Streptococcus-derived CpsA–CpsC fusion protein induces apoptosis in cervical cancer cells through in silico and in vitro methods

Abstract

Cervical cancer is a significant cause of cancer-related mortality in women, and the toxicity of conventional anticancer drugs has necessitated the development of novel therapeutic strategies. In the present study, we designed a fusion protein containing CpsA and CpsC proteins encoding capsular polysaccharides in Streptococcus agalactiae and evaluated its potential anticancer activity. Physicochemical properties, 3-dimensional (3D) structure, and molecular dynamics (MD) simulation were analyzed using in silico tools. Then, the recombinant gene was cloned, expressed in a bacterial host, and purified successfully. Finally, the anticancer effects of the CpsA-CpsC fusion protein on cervical cancer cell line (HeLa) were evaluated through MTT, Real-Time PCR, and flow cytometry techniques. Functional assays demonstrated that the CpsA–CpsC fusion protein exerted time- and dose-dependent cytotoxic effects on HeLa cells, with an IC50 value of 68.1 µg/mL at 48 after 48 h of treatment. Flow cytometric analysis using Annexin V/PI staining revealed a significant increase in apoptotic cell populations (17.5%). Furthermore, gene expression analysis revealed marked upregulation of the pro-apoptotic genes, Bax (22.62-fold) and Caspase-3 (22.06-fold), along with a downregulation of the anti-apoptotic gene, Bcl-2 (0.235-fold), supporting apoptosis as a potential mechanism of cell death. These findings suggest that the CpsA–CpsC fusion protein may exert apoptotic effects in cervical cancer cells and provide a basis for further investigation of bacterial-derived components as candidate anticancer agents.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-55464-2.

Introduction

Although cervical cancer ranks as the fourth leading cause of cancer-related deaths among women worldwide, it is the deadliest malignancy in 37 countries, mainly in low- and middle-income regions where screening and vaccination programs are limited¹. Cervical cancer remains a significant global health challenge, prompting the World Health Organization to launch a global attempt to eliminate it as a public health problem by reducing its incidence below 4 per 100,000 women-years². Surgery, chemotherapy, and radiation therapy are conventional approaches against cancer, but in many cases, they fail to destroy the malignant cells. It is because of the resistance of tumors to anticancer drugs that sometimes imposes substantial side effects on the patients. These limitations underscore the urgency of developing novel, efficient, and safe anticancer options³. In this regard, researchers investigated several compounds to find new drugs. Among various bioactive agents, bacterial derivatives have emerged as alternative anticancer candidates due to their tumor-targeting capacity and low toxicity. For instance, Clostridium novyi-NT spores selectively germinate in hypoxic tumors and induce oncolysis while sparing normal tissues⁴. Moreover, modified engineered bacterial toxins, Pseudomonas exotoxin A, fused with targeting ligands, have shown safety and efficacy in clinical trials⁵. These examples illustrate that specific bacteria or their products can serve as a promising therapeutic option with improved selectivity and low toxicity when appropriately engineered. Nonetheless, these properties are not universal among all bacterial derivatives and depend heavily on their engineering, dosage, and tumor context⁶.

Group B streptococci (GBS) are Gram-positive cocci colonized in the vagina and rectum⁷. This bacterium produces capsular polysaccharides (CPS), the main factor attributed to surviving in hosts. A cluster of genes located in the cps operon is responsible for the biosynthesis of the GBS capsular polysaccharides⁸. This operon is composed of three main regions: the first part (cpsA–D) involved in regulating the final steps of capsule biosynthesis, the central part (cpsE–L) encoding the glycosyltransferases and a polymerase. The last part (neuA–neuD) encodes enzymes implicated in sialic acid synthesis⁹. A key component of the GBS capsule is sialic acid, which contributes to immune evasion by mimicking host glycans and engaging inhibitory Siglec receptors on immune cells, thereby reducing phagocytosis¹⁰. At the same time, although capsular components stimulate protective antibody responses, CPS surrounding the bacteria prevents complement deposition and opsonophagocytosis⁹. Some studies provide convincing documents of the anticancer activity of GBS^11,12. Streptococcal sialic acids compete with cancer cells for binding to endothelial cells, an essential step for tumor migration. Sialic acid is a nine-carbon sugar originating from N-acetylneuraminic acid (Neu5Ac). These are similar to the sialyl-Lewis carbohydrates on the surface of mammalian cells¹³. Carcinogenesis is closely related to the excessive deposition of sialyl-Lewis on the surface of cancer cells. These accelerate the adhesion of malignant cells to endothelial cells expressing the sialyl-Lewis receptor¹⁴.

The similarity of GBS capsule and surface of malignant cells is the presence of NeuNAc-Gal-GlcNAc at the end of the side chain (Fig. 1). Accordingly, these polysaccharides compete with tumor cells to attach the ELAM- 1 on endothelium, hampering cancer metastasis¹¹. Besides, these polysaccharides induce an inflammatory response leading to tumor necrosis and proliferation inhibition¹⁵. Hence, we selected the cpsA and cpsC genes, which are involved in the last steps of CPS biosynthesis⁹, and designed a new recombinant fusion protein by applying bioinformatics tools, which opened up a new insight into biology. In the other step, we determine the anticancer properties of the recombinant protein by MTT, Real-Time PCR, and flow cytometry methods.

Fig. 1.

Interaction between sialyl Lewis carbohydrates (sialyl Le a and Le x) of cancer cells with ELAM- 1 on endothelium, facilitating tumor metastasis. This interaction can be inhibited by the polysaccharide capsule of GBS. These polysaccharides consist of a α-NeuNAc-(2→3)-β-Gal-(1→3)-β-GIcNAc trisaccharide unit, which is a structure specific to sialyl Lewis carbohydrates. The figure was created using Adobe.Illustrator.CC.25.4.1.498.0.

Results

To explore the therapeutic potential of a novel recombinant fusion protein derived from Group B Streptococci, we adopted the in silico and experimental approach. Initially, the fusion construct was designed and validated using various bioinformatics tools to assess its structural stability. This was followed by molecular dynamics (MD) simulations to evaluate the dynamic behavior and conformational flexibility of the modeled protein under physiological conditions. In the experimental phase, the anticancer properties of the recombinant protein were assessed against HeLa cervical cancer cells. Cellular assays included the MTT assay to determine cytotoxicity, Annexin V/PI staining to evaluate apoptosis induction, and quantitative Real-Time PCR to analyze the expression of pro- and anti-apoptotic genes (bax, caspase-3, and bcl-2). The following sections present the results from each of these sequential steps.

Design and bioinformatics analysis of construct

Anticancer sequences of CpsC and CpsA proteins were selected and connected by a flexible linker, maintaining the stability and flexibility of the fusion protein (Fig. 2a). Based on the iACP server estimation, the anticancer probability of the current fusion protein was 96.1%, while CpsC (0.24%) and CpsA (0.058%) sequences alone did not show anticancer properties. The recombinant protein contained 134 amino acids with a molecular weight of approximately 13.64 kDa. After codon usage, the codon adaptation index (CAI) increased from 0.74 to 0.96, which means the protein had higher expression in E. coli BL-21. At physiological pH (7.4), the fusion protein carries a net positive charge of approximately + 5.7, due to a higher proportion of positively charged residues (mainly Lysine and Arginine) compared to negatively charged ones (Aspartic acid and Glutamic acid), reflecting to anticancer activity of the current fusion protein. Physicochemical properties, namely theoretical pI, instability index, and GRAVY index, were estimated at 8.8, 41.78, and 0.4, respectively (Table 1). The physicochemical parameters of the recombinant protein calculated using the ProtParam tool provided predictive insights into its behavior during expression, purification, and in vitro stability. The pI reflects the pH at which the protein carries no net charge. The predicted pI (8.8) suggests the protein is moderately basic, which can influence its solubility and behavior in different pH environments. While the calculated pI is slightly above the average range for cytosolic proteins (typically pI 5–8), this value remains well within physiological relevance and is not unusual for secreted proteins. The instability index predicts the in vitro stability of a protein. According to Guruprasad et al. (1990), an instability index above 40 suggests the protein may be unstable in vitro. However, values in the range of 40–45 are considered borderline and often still compatible with successful expression and purification, especially when expression is performed in a controlled bacterial system such as E. coli¹⁶. The GRAVY score reflects the hydrophobicity or hydrophilicity of the protein. Generally, negative values indicate a hydrophilic nature, often correlating with higher solubility, while positive values suggest hydrophobic characteristics. A value of 0.4 indicates moderate hydrophobicity, suggesting membrane association or limited solubility. However, GRAVY values between − 0.5 and + 0.5 are common among functional recombinant proteins and do not necessarily indicate expression issues¹⁷. Taken together, these values are within ranges commonly reported for functional recombinant or secreted proteins and are supportive of the design and behavior of the construct.

Fig. 2.

First and secondary structures of fusion protein. (a) The fusion protein sequence. Green sequence: CpsC; Red sequence: linker; Blue sequence: CpsA (b) The secondary structure of fusion protein: blue, red, and purple lines are helix (19.40%), extended strand (36.57%), and random coil (44.03%), respectively.

Table 1.

Physicochemical properties of the fusion protein.

Parameters	MW	pI	Instability index	GRAVY	CAI	Net charge at physiological pH
Fusion protein	13.64	8.8	41.78	0.4	0.96	5.7

The secondary structure mainly consists of random coils (44.03%) between turns (Fig. 2b). The AlphaFold prediction for this construct showed that the majority of the structured core, particularly the α-helical regions, has very high confidence (predicted local distance difference test (pLDDT) > 90), indicating strong local structural stability. Flexible terminal and loop regions with lower pLDDT scores are expected to be intrinsically disordered and do not compromise the stability of the folded domain. Based on the Ramachandran analysis, 66.9%, 31.4%, and 1.7% of the residues were in favored, allowed, and disallowed regions, respectively, suggesting the quality of the 3-dimensional (3D) predicted protein model (Fig. 3a). The calculated Z-score of the modeled structure was − 0.94, which falls well within the range typically observed for experimentally determined native protein structures of similar size (usually between − 2 and + 2). This range is considered acceptable based on the ProSA-web evaluation criteria, and suggests that the model has no significant overall structural abnormalities. Z-score is commonly used to assess the quality of predicted models in comparison to high-resolution structures in the PDB¹⁸. The Z-score plot is illustrated in Fig. 3b.

Fig. 3.

Validation of fusion protein. (a) Ramachandran plot of fusion protein drawn by SAVE server. 98.3% of the residues are in the favored and allowed regions. (b) Z-score plot of the predicted tertiary structure of the fusion protein provided by ProSA server. The Z-score reflects the quality of the model compared to experimentally determined structures in the Protein Data Bank (PDB). The Z-score dot (black dot) located on the NMR (dark blue) regions suggested the highest accuracy of the simulated model. Note: This evaluation was performed in silico and serves as a computational quality check; no experimental structural validation has been conducted at this stage.

MD analysis results also revealed that the protein maintains its stability during simulation under physiological conditions. In other words, protein folding was intact in the cell’s aqueous conditions. The backbone root mean square deviation (RMSD, in Angstrom (Å)) of the fusion protein increased gradually during the initial phase of the MD simulation and reached approximately 1.6 nm around 20 ns, reflecting the relaxation and equilibration of the system. Following this adjustment, the RMSD remained stable with only minor fluctuations for the remainder of the 100 ns simulation, indicating that the construct preserved its overall structural stability without undergoing major conformational deviations. The RMSD plot is presented in Fig. 4a. The root mean square fluctuation (RMSF, in Å) profile of the current fusion protein did not reveal any unusual fluctuations across most residues, indicating overall structural stability. To further illustrate the conformational dynamics, representative snapshots of the structure at 10 ns intervals are provided in Supplementary Fig. S1, where RMSF values are mapped as B-factors onto the structure to highlight the most flexible regions. This increased flexibility is mainly attributed to the presence of a long loop in this region (the N-terminal region), whereas the rest of the structure remains relatively stable. Such localized fluctuations are expected in loop segments, as they are generally more dynamic and flexible compared to structured regions like α-helices and β-sheets. This intrinsic flexibility may contribute to proper folding and functional adaptability of the construct. The structural compactness of the fusion protein was assessed by monitoring the radius of gyration (Rg in Å) throughout the 100 ns MD simulation. The initial Rg value at the first frame was 2.50 nm, which gradually decreased during the first few nanoseconds and reached a plateau around 10 ns. At the end of the simulation, the Rg stabilized at approximately 1.498 nm. This pronounced reduction followed by stabilization indicates a transition of the construct from a relatively extended conformation to a more compact and stable form. After the initial relaxation phase, only minor fluctuations in Rg were observed, suggesting that the fusion protein structure maintained its compactness and overall stability during the remainder of the simulation.

Fig. 4.

Molecular dynamics (MD) simulation analysis of the CpsA–CpsC fusion protein. (a) RMSD plot for all the atoms of the fusion protein over a 100 ns simulation shows that the protein reached a stable conformation after 20 ns, indicating structural stability. (b) RMSF plot reveals flexibility variations among amino acid residues. (c) The Radius of gyration (Rg) plot displays minor fluctuations over time, suggesting compactness and conformational stability of the protein during the simulation. (d) Hydrogen bond analysis was performed to evaluate the intramolecular stability of the multi-epitope vaccine during the MD simulation. The probability mass function (PMF) of the hydrogen bond distribution exhibited a bell-shaped curve with a clear peak at ~ 80 hydrogen bonds. Notes: All simulations were performed for a single system (n = 1). Axis labels include units where applicable to enhance reproducibility.

The probability mass function (PMF) of the hydrogen bond distribution exhibited a bell-shaped curve with a clear peak at ~ 80 hydrogen bonds. This indicates that, for the majority of the simulation time, the fusion protein structure maintained approximately 80 intramolecular hydrogen bonds. The consistency of this distribution reflects the structural integrity and stable hydrogen bonding network of the construct throughout the simulation. RMSF and RG plots and Probability mass function of hbonds are illustrated in Fig. 4b, c, and d. The 3D structure predicted using AlphaFold2 is shown in Fig. 5. The energy decomposition showed that the short-range electrostatic (Coul-SR) interactions were strongly negative (average − 2135010 ± 16 KJ/mol), indicating dominant stabilizing forces in the system. In contrast, the short-range Lennard–Jones (LJ-SR) term exhibited positive values (average 339330 ± 8 KJ/mol), which is expected due to steric repulsions arising from close atomic packing in the solvated vaccine construct. Overall, the negative Coul-SR values outweighed the LJ-SR repulsions, confirming the structural stability of the fusion protein during the 100 ns simulation.

Fig. 5.

Final 3D conformation of the fusion protein visualized using ChimeraX, highlighting its overall folded structure and potential surface-exposed regions.

Although no experimental negative control with a tag-only construct was included, ProtParam analysis showed no significant impact of the His-tag on the overall physicochemical properties of the fusion protein. Additionally, no functional domains related to anticancer activity were identified in the His-tag or linker sequences using InterProScan and Pfam databases. Previous studies also showed that His-tags do not generally interfere with the protein’s biological activity or structure¹⁹.

Cloning, expression, and protein purification

Plating the bacterial suspension attained from the transformation step on Luria Bertani (LB) agar medium supplemented with ampicillin led to the selection of the bacteria-harboring recombinant plasmid (Supplementary Fig. S2). A single band on agarose gel after amplifying a 124 bp fragment from the recombinant gene indicated the successful transformation procedure (Fig. 6a). The presence of a distinct band corresponding to the expected size predominantly in the supernatant confirmed that the protein was mainly expressed in a soluble form. According to SDS-PAGE analysis, a concentration of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) induced protein expression. Moreover, the expression levels increased after the fourth hour of induction (Fig. 6b). As the fusion protein with a molecular weight of 13.64 kDa was in soluble form, we set the purification process under native conditions. The successful purification under native conditions using Ni-NTA affinity chromatography further supported the soluble expression of the target protein. In Fig. 6c, we can see an intense band in western blot analysis, proposing correct purification. The protein concentration was 390 µg/ml. Densitometric analysis of coomassie-stained SDS-PAGE indicated an estimated purity of ~ 59% for the final protein (Supplementary Fig. S3). The total yield of the purified fusion protein was approximately 2.1 mg from 3 L of bacterial culture, equating to a yield of ~ 0.72 mg/L. Although the overall yield was modest, the obtained amount was sufficient to conduct all downstream in vitro assays, including cytotoxicity, apoptosis, and gene expression analyses. The LPS level of the purified proteins was less than 1 EU/ml, which is acceptable range for in vitro studies.

Fig. 6.

Confirmation of transformation, expression, and purification. (a) PCR amplification confirms the presence of the fusion gene (cpsA–cpsC) (124 bp) with specific primers. Lane M: 100 bp DNA ladder; Lane 1: positive clone showing expected band, (b) SDS-PAGE analysis of the purified fusion protein after Ni-NTA affinity chromatography using the 12% polyacrylamide gel. Lane M: protein marker; Lane 1: purified protein showing a distinct band at 13.64 kDa. Protein was purified via Ni-NTA affinity chromatography, (c) Western blot detection of the expressed His-tagged fusion protein using anti-His antibody, Lane M: protein marker; Lane 1: purified fusion protein. A single band at 13.64 kDa successfully confirms expression of the target protein.

Cytotoxicity of CpsA-CpsC protein on cancerous cells

The quantitative anti-proliferative activity of the recombinant protein on cancer cells showed a dose- and time-dependent decline in cell viability. As shown in Fig. 7a, the number of viable cells did not drop significantly after 24 h (P-value = 0.1120). In contrast, increasing protein concentrations resulted in marked viable cell percentages, with an IC₅₀ value of 68.1 µg/ml after 48 h, as calculated from the fitted dose–response curve using GraphPad Prism. The growth of cells without treatment was constant. Morphological evaluation under a standard inverted light microscope (10×) demonstrated hallmark features of apoptosis in HeLa cells, including noticeable shrinkage, membrane blebbing, and detachment from the culture surface. While untreated control cells maintained normal morphology. No additional staining or labeling was performed (Fig. 7b, c).

Fig. 7.

Effects of fusion protein on the viability of HeLa cells. (a) Cells were treated with various concentrations of chimeric protein for 48 h. Notes: Data represent the mean ± SD of three independent biological replicates. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 compared to control group. (b, c) Morphological evaluation of HeLa cells after 48-hour treatment with IC₅₀ concentration (68.1 µg/ml) of the fusion protein. Images were captured using a standard inverted light microscope (10×). Treated cells (c) exhibited characteristic apoptotic morphology, including cell shrinkage, membrane blebbing, and detachment from the culture surface, in contrast to untreated control cells (b), which retained normal morphology. No staining was applied before imaging.

Gene expression assessment

The investigation of the expression level of apoptosis-related genes (bax, bcl-2, and caspase-3) was applied to recognize the death mechanism. These genes encode the pro- and anti-apoptotic proteins, facilitating apoptosis induction or prevention. As we expected, the expression of the bax gene was 22.62-fold higher than control cells (P-value = 0.0028) (Fig. 8), whereas the bcl-2 gene expression level was down-regulated, accounting for 0.235-fold (P-value = 0.0218) (Fig. 8). These changes were statistically significant compared to the control group that did not receive treatment. As we can see in Fig. 8, the expression level of the caspase-3 gene involved in intrinsic and extrinsic mitochondrial apoptosis pathways elevated 22.06-fold in treated cells (P-value = 0.0003). The housekeeping gene, human glyceraldehyde-3-phosphate dehydrogenase (gapdh), expression profile was constant after and before treatment. Therefore, the apoptosis pathway activation was one of the probable mechanisms of cervical cancer cell death. Melting curves are provided in Supplementary Fig. S4.

Fig. 8.

mRNA level of apoptosis-related genes (bax, bcl-2, and caspase-3) using RT-PCR technique. The expression level of the bax and caspase-3 genes increased 22.62-fold and 22.06-fold, respectively, compared to control. The expression level of the bcl-2 gene decreased 0.235-fold compared to the control. Notes: Data represent the mean ± SD of three independent biological replicates. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 compared to control group. Raw CT values are provided in Supplementary Table. S1.

Annexin V-FITC/PI-staining

In the present study, to explore the percentage of apoptotic, necrotic, and live cells, double staining with Annexin V/PI was performed. Annexin V is a calcium-dependent protein that can bind to phosphatidylserine exposed to the surface of cells. On the other hand, propidium iodide (PI) interacts with DNA after decreasing the membrane integrity and entering the cytosol. As illustrated in Fig. 9, following exposure of the cells to recombinant protein, the percentage of live cells reduced from 95.95% to 75%, while the early (from 0.4% to 10.5%) and late (from 0.621% to 6.84%) apoptotic cells increased significantly. The proportion of necrotic cells had no significant difference (P-value = 0.2500) between the treated and control groups. Overall, apoptotic cell percentage (early and late) increased by approximately 17.5% (P-value = 0.0050). Our findings suggested that treatment of HeLa cells with CpsA-CpsC recombinant protein made cells vulnerable to apoptosis.

Fig. 9.

Apoptosis detection by Annexin V/PI staining. (a) Untreated (control) cells, (b) Treated cells with IC50 concentration of fusion protein. Live cells (Q4), early apoptotic cells (Q3), late apoptotic cells (Q2), and necrotic cells (Q1) are located in each of the four quadrants of the plot. (c) Percentage of apoptotic and necrotic cells shown in a column graph. Note: Data represent the mean ± SD of three independent biological replicates. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 compared to control group.

Discussion

Cervical cancer remains a significant global health challenge in women due to its poor prognosis. We designed a novel recombinant fusion protein derived from GBS using bioinformatics tools and validated its anticancer activity experimentally. Our results revealed that the recombinant protein reduced the HeLa cell viability in a dose- and time-dependent manner, with an IC50 of 68.1 µg/ml. In 1993, Hellerqvist et al. showed the anticancer effect of GBS extracellular toxin. This exotoxin triggers inflammatory responses in infants’ lungs by targeting endothelial cells. In mice bearing lung tumors, treatment with this toxin induced tumor tissue disruption characterized by vasodilation, necrosis, and hemorrhagic lesions¹⁵. Although this polysaccharide exotoxin differs in structure and mechanism from our recombinant protein, both studies highlight the potential therapeutic value of GBS-derived products. These findings further support GBS as a source of bioactive molecules with therapeutic potential. However, the translation of these candidates into clinically applicable drugs requires extensive optimization and safety assessment. To overcome these challenges, computational methods have emerged as powerful tools in drug discovery, allowing researchers to efficiently predict, refine, and evaluate the biological activity and physicochemical properties of candidate agents^20,21. Unlike the solely experimental approach of Hellerqvist et al., our study combines in silico and experimental methods to enhance target specificity and safety.

Following the introduction of prediction tools, AntiCP, numerous bioinformatics platforms have emerged to identify and characterize anticancer peptides (ACPs)²². The iACP tool has gained popularity due to its user-friendly interface and robust performance metrics; therefore, we initially used it to assess the anticancer potential of the CpsA–CpsC fusion protein²³. Despite discrepancies in reported performance metrics (accuracy, sensitivity, specificity, and Matthews correlation coefficient (MCC)) across different studies, the high-confidence output of iACP provided sufficient rationale for its experimental evaluation. These discrepancies are likely due to differences in benchmark datasets, validation protocols (e.g., cross-validation vs. independent test sets), and model optimization strategies^24–26. Nevertheless, iACP was used in this study primarily as a screening tool, and its predictions were validated by in vitro experiments. The consistency between the predicted anticancer potential and the observed apoptotic effects in vitro (upregulation of bax and caspase-3) highlights the value of combining bioinformatics with experimental approaches. It should be emphasized that using bioinformatics tools alone are insufficient for pharmacopeial studies, and experimental validation is necessary to complement computational predictions. For example, Jalalvand et al. and Fathizadeh et al. predicted anticancer sequences with these tools and then performed experimental tests. The computational analysis was consistent with laboratory tests^27,28.

The aberrant expression of sialic acid on cancer cells contributes to tumor progression and metastasis. Blocking enzymes involved in the biosynthesis of glycoconjugate chains containing N-acetylneuraminic acid has been proposed as an early strategy to control metastasis in tumors expressing sialylated antigens²⁹. The GBS capsule mimics these structures, and the genes cpsA and cpsC are involved in the biosynthesis and regulation of the NeuNAc–Gal–GlcNAc trisaccharide units forming the capsule⁹. Considering the structural resemblance to tumor-associated sialylated antigens, these proteins may hypothetically interfere with tumor-associated sialylation processes. However, this remains a theoretical mechanism inspired by previous findings on GBS polysaccharides (Miyake et al., 1996), and no direct biochemical or binding evidence was obtained in the present study. Future investigations using migration and invasion assays, glycan-binding studies, or in vivo models are warranted to evaluate this potential role¹¹. Importantly, the initial selection of CpsA and CpsC was supported by in silico predictions using the iACP tool, which identified these proteins with anticancer activity. In silico analyses of physicochemical properties and molecular dynamics further supported the experimental findings. These analyses confirmed that the designed protein is structurally stable and soluble, properties critical for maintaining biological activity and enabling its anticancer effects.

Apoptosis, mediated by dysregulation of apoptosis-related genes and activation of the caspase cascade, is a key mechanism of cell death³⁰. In the current study, flow cytometric analysis indicated that apoptosis was the predominant mode of cell death. Notably, a slight increase in PI-positive/Annexin V-negative cells in the treated group (~ 7.5%) compared to control (~ 3%) suggests that, in addition to apoptosis, a small fraction of necrotic or membrane-compromised cells may have contributed to the observed cytotoxicity. However, further studies are required to confirm this observation and investigate the involvement of other cell death mechanisms. Additionally, the observed dysregulation of apoptosis-related genes (Bax and Bcl-2) and activation of the caspase cascade corroborated the flow cytometry findings.

The net positive charge of this fusion protein, due to the abundance of Lysine and Arginine residues, may enhance its attachment to cancer cells expressing excessive negatively charged antigens (phosphatidylserine, O-glycosylated mucins, sialylated gangliosides, and heparin sulfates)³¹. This electrostatic attraction is a known mechanism exploited by several anticancer peptides and proteins, leading to membrane disruption or internalization, followed by apoptotic signaling^32,33. Additionally, the presence of flexible amino acids (Alanine and Glycine), which may increase membrane fluidity, could facilitate protein penetration into the cell, potentially inducing cancer cell death via intracellular signaling pathways³⁴. However, this remains a hypothetical mechanism, as membrane binding affinity and charge neutralization were not experimentally evaluated. Further investigations, including direct binding assays or surface interaction studies, are required to confirm whether such interactions contribute to the observed cytotoxic effects. In silico prediction indicated a potential allergenic risk for the designed protein; however, such computational estimates are not always definitive and require validation through in vitro and in vivo studies. Moreover, such risks can be mitigated by rational protein engineering (e.g., epitope modification). The intended local administration (e.g., vaginal gel) may reduce systemic exposure, but comprehensive immunological assessments will be essential before clinical translation.

Limitations of the study

This study has several limitations. First, the predicted 2D and 3D structures of the recombinant fusion protein were not experimentally validated. While we employed established computational tools to assess protein quality, these approaches cannot replace experimental validation. Protein concentration was determined using the Bradford assay, which has limited precision compared to spectrophotometric methods. Nonetheless, as all assays were normalized to the same protein batch, this limitation likely had minimal impact. Due to limited access to structural biology facilities such as circular dichroism (CD) spectroscopy, size exclusion chromatography (SEC), X-ray crystallography, or NMR spectroscopy, experimental confirmation could not be performed.

Second, additional controls, including BSA treatment, His-tag only, and experiments on normal cell lines (HEK293), were not included. The absence of these controls may restrict conclusions about its overall safety and specificity. In this study, protein-only wells were not included in the MTT assay, so a potential interaction between the fusion protein and the MTT reagent cannot be fully excluded. However, the optical density values in untreated (cell-only) controls were consistent, and no abnormal color changes were observed in protein-treated wells, suggesting that the obtained cytotoxicity data are reliable. Future studies using complementary assays such as LDH release or trypan blue exclusion are recommended.

While no significant effect on cell viability was observed at 24 h, the 48-hour exposure revealed the protein’s time-dependent cytotoxic potential. Longer incubation may be needed to fully characterize kinetic effects. Additionally, while SDS-PAGE and densitometric analysis confirmed successful purification of the recombinant fusion protein, the estimated purity (~ 59%) indicates that minor impurities remained in the final preparation. These impurities may have partially contributed to cytotoxicity assays, though the reproducibility across replicates supports overall data reliability. Future optimization of purification and chromatographic conditions will aim to improve both yield and purity.

Functional assays (e.g., caspase activity or mitochondrial potential) were not performed, which could provide further mechanistic insight into apoptosis induction. Our study focused on the anticancer potential of the recombinant construct rather than quantifying capsule unit assembly. Future studies should investigate whether linker length or the independent activity of CpsA and CpsC influences polysaccharide assembly.

Conclusion

In this study, we successfully designed, expressed, and purified a novel CpsA–CpsC fusion protein derived from Streptococcus agalactiae. Computational predictions and molecular dynamics simulations indicated structural stability and physicochemical properties compatible with biological activity. In vitro assays demonstrated a dose- and time-dependent reduction in the viability of HeLa cells, with an IC50 of 68.1 µg/ml after 48 h. Flow cytometric analysis revealed a substantial increase in apoptotic cells (~ 17.5%) compared to untreated controls (< 1%), indicating more than a 17-fold rise in apoptosis (P-value = 0.0050). This significant increase, together with the upregulation of bax and caspase-3 and the downregulation of bcl-2, supports the activation of the apoptotic pathway rather than general cytotoxic stress. These findings suggest that apoptosis is likely the predominant mode of cell death, although minor contributions from necrotic pathways cannot be excluded. Further investigations involving additional cancer cell lines, normal cell models, functional assays such as caspase activity and mitochondrial membrane potential analyses will be required to confirm the mechanistic details and the specificity.

Materials and methods

Materials

The E. coli-BL-21 strain was kindly donated by the Baqiyatallah University of Medical Sciences, Tehran, Iran. LB agar and broth medium and Dimethyl sulfoxide (DMSO) were purchased from Merck Company (Germany). Ampicillin powder, IPTG, imidazole, MTT reagent, Triton X-114, and PMSF were obtained from Sigma-Aldrich (USA). Ni-NTA chromatography column acquired from Pars tous, Iran. HRP anti-His Tag antibody was purchased from Cyto Matin Gene, Iran. Diaminobenzoic acid (DAB) was obtained from Bio-Rad, USA. Limulus Amebocyte Lysate (LAL) kit was donated from Lonza Company (USA). The Real-Time PCR master mix was obtained from Amplicon, Denmark. Bradford and Annexin V/PI staining kits were purchased from DNAbiotec (Iran) and Biolegend (England), respectively. Both RNA extraction and cDNA synthesis kits were obtained from the Favorgen (Taiwan) Company. DMEM medium and FBS were obtained from Gibco, USA. Primer sequences were synthesized by Bioneer, Korea. DNA and protein size markers were procured from Sinaclon, Iran. HeLa cell line was obtained from the Iranian Biological Resource Center (IBRC), Iran.

Construct design and analysis

To design the recombinant protein, the complete protein sequences of CpsC (Accession No. P0C0T8) and CpsA (Accession No. Q1G5T1) were obtained from UniProt (https://www.uniprot.org/). To computationally evaluate the anticancer potential of each sequence and the designed fusion protein with parts of two proteins, we employed the sequence-based iACP tool (http://lin.uestc.edu.cn/server/iACP), which achieves an overall accuracy of approximately 94% and the MCC of ~ 0.89²³. This tool classifies input sequences as anticancer or non-anticancer and provides a confidence score based on a support vector machine (SVM)-based model. Initially, the amino acid sequences of the CpsA and CpsC proteins were submitted separately in FASTA format, and their anticancer prediction scores were recorded. Subsequently, the full-length fusion protein sequence (CpsA–CpsC) was analyzed using the same server. These in silico predictions served as a preliminary screening step before experimental validation. The flexible linker Gly-Gly-Gly-Gly-Ser (GGGGS) was chosen due to its frequent use in recombinant protein engineering, where it provides high flexibility, low immunogenicity, and has been widely validated as a standard linker in antibody and fusion protein design³⁵. In this construct, the GGGGS linker was particularly selected to minimize steric hindrance between CpsA and CpsC and to allow independent folding of both domains. The physicochemical features and the secondary structure of the recombinant protein were computed by (http://web.expasy.org/protparam/) and the GOR4 (https://npsa-prabi.ibcp.fr/cgibin/npsa) servers, respectively. The net charge of the fusion protein at physiological pH (7.4) was estimated using the PepCalc online tool (Innovagen, https://pepcalc.com), as described by Lear & Cobb³⁶. The 3D structure of the recombinant protein was predicted using the AlphaFold2 algorithm (https://alphafold.ebi.ac.uk/), which employs deep learning and attention-based neural networks for protein structure modeling. AlphaFold2 predicted the atomic coordinates of the fusion protein and provided per-residue confidence scores (pLDDT) to assess model reliability. The 3D model was validated using the ProSA (https://prosa.services.came.sbg.ac.at/prosa.pHp) and PROCHECK (http://services.mbi.ucla.edu/PROCHECK) servers. PROCHECK provided a Ramachandran plot through the SAVE server (http://services.mbi.ucla.edu/SAVES), showing the percentage of residues in the favorable region. The ProSA web server calculated the Z-score. The sequence was subjected to the GeneScript (http://www.genscript.com/) tool to optimize and yield the appropriate expression in the prokaryotic host (E. coli BL-21). Finally, Biomatik (Canada) chemically synthesized the recombinant gene into the pET22b(+) vector, which includes a C-terminal His₆-tag for purification and PelB signal peptide encoding genes (Supplementary Fig. S5). Using the pET22b(+) vector containing a PelB signal sequence is a strategy to enter the recombinant protein into the solution phase. This sequence transports the protein into the periplasmic space. Since the His₆-tag did not cleave from the recombinant protein, to ensure that the anticancer effects are not due to the His-tag, sequence-based prediction tools (Pfam and InterProScan) were used.

Molecular dynamics simulation

To assess how cell conditions affect the recombinant protein structure, MD simulations were conducted using GROMACS 2024.2. This analysis led to the refinement of the structures obtained from AlphaFold2. System topologies were generated based on the AMBER03 force field. In summary, each complex was centered in a dodecahedral simulation box approximately 8 × 8 × 8 nm³ containing ~ 40,000 atoms (protein, water, and ions), and solvated with TIP3P water.

The system was neutralized by replacing H₂O molecules with Na⁺ or Cl⁻ ions randomly and followed by energy minimization using the steepest descent algorithm to eliminate unfavorable contacts. The systems were equilibrated in two phases: a 100 ps NVT ensemble at 298 K followed by a 100 ps NPT ensemble at 1 bar pressure. Long-range interaction computation, like electrostatic interactions, was done by the Particle-mesh Ewald (PME) method. The Linear Constraint Solver (LINCS) algorithm was applied to constrain all bonds in recombinant protein during the simulation. Finally, the protein was run into software for 100 ns without restraint. The outputs of GROMACS were RMSD, RMSF, Rg plots, and hydrogen bond analysis, demonstrating the stability of the protein³⁷.

RMSD measures the average displacement of atomic positions of the protein backbone over time with respect to the initial structure. A low and stable RMSD value during the simulation suggests that the protein maintains its global structural integrity and does not undergo large conformational changes. It is a primary indicator of structural stability in MD simulations³⁷. RMSF quantifies the flexibility of individual residues throughout the simulation. Higher RMSF values indicate greater flexibility, often corresponding to loop regions or terminal residues, while lower values suggest rigid and stable secondary structure elements such as α-helices and β-sheets³⁸. The radius of gyration assesses the overall compactness of the protein structure during simulation. A consistent Rg value indicates that the protein maintains a stable folded conformation, whereas large fluctuations may suggest unfolding or structural instability³⁹. Hydrogen bond analysis was performed to evaluate the intramolecular stability of the fusion protein during the MD simulation. The LJ potential energy of the fusion protein was calculated to estimate van der Waals interactions and short-range repulsion forces. LJ energies were obtained using the default force field parameters following energy minimization. The results provided a qualitative measure of non-bonded interaction stability, complementing structural validation.

Cloning, expression, and protein purification

E. coli BL-21 strain was recruited as the host for protein expression. In brief, the cloned gene in the pET22b(+) vector was transformed into competent cells using the heat shock process, followed by plating the transformed cells on the LB agar medium containing ampicillin (100 µg/ml) to phenotypically confirm transformation. The DNA from suspicious colonies was extracted and amplified by PCR with specific primers. The expression was induced by adding IPTG (1mM) to the logarithmic phase of bacterial suspension cultured on the LB broth supplemented with ampicillin. After IPTG induction, the bacterial cells were harvested and lysed by sonication and lysis buffer comprising antiprotease (PMSF). The lysate was centrifuged at 15,000 ×g for 30 min at 4 °C to separate the soluble (supernatant) and insoluble (pellet) fractions. Both fractions were analyzed by the sodium dodecyl SDS-PAGE system (Bio-Rad, USA) to assess the expression level and solubility of the fusion protein (Supplementary Fig. S6). The recombinant protein with a 6His tail at the carboxyl end was purified using Ni-NTA chromatography. In brief, the filtered supernatants containing the released protein were added to the equilibrated column with the binding buffer (0.15 M NaCl, 10% glycerol, 0.5% Triton X-100, 50 mM sodium phosphate buffer). Initially, non-specific proteins bound to the column were removed by washing with low concentrations of imidazole (10, 20, 100, and 200 mM). Using higher concentrations of imidazole (500 mM) resulted in protein elution. Purified fractions were run on SDS-PAGE. The protein band intensity was quantified by densitometric analysis using ImageJ software (version 1.53k), and percent purity was calculated as (area of target band / total lane area) × 100. Lipopolysaccharide (LPS) was removed from purified recombinant proteins by phase separation with Triton X-114, followed by quantification of the LPS level using the Limulus Amebocyte Lysate (LAL) assay, according to the manufacturer’s protocol. Eluted fractions were dialyzed with phosphate-buffered saline (PBS) for 24 h to remove imidazole or other components of buffers¹². The His-tag was retained throughout the purification process, and no proteolytic cleavage was performed.

Western blotting

Purified proteins were subjected to western blot analysis to confirm purification accuracy. First, the SDS-PAGE-separated proteins were blotted on the polyvinylidene fluoride (PVDF) membrane employing the semi-dryer system (Bio-Rad, USA). The blotting reaction was blocked using 5% non-fat dry milk for 1 h. Afterward, the membrane was washed with PBS containing Tween-20, and then the HRP anti-His Tag antibody was added. DAB was employed to visualize a protein-specific band. The extracted protein concentration was determined using the Bradford assay⁴⁰. The total yield of purified fusion protein (mg/L of bacterial culture) was calculated as:

Cytotoxicity assay

Antiproliferative effects of the recombinant protein were determined using the MTT reduction assay on the HeLa cell lines. Briefly, 10⁴ cells were grown per well in a 96-well plate with a complete medium (DMEM medium containing 5% FBS and 1% penicillin/streptomycin). Cells were incubated in an atmosphere with 5% CO2 for 24 h at 37 °C. After incubation, the medium was aspirated, and cells were treated with various concentrations of the recombinant protein (195, 97.5, 48.75, 24.37, 12.18, 6.09 µg/ml). The exposure time of cells with recombinant protein was 24 and 48 h. The process continued by adding the MTT reagent (5 mg/ml to each well). The MTT-containing medium was replaced with DMSO to dissolve the formazan crystals. The last step of the test was reading the optical density at 570 nm. The cell viability and IC₅₀, defined as the concentration of drug required for inhibition of 50% of the biological process, were determined by nonlinear regression analysis of the dose–response curve using GraphPad Prism, version (8). Cells without treatment were the negative control⁴¹.

Quantitative real-time PCR assay

To further study cell death mechanisms, the expression of the panel of apoptosis-related genes (bax, bcl-2, and caspase-3) was investigated by the quantitative Real-Time PCR technique. For this purpose, after the growth of 2 × 10⁶ cells in a 6-well plate, cells were exposed to IC50 concentration (68.1 µg/ml) of recombinant protein, and then the plates were incubated for 48 h. Total RNA extraction was performed based on the instructions of the RNA extraction kit. The quantity and quality of RNA were checked by measuring absorbance at 260 nm by the Nanodrop spectrophotometer (Thermo Scientific, USA) and electrophoresis on agarose gel, respectively. The extracted RNA was reverse-transcripted to complementary DNA (cDNA) by applying the cDNA synthesis kit. The Real-Time PCR mixture included 10 µl of 2X SYBR green master mix, 1 µl of each primer (Table 2), 2 µl of cDNA, and 4 µl of deionized water. The thermal cycling was programmed as follows: 95 °C (10 min), 40 cycles at 95 °C (15 s), and 60 °C (60 s). The gapdh gene was considered an internal control. The gene expression level was determined using the 2^−ΔΔCt method. Cells without treatment were the negative control¹².

Table 2.

Primer sequences used in the present study.

Gene	Primer sequences	References
caspase-3	5’-ATGGAAGCGAATCAATGGA-3’ 5’-TGTACCAGACCGAGATGTC-3’	42
bcl-2	5’-AGGAAGTGAACATTTCGGTGAC-3’ 5’-GCTCAGTTCCAGGACCAGGC-3’	43
bax	5’-AGGGTTTCATCCAGGATCGAGCAG-3’ 5’-ATCTTCTTCCAGATGGTGAGCGAG-3’	12
gapdh	5’-GAA GGT GAA GGT CGG AGT CA-3’ 5’-GAA GAT GGT GAT GGG ATT TC-3’	28

Annexin V-FITC/PI-staining

Apoptosis was determined using an Annexin V/PI double staining assay as recommended by the manufacturers of the Apoptosis Detection Kit. 200,000 cells/well were seeded in a 6-well plate and allowed to reach 80% confluency. For apoptosis detection, the IC50 concentration (68.1 µg/ml) of recombinant protein was inoculated into wells, followed by 48 h incubation at 37 °C. Harvested cells were subsequently washed twice with cold BioLegend Cell Staining Buffer with gentle agitation and then resuspended in 200 µl Annexin V Binding Buffer (1X). After addition of 5 µl FITC Annexin V and 10 µl PI into the cell suspension, cells were vortexed lightly and then incubated for 15 min in the dark. The volume of each tube reached 500 µl with the addition of the Annexin V binding buffer. Samples were subjected to the flow cytometer (BD Biosciences, USA) to analyze the percentage of live and dead cells. The results were analyzed using Flowjo software to determine different cell populations: live cells (Annexin V⁻/PI⁻), early apoptotic cells (Annexin V⁺), late apoptotic cells (Annexin V⁺/PI⁺), and necrotic cells (Annexin V⁻/PI⁺). Untreated cells were the negative control.

Statistical analysis

All tests were repeated in triplicate independently. The data were analyzed using the T-test and one-way ANOVA tests by GraphPad Prism software (8.0.1). Results were statistically significant when the P-value was below 0.05.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Z. M, [M.S, ] Z S, M A, DE contributed to the design of the study. Z. M, [M.S, ] DE. contributed to the general supervision of the study. Z. M, [M.S, ] DE collaborated in data collection. Z. M, DE collaborated in the data analysis, developed the first version of the manuscript Z. M. and. Z S, M A, made the first revision of the manuscript. All authors revised the different versions of the manuscript, read, and agreed to the submitted version of the manuscript.

Funding

This study was financially supported by Hamadan University of Medical Sciences, Hamadan, Iran [Grant Number: 140306204966].

Data availability

All relevant data supporting the findings of this study, including raw Ct tables and melting curves from the real-time PCR analysis, unprocessed SDS-PAGE gel images, the cloning vector map, phenotypic confirmation of transformation, SDS-PAGE densitometric analysis and protein band intensity quantification, and accession links to the public databases used, are included within the article and its supplementary file.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

This project was approved by the ethical committee of Hamadan University of Medical Sciences, Iran [IR.UMSHA.REC.1403.375].