,Gadd45b,Irs2,Prkag2,Plk3,Egfr, Bcl2l11,Insr,Foxo3,Bnip3,Gabarapl1, Gadd45g,Akt1,Cdkn1a	Igf1r,Gadd45b,Prkag2, Gabarapl1
mmu04115	p53 signaling pathway	Igfbp3,Gadd45b,Bcl2l1,Thbs1,Siah1a,Sesn2, Sesn1,Cd82,Gadd45g,Fas,Cdkn1a	Gadd45b,Cd82
mmu04137	Mitophagy—animal	Gabarap,Ulk1,Atf4,Map1lc3b,Optn,Bcl2l1, Ubc,Nbr1,Tax1bp1,Bnip3l,Sqstm1,Ubb, Tbc1d15,Foxo3,Bnip3,Gabarapl1	Ubc,Gabarapl1
mmu04140	Autophagy—animal	Gabarap,Ulk1,Igf1r,Pik3r1,Nrbf2,Map1lc3b, Ddit4,Pik3ca,Zfyve1,Bcl2l1,Ern1,Irs2, Mtmr14,Smcr8,Uvrag,Rb1cc1,Dapk1, Sqstm1,Bnip3,Atg13,Gabarapl1,Ctsl,Akt1	Igf1r,Smcr8,Gabarapl1
mmu04150	mTOR signaling pathway	Ulk1,Lpin1,Cab39,Eif4ebp1,Igf1r,Pik3r1, Atp6v1b2,Atp6v1h,Slc7a5,Ddit4,Pik3ca, Atp6v1a,Sesn2,Insr,Slc3a2,Castor1,Flcn, Lpin3,Akt1	Igf1r,Atp6v1b2,Slc7a5
mmu04152	AMPK signaling pathway	Ulk1,Cab39,Eif4ebp1,Igf1r,Pik3r1,Foxo1, Pik3ca,Irs2,Prkag2,Insr,Foxo3,Ppargc1, Pfkfb3,Akt1	Igf1r,Prkag2,Ppargc1a

We focused on the 'FOXO signaling pathway’, 'Autophagy-animal’ and 'Mitophagy-animal’ the three pathways most directly related to autophagy, and mapped the MM vs MN-enriched gene network (Fig. 4A). These genes were up-regulated in MM vs MN, of which only Gabarapl1 and Igf1r (Table 1) were down-regulated in the MP vs MM group (Fig. 4A). But this could not paint a complete picture of the role of APS. We observed that the MM vs MN group was enriched in high-frequency genes on these pathways, including Irs2, Pik3ca, Akt1, Pik3rl, Igf1r, Gabarapl1, Gabarap, Bnip3, Sqstm1, Ulk, Bcl2l1, Map1lc3b, and Foxo3. Upon further analysis, we found that not all genes exhibited significant differential expression between the MP and MM groups. However, they still demonstrated a ‘low–high–low’ expression trend (Fig. 4B). All expressed genes in MM vs MN and MP vs MM were separately ranked in descending order based on logFC, and GSEA enrichment analysis was performed on all genes involved in the KEGG pathways 'FOXO signaling pathway’, 'Autophagy–animal’ and 'Mitophagy–animal’. The differential expression of genes in the 'MM vs MN’ was predominantly localized to the side indicating activation of those three pathways, whereas in the 'MP vs MM’, the differentially expressed genes were predominantly localized to the suppressed side (Fig. 4C–H). Taken together, we hypothesized that autophagy and mitochondrial autophagy are over-activated in cancer-induced malignant muscle tissues, and that APS modulates the expression of autophagy- and mitochondrial autophagy-related transcripts in cancer cachexia muscles, reducing this activity.

Fig. 4.

Further Analysis of Transcriptomics. (A) Network diagram of KEGG pathways in the MM vs MN group, including the 'FOXO Signalling Pathway’, 'Autophagy-Animal’, and 'Mitochondrial Autophagy-Animal’. (B) Heatmap of the expression of high-frequency genes enriched in the 'FOXO signalling pathway’, 'Autophagy-animal’, and 'Mitophagy-animal’ pathways in the MM vs MN group. (C) GSEA of 'Autophagy-animal’ in the MM vs MN group. (D) GSEA of 'Mitophagy-animal’ in the MM vs MN group. (E) GSEA of 'Foxo signalling pathway’ in the MM vs MN group. (F) GSEA of 'Autophagy-animal’ in the MP vs MM group. (G) GSEA of 'Mitophagy-animal’ in the MP vs MM group. (H) GSEA of 'Foxo signalling pathway’ in the MP vs MM group. Bioinformatic analysis was performed using the OECloud tools at https://cloud.oebiotech.com.

Metabolomic profile of cancer cachexia muscle after APS treatment

To analyze the changes in muscle metabolites after APS treatment, we assessed metabolite expression levels across 24 samples (with 8 samples per group). OPLS-DA analysis revealed clear separation of metabolic states among the MN, MM, and MP groups in both POS and NEG (Fig. 5A–D). Differential metabolites were screened according to VIP > 1 and P < 0.05. In the positive ion mode, 65 differential metabolites were identified in the MM vs MN comparison and 28 in the MP vs MM group, with 11 metabolites being differentially expressed in both comparisons. In NEG, 123 metabolites were differentially expressed in the MM vs MN group and 35 in the MP vs MM group, with 16 metabolites shared between the two comparisons. Sixteen metabolites showed differential metabolism in both groups and were expressed according to the trends of 'high–low–high’ and 'low–high–low’ (Fig. 5E,F). The 27 metabolites were categorized and plotted in a pie chart (Fig. 5G), with amino acid metabolites and derivatives accounting for the largest proportion, followed by amino acids, tricarboxylic acid cycle metabolites, coenzymes, and bile acids. MetaboAnalyst 6.0 (https://www.metaboanalyst.ca) was used to perform pathway analysis (Fig. 5H), which showed that the significant metabolic pathways included “Riboflavin metabolism”, “Arginine biosynthesis”, “Nitrogen metabolism”and "Citrate cycle (TCA cycle)". A total of 17 metabolic pathways were plotted (Fig. 5H) (Table 2). Based on metabolic pathway analysis, we hypothesize that the metabolic effect of APS on skeletal muscle may be related to the metabolic process of nitric oxide-citrulline cycle through NOS to NO. The enriched metabolites were fumarate, flavin mononucleotide (FMN), flavin adenine dinucleotide (FAD), and L-glutamic acid (Table 2) (Fig. 6A). APS regulate NO metabolism and alleviate muscle atrophy by regulating the above processes.

Fig. 5.

Metabolomics Analysis. (A) Visual representation of OPLS-DA scores for samples from the MM and MN groups in NEG. (B) Visual representation of OPLS-DA scores for samples from the MP and MM groups in NEG. (C) Visual representation of OPLS-DA scores for samples from the MM and MN groups in POS. (D) Visual representation of OPLS-DA scores for samples from the MP and MM groups in POS. (E) Heatmap of differential metabolite expression for both the MM vs MN and MP vs MM groups in NEG. (F) Heatmap of differential metabolite expression for both the MM vs MN and MP vs MM groups in POS. (G) Metabolite categorisation circle plots. (H) Plot of differential metabolite pathway analysis.

Table 2.

Metabolic pathways and their enrichment metabolites.

Pathway	Total	Hits	Raw p	-lg(p)	Holm adjust	FDR	Impact	Hits Cpd
Riboflavin metabolism	4	2	0.000337	3.4725	0.026954	0.026954	0.5	Flavin mononucleotide,Flavin adenine dinucleotide
Glutathione metabolism	28	3	0.001089	2.9631	0.086011	0.04355	0.33242	Glutathione,L-Glutamic acid,Cysteinylglycine
Tyrosine metabolism	42	3	0.003584	2.4456	0.27954	0.095569	0.10104	Epinephrine,Homogentisic acid,FUMARATE
Arginine biosynthesis	14	2	0.004891	2.3106	0.37661	0.09782	0.11675	L-Glutamic acid,FUMARATE
Histidine metabolism	16	2	0.006393	2.1943	0.4859	0.1023	0	L-Glutamic acid,Imidazoleacetic acid
Alanine, aspartate and glutamate metabolism	28	2	0.019109	1.7188	1	0.25479	0.19952	L-Glutamic acid,FUMARATE
Nitrogen metabolism	6	1	0.046309	1.3343	1	0.52925	0	L-Glutamic acid
Butanoate metabolism	15	1	0.1121	0.95041	1	1	0	L-Glutamic acid
Ubiquinone and other terpenoid-quinone biosynthesis	18	1	0.13308	0.87588	1	1	0	Homogentisic acid
Pentose and glucuronate interconversions	19	1	0.13998	0.85394	1	1	0.12048	D-Xylulose 5-phosphate
Citrate cycle(TCA cycle)	20	1	0.14682	0.83321	1	1	0.02981	FUMARATE
Pyruvate metabolism	23	1	0.16705	0.77714	1	1	0	FUMARATE
Pentose phosphate pathway	23	1	0.16705	0.77714	1	1	0.0614	D-Xylulose 5-phosphate
Glyoxylate and dicarboxylate metabolism	32	1	0.22514	0.64754	1	1	0	L-Glutamic acid
Arginine and proline metabolism	36	1	0.24975	0.6025	1	1	0	L-Glutamic acid
Primary bile acid biosynthesis	46	1	0.30816	0.51123	1	1	0	Cholic acid
Purine metabolism	71	1	0.43643	0.36009	1	1	0.01316	Guanine

Fig. 6.

Metabolomics Mechanism Map and Joint Multi-Omics Analysis. (A) Nitric oxide-citrulline cycle mechanism. Red represents DEMs. Arginine succinate is converted to L-arginine, generating fumarate. L-arginine produces NO via NOS. Electrons within NOS are generated through the oxidation of NADPH and transferred to haemoglobin, promoting NO production from L-arginine through the regulation of FAD and FMN. L-glutamate regulates the opening of cytosolic Ca²⁺ receptors, encouraging more Ca²⁺ to enter the cell and bind with CaM, thereby increasing NOS activity. This figure was drawn in the Generic Diagramming Platform. (B) Correlation diagram of DEGs and DEMs, where red denotes a positive correlation and blue signifies a negative correlation. A deeper hue indicates a stronger relationship. (C) Differential Gene-Differential Metabolite-Pathway Network Diagram. Orange dots represent differential genes, blue dots represent DEMs in positive ion mode, green dots represent DEMs in NEG, and the larger the dot, the higher the node degree. Pink arrows indicate pathways constituting the network with differential genes, while pink diamonds represent pathways constituting the network with differential metabolites. The hexagonal shape represents the 'Arginine and proline metabolism’ pathway, which is jointly enriched by differential genes and metabolites.

Comprehensive multi-omics analysis

To deepen our knowledge of the interconnection between metabolomics and transcriptomics, we applied the PEARSON algorithm to calculate the correlation coefficients between key differential genes and key metabolites, and plotted a clustered heat map (Fig. 6B). Additionally, we constructed a differential gene-differential metabolite-pathway network diagram based on the correlations between metabolites and genes (Fig. 6C). As shown in the figure, arranged by node degree, the key genes associated with mitochondrial autophagy were Gabarapl1, Igf1r, and Gadd45b, while the key metabolites were L-glutamic acid, fumarate, and FMN. Combined correlation analysis revealed that L-glutamic acid, fumarate and FMN were positively correlated with Gabarapl1, but negatively correlated with Igf1r and Gadd45b. ‘Arginine and proline metabolism’ was identified as a common enriched pathway in both transcriptomics and metabolomics, with L-glutamic acid as the enriched metabolite.

The nitric oxide-citrulline cycle is an important pathway for generating NO, initiated by citrulline and aspartate, and involving the argininosuccinate synthase (ASS) and argininosuccinate lyase (ASL) to produce arginine and fumarate. The metabolically produced arginine is then reconverted to citrulline through the action of NOS to produce NO (Fig. 6A)²⁷. NOS is the basis for nitric oxide production, with three isoforms: neuronal nitric oxide synthase (nNOS), inducible nitric oxide synthase (iNOS), and endothelial nitric oxide synthase (eNOS), all of which are present in skeletal muscle, with nNOS and eNOS being more widely expressed^28,29. The monomeric form of NOS is inactive. The main body consists of a reductase consisting of FMN and FAD, which connects to an oxygenase structural domain. This connection relies on the tight junctions of the oxygenase structural domains to form a dimer that produces activity. The active NOS transfers electrons from Nicotinamide Adenine Dinucleotide Phosphate (NADPH) to the oxygenase structural domain (which includes tetrahydrobiopterin, heme, and arginine binding sites) via FMN and FAD, oxidizing the nitrogen atom of arginine to produce NO^30,31. Comprehensive omics analysis has unveiled the interconnections between autophagy and the nitric oxide-citrulline cycle, as well as the metabolites associated with the NOS composition.

Experimental validation

Experiments were conducted to assess the key indicators from the results mentioned above to validate the findings. Firstly, we used transmission electron microscopy to observe changes in gastrocnemius muscle fibers and mitochondria. At a magnification of 5000 × , the muscle fibers in the control group exhibited clear bright and dark bands, well-organized myofilaments, and normal mitochondria. In contrast, the model group displayed blurred bright and dark bands, degraded M-lines, indistinct and fragmented Z-lines, and a reduced number of intact mitochondria. The bright bands showed better recovery, Z-lines were clearer, and the number of mitochondria was restored following high-dose APS treatment. Mitochondria and autophagosomes were examined under a 20,000 × microscope. The model group exhibited an increased number of autophagic lysosomes and mitochondrial autophagosomes, with the mitochondrial cristae absent due to severe damage. In contrast, mitochondria in the normal and APS groups appeared to be in better condition (Fig. 7D).

Fig. 7.

Experimental Validation. (A–C) Western blot detection of muscle atrophy-related proteins Atrogin-1 and MuRF1. Atrogin-1, Significant differences by one-way ANOVA (P < 0.0001) and Tukey’s test: N vs. M ***P = 0.0002, M vs. APS-H ***P = 0.0002, n = 3. MuRF1, Significant differences by one-way ANOVA (P < 0.05) and Tukey’s test: N vs. M *P = 0.044, M vs. APS-H *P = 0.492, n = 3. (D) Electron microscope images of the control group (Group N), model group (Group M), and high-dose APS group (Group APS-H). The left images show muscle tissue observed at 5000 × magnification, high-contrast mode, with an accelerating voltage of 80.0 kV. The right images show mitochondria and autophagosomes observed under the same conditions at 20,000 × magnification. Blue arrows indicate mitochondria, and yellow arrows indicate autophagosomes. (E–I) Western blot detection of autophagy-related proteins BNIP3, LC3-I/II, GABARAPL1, and SQSTM1/p62. BNIP3, Significant differences by one-way ANOVA (P < 0.01) and Tukey’s test: N vs. M **P = 0.0035, M vs. APS-H **P = 0.003, n = 3. LC3-I/II, Significant differences by one-way ANOVA (P < 0.01) and Tukey’s test: N vs. M **P = 0.005, M vs. APS-H *P = 0.0399, n = 3. GABARAPL1, Significant differences by one-way ANOVA (P < 0.05) and Tukey’s test: N vs. M *P = 0.0204, M vs. APS-H *P = 0.049, n = 3. p62, Significant differences by one-way ANOVA (P < 0.01) and Tukey’s test: N vs. M **P = 0.0027, M vs. APS-H *P = 0.03, n = 3. (J–L) Western blot analysis of the transcriptional regulator FOXO3 and p-FOXO3. FOXO3, Significant differences by one-way ANOVA (P < 0.05) and Tukey’s test: N vs. M *P = 0.0152, M vs. APS-H *P = 0.0263, n = 3. p-FOXO3, Significant differences by one-way ANOVA (P < 0.001) and Tukey’s test: N vs. M ***P = 0.0005, M vs. APS-H *P = 0.0347, n = 3. (M–P) Western blot detection of eNOS and nNOS. eNOS, Significant differences by one-way ANOVA (P < 0.001) and Tukey’s test: N vs. M ***P = 0.0002, M vs. APS-H **P = 0.0034, n = 3. nNOS, Significant differences by one-way ANOVA (P < 0.05) and Tukey’s test: N vs. M *P = 0.0116, M vs. APS-H *P = 0.0384, n = 3.

We then validated these two marker proteins for muscle atrophy. Atrogin-1 and MuRF1 are ubiquitin ligases specific to muscle tissue, facilitating protein degradation through the regulation of ubiquitination processes^32,33. Based on literature studies, we verified the expression of these two proteins to assess the success of the modeling and the regulatory effect of APS on muscle atrophy. Western blot analysis demonstrated a significant increase in the abundance of Atrogin-1 and MuRF1 in the model group relative to the control group (P < 0.05), suggesting the occurrence of muscle atrophy in our experimental model. In contrast, the expression of these two proteins significantly decreased in the muscles after treatment with high doses of APS (P < 0.05) (Fig. 7A–C). Combined with HE-stained sections and electron microscopy, these findings suggest that high doses of APS have a mitigating effect on muscle atrophy.

Although our integrated omics analysis identified only one Gabarapl1 gene associated with autophagy, transcriptomics revealed several additional autophagy markers. The protein expression of these markers was subsequently verified in follow-up experiments. The autophagy marker LC3b (MAP1LC3B) exists in two forms, LC3b-I and LC3b-II. Normalization using the LC3b-II/I ratio revealed that it demonstrated a pronounced increase in the model group compared to the control group (P < 0.05) and decreased in the high-dose APS group relative to the model group (P < 0.05) (Fig. 7F,I), revealing that autophagic flows were enhanced in the model group and subsequently normalized with high-dose APS. The gene Gabarapl1, which was differentially expressed in all three groups, is also an autophagy marker gene. Validation confirmed that the protein expression followed a similar trend as the gene expression, exhibiting a ‘low–high-low’ pattern across the control, model, and high-dose APS groups, with significant differences (P < 0.05) (Fig. 7G,I). The expression of BNIP3 showed a consistent trend with that of GABARAPL1, also exhibiting significant differences (P < 0.05) (Fig. 7E,I). The smooth degradation of p62 implies the effective operation of the autophagy process³⁴. We assessed p62 expression in muscle and reported a notable reduction in the model group versus the control and high-dose APS groups (P < 0.05) (Fig. 7H,I). Finally, we validated the regulatory factor FOXO3 identified in the histological results and found that the expression of FOXO3 was higher in untreated malignant muscle compared to the control group (P < 0.05), and also elevated relative to the treated group (P < 0.05) (Fig. 7J,K). In contrast to FOXO3, the expression of p-FOXO3 was inversely correlated among the three groups, with phosphorylation of FOXO3 at the S253 site being reduced in the model group and subsequently restored following treatment (P < 0.05) (Fig. 7J,L).

Finally, we validated the protein expression of eNOS and nNOS across the three groups and found that their expression was upregulated in the model group, while treatment with the drug normalized their expression levels (P < 0.05) (Fig. 7M–P). In conjunction with the metabolomics data and the results presented in Fig. 6A, we observed that the products of the nitric oxide-citrulline cycle, including fumarate, the Calmodulin (CaM)-regulated metabolite L-glutamic acid, and the NOS constituents FMN and FAD, all exhibited a “low–high–low” expression pattern in the metabolomics analysis. The experimental validation confirmed that the expression trends of at least eNOS and nNOS were consistent with the aforementioned metabolites. This indicates that the nitric oxide-citrulline cycle is indeed operative in cancer cachexia muscles and that the therapeutic effects of APS are associated with this cycle, particularly with the expression of eNOS and nNOS (as denoted by NOS in other parts of the article).

Discussion

The ubiquitin–proteasome system and the autophagy–lysosome system are the primary mechanisms of skeletal muscle atrophy. While studies have established the participation of the ubiquitin–proteasome system in muscle atrophy, the mechanism of the autophagy–lysosome system in malignant cancer models is less well studied, and corresponding therapeutic measures are lacking^35,36. Autophagy is considered a "double-edged sword". On the one hand, autophagic degradation eliminates damaged mitochondria and prevents cysteine-dependent cell death³⁷. On the other hand, stress-induced autophagy fails to alleviate muscle atrophy; instead, excessive autophagy increases protein and mitochondrial degradation, and silencing autophagic flux can inhibit FOXO3-mediated muscle loss^12,38. Mitochondrial autophagy, a selective form of autophagy, is triggered by the massive loss of mitochondria, leading to dysregulation of energy metabolism, and ultimately cell death. Excessive or prolonged mitochondrial autophagy can result in enzyme leakage, which activates the cell death pathway^39,40. Previous studies have shown that autophagy, including mitochondrial autophagy, is exacerbated in CC⁴¹. Electron microscopy further confirmed the presence of multiple autophagic lysosomes in the model group, where autophagic vesicles engulfing mitochondria were observed, accompanied by more severe muscle atrophy. Our GSEA results corroborate that autophagy and mitophagy are highly active in the model group. Western blot analysis validated the increased expression of autophagy-related proteins in the model group, with elevated LC3-II/I ratio and decreased p62 protein expression. These findings suggest that excessive autophagy may be one of the key factors contributing to muscle atrophy and functional impairment in the cancer cachexia muscle model.

LC3 (MAP1LC3) belongs to the LC3/GABARAP family, with LC3B (MAP1LC3B) being the most commonly studied protein in this family. The Atg5-Atg12-Atg16 complex mobilizes LC3 to phospholipid membranes. LC3-I is initially activated by the E1 enzyme, and then passed to the E2 conjugating enzyme, where it undergoes conjugation with phosphatidylethanolamine (PE) to generate LC3-II. This step, known as LC3 lipidation, connects LC3 to the phospholipid membranes of autophagosomes, which is crucial for autophagosome formation and membrane elongation^42,43. LC3 also recognizes autophagy receptors, recruits autophagic cargo to the autophagosome membrane, and wraps the cargo during autophagosome maturation⁴⁴. GABARAPL1 is another member of the LC3/GABARAP family. Similar to LC3, GABARAPL1 exists in two forms and undergoes conversion from GABARAPL-I to GABARAPL-II, a process that is crucial for the elongation of autophagosome membranes^45,46. GABARAPL1 and LC3B act in concert during autophagy. Both proteins possess a common LC3-interacting region (LIR) motif, which enables them to bind receptors on the mitochondrial surface, thereby collectively facilitating mitophagy⁴⁷. The C-terminus of BNIP3, which functions as a mitochondrial surface receptor, contains a mitochondria-localized transmembrane domain (TMD) that enables BNIP3 to be directly targeted to the mitochondria⁴⁸. The LIR, located at the N-terminus of BNIP3, can bind directly to mature LC3 without the involvement of ubiquitin chains or p62⁴⁹. Under normal conditions, BNIP3 has minimal effect; however, it has been shown that BNIP3 is a key protein inducing mitochondrial autophagy under hypoxic conditions^50,51. p62 (SQSTM1) contains both a ubiquitinassociated (UBA) domain and a LIR, and its dual role in the autophagy process relies on these two structures. The UBA domain binds ubiquitinated autophagic cargoes, while the LIR region interacts with Atg8-family proteins, such as LC3 and GABARAPL1, enabling autophagosomes to recognize and encapsulate the cargo^52,53. p62, along with the cargo, is engulfed by the autophagosome and degraded after fusing with lysosomes. The protein level of p62 serves as a marker of autophagic flow, with decreased p62 levels indicating smooth autophagy progression, and increased p62 levels suggesting an interruption in the process^34,54. In our study, the transcriptional expression of p62/Sqstm1 was observed to be elevated in the model group but decreased following APS treatment. However, Western blot analysis revealed results contrary to the transcriptional expression. This discrepancy suggests that the active autophagy in cancer cachexia muscles leads to increased transcriptional expression of p62, while p62 protein is efficiently degraded during the autophagic process. The use of APS also altered this process in the muscles.

Through transcriptomics, we identified significant enrichment of Foxo3, Bnip3, Gabarapl1, p62/Sqstm1, and Map1lc3b in cancer cachexia muscles. These findings were further validated at the protein level by Western blot analysis. S253 is a common phosphorylation site of FOXO3, and the dephosphorylation of this site facilitates the nuclear localization of FOXO3 and enhances its transcriptional activity. Conversely, FOXO3 is inhibited in its transcriptional activity through its interaction with 14–3–3^55–57. Acting as a transcription factor, FOXO3 undergoes dephosphorylation in the cytoplasm, which promotes the transcription of autophagy-related genes, including Bnip3, Gabarapl1, and Map1lc3b^58,59. Chromatin immunoprecipitation (ChIP) sequencing identified several potential FOXO3 binding sites in the Bnip3 promoter, with increased binding observed in nutrient-deprived muscles³⁸. Activation of cytoplasmic FOXO3, followed by its translocation into the nucleus, enhances Gabarapl1 transcription, which is crucial in muscle atrophy and is associated with mitochondrial dysfunction and excessive autophagy⁶⁰. FOXO3 also promotes the lipidation of LC3, increasing the transformation of LC3-I to LC3-II, aids in the restoration of p62 degradation^38,61. A study has demonstrated that the transcriptional regulation of Gabarapl1, Bnip3, and Map1lc3b by Foxo3 plays a crucial role in excessive muscle autophagy, corroborating our findings⁶⁰. In cancer cachexia mice treated with APS, the levels of relevant proteins in the muscles were reduced, indicating that APS exerts a regulatory effect on the processes of autophagy and mitophagy. Interestingly, we found that the transcription factor FOXO3 also exerts transcriptional regulation on proteins associated with the ubiquitin–proteasome system. FOXO3 can directly target the promoter region of the Atrogin-1 gene in the nucleus, resulting in increased levels of Atrogin-1 mRNA and, consequently, increased expression of the Atrogin-1 protein. This leads to the degradation of muscle fiber proteins and a decrease in the cross-sectional area of muscle fibers⁶². In cardiac myocytes, FOXO3 accumulation in the nucleus is associated with elevated Murf1 expression⁶³. In our experimental validation, we observed that the levels of Atrogin-1 and MuRF1 were indeed elevated in muscles associated with CC, and it cannot be ruled out that this is related to the transcriptional regulation by FOXO3. Consistent with the latest research, autophagy and mitophagy in cancer cachexia are initiated through multiple pathways and mechanisms, exhibiting complex and diverse characteristics rather than occurring via a single pathway⁶⁴. Correspondingly, APS treatment has been shown to reduce autophagy levels across multiple pathways. Figure 8 illustrates the mechanisms described above.

Fig. 8.

Transcriptomic mechanism map. (A) Autophagy-lysosomal pathway and mitochondrial autophagy. The transition of LC3-I to LC3-II promotes the formation of autophagosomal membranes. GABARAPL1-I/II are functionally analogous to LC3-I/II. The p62 receptor labels ubiquitinated proteins and mitochondria, while the BNIP3 receptor directly targets mitochondria. The autophagosomal membrane encapsulates cargo such as proteins and mitochondria, which are then recognized by lysosomes for degradation. (B) Ubiquitin–proteasome pathway. MuRF1 and Atrogin-1 are both E3 ubiquitin ligases that tag proteins for further ubiquitination. These ubiquitinated proteins are recognized by the proteasome, leading to their degradation. (C) FOXO3 is involved in the transcriptional regulation of associated proteins.

As a metabolite of the nitric oxide-citrulline cycle, fumarate levels were heightened in the model group and decreased in the high-dose group of APS in the above metabolomics study. This implies that activation of the nitric oxide-citrulline cycle may be enhanced in the malignant muscles of cancer, potentially leading to increased production of excess NO. Excess NO induces nitrosative stress through the formation of potent oxidants that promote protein nitration, thereby affecting protein structure and function and even activating the ubiquitin–proteasome pathway⁶⁵. NOS is a key enzyme in the nitric oxide-citrulline cycle, incorporating FMN and FAD (Fig. 6A). Inhibition of NOS can alleviate muscle atrophy and dedifferentiation in cachexia models⁶⁶. CaM regulated electron transfer from the FMN/FAD reductase to heme is crucial for NOS activity, and this process is enhanced by an increase in intracellular Ca²⁺ concentration, which is stimulated by glutamate through NMDA receptor activation and subsequent Ca²⁺ binding to CaM^67,68. The expression levels of FMN and FAD varied among all three groups, exhibiting a fluctuating trend in a low–high–low pattern. The increase in FMN and FAD may enhance the efficiency of electron transfer. In the model group, elevated glutamate levels promoted the influx of intracellular calcium ions, thereby modulating the enzymatic activity of NOS. In the treated group, glutamate levels were lower than those in the model group, suggesting that APS may exert regulatory effects on glutamine content. Western blot analysis confirmed that the expression levels of two of the three NOS isoforms were elevated in the model group muscles and normalized following treatment, which is consistent with the expected results from our metabolomics study. NOS can generate NO, which combines with reactive oxygen species, further promoting muscle atrophy^65,69,70. NO can trigger the translocation of PRKN to mitochondria in PINK-deficient cells, inducing the activation of PINK1-PRKNmediated mitochondrial autophagy⁷¹. In vitro induction of cells with the NO donor sodium nitroprusside increased the levels of p62 and LC3-II and increased the number of autophagic lysosomes⁷². This provides a foundation for future research into the mechanisms linking NOS activity, NO production, and mitophagy in cachectic muscles, and suggests that APS may exert therapeutic effects not only by alleviating excessive autophagy but also through multiple other potential pathways.

A considerable body of preliminary research supports the use of herbal extracts in managing cancer and cachexia. The compound paeoniflorin (Pae) extracted from the root of Paeonia lactiflora inhibits TLR4/NF-κB signaling and activates the AKT/mTOR pathway to ameliorate muscle atrophy in malignant diseases⁷³. Atractylenolide I, an active ingredient extracted from the traditional Chinese herb BaiZhu (Atractylodes macrocephala Koidzumi), has been shown to ameliorate body weight loss, fat depletion, and muscle atrophy in CC⁷⁴. APS, on the other hand, can inhibit the ratio of M2-type macrophages while increasing the presence of M1-type macrophages by regulating the polarization state of tumor-associated macrophages (TAMs), which inhibits tumor growth⁷⁵. Our group has previously explored the role of astragalus polysaccharide in cachexia, finding that in cachexia caused by heart failure, it can reduce fat consumption and brown fat thermogenesis, as well as exhibit anti-inflammatory effects and improve systemic energy metabolism²⁰.

Study strengths and limitations

This study explores, for the first time, the impact of APS on muscle atrophy associated with CC, while integrating transcriptomics and metabolomics; this study proposes that the ameliorative effects of APS on excessive autophagy are mediated through multiple pathways. The study also found that APS treatment reduced the overexpression of multiple metabolites in cancer cachexia muscles and decreased the elevated expression of NOS protein. This has led to the hypothesis that nitric oxide metabolism may be related to the mechanism of mitophagy. This hypothesis could serve as a prospective research direction in the future. Based on this study, further experiments could be designed using gene knockout mice to identify the precise targets of APS. Additionally, experiments could be designed to block lysosomal degradation to confirm autophagic flow. The muscle samples in this study were homogenized from the entire leg muscles, representing an average result without considering the potential influence of different muscle fiber types. The results of this study are based on in vivo experiments but lack in vitro experiments and clinical observations. Due to the time constraints associated with sequencing, the extended preservation time of our mouse tissues has imposed limitations on the study of nitric oxide metabolism and NOS activity. Based on previous research, theoretically, the inflammatory and oxidative stress environment in cancer cachexia muscles would force iNOS to produce large amounts of NO, which would have adverse effects²⁸. However, we were unable to effectively detect the expression of iNOS in the muscles, and could only confirm the expression of the other two NOS isoforms. Future studies could be designed to further investigate this by using fresh tissues for the quantification of NO and the detection of NOS.

Conclusion

In summary, this study employs a comprehensive approach combining transcriptomics and metabolomics to propose that excessive activation of autophagy and the ubiquitin-proteasome pathway in cancer cachexia muscles have detrimental effects on muscle mass and function. We have identified APS as a potential herbal extract that protects muscles in cancer cachexia by reducing autophagy and the ubiquitin pathway. Additionally, we found that the expression and activity of certain NOS isoforms are enhanced in cancer cachexia muscles, which may have adverse effects on muscle tissue. APS also attenuated the expression of NOS.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

APS

Astragalus polysaccharides

ASL

Argininosuccinate lyase

ASS

Argininosuccinate synthase

ATP

Adenosine triphosphate

BCA

Bicinchoninic acid

BMI

Body mass index

CaM

Calmodulin

CC

Cancer cachexia

ChIP

Chromatin immunoprecipitation

DEGs

Differentially expressed genes

DEMs

Differentially expressed metabolites

ECL

Enhanced chemiluminescence

eNOS

Endothelial nitric oxide synthase

FAD

Flavin adenine dinucleotide

FMN

Flavin mononucleotide

gf

Grams-force

GSEA

Gene set enrichment analysis

H&E

Hematoxylin–eosin

HCD

High-energy collisional dissociation

iNOS

Inducible nitric oxide synthase

KEGG

Kyoto encyclopedia of genes and genomes

LC–MS

Liquid chromatography–mass spectrometry

LIR

LC3-interacting region

NADPH

Nicotinamide adenine dinucleotide phosphate

NEG

Negative ion modes

nNOS

Neuronal nitric oxide synthase

NO

Nitric oxide

NOS

Nitric oxide synthase

OPLS-DA

Orthogonal partial least squares-discriminant analysis

Pae

Paeoniflorin

PCA

Principal component analysis

PE

Phosphatidylethanolamine

POS

Positive ion modes

QC

Quality control

ROS

Reactive oxygen species

SD

Standard deviation

TAMs

Tumor-associated macrophages

TCM

Traditional Chinese Medicine

TIC

Total ion current

TMD

Transmembrane domain

UBA

Ubiquitin-associated

VIP

First principal component

Author contributions

Zhihan T, Yong W and Dufang M contributed to the study design and supervised data analysis. Zhihan T drafted and revised the manuscript. Experimental studies were conducted by Zhihan T, Xue L, and Xin X, with technical guidance provided by Yiwei Q and Shuai L. Ziyuan L and Xiaoyu S conducted the literature review. Dufang M critically reviewed and refined the manuscript. All authors have read and approved the final version of the manuscript and have agreed to their respective contributions.

Funding

This project was supported by grants from the National Natural Science Foundation of China (Nos. 82004280, 82374376), the China Postdoctoral Science Foundation (Certificate No. 2023M732136), and the Taishan Scholar Foundation of Shandong Province (No. tsqn202408382).

Data availability

The sequence data from transcriptome sequencing generated and analysed during the current study are available in the Gene Expression Omnibus (GEO) repository, GSE301467. The metabolomics matrix data are provided in the supplementary information files, named ‘Supplementary Material’.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval and consent to participate

The animal study was approved by the Animal Ethics Committee of Affiliated Hospital of Shandong University of Traditional Chinese Medicine (Approval number: SDSZYYAWE20241118002). This study adhered to the ARRIVE guidelines and complied with the National Science and Technology Commission’s "Regulations for the Administration of Experimental Animals" and "Detailed Rules for the Administration of Experimental Animals in Medical Research", as well as the relevant regulations of the Animal Ethics Committee of the Affiliated Hospital of Shandong University of Traditional Chinese Medicine. The personnel and equipment conditions met the requirements.