The Smad4-intact SW480 [24] and Smad4-mutated HT29 [25] colon cancer cell lines were acquired from the American Type Tissue Collection (ATCC; VA, United States). Biologically active Act-A (#338-AC-050/CF), Act-B (#659-AB-005/CF), Act-AB (#1066-AB-005/CF), and FS (#669-FO-025/CF) were obtained from R&D Systems (MN, United States). SW480 and HT29 cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM; #31966047; Thermo Fisher) and Roswell Park Memorial Institute 1640 media (RPMI; #61870044; Thermo Fisher), respectively; both media were supplemented with 10% Foetal Bovine serum (FBS; #A3160802; Thermo Fisher) and 1% Antibiotic-Antimycotic solution (#15240062; Thermo Fisher), and the cells grown in a humified incubator at 37°C and 5% CO₂.

For cell cycle analysis, 2 × 10⁵ SW480 and 3 × 10⁵ HT29 cells were seeded in 6-well plates for 24 h, whereas for immunofluorescence, cells were seeded in 8-well chamber slides (#229168; Celltreat™ Scientific Products; MA, United States) at 8,000 (SW480) and 16,000 (HT29) cells/well, also for 24 h. Concentrations (IC50) of Act-A (100 ng/ml), Act-AB (200 ng/ml) and FS (100 ng/ml) were determined using the 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) cytotoxicity assay (data not shown). Act-B had little effect at the highest tested dose (500 ng/ml), and was therefore used at 200 ng/ml, matching the highest concentration of the activins, described above.

All variables were assessed for normality by the Kolmogorov and Smirnov’s test and homogeneity by Levene test using SPSS version 25. Ordinal and discontinuous data are presented as numbers and percentages, and cross-tabulation followed by Chi-square (χ²) test were used for frequency analysis. Independent Student’s t or Mann-Whitney U tests were used to compare between two groups based on normality. Continuous data are shown either as mean ± standard deviation (SD) or median with interquartile range (IQR; 25th–75th percentiles), depending on data normality. Correlations were determined by Pearson’s or Spearman’s tests according to data normality. p value <0.05 was considered significant.

The FFPE (n = 90) and fresh (n = 40) tissue cohorts comprised 69 males (53.1%) and 61 females (46.9%), with average age comparable between both genders (63.3 ± 11.3 and 61.2 ± 12.7 years, respectively). The tumours were staged as T1 in six (4.6%), T2 in 21 (16.2%), T3 in 64 (49.5%) and T4 in 39 (30%) patients. Metastasis in regional lymph nodes was detected in 57 (43.9%) cases, whereas 18 patients (13.8%) had distant metastasis (stage M1a) in liver. Adenocarcinoma was the most common histology type (n = 108; 83.1%) and the remainders were mucinous carcinomas. Moreover, 15 (11.5%), 91 (70%) and 24 (18.5%) cases were diagnosed with poorly, moderately, and well-differentiated tumours, respectively. As per the TNM staging system, 60 (46.2%) patients were clinically classified as early (stages I/II) and the remainder (53.8%) as advanced (stages III/IV) cancers. Furthermore, the clinico-pathological features of tumours were similar between the fresh and FFPE sets (Table 1).

Right-sided tumours were less prevalent (n = 39; 30%) in the FFPE and fresh samples and had markedly larger sizes alongside higher rates of advanced T stages (T3 and T4), distant metastasis, mucinous carcinoma, and poorly differentiated histology than the LSC group (Table 2). However, there was no significant difference between the RSC and LSC groups in relation to age, gender, N stages and the rates of early (I/II) and advanced (III/IV) cancer stages (Table 2).

The Smad4 mRNA was markedly decreased in the early-stage (3.2-fold) and advanced-stage (11.1-fold) neoplastic tissues relative to their corresponding normal specimens (p < 0.001 for both), and the levels were lowest in the late-stage malignant tissues.

Smad4 protein was exclusively localised in the epithelial nuclei and the expression in normal tissues was comparable between the FFPE and fresh samples, anatomical sites, and cancer stages (Supplementary Figure S1). Nevertheless, Smad4 immunostain was markedly (1.4-fold and 1.9-fold, respectively) weaker in the tumorous tissues from early and advanced cancer stages compared with their corresponding normal specimens (Figure 1; p < 0.001). Smad4 protein in the malignant tissues from advanced stages was also significantly lower than the early cancer stages (p < 0.05). However, the expression of Smad4 was similar between the RSC and LSC during either the early or advanced stages of CRC (Figure 1).

The overall median of the Smad4 IHC scores in the FFPE and fresh malignant sites was 130.4 (96.2–157.64), and 26 patients (20%) showed IHC scores ≤ the 20th percentile cut-off value (87.44) and were classified as the L-S4 group. Moreover, the frequencies of low Smad4 protein were not significantly different between the FFPE and fresh malignant specimens, and the age was comparable between the N-S4 and L-S4 groups (Table 3). However, the LS-4 group was associated with markedly larger tumour sizes, advanced T and N stages, distant metastasis, mucinous carcinoma, poorly differentiated histology, and RSC than the N-S4 group. Moreover, the late-stage tumours (III/IV) were significantly more prevalent in the L-S4 group (Table 3).

The mRNA levels of the targeted genes in normal colonic tissues, from the FFPE and fresh cohorts of specimens, were similar between the RSC and LSC as well as between the early and late malignancy stages within each set of specimens (Figure 2). While the βA-subunit mRNA was significantly (8.5-fold) higher, the βB-subunit (2.3-fold) and FS (2.1-fold) decreased markedly in the neoplastic tissues of the early stages compared with their corresponding normal samples (p < 0.001 for all genes). Although the gene expression of β-subunits in the early-stage malignant tissues was equal between the RSC and LSC groups, the latter showed marked (21-fold) increases in the βA-subunit alongside (6.2-fold) declines in the βB-subunit genes in the late stages of CRC (Figure 2; p < 0.0001). Interestingly, the advanced-stage cancer specimens showed significant and paradoxical (9.6-fold) increases in FS mRNA, relative to their corresponding normal tissues, and were also substantially higher than the early cancer tissues (p < 0.0001). Moreover, the RSC samples from late stages showed the greatest rise in the βA-subunit and FS gene expression alongside the lowest βB-subunit mRNA levels compared with all other tissues (Figure 2).

The β-subunits and FS mRNAs were also equal between the normal specimens corresponding to the N-S4 and L-S4 cancerous tissues. However, the βA-subunit (34.9; IQR: 13.52–39.13) and FS (12.57; IQR: 8.35–14.88) mRNAs were markedly higher in the L-S4 malignant specimens relative to the N-S4 tumours (9.85; IQR: 7.64–14.35 and 0.78; IQR: 0.48–8.85, respectively; p < 0.0001). In contrast, the βB-subunit gene expression was significantly reduced in the malignant tissues with low Smad4 (0.16; IQR: 0.1–0.21) than the cancerous samples with normal Smad4 expression (p < 0.0001).

The antibodies against the β-subunits and FS clearly labelled the cytoplasm of normal luminal and glandular colonic epithelia (Supplementary Figure S1). Moreover, there were no significant differences in the targeted proteins between the normal specimens according to sample sources, anatomical sites, and cancer stages (Figure 3).

In cancerous tissues, the βA-subunit was mainly localised in the cytoplasm of carcinomas with scarce stromal staining, and the early and late cancer stages showed markedly (1.4-fold and 2.3-fold, respectively) higher expression relative to their corresponding normal tissues (p < 0.001). Furthermore, the βA-subunit stain intensity in the malignant tissues was stronger in late than early stages of cancer (Figure 3; p < 0.01). In contrast, the βB-subunit (1.7-fold and 2.6-fold, respectively) diminished markedly in the tumorous tissues from early and advanced stages of CRC compared with their corresponding normal samples (p < 0.001). Moreover, the βB-subunit was almost undetectable in the late-stage cancer tissues and was substantially weaker than the early stages of malignancy (Figure 3; p < 0.01). Although FS (1.3-fold) declined significantly in the early stages of cancer, and was mostly located in the stroma surrounding tumours, the protein (2.5-fold) escalated markedly and showed strong cytoplasmic staining in advanced cancers than their corresponding normal tissues (Figure 3; p < 0.001). Follistatin IHC scores were also higher in the advanced than early stages of cancer (p < 0.0001). According to tumour sidedness, there were no significant differences in the expression of activin β-subunits and FS during the early stages of malignancy between the RSC and LSC groups. However, the RSC group showed (1.4-fold and 1.3-fold) higher βA-subunit and FS proteins alongside (1.7-fold) lower βB-subunit in the late cancer stages (Figure 3).

Moreover, the βB-subunit protein was equal in the normal specimens from the N-S4 (188.2; IQR: 166.8–210.8) and L-S4 (176.8; IQR: 159.8–187.6) groups. However, the βA-subunit (209.7; IQR: 185.9–237.1 vs. 219.8; IQR: 202.7–253, respectively; p = 0.03) and FS (248.4; IQR: 229.8–270.7 vs. 233.1; IQR: 226.4–254.2, respectively; p = 0.007) were higher in the normal tissues from the L-S4 than the N-S4 group. The L-S4 tumours also had higher βA-subunit (621.2; IQR: 433.9–701.6 vs. 340.2; IQR: 287.6–442.4) and FS (688.4; IQR: 479.9–773.3 vs. 237.4; IQR: 287.6–631) alongside lower βB-subunit (61.4; IQR: 42.7–82.3 vs. 99.03; IQR: 83.4–118.1) proteins relative to cancers with normal Smad4 (p < 0.0001 for all).

The Act-A, Act-AB, Act-B, and FS concentrations were similar between the right and left-sided normal specimens as well as between normal tissues obtained from the early and late stages of CRC (Figure 4). The intratumoral Act-A levels were markedly higher than their corresponding normal tissues, showed steady increases with cancer progression, and the highest levels were detected in the advanced malignancy stages from the RSC group (Figure 4). In contrast, both Act-AB and Act-B decreased markedly in the malignant specimens than their corresponding normal tissues. While Act-AB levels were significantly lower in the tumorous samples from late compared with early cancer stages, the lowest amounts were seen in the late-stage RSC specimens (Figure 4). Intratumoral Act-B, however, was equal between the early and late stages of cancer, as well as, between the RSC and LSC samples. Contrariwise, FS was markedly lower in the early-stage cancer tissues compared with their corresponding normal samples, whereas the levels surged drastically in late-stage tumours relative to the normal and early-stage cancer tissues, and the highest amounts were detected in the right-sided malignancies (Figure 4).

The normal tissues from the N-S4 and L-S4 had equal levels of Act-A (299.3; IQR: 260.7–333.1 vs. 291; IQR: 255.6–321 pg/ml), Act-AB (268.6; IQR: 187.5–304.5 vs. 268.2; IQR: 218.1–287.1 pg/ml), Act-B (304.6; IQR: 274–363.1 vs. 328.4; IQR: 277–363.1 pg/ml), and FS (4.2; IQR: 3.8–5.6 vs. 4; IQR: 3.6–4.8 ng/ml). In contrast, the L-S4 malignant tissues had markedly higher Act-A (1332.5; IQR: 791–1638.8 vs. 649.2; IQR: 535.9–794 pg/ml; p = 0.001) and FS (28.7; IQR: 24.6–32.1 vs. 14.4; IQR: 2.2–24.4 ng/ml, p < 0.01) than the N-S4 cancerous specimens. By contrast, the neoplastic tissue Act-B amounts (119.8; IQR: 75.1–197 vs. 201.4; IQR: 144.5–233 pg/ml, respectively; p < 0.01) and Act-AB (62.1; IQR: 53.1–96.7 vs. 123.8; IQR: 75.9–153.7 pg/ml; p < 0.001) were substantially lower in the L-S4 group.

Act-A and FS intratumoral levels correlated together directly (r = 0.722; p < 0.0001), whereas they showed inverse associations with Act-AB levels (r = −0.728 and r = −0.641, respectively; p < 0.0001). However, Act-B levels in cancerous tissues did not correlate with Act-A (r = −0.249), Act-AB (r = 0.087), and FS (r = −0.027). Moreover, Act-A and FS in malignant tissues linked positively, whereas Act-AB associated negatively, with tumour size, sidedness, histology, tumour grade, numbers of positive lymph nodes, TNM stages, and Smad4 status (Table 4).

Smad4 protein was detected in the untreated SW480 cells and showed perinuclear localisation, whereas it was negligible in the HT29 cells (Supplementary Figure S1). Act-A and Act-AB markedly increased the numbers of SW480 cells in the sub-G1 phase of cell cycle relative to untreated cells, where Act-A induced the highest numbers (Figure 5A). The percentages of cells in the G0/G1-phase also increased significantly, with declines in the S and G2/M phases, after treatment with the same two proteins relative to untreated controls cells. In contrast, the different cell cycle phases were unaffected by the addition of either Act-B or FS, compared to non-treated SW480 cells (Figure 5A).

On the other hand, Act-A significantly reduced the numbers of HT29 cells in the sub-G1 phase, whilst simultaneously increasing the percentages in the G0/G1 and S-phases, relative to control cells (Figure 5B). The Act-B effects on the HT29 cell numbers in the sub-G1 and G0/G1-phases mimicked those of Act-A but were much weaker. Moreover, Act-B increased the cell counts significantly in the G2/M-phase of cell cycle, not observed with Act-A treatment. Contrariwise, Act-AB showed the highest cell proportions in the Sub-G1 and G0/G1 phases, with concomitant declines in the S- and G2/M-phases, relative to control cells (Figure 5B). Although the HT29 cell numbers in the G0/G1-phase were also significantly higher following FS treatment compared with control, the numbers were significantly lower than Act-AB (Figure 5B). Moreover, FS markedly reduced the numbers of HT29 cells in the S and G2/M, but not sub-G1, phases of cell cycle relative to non-treated cells.

In the SW480 cells, CCND1 (Figure 6) and CCND3 (Figure 7) proteins declined markedly with Act-A and Act-AB treatments compared with control cells. Moreover, Act-A and Act-B increased the p27 and p21 proteins in SW480 cells relative to untreated cells (Figures 6, 7). Whilst Act-A and Act-AB also promoted the p27 and p21 protein expression alongside reduced CCND3 levels in the HT29 cells, the effects of Act-AB were significantly higher than the untreated and Act-A-treated cells. However, the expression of CCND1 in the HT29 cells only weakened significantly with Act-AB treatment (5-fold; Figure 6).

Interestingly, Act-B only diminished CCND1 in the SW480 cells (Figure 6), whilst enhanced the expression of p21 and lowered CCND3 in the HT29 cells (Figure 7); but its effects were weaker than the other activins. On the other hand, the cell cycle regulatory proteins were comparable between the FS-treated and untreated SW480 cells, whereas FS markedly increased p27 and p21 levels and concurrently reduced CCND3 and CCND1 proteins expression in the HT29 cells relative to control cells (Figures 6, 7).

The expression of survivin decreased, whereas Casp-8 increased, significantly in the SW480 cells with Act-A and Act-AB treatments compared with untreated cells (Figure 8). Likewise, Act-A and Act-AB lowered BCL2 and increased cleaved Casp-3 in the SW480 cells (Figure 9). However, Act-A showed no effect on the proliferation and apoptosis proteins in the HT29 cells, whereas Act-AB markedly lessened survivin and BCL2, alongside boosted Casp-8 and cleaved Casp-3 relative to control cells (Figures 8, 9).

Although Act-B in the SW480 cells only showed limited increases in Casp-8 protein relative to control, the expression of BCL2, survivin and Casp-3 were equal between both groups (Figures 8, 9). In the HT29 cells, Act-B only decreased survivin marginally and had no effect on BCL2, Casp-8, and Casp-3 proteins. In contrast, FS reduced survivin in both cell lines, whilst increased Casp-8 only in the HT29 cells. Moreover, the BCL2 and Casp-3 proteins were equal between the FS-treated and control cells in both cell lines (Figures 8, 9).