Helicobacter pylorus is one of the pathogens involved in various gastric diseases such as peptic ulcers, chronic gastritis, mucosa-associated lymphoid tissue lymphoma, and gastric cancer [1–3]. Infection with H. pylori is associated with an increased risk of gastric adenocarcinoma and has attracted attention as a cofactor in the pathogenesis of this malignant condition [4]. Moreover, the risk of developing cancer is related to the physiologic and histologic changes induced by a H. pylori infection in the stomach [5]. Despite a general decline in the incidence of gastric cancer, it remains the fourth most common cancer and second leading cause of cancer-related deaths worldwide [6]. However, most H. pylori infections do not cause cancer. The sporadic distribution of the disease caused by H. pylori looks to be dependent on host-related factors: the host (human individual) genetics controlling the inflammatory response, the age when the H. pylori infection was acquired, poor nutrition, storage of food, and the pattern of food consumption can be considered as host-related factors [7–9].

In addition, bacterial factors associated with the risk of gastric cancer are also emphasized, and molecular and cell biology approaches aimed at understanding the interaction between H. pylori and transforming epithelial cells have been carried out. Since H. pylori is a highly heterogeneous bacterial species, both genotypically and phenotypically, and is highly adapted for survival in the gastric niche, it is not easy to figure out the major bacterial factors that are directly associated with etiopathogenesis [10,11]. Based on the current knowledge, several virulence factors such as genes within the cag (cytotoxin-associated antigen) pathogenicity island, including the gene encoding the CagA protein, as well as polymorphic variation in the VacA vacuolating exotoxin and the blood group antigen binding adhesions, BabA and SabA, are regarded as possible bacterial factors [6,10,12]. A duodenal ulcer-promoting gene (dupA), located in the “plasticity region” of the H. pylori genome, was reported as a potential virulence marker [10,13]. Other bacterial factors such as peptidoglycan, lipopolysaccharide(LPS), γ-glutamyl trans-peptidase(GGT), and protease HtrA may be linked to pathogenicity [14].

Although a huge amount of biological data on H. pylori has been accumulated, enzymes or proteins of unknown function still make up more than a third of the open reading frames (ORF) of H. pylori. An unknown protein could be defined as a protein whose function has not yet been characterized, and a hypothetical protein could be defined as a protein that is supposed to exist in an organism although its existence has not been shown experimentally. Therefore, in a broad sense, hypothetical proteins could be included in unknown proteins. To completely understand the pathogenic mechanism of H. pylori, it is very important to elucidate the functions of these unknown proteins. To fill in the “missing parts list” is accordingly one of the greatest challenges for post-genomic biology, and a tremendous opportunity to discover new biological and pathogenic machinery in H. pylori.

The sequencing of the H. pylori genome started in 1997 with the H. pylori strain 26695 [15]. It was isolated from an English patient with chronic gastritis. The chromosome of strain 26695 is circular and composed of 1.67 mega base pairs (Table 1). The average G-C content is approximately 38.9% and the genome has 1590 open reading frames (ORF) that are possibly protein-coding loci [1], together with the RNA coding genes (2 copies of 16S rRNA and 23S rRNA genes, 36 tRNA genes). From the following analysis of the same genome, it was suggested that a smaller number of ORFs is in the sequence of strain 26695 [16].

Ongoing studies have found genes that were missing in previous analyses, as in the case of SecE. A general secretion machinery is widely present in bacteria, which functions in the secretion of outer membrane proteins to extracellular environments [18]. From the first annotation results, it was thought that strain 26695 had only a partial general secretion machinery because it lacked SecE [15]. A new small open reading frame between nusG and rmpG (HP1203–HP1204) in the genome sequences was found using an ab initio server, GeneMark, Glimmer, and BlastX [19]. It has a high homology and structural similarity to the SecE protein in related bacteria implying that strain 26695 has a complete general secretion machinery. In addition, small RNA genes are universally present in bacteria [20]. The tmRNA gene (ssrA) has been found in H. pylori, encoding a functional RNA molecule and a small peptide involved in the quality control of translation [21]. In addition, the H. pylori strain contains a sRNA gene encoding the RNA component of RnaseP and the 4.5S RNA gene which is involved in secretion [22,23].

In 2008, the adaptations of H. pylori to a rarely captured event in the evolution of its impact on a host biology were characterized by defining the impact of these adaptations on an intriguing but poorly characterized interaction between this bacterium and gastric epithelial stem cells [24]. H. pylori HPKX_438_AG0C1 and HPXK_438_CA4C1 were isolated from a single patient who progressed from ChAG (chronic atrophic gastritis) to adenocarcinoma using a population-based endoscopy study. ChAG-associated Kx1 and Cancer-associated Kx2 genomes were analyzed to examine the adaptation of H. pylori, respectively. Micro-arrays gave a comprehensive view of the genome diversity of the H. pylori pathogen. This was performed with information on the origin of the hspA together with glmM alleles revealing that H. pylori infection may be acquired by more diverse routes than previously expected [25]. According to cluster analysis, isolates from family D belonged to three different strains, those from family L consisted of two strains, and those from family A were grouped into at least 5 strains. Strains from family D and family L differed by the presence/absence of 24 to 42 CDSs (coding sequences). In family A, one strain was difficult to define due to the small differences in gene profiles between neighboring branches.

In 2009, the complete genome sequence of H. pylori G27 was reported [26]. The G27 strain was originally isolated from an endoscopy patient from Italy [27]. The genome consists of a single circular chromosome with about 1.65 mega base pairs (Table 1) that is AT rich (61.6%), contains 1515 ORFs, and is similar in size and composition to the other published H. pylori genomes of strains 26695, J99, and HPAG [15,16,28]. The G27 strain contains 58 genes that are not found in 26695, J99, or HPAG, as defined by a blastp hit. The majority of these G27-specific genes are predicted to encode hypothetical proteins [26].

In the same year, the genome sequences of two H. pylori strains were analyzed [29]. H. pylori strain 98-10 was isolated from a patient with gastric cancer and strain B128 was isolated from a patient with gastric ulcer disease. Strain 98-10 was most closely related to H. pylori strains of East Asian origin and strain B128 was most closely related to strains of European origin. Strain 98-10 contained multiple features characteristic of East Asian strains, including a type s1c vacA allele and a cagA allele encoding an EPIYA-D tyrosine phosphorylation motif.

Very recently, several genome sequences of different strains were reported accelerating H. pylori genomic and proteomic research [30–38]. Strain 908 is a close relative strain of J99 [39] and was isolated from an African patient living in France, who suffered from duodenal ulcer disease [40]. The B8 strain consists of about 1.67 mega base pairs and a small plasmid of about 6000 base pairs carrying nine putative genes. Interestingly, the B8 strain contains coding sequences, 293 of which are strain-specific, coding mainly for hypothetical proteins with unknown functions [31]. Similarly, the P12 strain contains plasticity zones, encoding for the type IV secretion system and having the typical properties of genomic islands [32]. Another sequenced genome, the Shi470 strain known as the Shiimaa village strain was more Asian- than European-like genome-wide, indicating Amerind ancestry. This strain contains two unique cagA virulence genes and a novel allele of gene hp0519 encoding host tissue interaction protein [33]. There are several H. pylori populations such as hpAfrical, hpEurope, hspEAsia, and hspAmerind because this bacterium has colonized the stomach since early in human evolution and diverged with ancient human migrations [41–43]. One of these populations, the hspAmerind strain V225d, was cultured from a Venezuelan Piaroa Amerindian subject and identified. The V225d strain is cag-positive encoding a multifunctional effector protein injected into host cells by the cag type IV secretion system [34]. Two strains, 2017 and 2018, are the chronological subclones of strain 908 and cultured from the antrum and corpus, respectively. Using comparative genomic analysis [35,37], these two strains are almost identical and descended from the genome of strain 908 [30,36]. The B45 strain was sequenced from a gastric mucosa-associated lymphoid tissue (MALT) lymphoma patient and induced an integrated prophage in this strain by UV irradiation [38].

The Comprehensive Microbial Resource (CMR) is a free tool that allows researchers to access all of the publicly available bacterial genome sequences completed to date [44] (Figure 1). Currently, it provides genomic sequences of three strains of Helicobacter pylori (26695, HPAG1, J99).

As in the case of other genomic research, Structural Genomics Initiatives are mainly responsible for determination of H. pylori protein structures. These initiatives, together with the structure determination of known proteins, have made enormous strides in the elucidation of unknown protein structure of H. pylori [15,16,24–26,28–38,45–47]. The available structural data have already led to the identification of potentially new drug targets [48] and has been helpful in assigning functions to proteins of which the functions were previously unknown [49,50].

The increase in structure determination for H. pylori has been triggered by the sequencing of the H. pylori 52 and 26695 genomes [15,25,45,47]. The genome sequences and their protein structures yielded many clues to help understand the pathogenesis of H. pylori. Approximately 14% of Lyase structures have been determined and represent the largest proportion of any functional class of which the structures have already been solved (Table S1).

The sequencing of the genome led to a dramatic increase in the number of known structures for H. pylori proteins deposited in the Protein Data Bank (PDB) (Figure 2). The first H. pylori protein structure was determined in 2001 (PDB ID: 1G6O) [51]. In the following four years, 32 more structures were reported (Figure 2). After several sub-species genome sequences of H. pylori became publicly available, the number of structures determined after 2005 increased sharply and at an increasing rate.

Usually, protein solubility is one of the main bottlenecks in structure determination [53]. In the case of H. pylori, methods have already been developed that remedied this problem, such as the development of customized expression strategies for H. pylori proteins in Escherichia coli [54]. The increase in determined structures is also due to the development of improved methods for high-throughput X-ray crystallography. However, the major driving force for this increase was the availability of genome-wide sequence data in the early 2000s.

There are currently 79,356 structures in the PDB as of 14 February 2012, of which 0.35%, a total of 279, are structures of H. pylori proteins. Of these proteins, 28 are unknown in function, which represents 10.03% of the determined H. pylori structures (Table 2).

A complete list of H. pylori protein structures deposited in the PDB is given in the Supporting Information Table S1. The predominant method used to determine these structures was X-ray crystallography, which accounts for 261 of the total number of H. pylori structures currently determined (Figure 2). A further 18 were elucidated by solution-state NMR spectroscopy. Most structures are of individual proteins, although many are bound by small molecule ligands such as substrate analogues and only 11 protein-DNA complexes have been determined (Figure 3, Table S1).

The most typical approach of predicting the function of an unknown protein is to use sequence similarity by finding a similar protein of known function [56]. Based on sequence-similarity, a predictor assigns the known function to the inferred protein. Actually, the functions of enzymes tend to be conserved if they share more than a 40%–50% sequence identity. The sequence-based approach is reasonable, however, approximately 50% of the unknown proteins from a newly sequenced genome could not be assigned to their function using only sequence-similarity approaches [57] (Figure 1). The low efficiency of the sequence-similarity search may be partly caused by gene sequences that are quite new and genes that have not yet been characterized in other bacterial systems. To overcome the weakness of sequence-similarity searches, several trials were employed using so called “similarity free” methods [57]. The methods use physicochemical properties and secondary structure of proteins. Bioinformatics developed the methods and there have been successful cases for characterizing function or structure [58–60]. However, the methods need to be improved since similarity-free methods still depend to a certain extent on similarity.

Another approach to identify function is to use 3D structures. This approach often succeeds in cases where sequence-based methods fail. This may be due to the idea that in many cases evolution retains the folding pattern long after the sequence similarity becomes undetectable. Structural similarity searches use the global fold of the protein [61–64] or detect the functionally important regions of the protein [65–69]. Since structures diverge more slowly than sequences, a sequence comparison may be less sensitive than a structure comparison [70]. However, the structural comparison still has the limitation of false positives being reported and needs to be improved to overcome overestimation of statistical significance like sequence-similarity searches [70]. This means that experimental confirmation is still required for exact assignment of function to an unknown protein.

Some examples of functional elucidation of unknown proteins from H. pylori are provided below. For estimation, we generally conducted four steps: (i) structure determination; (ii) sequence homology search using PSI-BLAST [71]; (iii) structural homology search using the web server DALI [62]; and (iv) experimental confirmation of the function.

Bioinformatics tools have been remarkably developed, providing biologists valuable information for functional elucidation. Nevertheless, prediction of protein function from sequence and structure is a difficult problem, because homologous proteins often have different functions. In addition, the protein that is classified into a known protein, based on the sequence homology, often shows some functional ambiguity since the composition of the operon is quite different from that of the known system. In addition, some of the proteins, which are considered to be well characterized, may have additional functions beyond their listed function [102]. In this regard, it is still worth investigating known proteins from a newly sequenced genome for their cell and biological functions. Here, we present two examples of well-defined proteins that have different characteristics compared to the homologues.

Copper metabolism by copper chaperones has been studied extensively in both eukaryotes and bacteria. In the gram-positive bacterium, Enterococcus hirae, the cop operon is composed of four proteins: two integral membrane P-type ATPases, CopA, and CopB which transport Cu(I) into cells under Cu(I) limiting conditions and eliminate Cu(I) under conditions of high Cu(I) levels, respectively [103,104]. The imported copper ions are transferred from CopA to the CopZ chaperone [105–107] and CopY, a gene repressor, is released from the cop operon promoter when Cu(I) is delivered to CopY by the copper chaperone, CopZ (Figure 8A). In the case of the gram-negative bacterium, H. pylori, copper homeostasis seems to be maintained by only two proteins CopA and CopP (HP1073). The H. pylori cop operon (Figure 8A) is included in a novel stress-responsive operon (sro), which encodes the flagellar motor switch protein CheY, the putative methyltransferase Hsm, the cell division protein FtsH, the putative phosphatidyltransferase Ptr, the heavy metal-binding proteins CopA and CopP, and an open reading frame of unknown function [108]. CopA is a member of the bacterial copper ion ATPase family, and CopP, which is homologous to E. hirae CopZ, is a putative copper binding regulatory protein of 66 amino acids [104,108]. CopA of H. pylori was identified as a Cu(II) export ATPase [109], which shows that its biological role is more similar to that of E. hirae CopB, rather than CopA [110]. Moreover, the CopP gene resides immediately downstream of the CopA gene, while the E. hirae CopZ gene resides upstream of the CopA gene. Therefore, the cop operon organization seems to be evolutionarily modified in each bacterium.

Generally, CopZ proteins share a conserved structure, βαββαβ with a similar metal binding region. Interestingly, HpCopP adopts the βαββα fold with a missing C-terminal β strand [111]. The overall topologies of the secondary structural components are very similar between the CopZs and HpCopP, while some variations in the loop regions appear (Figure 8). The relationship between the unusual fold and the copper specificity was evaluated [111]. We showed that HpCopP was not adequate for Cu(II) binding since the fold stability decreased in the presence of Cu(II) ion, suggesting that the structure of HpCopP is optimized for the transfer of toxic Cu(I). The absence of the C-terminal β-strand may lead to decreased conformational stability of loop I including the CXXC motif (Cu binding motif), which probably contributes to the disulfide bond formation between the two cysteine residues in the presence of Cu(II) ion. These findings should be helpful in evaluating the copper metabolism related with HpCopA and HpCopP in H. pylori.

Acyl carrier protein (ACP) found in bacteria is a monofunctional protein, that is, a type II enzyme in fatty acid biosynthesis. All the ACPs are decorated by acyl carrier protein synthase (ACPS) with fatty acids, which are covalently attached as thioesters to the 4′-phosphopantetheine prosthetic group at highly conserved Ser 36 [112]. Fatty acid binding has little influence on ACP conformation under physiological conditions [113], but it stabilizes ACP against denaturation at alkaline pH [114].

H. pylori ACP (HP0559) is composed of 78 amino acids with a pI value of 3.9, and its primary structure is similar with those of homologous ACPs. Like other ACPs, HpACP forms a helical bundle structure through hydrophobic contacts between the helices (Figure 9). However, we found an unusual behavior of HpACP at neutral pH [115]. HpACP exists as a partially unfolded state at neutral pH, which is a unique characteristic of HpACP (Figure 9). In contrast, the overall helical structure of E. coli ACP was maintained at pH 7 [116] and Vibrio harveyi ACP exhibited a random coil-like conformation at pH 7 [117].

The pH dependent-conformational change of a protein from H. pylori is a very interesting feature, considering that the environment of the stomach has a low pH. A few studies showed the relationship between the mutation of various residues and the pH-dependent structural stability. The mutation of Val 43 to Ile in E. coli ACP increases the stability to pH-induced expansion in electrophoretic systems, concomitantly inducing more compact folding [118]. The mutants F50 A and I54 A of V. harveyi are incapable of adopting a native conformation with increased hydrodynamic radius at neutral pH [117]. In addition, a few basic residues scattered near the N- and C-termini, for example, His 75 of E. coli ACP, are necessary for ACP to maintain a native conformation at neutral pH [119]. Through our structural analysis, we found that several hydrophilic residues (Glu 47, Asn 75, and Lys 76) play an important role in structural stability. Therefore, we could suggest that, unlike other ACPs, the helical bundle of H. pylori ACP is maintained by, not only hydrophobic interactions, but also by hydrophilic interactions and these interactions may be weakened by elevation of the pH because the exchange rate of protons attached to the side chain amide of Asn and Lys may increase [115].

Mass genomic sequencing has been yielding many protein sequences that cannot be annotated, and structural genomics projects are yielding many protein structures that have unknown functions. Unknown proteins represent up to about half of the proteins in prokaryotic genomes, and much more than this in higher plants and animals [120]. In bacteria such as H. pylori, 30–40% of the proteins encoded by typical bacterial genomes have no clear known function [121]. Thus, a major issue of genomic studies may be to narrow the gap between the richness of sequences (and/or structures) and functional characterization as subsequent experimental investigation is costly and time-consuming [122]. Actually, only 54% of E. coli gene products have been experimentally investigated so far [123]. Therefore, more robust bioinformatic methods or approaches may be necessary to overcome this situation. Here, we showed several examples of successful cases for elucidating the function of H. pylori unknown proteins based on their structural information, which supports the potential of structural comparison for functional identification. It is hoped that the structural comparison can at least act as a guide to the possible function, even though all structures cannot elucidate the actual function.