Abstract

Background:

Colorectal cancer (CRC) initiating/stem cells (CICs/CSCs) represent a rare tumor subpopulation with self-renewal capacity that drives tumor progression, recurrence, therapeutic resistance, and immune evasion. Despite extensive efforts to define CSCs using surface and functional markers, no universally accepted marker exists for CSC isolation and enrichment. Moreover, the molecular mechanisms underlying CSC-associated phenotypes remain incompletely characterized, highlighting the need for unbiased proteome-wide molecular profiling to better define CSC states and identify candidate biomarkers and therapeutic targets.

Methods:

CSC-enriched spheroids were generated from two colorectal cancer cell lines (SW620 and HCT-116) using three-dimensional, serum-free culture conditions and compared with their corresponding parental adherent cells. Comparative proteomic profiling was performed using mass spectrometry-based label-free shotgun proteomics. Differentially abundant proteins were analyzed using Ingenuity Pathway Analysis (IPA) to identify overrepresented canonical pathways and predict upstream regulators. Selected differentially abundant proteins and predicted upstream regulators were validated by RT-qPCR and/or Western blotting.

Results:

Comparative proteomic profiling showed that CSC-enriched spheroids shared convergent pathway-level alterations despite cell line-specific differences in individual protein abundance. IPA pathway and functional analyses predicted activation of metabolic reprogramming, invasion, and hypoxia adaptation, along with predicted suppression of apoptotic pathways. Notably, HMGCS1, a key mevalonate-pathway enzyme, was strongly upregulated at both mRNA and protein levels in CSCs from both cell lines. MYC, MLXIPL, EGF/EGFR, VEGFA, and HIF-related signaling were among the top predicted upstream regulators shaping these alterations. In addition, altered expression of proteins involved in immunosuppressive signaling was observed in CSC-enriched spheroids, with TGF-β signaling emerging as a prominently activated upstream regulator, potentially contributing to CSC-associated epithelial-mesenchymal transition and immunomodulation.

Conclusion:

In summary, this study provides a better understanding of key dysregulated pathways and proteins in CRC CSCs, highlighting potential biomarkers and regulatory programs with relevance to stemness, immune modulation, and therapeutic resistance.

1 Introduction

Colorectal cancer (CRC) is the third most commonly diagnosed malignancy and the second leading cause of cancer-related mortality worldwide (Global Cancer Observatory GCO, 2025). Despite major advances in early screening and the introduction of targeted therapies that have improved overall survival (OS), a substantial proportion of patients exhibit either primary resistance or develop acquired resistance during treatment, ultimately progressing to advanced, metastatic disease (SEER, 2025; Tang et al., 2023). In contrast to localized CRC, metastatic CRC is associated with a poor prognosis, with an approximate 5-year survival rate of about 16.2% (SEER, 2025). Accordingly, identifying molecular drivers of tumor progression and therapeutic resistance remains essential to improving long-term clinical outcomes.

Tumors are composed of heterogeneous cell populations with distinct phenotypic and functional properties. Within this diversity, cancer stem cells (CSCs), also referred to as cancer stem-like cells (CSLCs) or cancer-initiating cells (CICs), represent a minor subpopulation endowed with “stemness” features, including self-renewal capacity, the ability to enter quiescence, and pluripotency (Maccalli et al., 2018; Lee H. et al., 2025). CSCs are increasingly recognized as key contributors to tumor initiation, metastasis, and resistance to conventional therapies (Jiang et al., 2025; Chu et al., 2024). Moreover, emerging evidence indicates that CSCs may facilitate immune evasion and influence responses to immunotherapies (Hussein et al., 2025; Ravindran et al., 2019; Gupta et al., 2023). Together, these findings underscore the importance of elucidating CSC-associated molecular programs as a prerequisite for developing more effective and durable therapeutic strategies and biomarkers.

Despite their recognized biological and clinical relevance, CSCs remain challenging to define and isolate due to their low abundance in tumor tissue, pronounced heterogeneity, and phenotypic plasticity. Although multiple surface markers (e.g., CD44, CD133, EpCAM, CD166, and CD44v6), as well as intracellular and functional markers (e.g., ALDH and LGR5), have been used to identify and enrich CRC stem-like populations (Pashirzad et al., 2022; Maccalli and De Maria, 2015), these markers exhibit limited specificity as they are also expressed by normal cells and non-CSCs, often at lower levels (Chu et al., 2024). Furthermore, the expression of these markers may vary depending on proliferative and differentiation states, reflecting CSC plasticity and substantial intra- and inter-tumoral heterogeneity (Tout et al., 2025). These limitations complicate the isolation of well-defined CSC populations and hinder comprehensive molecular characterization using marker-based strategies alone.

To address these challenges, cancer cell lines have been shown to provide an alternative, reproducible source for CSC research. Three-dimensional (3D) spheroid culture under non-adherent, serum-free conditions is widely used to enrich for cancer stem or stem-like populations by exploiting key CSC-associated properties, such as resistance to anoikis, stress tolerance, and self-renewal capacity (Weiswald et al., 2015). Large-scale proteomic and proteogenomic studies have identified dysregulated signaling pathways, biomarkers and drug targets (Vasaikar et al., 2019). In our recent work, we successfully enriched CSCs from four CRC cell lines, with HCT-116- and SW620-derived spheroids exhibiting the highest self-renewal potential, as evidenced by sustained spheroid formation across multiple passages (Hussein et al., 2026). In order to deepen the understanding of the molecular features of CSCs, a proteomic approach was employed.

Recent advances in mass spectrometry–based proteomics have significantly expanded our understanding of CRC biology (Zhang B. et al., 2014). Large-scale proteomic and proteogenomic studies have identified dysregulated signaling pathways, biomarkers and drug targets (Vasaikar et al., 2019; Wiśniewski et al., 2015). Importantly, proteomic profiling provides complementary insights beyond transcriptomic analyses by capturing post-transcriptional regulation and functional pathway activity that cannot be reliably inferred from gene expression alone. Nevertheless, proteome-level alterations distinguishing colorectal differentiated tumor cells from their stem-like counterparts remain poorly characterized. In this study, we performed a comparative proteomic analysis of colorectal CSC-enriched spheroids vs. parental bulk cancer cells to gain deeper insights into differentially regulated proteins and pathways, with the aim of identifying CSC-associated molecular signatures that may guide biomarker discovery and therapeutic targeting.

2 Materials and methods2.1 Cell culture of cancer cells

Two colorectal cancer cell lines, HCT-116, that has epithelial morphology, and SW620, that has fibroblast-like morphology, were obtained from the American Type Culture Collection (ATCC). Cells were cultured in high-glucose Dulbecco’s Modified Eagle’s Medium (DMEM) containing GlutaMAX (Gibco; cat. 31,966–047), supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% Antibiotic-Antimycotic (Gibco; cat. 15,240–062) and maintained at 37 °C in a humidified incubator with 5% CO2.

2.2 CSC-enriched spheroid culture

To enrich for CSC populations, bulk HCT-116 and SW620 cells were harvested and seeded at a density of 50,000 cells/mL in ultra-low-attachment T-75 flasks (Nunclon Sphera, Thermo Fisher Scientific). Cells were cultured in serum-free DMEM/F12 (StemFlex medium; Gibco; cat. A3349401) supplemented with 1× StemFlex supplement and 1× Antibiotic-Antimycotic. Spheroids were maintained through sequential passaging to further enrich for CSCs. Briefly, spheroids were collected by gravitational sedimentation for 10 min, enzymatically dissociated into single cells using 1× TrypLE Express (Gibco, 12,605–028), and mechanically by gentle pipetting. The resulting single-cell suspension was then reseeded at the same density under the same culture conditions (Rybak et al., 2011; Mukherjee et al., 2021).

2.3 Protein extraction for the whole proteome analysis

Protein extraction was performed as previously described (Therachiyil et al., 2024; Dhulkifle et al., 2024). Briefly, cells were washed with ice-cold PBS, and total proteins were extracted using RIPA lysis buffer containing 1× Halt™ Protease Inhibitor Cocktail (Thermo Scientific; cat. 78429). Protein concentrations were determined using the Pierce™ Rapid Gold BCA Protein Assay Kit (Thermo Scientific; cat. A53225) following the manufacturer’s instructions.

2.4 Sample preparation for mass spectrometry analysis

Proteins from three independent replicates for each cell line were reduced with 5 mM dithiothreitol for 40 min at RT and alkylated with 10 mM iodoacetamide for 30 min at RT in the dark. Samples were then digested overnight at 37 °C in ABC buffer using Mass Spec Grade Trypsin/Lys-C Mix (Promega; Madison, WI, United States of America; 12.5 ng/μL) at an enzyme-to-protein ratio of 1:25. Digestion was quenched by acidifying the reaction mixture with 1% formic acid (FA). Peptides were subsequently desalted and concentrated by solid-phase extraction using Bond Elut C18 cartridges (Agilent Technologies) and eluted with stepwise gradients of acetonitrile (ACN) in 0.1% FA. The eluate was dried in a SpeedVac concentrator to obtain a peptide pellet, reconstituted in 2% ACN containing 0.1% FA, and subsequently analyzed by mass spectrometry (Therachiyil et al., 2024; Rappsilber et al., 2007).

2.5 Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis

Trapped ion mobility mass spectrometry (TIMS) analyses were performed on a timsTOF Pro mass spectrometer (Bruker Daltonics, Germany) coupled to a nanoElute nano-liquid chromatography system (Bruker Daltonics). For each run, 10 μL of peptide solution corresponding to 2 μg of total peptides were injected into the LC-MS system. Desalted peptides were separated by reversed-phase chromatography on a C18 column with an integrated CaptiveSpray emitter (25 cm × 75 μm, 1.6 µm; IonOpticks, Australia) at a flow rate of 400 nL/min using mobile phase A (ultrapure water containing 0.1% FA) and mobile phase B (ACN containing 0.1% FA). The gradient was run from 2% to 80% B over 110 min, with a total run time of 120 min. MS data were acquired in PASEF data-dependent acquisition (PASEF-DDA) mode, selecting the top 10 most abundant precursor ions from each full MS survey scan (m/z 400–1800) for fragmentation and MS/MS analysis. Precursors with a charged state of +1 were rejected, and the dynamic exclusion duration was set to 25 s.

2.6 MS and MS/MS data processing and analysis

Raw MS/MS data were processed in MaxQuant (v2.1.4.0) using the integrated Andromeda search engine, following the standard workflow (Tyanova et al., 2016a; Cox et al., 2011). Protein identification was performed by searching against the UniProtKB/Swiss-Prot human reference proteome (UP000005640_9606) database. Methionine oxidation was set as a variable modification, whereas carbamidomethylation of cysteine was defined as a fixed modification. Mass tolerances were set according to instrument-specific default settings in MaxQuant (first search 20 ppm; main search 10 ppm; MS/MS tolerance default). A maximum of one missed cleavage was allowed for tryptic digestion. Peptide-spectrum matches and protein identifications were filtered at a 1% false discovery rate based on a reverse-sequence decoy database.

Label-free quantification (LFQ) was performed using the MaxLFQ algorithm in MaxQuant, as previously described (Cox et al., 2014) with an LFQ minimum ratio count of 2. Retention-time alignment and the match-between-runs feature were enabled (match time window 0.7 min; alignment time window 20 min, default). Protein abundance was inferred from LFQ intensity values.

Downstream data processing was performed in Perseus (v2.0.7.0) (Therachiyil et al., 2024; Tyanova et al., 2016b). Proteins annotated as potential contaminants, reverse-sequence matches, and those identified only by site were excluded. LFQ intensities from the MaxQuant were Log2-transformed, and proteins consistently quantified in at least two biological replicates of at least one experimental condition were retained for subsequent comparative analysis. Missing values were imputed in Perseus by normal distribution according to default settings (width = 0.3; downshift = 1.8). Differential protein abundance was assessed using a two-tailed Student’s t-test with a permutation-based false discovery rate control set to 5% and absolute Log2 fold-change (Log2FC|) ≥ 1.5. Principal component analysis (PCA) was performed to evaluate the reproducibility and clustering of biological replicates. The raw mass spectrometry proteomics data have been deposited in the PRIDE repository via the ProteomeXchange Consortium under the dataset identifier PXD075150. A complete list of differentially abundant proteins is provided in the Supplementary Material.

2.7 Ingenuity Pathway Analysis

Ingenuity Pathway Analysis (IPA; QIAGEN) was used for exploratory functional enrichment and pathway analysis. Lists of proteins, together with their Log2FC values and p-values, were uploaded into the IPA application and analyzed. Proteins with a p-value <0.05 for SW620 and adjusted p-value <0.05 for HCT-116 together with an effect-size cutoff of |Log2FC| ≥ 1 were included in the IPA analysis. This threshold was used to retain a sufficient number of proteins for pathway enrichment analysis. IPA was used to identify over-represented canonical pathways, upstream regulators, and disease and biological function annotations based on the Ingenuity Knowledge Base, which is curated from published experimental literature. For canonical pathway enrichment, IPA calculates statistical significance using a right-tailed Fisher’s exact test, and pathway significance is presented as -Log10 (p-value). Predicted pathway activation or inhibition was assessed using the IPA activation z-score algorithm, which incorporates the direction of protein abundance changes.

2.8 Quantitative real‐time PCR (RT-qPCR)

Total RNA was extracted from cells using PureLink™ RNA Mini Kit (Invitrogen, cat. 12183025) according to the manufacturer’s instructions. RNA purity and concentration were determined by NanoDrop™ 8000 Spectrophotometer. Reverse transcription was performed using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, cat. 4374966). qPCR was performed on a QuantStudio™ 12K Flex Real-Time PCR System using PowerUp™ SYBR™ Green Master Mix (Applied Biosystems, cat. A25742). Relative gene expression was calculated using the comparative ΔΔCt method, with GAPDH as the reference gene. Primer sequences are provided in Supplementary Table S1.

2.9 Western blot analysis

Proteins (30 μg) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by immunoblotting as previously described (Therachiyil et al., 2022). Bands were visualized using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (Thermo Scientific, cat. 34580) and imaged on a ChemiDoc Imaging System (Bio-Rad). Densitometry analysis was performed using ImageJ (https://imagej.net/ij/). The protein bands in the figures are a composite of different blots and are representative blots for the indicated proteins and the loading control. Lists of the antibodies used are provided in Supplementary Table S2.

2.10 Statistical analysis

For proteomics, differential protein abundance was assessed using a two-tailed Student’s t-test with a permutation-based FDR of 5% to control for multiple testing in Perseus. For RT-qPCR and Western blot assays, significance between two groups was evaluated using an unpaired two-tailed Student’s t-test by GraphPad Prism (version 10) with p < 0.05 considered statistically significant. Data are presented as mean ± SEM from at least three independent experiments.

3 Results3.1 LC-MS/MS analysis identified differentially abundant proteins in colorectal CSCs

To compare the global protein abundance profiles between CSC-enriched spheroids and their corresponding adherent parental colorectal cancer cells (HCT-116 and SW620), label-free quantitative proteomic analysis was performed.

Using a significance threshold of p < 0.05 together with an effect-size cutoff of |Log2FC| ≥ 1.5, SW620 CSCs exhibited 245 differentially abundant proteins (240 upregulated and 5 downregulated), whereas HCT-116 CSCs exhibited 201 differentially abundant proteins (92 upregulated and 109 downregulated) relative to their parental cancer cells (Figures 1A,B). Significant differentially abundant proteins between CSC-enriched spheroids and bulk cancer cells are visualized in the heatmaps (Figures 1C,D). Additionally, proteins showing the largest magnitude of change (|Log2FC|) are listed in Tables 1,2.

Differential proteomic profiling of colorectal CSCs. (A,B) Volcano plots of differentially abundant proteins between spheroid-enriched CSCs and bulk cancer cells from SW620 (A) and HCT-116 (B) cell lines. Log2FC (x-axis) is plotted against -Log10 (p-value) (y-axis). Each point represents an individual protein and is color-coded in green or red when up- or downregulated in CSCs, respectively, and in gray when non-significant. Differentially abundant proteins were defined using a statistical threshold of p < 0.05 together with an effect-size cutoff of |Log2FC| ≥ 1.5. (C,D) Heatmaps showing differentially abundant proteins in CSCs derived from SW620 (C) and HCT-116 (D). (E,F) Venn diagrams showing the overlap between CSC-upregulated proteins and proteins listed in the ONGene oncogene database and the BCSCdb CSC biomarker database. Shared CSC-associated proteins for each comparison are listed.

Protein IDGene nameProtein nameLog2 FCAdjusted p-valuep-valueP53999SUB1Activated RNA polymerase II transcriptional coactivator p156.530.0177<0.0001P12004PCNAProliferating cell nuclear antigen5.590.0177<0.0001P14314PRKCSHGlucosidase 2 subunit beta5.530.0177<0.0001P62328TMSB4XThymosin beta-4; hemo regulatory peptide AcSDKP5.480.01870.0001P46776RPL27ALarge ribosomal subunit protein uL155.420.04310.0006P61586RHOATransforming protein RhoA5.250.0177<0.0001Q58FF8HSP90AB2PPutative heat shock protein HSP 90-beta 25.200.01870.0001P62873GNB1Guanine nucleotide-binding protein G(I)/G(S)/G(T) subunit beta-15.040.01870.0001Q13185CBX3Chromo box protein homolog 35.010.01870.0001P16949STMN1Stathmin4.690.02250.0002Q13435SF3B2Splicing factor 3B subunit 24.670.02040.0001P63313TMSB10Thymosin beta-104.360.02250.0002Q14764MVPMajor vault protein4.350.03040.0003Q01581HMGCS1Hydroxy methyl glutaryl-CoA synthase, cytoplasmic4.330.03040.0003P00387CYB5R3NADH-cytochrome b5 reductase 34.290.02940.0002P06396GSNGelsolin4.290.03270.0004Q16543CDC37Hsp90 co-chaperone Cdc37; Hsp90 co-chaperone Cdc37, N-terminally processed4.260.03110.0003Q5JRX3PITRM1Presequence protease, mitochondrial4.240.04310.0006P26373RPL13Large ribosomal subunit protein eL134.160.04310.0007Q9Y2Z0SUGT1Protein SGT1 homolog4.130.03110.0004

Differentially expressed proteins with the largest magnitude of change in SW620 CSC-enriched spheroids versus bulk cancer cells.

Top 20 differentially expressed proteins, demonstrating the highest intensity of change in SW620 CSC-enriched spheroids compared with bulk cancer cells. The proteomic dataset was filtered to include proteins meeting p < 0.01 and adjusted p < 0.05 and absolute Log2FC > 1.5. The 20 proteins with the greatest absolute magnitude of change were selected and ranked according to their Log2FC., The table reports UniProt protein ID, gene symbol, protein name, Log2FC, adjusted p-value, and p-value.

Protein IDGene nameProtein nameLog2 FCAdjusted p-valuep-valueP08670VIMVimentin7.860.00270.0002P07203GPX1Glutathione peroxidase 15.110.0010<0.0001P08134RHOCRho-related GTP-binding protein RhoC4.170.00170.0001Q9H1E3NUCKS1Nuclear ubiquitous casein and cyclin-dependent kinase substrate 13.980.0001<0.0001Q8WVX9FAR1Fatty acyl-CoA reductase 13.770.0005<0.0001Q9Y394DHRS7Dehydrogenase/reductase SDR family member 73.500.00410.0004Q4V328GRIPAP1GRIP1-associated protein 1; GRASP-1 C-terminal chain3.490.00920.0016P00374DHFRDihydrofolate reductase−3.450.0003<0.0001Q99988GDF15Growth/differentiation factor 15−3.450.0014<0.0001Q9Y276BCS1LMitochondrial chaperone BCS1−3.550.01430.0033Q13257MAD2L1Mitotic spindle assembly checkpoint protein MAD2A−3.580.00190.0001Q6P1X6C8orf82UPF0598 protein C8orf82−3.770.00230.0001Q92572AP3S1AP-3 complex subunit sigma-1−3.880.0007<0.0001Q9NPH2ISYNA1Inositol-3-phosphate synthase 1−4.030.0012<0.0001Q96TA2YME1L1ATP-dependent zinc metalloprotease YME1L1−4.520.01680.0041Q9UEW8STK39STE20/SPS1-related proline-alanine-rich protein kinase−4.530.0010<0.0001Q9H4G0EPB41L1Band 4.1-like protein 1−4.610.0015<0.0001Q9NY61AATFProtein AATF−4.750.0015<0.0001P47895ALDH1A3Retinaldehyde dehydrogenase 3−4.780.0001<0.0001O00483NDUFA4Cytochrome c oxidase subunit NDUFA4−4.940.0013<0.0001

Differentially expressed proteins with the largest magnitude of change in HCT-116 CSC-enriched spheroids versus bulk cancer cells.

Top 20 differentially abundant proteins showing the largest magnitude of change in HCT-116 CSC-enriched spheroids compared with bulk cancer cells. The proteomic dataset was filtered to include proteins meeting p < 0.01 and adjusted p < 0.05 and an effect-size cutoff of |Log2FC| ≥ 1.5. The 20 proteins with the greatest absolute magnitude of change were selected and ranked according to their Log2(FC). The table reports UniProt protein ID, gene symbol, protein name, Log2(FC), adjusted p-value, and p-value.

Among key upregulated proteins (adjusted p < 0.05) in SW620 CSC-enriched spheroids were SUB1, PCNA, PRKCSH, and TMSB4X, whereas VIM, GPX1, RHOC, and NUCKS1 were among the upregulated proteins in HCT-116 CSC-enriched spheroids. Notably, HMGCS1, ACAT2, CTSD, LAMP1 and CORO1B were among the proteins commonly upregulated in both cell lines (p < 0.05). Given that CSC phenotypes are frequently associated with oncogenic signaling, CSC-upregulated proteins were matched against entries in the ONGene oncogene database and with CSC-related biomarkers curated in BCSCdb. Figures 1E,F summarize these overlaps, highlighting a subset of CSC-upregulated proteins with prior evidence of oncogenic and CSC relevance (e.g., VIM, NRAS, and S100A4).

3.2 Functional annotation and pathway enrichment of CSC-associated proteins

IPA was employed to investigate the biological relevance of proteins differentially abundant between CSC-enriched spheroids and bulk cancer cells, with a focus on canonical pathways and associated disease and functions.

As shown in Figure 2A, the top enriched canonical pathways in SW620 CSCs that were predicted to be activated were primarily related to translational regulation and stress-adaptive responses (e.g., Eukaryotic Initiation Factor 2 [eIF2] signaling, ribosomal quality control, eukaryotic translation elongation), metabolic reprogramming (e.g., cholesterol biosynthesis) together with hypoxia-associated signaling (e.g., HIF-1α signaling) and oncogenic signaling (e.g., VEGF signaling and Erb-B2 receptor tyrosine kinase 2 [ERBB2/HER2] signaling). In HCT-116 CSCs, canonical pathways predicted to be activated were predominantly linked to metabolic reprogramming (e.g., super pathway of cholesterol biosynthesis, cholesterol biosynthesis I, SREBF-mediated gene expression, and the mevalonate pathway) (Figure 2B). Consistent with these pathway-level changes, IPA disease and function analysis indicated that CSC-associated proteins in both cell lines were strongly linked to cancer-relevant functional programs (Figures 2C,D). Overall, the top activated functions were related to cell survival/viability, proliferation, cellular movement, migration, invasion and carcinoma/tumor development, together with metabolic processes. Conversely, the top inhibited functions were predominantly associated with apoptosis, cell death, and cellular sensitivity, consistent with a CSC phenotype favoring tumor persistence and progression. An illustrative summary generated using IPA is provided in Supplementary Figures S1-S3

Ingenuity Pathway Analysis (IPA)-based functional and pathway enrichment analysis of colorectal cancer stem cell (CSC)-associated proteomic profiles. Spheroid-enriched CSCs derived from SW620 and HCT-116 cells were compared with the corresponding bulk (parental) cancer cells from the same cell line. (A,B) Top enriched canonical pathways associated with SW620 CSCs (A) and HCT-116 CSCs (B). (C,D) Top enriched diseases and functions associated with SW620 CSCs (C) and HCT-116 CSCs (D). Bars represent pathway/functional enrichment significance shown as -Log10(P-value). Bar color indicates IPA activation z-score: orange = positive z-score (predicted activation) and blue = negative z-score (predicted inhibition). The black line indicates the number of proteins mapped to each pathway/function.

3.3 Functional categorization and mRNA-level validation of CSC-associated proteins

The highly expressed proteins that were significantly altered in CSC-enriched spheroids were further categorized according to their predominant biological functions, including apoptosis inhibition, invasion and metastasis, metabolism, angiogenesis, and immune modulation, as illustrated in Figures 3A–E; Supplementary Tables S3-S5. To further validate the differential expression of CSC-associated markers, the mRNA expression of selected functionally relevant proteins was quantified by RT-qPCR. Figures 3F–P, representing key functional categories, including apoptosis inhibition (PRKCSH, HMGCS1, PCNA), invasion and metastasis (CORO1B, HMGCS1, NRAS), angiogenesis (NRAS, RHOA), metabolic reprogramming (HMGCS1, PCNA, CTSD) and immune modulation (ITGB4, PKM, CDC42, RHOA, CTSD), were upregulated in CSCs, that were consistent with the proteomic results. Remarkably, HMGCS1 was significantly upregulated at the mRNA level in the CSCs of both SW620 and HCT-116 cells (34.9-fold and 14.6-fold, respectively), consistent with corresponding increases at the protein level (20.2-fold and 11.1-fold, respectively) relative to bulk cancer cells.

Functional categorization and validation of selected differentially abundant proteins in colorectal CSCs. (A–E) Plots showing Log2 LFQ intensity (mean ± SEM) of selected proteins in SW620-derived CSC-enriched spheroids (red) vs. bulk SW620 cancer cells (blue), grouped by their reported involvement in (A) Apoptosis inhibition, (B) invasion and metastasis, (C) metabolism, (D) angiogenesis and (E) immune modulation. (F–P) RT-qPCR validation of selected targets in SW620 and HCT-116 CSC-enriched spheroids vs. their corresponding bulk cancer cells. mRNA levels were normalized to the expression of the housekeeping gene GAPDH. Data are presented as mean ± SEM from at least three independent experiments relative to bulk cancer cells of the same cell line. *P < 0.05, **P < 0.01, ***P < 0.001.

3.4 Upstream regulator analysis of CSC-associated proteomic signatures

Upstream regulator analysis was performed using IPA to infer potential molecular drivers underlying the CSC-associated proteomic changes based on the coordinated expression patterns of downstream target proteins. In SW620 CSC-enriched spheroids, IPA predicted activation of multiple regulators implicated in stemness maintenance (MYC, CD44), metabolic reprogramming (MLXIPL, XBP1), hypoxia adaptation (HIF1 complex), and oncogenic signaling (EGF/EGFR-related signaling, VEGFA, KRAS). In contrast, a subset of regulators linked to tumor-suppressive or differentiation-related programs, such as TXNIP, OVOL2, and miR-122-5p, were predicted to be inhibited (Figure 4A). Similarly, in HCT-116 CSC-enriched spheroids, IPA predicted activation of upstream regulators linked to growth factor signaling (CSF1), hypoxia adaptation (HIF-related signaling), TGF-β signaling (SMAD3/TGF-β-associated regulators), metabolic regulation (SREBF2) and oncogenic pathways (KRAS) (Figure 4B). To experimentally validate these in silico predictions, RT-qPCR analysis was performed. CSC-enriched spheroids from both cell lines exhibited significantly increased mRNA expression of selected predicted upstream regulators, including MYC, MLXIPL, EGF, VEGFA, HIF-1α, and HIF-2α, compared with bulk cancer cells (Figures 4C–H). Together, these findings support the involvement of these regulators in shaping CSC-associated molecular programs.

Predicted upstream regulator analysis of differentially expressed proteins in colorectal CSCs. (A,B) Ingenuity Pathway Analysis (IPA) Upstream Regulator Analysis based on differentially abundant proteins in