Introduction
Peptide-based drugs have attracted attention in recent years for their potential application in anticancer and antimicrobial therapies. They offer greater specificity, fewer off-site side effects, and increased potency compared with current cancer therapies [
1]. Protein-based biologics are complex peptide molecules with a typical molecular weight exceeding 5 KDa; synthesizing these molecules is expensive and challenging [
2]. On the other hand, conventional small-molecule drugs with molecular weights less than 0.5 KDa, tend to exhibit lower specificity and often induce off-target side effects [
3]. Peptide drugs fall within the molecular weight range of 0.5 to 5 KDa, offering a unique combination of biologics' specificity, potency, and low toxicity, along with the advantages of easy production and metabolic stability of small-molecule drugs [
2]. Anticancer peptides (ACPs) comprise one class of peptide drugs and are relatively small molecules (5 to 50 amino acids) with cationic and amphiphilic properties [
2]. Due to these properties, ACPs specifically target relatively anionic cancer cells with low reactivity with relatively neutral normal mammalian cells [
4,
5]. This targeting strategy is rarely used in current cancer therapeutics. Chemotherapeutic agents do not differentiate between normal proliferating cells and cancer cells, thus cannot target indolent or dormant cancers [
6]. The development of a chemoresistant phenotype further reduces the therapeutic utility of chemotherapy [
7], as current cancer therapy regimens are unable to overcome multidrug resistance, which is slowly acquired via progressive tumor mass exposure [
6].
The anticancer properties of ACPs are established through membranolytic (direct-acting ACPs) and/or non-membranolytic modes of action (indirect-acting, programmed cell death) [
8]. In terms of membranolytic mechanisms, ACPs can disrupt the cellular, mitochondrial, or lysosomal membranes by either the carpet or barrel-stave mechanism. Small peptides can aggregate via hydrophobic interactions to form a structure through the plasma membrane resembling a traditional ion channels, or can alternatively pass through the plasma membrane and permeate the mitochondrial membrane. Here they induce swelling of the mitochondria, causing the release of cytochrome c and subsequent activation of caspases 9 and 3, or can modify the lysosomal membrane, resulting in acidification of the cytosol. The non-membranolytic mechanisms of ACP action include activation of calcium channels and consequent calcium ion influx, augmentation of proteasome activity, inhibition of pro-survival genes, or cell-cycle arrest [
5,
9]. Thus, ACPs exert a non-systemic targeted effect compared with conventional chemotherapy [
10]. Additionally, ACPs have been shown to promptly induce the death of cancer cells, hindering the development of resistance [
11].
Some ACPs are inherently antimicrobial in nature as presented in antimicrobial peptides (AMPs) [
12], and have been classified according to their secondary/tertiary structure as α-helical or β-sheets [
13]. This dual property is predicted to make ACPs superior to conventional chemotherapeutic drugs by minimizing the development of cellular acquired resistance in cancer cells [
10].
ACPs exhibit promising therapeutic potential against cancer, with their ability to inhibit cancer cell proliferation, migration, and angiogenesis. Moreover, ACPs possess the added advantage of being cost-effective to produce, positions them as preferable alternative to conventional chemotherapy, immunotherapy, or radiotherapy for cancer treatment.. However, the substantial cytotoxicity and poor targeting of ACPs have impaired their application. Characteristics, such as the number, charge, and sequence of the specific amino acids are altered via reconstruction mechanisms to overcome the shortfalls of the targeting effectiveness of ACPs. This reconstruction process involves modifying the amino acids that constitutes the main and/or side chains employing three distinct machine learning (ML) approaches: supervised, unsupervised, and reinforced learning.
The present study aimed to screen the American University in Cairo (AUC) Red Sea metagenomics data library, generated during AUC/KAUST Red Sea Microbiome Project, to identify a novel antimicrobial peptide with potential anticancer activity. The metagenomic samples primarily came from three locations in the Red Sea: Atlantis II Deep, Discovery Deep, and Kebrit Deep [
14]. Scientists studying the chemical properties of the brine pools in Atlantis II Deep and Discovery Deep observed those of the former to be predominantly sulfur enriched, while those of the latter were carbon and nitrogen enriched. This difference in chemical composition results in unique microorganism diversity, specifically observed at the deepest soil layers in each site; specializing in metabolizing the compounds dominant in their respective sites. We identified a novel 37-residue antimicrobial peptide from the brine pool of Atlantis II Deep and modified the amino acid sequence to increase its hydrophobicity and anticancer activity. The modified 37-mer peptide was tested in a dose-dependent cytotoxicity assay on grade I, II and III/IV hepatocellular carcinoma cell lines, (HepG2 and SNU449, respectively), an ovarian cancer cell line (SKOV3), and on HeLa cervical cancer cells. The anticancer properties of the peptide were evaluated by analyzing changes in cell viability, cell morphology, and inhibition of cellular migration. Mode of cell death was also investigated, and the targeted selectivity was validated via screening of human erythrocytes. The peptide was also assessed for antimicrobial activity on both gram-positive (
Staphylococcus aureus) and gram-negative (
Escherichia coli) bacteria to verify its antimicrobial potential.
Discussion
We have developed an SVM model to identify anticancer peptides from the AUC Red Sea metagenomics library. We generated a list of 59 potential anticancer peptides, each with a model-performance score, using our SVM. We then selected the best candidate peptide using cationicity and HMM domain as the main criteria and introduced serial modifications outside of the homeodomain region to enhance the model-performance score. The modifications did not significantly decrease the predicted blood-borne half-life.
It is likely that our SVM model did not initially determine any peptides to be potential ACPs because the variations between ACP and non-ACP groups were minimal. Thus, although the model could discern ACPs from AMPs, due to the distinct properties of ACPs, a larger and more diverse dataset was required to produce a clear distinction between these groups and to train the SVM to recognize anticancer peptides from a pool of random peptides.
The DELTA-BLASTp and HMMER alignments showed that our peptide aligned with the homeodomain region of several homeoproteins from the MEIS, PBX, and SIX families, corroborating the previous finding that our peptide contained a homeodomain, suggesting that the peptide would exhibit similar behavior to homeoproteins.
The predicted secondary structure of our peptide revealed two α-helical regions with architecture consistent with the DNA binding region of the homeodomain. Ligand-binding prediction further supported that our peptide would bind to DNA fragments and GO-term annotation implicated the peptide in pathways involving direct sequence-specific binding of DNA and modulation of gene transcription.
The MTT assay revealed our peptide to elicit a cytotoxic response in SNU449, HepG2, and SKOV3 cells. Treatment of HeLa cells caused the formation of numerous large vacuoles, with partial cytotoxicity, while human noncancerous 1BR-hTERT cells exhibited a cytotoxic effect in response to the peptide, although this was attenuated compared with SNU449 or HepG2 cells. The high IC50 value calculated from 1BR-hTERT cell assays demonstrates that our peptide may act selectively toward cancer cells, with a minimal effect on non-cancer cells. The Hill-slope values were all above 1, suggesting that a positive cooperative feedback loop of peptide binding existed.
Cisplatin is an anticancer control drug and was used as a positive control in the present study. The cytotoxic effect of cisplatin toward SNU449 cells was significantly greater than that of our peptide. The IC50 for the SNU449 cells was calculated to be 88.4 μM, compared with 12.8 μM for cisplatin. Considering treatment of SNU449 cells, the higher peptide IC50 at 48 h compared with 24 h suggested that peptide molecules were consumed by cells at 24 h and surviving cells were able to recover and proliferate. Thus, higher concentrations of the peptide would be required to elicit an effective dose–response.
The peptide caused an irreversible morphological change in SNU449, HepG2, SKOV3, and 1BR-hTERT cells after 24 h of exposure. SNU449 cells transitioned from an epithelial-diffuse cell structure to a more compact circular structure, with cells releasing vacuoles resembling autophagosomes and partial rupture and detachment observed. The changes in cellular morphology preceded cell death, which occurred via multiple death pathways. Vacuole formation was also observed in treated HepG2 cells. A major loss in the normal cellular topology was identified in treated cells, most dramatically in SKOV3 cells, which became significantly more fragmented and deformed.
The IC50 values of our peptide were compared to other conventionally used chemotherapeutics. However, chemotherapeutic drugs were extremely cytotoxic to noncancerous drugs; thus, administration at low concentrations was vital to maintain cell survival. Other experimentally validated peptide drugs have shown a wide array of concentrations [
35] that also have an antimicrobial slant. Our results have shown that peptide treatment was significantly more effective in stimulating a cytotoxic effect on cancer cells, between 14 – 200% depending on cell type, compared to noncancerous fibroblasts.
A low concentration gradient was tested and optimized on HepG2 cell lines (with IC50 calculated in (Fig.
4). Knowing the invasive properties of SNU449 compared to HepG2 cell lines based on previous studies [
36,
37], a higher concentration gradient was developed to characterize the peptide mode of action on SNU449 mesenchymal-like subgroups of hepatocellular carcinoma when compared to HepG2, the less invasive hepatoblast-like subgroup of HCC. Using this higher gradient allowed us to evaluate the peptide cytotoxic effect and potential mode of action. This gradient was also tested on 1Br-hTERT cell lines representative of normal cell lines, showing minimal cytotoxicity.
We determined the primary cell lines of interest based on the peptide dose-dependent response, with the IC50 values of SNU449 and SKOV3 cells being the highest and lowest, respectively, compared with other tested cells. The peptide caused drastic changes in cellular morphology and shrinkage among cells on the sides of the scratches and inhibited SNU449 cell migration, resulting in the enlargement of scratches compared with untreated controls at both 24 h and 48 h of exposure. Additionally, SKOV3 wound closure was severely reduced following peptide treatment. Thus, treatment with the peptide at the IC50 concentration inhibits cellular migration and proliferation of SNU449 and SKOV3 cells with expression of
KI67, a proliferation marker, decreased in the latter. The unchanged EMT and autophagic genes transcription profiles of SNU449 suggested that changes occurred at the protein level, including epigenetic regulation of these genes [
30,
31,
38]. In contrast, expression of
KI67 and
B-catenin (adhesion marker) was increased in SKOV3 cells following peptide treatment, as was expression of the EMT marker
vimentin and autophagic markers
ATG5 and
ATG6. No expression of
ATG7 was detected. These differences may indicate that exposure of SKOV3 cells to the peptide reduced gene expression associated with proliferation, cell adhesion, and autophagy while increasing
vimentin expression [
39]. Further study of the expression of EMT markers and other relevant downstream target proteins is warranted to verify the implications of our findings in relation to the treatment of cancer cells.
Apoptosis is characterized by morphological changes, nucleic acid fragmentation, and loss of membrane asymmetry [
40]. The annexin V assay revealed a clear bias toward early apoptosis for peptide treated SNU449 cells, and found to be highly presented mode of cell death upon peptide treatment. In treated cells, 51% of cells were in early apoptosis in comparison to the untreated control displaying only 6% of the cells in early apoptosis, while 18% of the cells were in late apoptotic/necrotic. This indicated that peptide treatment had high percentage of different stages of apoptosis, suggesting apoptosis as an important cellular death pathway to further explore. The differences between the fields of untreated control and treated cells show that peptide treatment drives cells into apoptotic programmed cell death after 24 h exposure, due to cellular presentation of phosphatidylserine (PS) on the outer plasma membrane. Impact of peptide treatment on cellular viability, morphology and migration; demonstrated anticancer properties on SNU449 cells. The proportion of cells in each cellular death phase, early and late apoptosis, confirmed that apoptosis was the most prevalent mode of cellular death following treatment [
41].
Hemolytic analysis indicated that our novel peptide did not cause normal erythrocytes to rupture during the incubation period, in agreement with other reports of similar peptides [
42], suggesting that the peptide’s cytotoxicity was higher in relatively anionic cancer cells than normal neutral cells. Therefore, we propose intravenous administration would be appropriate for in vivo investigations. The blood-borne predicted half-life of the peptide is similar to the 20—30 min half-life of the platinum-based drug, cisplatin [
43], indicating that it would persist within the bloodstream. The Hemolysis Assay was used to validate the potential application of the peptide in vivo through intravenous application and subsequently identify itseffect on RBCs as representatives for normal primary cells. Thr IC50 and the IC25 showed negligible cytotoxic effects of the peptide (Fig.
8). The promising results from cell viability, proliferation, morphology, and migration analyses, and the minimal hemolytic activity of our peptide support its application as a novel anticancer therapeutic [
44].
Antimicrobial peptides (specifically, host defense peptides) are part of natural host defense mechanisms against pathogens in the innate immune system of all species [
39,
40,
45]. Most previously studied AMPs elicit broad-spectrum effects on all microbes including bacteria, viruses, and fungi [
10]. The antibacterial effects of our novel 37-mer peptide toward
S. aureus [
46] and
E. coli [
47] which are part of the normal human flora, colonizing the skin and most of the mucosal membrane microbiota and the gastrointestinal tract, respectively [
41,
42] highlight a potential consequence of the peptide upon administration to humans. However, the IC50 of our peptide was much lower in all tested cell lines than the concentrations we tested on both bacterial strains; thus use of the peptide as a therapeutic agent may have a marginal effect on the normal flora. The negligible hemolytic activity of the peptide toward normal erythrocytes suggests that intravenous administration would be the preferable mode of administration.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.