Clostridium difficile is a multi-strain, spore-forming, Gram-positive anaerobic bacterium posing a global threat to post-operative individuals. Infamously known for antibiotic-associated diarrhoea, it is one of the most important causes of healthcare-associated infections worldwide, leading to a quarter of reported cases of infectious diarrhoea and a broad spectrum of gastrointestinal complications, including sepsis and pseudomembranous colitis (Barbut and Petit, 2001). Studies have suggested that it is a crucial part of healthy human gut flora as it overgrows and imbalances intestinal microflora with unnecessary antibiotic therapies (Abt et al., 2016). With the progression of antibiotic-based therapeutics accompanied by sub-standard hygiene in hospitals, the incidence of C. difficile infection (CDI) has significantly increased since the 20th century (Czepiel et al., 2019). Being a major causative pathogen, C. difficile contributes to almost half a million cases with 29,000 deaths per annum in the United States alone and impacting Latin America, Europe, and the Asian regions (Goudarzi et al., 2014; Lessa et al., 2015). Whereas in India, the incidence and prevalence rates of CDI-associated diarrhoea in hospitalised patients ranges from 3 to 29% and 7.1–26.6%, respectively (Segar et al., 2017).

C. difficile possesses a huge, diversified pangenome with high levels of evolutionary plasticity accumulated over time due to gene flux and recombination in response to environmental changes. Literature suggests the evolutionary rate of C. difficile to be 3.2 × 10^–7 mutations per nucleotide per year, resulting around 1.4 mutations per genome per year that drives and reshapes the genetic diversity of the pathogen (Didelot et al., 2012). Additionally, the ratio of the nucleotide substitution rate to result of mutation (r/m) has been estimated around 0.2 or higher. These rates are comparatively lower to other guts pathogens (He et al., 2010). C. difficile infection involves an opportunistic colonisation of the intestinal tract leading to nosocomial, antibiotic-associated severe diarrhoea with or without colitis, fever with chills, and abdominal pain (Bartlett, 2002; Korman, 2015). C. difficile infection occurs via transmission of spores that are resistant to acid, heat and antibiotics. Antibiotics like metronidazole and oral vancomycin have been recommended as a cure for the acute infection. Other antibiotics including penicillin, tetracycline, and clindamycin, have been employed to control CDI, but studies have suggested that imprudent overuse has led to the development of antimicrobial resistance in C. difficile strains (Nelson et al., 1994). Current treatments for CDI consist of supportive care, discontinuation of unnecessary antibiotics and specific antimicrobial therapies.

Furthermore, novel methodologies including fidaxomicin therapy, and faecal microbiota transplantation-mediated therapy have shown prominent results. Faecal microbiota transplantation has shown significant efficacy to overcome CDI and reduce its recurrence (Goudarzi et al., 2014). The appearance of hyper-virulent antibiotic-resistant strains with the production of antimicrobial peptides from activated immune cells and inflamed epithelial cells allows the residual C. difficile to re-expand, following the end of treatment (Vindigni and Surawicz, 2015). Growth and development of the bacteria can be prevented at the genetic level effectively, by reducing the prevalence of C. difficile and also limiting the rates of recurrence (Leber et al., 2017). The pathophysiology of the disease, including the transmission and physicochemical pathways employed by C. difficile, has aroused several researchers in the past few years to investigate the proteins involved in their virulence (Rineh et al., 2014; Smits et al., 2016).

Advanced high-throughput technologies like genome sequencing, gene editing, and functional annotation might be helpful to understand the biology of C. difficile and its genomic composition. In the recent past, domains like genomics, transcriptomics and proteomics studies have assisted in gaining insights to the mechanism of microbial adaptation (Sebaihia et al., 2006; Stabler et al., 2009; Boetzkes et al., 2012; Cafardi et al., 2013). Moreover, there are still challenges while decoding these mechanisms, and the bioinformatics tool aids in our understanding via functional annotation, protein-protein interactions, and pathway analysis. Functional annotation of uncharacterized proteins is a crucial step in deciphering the role of proteins. An uncharacterized or hypothetical protein (HP) is defined as the one that is predicted to be expressed in an organism but no proper function is known (Suravajhala et al., 2015). Most of these hypothetical proteins are expected to play essential roles and their annotation can unveil novel functional pathways. The utilisation of in silico approaches to predict annotations of HPs has been successful in numerous bacterial species (Singh et al., 2015; Varma et al., 2015; Prabhu et al., 2020). Additionally, there are still several challenges in annotating such proteins, given the scanty organelle information known and the pervasive nature of the subcellular location of these proteins. Earlier, methods to annotate HPs by us (Ijaq et al., 2019) could be useful, but given the bacterial system, a coherent need for employing several computational tools, viz. determine the conserved domain, subcellular localization, secretory nature, physicochemical characterization, identification of prophage sequence and CRISPR-Cas9 system, detection of antimicrobial resistance, comparative homology analysis, virulence factors, antigenicity analysis, allergenicity determination along with structure prediction and binding-site analysis would allow us to annotate the possible functions for the HPs. Deciphering the role of complete gene coding regions in the genome is crucial to merge the gaps in the proteome to fully understand the pathogenicity. This study aims to determine the functional annotations of HPs of C. difficile to have a clear implication.

Understanding the genomic epidemiology and annotating proteins remains the most impactful strategies to detect, characterise, and monitor pathogens that impact human health. Though traditional biochemical and molecular experiments can be used to assign proper functions for genes, they are expensive, tedious, and have resulted in only 50–60% gene annotations (Sivashankari and Shanmughavel, 2006). Despite continuous research efforts, a large portion of the proteome is still represented by uncharacterized proteins designated as HPs. They are predicted from the nucleic acid sequences and have unknown functions (Varma et al., 2015). Thus, automated gene/protein annotation using bioinformatics tools can overcome this challenge of the post-genome era where complete genome sequences of several organisms are available. Several researchers have gradually worked on structural or functional annotation of HPs from different microbes like Staphylococcus aureus, Vibrio cholera, Blumeria graminis and Serratia marcescens and others (Islam et al., 2015; Varma et al., 2015; da Costa et al., 2018; Prabhu et al., 2020). But still, they are shrouded in the mystery and there is a dire need to expedite the process (Omeershffudin and Kumar, 2019). Some of them include deriving information from sequence similarity analysis, interaction of proteins with other proteins or ligands, gene expression profiles, conserved domains/motifs, phylogenetic analysis, phosphorylation regions and active site residue similarity analysis. The most orthodox way of speculating protein function involves sequence similarity analysis using BLAST tool (Varma et al., 2015).

Quite a few studies have elucidated the genomic epidemiology of C. difficile, but none of them have focused on the HPs (Cafardi et al., 2013; Ezhilarasan et al., 2013; Basak et al., 2021). Recently, Basak et al., have characterised an in silico vaccine using the immunoinformatics approaches via utilising CotE, SlpA and FliC proteins, which were responsible for gastrointestinal tract colonisation, TLR4 interaction, cytokine production, and plays a major role in the adherence of the bacterial cell, which ultimately triggers the innate immune response (Péchiné et al., 2005; Hong et al., 2017; Mori and Takahashi, 2018). Another study has also emphasised on extracellular factors involved in the pathogenesis of C. difficile. A unique HP named, CD630_28,300 was found to share sequence similarity with zinc metallopeptidase, which demonstrated the binding of zinc with CD630_28,300 and its ability to disrupt the human fibronectin network by cleaving the fibronectin and fibrinogen in vitro in a zinc-dependent manner (Cafardi et al., 2013). Researchers have also employed similarity searches between pathogen and host, essentiality analysis, metabolic functional association, and choke point analysis. They identified 19 promising drug targets which were non-homologous to host proteins, and participated in four pathogen specific pathways, of which the peptidoglycan biosynthesis was found to be the highest contributor to the list of potential target proteins. MurG enzyme from the peptidoglycan biosynthesis pathway was found as one of the potential targets (Ezhilarasan et al., 2013).

In this study, the C. difficile HPs identified from NCBI database were subjected to various in silico experiments like, physicochemical properties, subcellular localisation identification, transmembrane helices detection, comparative homology analysis, antigenicity analysis, allergenicity analysis, secretory nature detection, and AMR identification. Furthermore, the top antigenic HPs were shortlisted and subjected to structure prediction and binding site analysis. The analysis of six Clostriodioles strains resulted in identifying approximately 11% of HPs from around 3,500 proteins coded by nearly 4.1 Mb genome size of each strain. The physicochemical properties of HP from 6 C. difficile strains, namely, BR81, R20291, CF5, M120, 196, and 2,007,855, were analysed. In all these strains of C. difficile, approximately 70% of the HPs were found to be stable as they had instability index (II) value below 40. Isoelectric point (pI) and grand average of hydropathicity (GRAVY) value were other important physicochemical parameters that were determined. The pI of HP ranges from 4.05 to 11.99 in all the strains. Isoelectric point (pI) is that pH where the amino acid of protein has a net zero charge and hence does not move in a direct current electrical field. At pI solubility of protein is lowest and electro focussing system mobility is zero, thereby making proteins stable and compact at this pH. This information can be utilised to develop buffer system for protein purification by isoelectric focussing (Islam et al., 2015). The GRAVY number of proteins is the measure of its hydrophilicity or hydrophobicity which are combined in a hydropathy scale. A positive value indicates proteins are hydrophobic while a negative value indicates that they are hydrophilic (Chang and Yang, 2013). In present study, around 70% HP of these strains had negative GRAVY values, demonstrating that they are hydrophilic in nature.

Since proteins located on the cell membrane can act as potential vaccine targets and those in the cytoplasmic matrix can act as potential drug targets, therefore, knowledge from subcellular localization is an important parameter for functional characterization of a protein (Prabhu et al., 2020). Moreover, research suggests the role of cell surface proteins in Clostridial pathogenesis, yet not many cell surface or secreted proteins of the nosocomial pathogen C. difficile have been identified or functionally characterised (Cafardi et al., 2013). Protein subcellular localization of HPs from all six strains of C. difficile were examined by PSORTb tool which categorises them into four categories, namely, cytoplasmic, cytoplasmic membrane, extracellular and unknown, based on their location in the bacterial cell. Approximately, 32–38% and 23–27% HPs were localised in the cytoplasm and cytoplasmic membrane, respectively. Meanwhile, 1–3% and 37–40% of all HPs in these six strains were located in the extracellular space, or their location was unknown.

Prediction of signal peptides is a key feature to determine the transportation system of particular proteins and their cleavage site. All non-cytoplasmic proteins have signal peptides that facilitate the transport of proteins across the membrane to a designated cellular location or organelles (Prabhu et al., 2020). We have used SignalP 5.0 server and found that almost 87–92% of HPs from each strain did not have signal peptides while the remaining 8–13% proteins had signal peptides indicating their involvement in a secretory pathway. Membrane proteins are also involved in various biological processes like signalling, transport, energy transduction and pathogenesis and can act as potential drug targets. Thus, it is important to predict membrane proteins to develop potent drug molecules (Prabhu et al., 2020).

Consequently, to check whether these HPs C. difficile can act as potential vaccine targets, we evaluated their sequence homology with humans, virulence factor, antigenicity and allergenicity. For a protein to be considered a potential candidate, it should be non-homologous to human proteins to avoid cross-reactivity with them. Also, they should be antigenic, non-allergenic and can have presence of virulence factors, our study demonstrated that almost 99.7% HP from all strains were non homologous but only a minute fraction of them around 0.25–1.45% has virulent factor. While around 36–41% of all HP were antigenic and none of them were allergens. Additionally, AMR analysis identified that the glycopeptide resistance protein vanZ1 gene, glycosylating toxin TcdB gene, holin-like glycosylating toxin export protein TcdE, glycosylating toxin sigma factor TcdR, and CDD family class D beta-lactamase blaCDD genes were common in all strains according to the AMR and virulent factor data generated by AMRFinderPlus for WGS sequences, of which the vanZ1 and the blaCDD genes were AMR, rest of the above mentioned genes were virulent. The AMR genes identified by ResFinder in only the WGS strains of M120 and 2,007,855 were also identified as AMR genes by the AMRFinderPlus. In the protein sequence from RAST three genes vanZ1, tcdB, and blaR1 were found in all six chains, of which the tcdB is a virulence factor. In the HPs from RAST, the AMRFinderPlus was able to identify genes with virulence factors and the tcdR gene was the common gene in all six chains. The HPs play a role in virulence and could be used as potential targets for drug discovery or as antigen to develop vaccines. These properties of virulence, stability, polarity, presence in cytoplasmic and extracellular regions, minimal number of transmembrane helices present, non-homology with the human proteome, antigenicity and non-allergenicity which these HPs show, further experimental and computational studies can be done to assess the potentiality of these HPs as targets for drug discovery and as vaccine candidates.

Identifying protein functions is crucial for understanding various biological processes. Here, we implemented in silico approaches to predict the function of HPs from six strains of the C. difficile. While employing various tools to annotate and characterise HPs, characteristic predictions like subcellular localization, secretory nature and physicochemical properties were suitable to understand particular features. Further, identification of AMR genes, prophage sequences, CRISPR-Cas9 genes, and elucidation of immunoinformatics properties has allowed us to identify proteins that might play important roles in the complex mechanisms and biological processes. In conclusion, the pipeline used in this study allowed us to screen and characterise candidate HPs for assigning protein functions and further broaden up the possibilities for better downstream validation.