Consensus computational network analysis for identifying candidate outer membrane proteins from Borrelia spirochetes

Pathogenic spirochetes belonging to the genus Borrelia cause Lyme disease and relapsing fever, both of which are vector-borne illnesses. Lyme disease is caused by pathogenic spirochetes of the Borrelia burgdorferi sensu lato complex which are transmitted to humans through the bite of hard-bodied Ixodes ticks [1, 2]. The earliest manifestations of Lyme disease include a characteristic skin rash, termed erythema migrans, along with concomitant flu-like symptoms followed by disorders of the heart, nervous system, and joints [1]. Globally, most cases of Lyme disease can be attributed to three Borrelia genospecies, B. burgdorferi sensu stricto (hereafter referred to as B. burgdorferi), B. afzelii, and B. garinii. In recent years, however, the number of genospecies associated with Lyme disease has expanded to include other organisms such as B. bissettii, B. valaisiana, B. spielmanii, and B. lusitaniae [3–7]. A second group of Borrelia spirochetes including B. hermsii, B. recurrentis, B. duttonii, B. parkeri, B. crocidurae, B. miyamotoi, and B. turicatae, are the causative agents of relapsing fever, a disease characterized by recurring episodes of fever with muscle and joint aches [8, 9]. Relapsing fever Borrelia are generally transmitted to humans by a soft-bodied tick of the genus Ornithodoros; however, one genospecies, B. recurrentis, is transmitted by the body louse [2].

Similar to Gram-negative organisms, Borrelia spirochetes are dual-membrane organisms with both an inner membrane and an outer membrane (OM); however, Borrelia species lack lipopolysaccharide [10, 11]. Instead, the surface of Borrelia spirochetes is characterized by the presence of numerous surface-exposed lipoproteins that are attached to the outer leaflet of the OM via N-terminal lipid moieties [12]. Borrelial lipoproteins have been the focus of intense study for several decades and are known to be important in virulence and host-pathogen interactions [11, 13–23]. Many of the borrelial lipoproteins are plasmid-encoded and differentially expressed throughout the life cycle of the organism [23–28].

The OM of Gram-negative organisms contain membrane-spanning, outer membrane proteins (OMPs) that form amphipathic β-barrels that can typically form nonspecific or substrate-specific OM pores [29, 30]. Freeze-fracture electron microscopy of the B. burgdorferi OM confirmed that the borrelial OM also possess integral OMPs, although the number of OMPs in the borrelial OM is at least 10-fold reduced as compared to Escherichia coli [31, 32]. The Borrelia OMPs identified to date are implicated in nutrient acquisition, antibiotic resistance, host-pathogen interactions, protein transport and assembly, and pore formation [33–42]. These proteins are characterized by their OM-localization and/or surface exposure but are not lipid-modified lipoproteins. Unlike the majority of the surface-exposed lipoproteins, all of the borrelial OMPs identified thus far are encoded on the ~900 kB linear chromosome [11, 33, 39, 41, 43–46]. While it is known that the borrelial OM contains membrane spanning OMPs with β-barrel structure and/or pore-forming capabilities that are important in overall physiology and host interactions, fewer than 10 borrelial proteins have been identified as potential OMPs and only half of those have thus far been characterized [22, 33, 39, 41, 45–52]. Of the known borrelial OMPs, B. burgdorferi proteins BamA (BB0795), BesC (BB0142), DipA (BB0418), P66 (BB0603), and P13 (BB0034) have been shown to form a β-barrel, to form pores in the borrelial OM, or to be functional orthologs to known OMPs. Efforts to identify novel OMPs in Borrelia spirochetes have been hindered for several reasons, including the low abundance of borrelial OMPs in the Borrelia OM and the unique fragility of the borrelial OM. Moreover, few orthologs to well-characterized proteins from other bacterial organisms have been identified through sequence comparison analyses [11, 20].

Given the challenges of identifying borrelial OMPs using conventional biochemical and proteomic approaches, we created a bioinformatics approach that utilizes an algorithm for predicting OMPs based on their unique properties and secondary structures. Using the derived computer-based algorithm, we examined chromosomally encoded proteins from both Lyme disease and relapsing fever Borrelia to identify candidate proteins that were both conserved and predicted to be OMPs in the genus Borrelia. Known borrelial OMPs were identified on the final candidate OMP list; and, importantly, three of the previously uncharacterized B. burgdorferi proteins identified by this method (BB0838, BB0405, and BB0406) were confirmed to be OMPs, indicating the computer-based methodology could, in fact, predict novel borrelial OMPs. Specifically, we determined that the B. burgdorferi protein BB0838 is amphiphilic and has surface-exposed regions that were accessible to proteases. B. burgdorferi BB0405 was previously shown by our laboratory to be a surface-exposed protein that localized to the OM [22, 39], and BB0405 and BB0406 were detected in OM vesicles by Pal and colleagues in a study examining the overall protein and lipoprotein content of the borrelial OM [53]. Herein, we found that BB0406 was indeed amphiphilic and localized to the B. burgdorferi OM as shown previously for BB0405 and that both BB0405 and BB0406 are pore-forming proteins.

Given that Lyme disease and relapsing fever spirochetes are extracellular pathogens, the surface of these spirochetes and the proteins localized to their OM provide the interface between these pathogens and their various hosts during infection. It also has been established that humoral immunity provides protection against these infections; therefore, borrelial surface proteins have been the target of much study with regards to identifying new vaccine targets. To date, most vaccine studies have focused on the immunogenic surface lipoproteins [12, 25, 88–93] with much less emphasis being placed on integral OMPs localized to the surface of the organism. This is due in part to the fact that surface lipoproteins are highly immunogenic and highly abundant, which has made lipoproteins much easier to identify and characterize than the borrelial OMPs. In this study, we sought to identify novel OMPs from Borrelia. Given that computer-based algorithms do not always accurately predict protein structure, we employed a consensus strategy in which we relied on six different algorithms that predict both OM localization and β-barrel conformation to identify candidate borrelial OMPs [58–63]. By using this stratified consensus approach, we were able to prioritize candidate OMPs based on the number of programs predicting that a specific sequence encoded a novel OMP. We curated our list by eliminating lipoproteins using a lipoprotein prediction algorithm specific for spirochetes [55] and proteins containing transmembrane domains [56, 57] that are indicative of inner membrane proteins. Moreover, we only accepted sequences with an N-terminal signal peptide as would be required for translocation across the bacterial inner membrane, and we again instituted a consensus approach using four different programs to predict signal peptides [64–67]. Importantly, we identified several known borrelial OMPs by the computational framework utilized including BamA, BesC, P66, and DipA, and, of the well-characterized OMPs identified thus far, only the known borrelial OMP P13 was not identified in our study. In a previous study, chromosomally encoded proteins from all sequenced bacterial genomes including the Borrelia spirochetes were analyzed by the TMBB algorithm which aims to identify β-barrel, OMPs [94]. Using this single method, 26 candidate B. burgdorferi B31 OMPs were identified of which eight were orthologous to non-OMPs. The study presented herein, however, identified 41 candidate OMPs further supporting the notion that the consensus approach utilized in these studies is an improvement from relying on a single algorithm.

Notably, we were able to further prioritize our list of candidate OMPs by examining and comparing the output generated from the chromosomes of nineteen different Lyme disease and relapsing fever organisms. With this approach, candidate OMPs could be further verified by the overall likelihood that the sequence was predicted to be an OMP in numerous other related borrelial organisms. In fact, potential false-positive or false-negative predictions in a single spirochetal species or strain can easily be identified by comparing the OMP predictions for proteins in the same cluster from many different organisms as outlined here. Furthermore, this more global approach will allow future studies to focus on sequences predicted to be OMPs in numerous species that are considered viable vaccine candidates based on their overall sequence homology and tendency to cluster together according to the algorithm utilized in this study. The cluster analysis demonstrated that there is a large group of predicted OMPs that are shared among the genomes of all currently sequenced Borrelia spp. analyzed (Fig. 5, Group C), which potentially represent vaccine candidates that could protect against both Lyme disease and relapsing-fever infections. This is an important addition that could not only enhance vaccine development for various spirochete diseases, but this system could also be used for other groups of human pathogens given that one of the major caveats in vaccine development has been generating protective immune responses against multiple strains or species.

While it is likely that some of the candidates identified using the computational algorithm outlined herein are not OMPs, the physicochemical data provided for BB0405, BB0406, and BB0838 provide a robust proof of principle that the algorithm can accurately identify novel OMPs. Interestingly, a larger domain of the C-terminal region of BB0838 was protected from degradation when organisms were treated with trypsin as compared to proteinase K. This type of proteolysis data would be entirely consistent with a large extracellular loop that contains a trypsin susceptible lysine or arginine residue but is otherwise fully susceptible to proteinase K degradation. Consistent with this notion, the B. burgdorferi OMP P66 also has been shown in previous studies to be partially sensitive to both proteinase K and trypsin with a lysine residue in a surface loop that is uniquely susceptible to trypsin [51]. Interestingly, bb0838 was found to be co-transcribed with uvrA and uvrB. UvrA and UvrB along with UvrC and UvrD are members of the nucleotide excision repair (NER) pathway which was previously shown to be the only borrelial DNA repair pathway that functions to repair DNA in response to UV light damage [73, 74]. In fact, B. burgdorferi spirochetes appear to have a reduced number of DNA repair enzymes [11, 20]. Why a potential borrelial OMP such as BB0838 would be co-expressed with two components of the NER pathway is unclear. It is worth noting, however, that this gene arrangement and genomic organization was conserved among all Lyme disease- and relapsing fever-causing spirochetes examined. This conservation in gene organization and, by correlation, co-expression pattern of uvrAB and bb0838 and its various orthologs suggests this happened prior to the divergence of the various Borrelia spp. Whether BB0838 can incorporate into liposomes and/or forms a pore remains to be determined. Such experiments with BB0838 could not be performed due to the large size of the full-length native protein and our inability to generate a full-length recombinant protein that was not lethal when expressed in E. coli.

BB0406 also was observed to be a novel OMP that had not previously been identified from B. burgdorferi or any other Borrelia spp. to date. As previously suggested, BB0405 was confirmed to be a surface-exposed OMP [22, 39]. While we demonstrated that BB0405 and BB0406 are OM-localized and capable of forming pores, BB0406 was found to be resistant to proteinase K degradation. Complete and partial resistance to protease has been reported for many other bacterial OMPs [30, 39, 95–98], so it was not entirely surprising that a borrelial OMP also is protease resistant. Given that we have now shown that BB0406 is a potential porin, it seems most likely BB0406 has either very little surface exposure or the protein is possibly protected from degradation by interacting with other surface OMPs or the abundant surface lipoproteins as was previously shown for the known OMP P66 [99]. Borreliae are known to have a relatively small genome as compared to other bacterial organisms and are thus dependent on the uptake of nutrients from the host environment. This highlights the importance of proteins such as porins for the survival of the spirochete throughout the organism’s life cycle. Yet, a limited number of porins have been identified in B. burgdorferi [41, 50, 100] despite the observation that in black lipid bilayer experiments using borrelial OM fractions there are numerous pore-forming proteins that have not been identified [101]. While BB0405 and BB0406 do form pores, we have not shown that these proteins form β-barrels as was predicted by the computer algorithms utilized in this study. Crystal structures will need to be resolved in order to definitively determine whether these proteins are β-barrels as predicted by the algorithm. However, recent reports have examined the structural conformation of both the B. afzelii BB0405 ortholog (BaPKo_0422) [102, 103] and the B. garinii BB0406 ortholog (BG_0408) using small-angle X-ray scattering [103]. Notably, in these reports, it was demonstrated that both proteins form structures entirely consistent with an 8-stranded β-barrel further supporting the notion that BB0405 and BB0406 fold into β-barrels in the borrelial OM.

Among previously characterized pore-forming OMPs in B burgdorferi, P13 and P66 are both known to be immunogenic and P66 has been shown to be at least partially protective in mice challenged with B. burgdorferi B31 [104–106]. Prior studies have shown that OMPs P66 and BesC are also required for establishing infection in mice [33, 107]. Along these lines, we also previously observed that antibodies recognizing the putative OMP BB0405 were bactericidal and that nonhuman primates infected with B. burgdorferi elicit a specific antibody response against BB0405 [22]. In other studies, we also have determined that a mutant B. burgdorferi strain lacking BB0405 and BB0406 is unable to establish an infection in mice, suggesting that one or both of these proteins is also essential for mammalian infection (Shrestha, Kenedy, and Akins, unpublished observations). Whether the inability of the BB0405/406 mutant to infect mice is dependent on the porin function of these proteins is unknown at this time. Notably, BG0407 the B. garinii BB0405 ortholog and BAPK0422 the B. afzelii BB0405 orthlolg have both been shown to bind human Factor H [102, 108]. Binding of Factor H by bacteria inhibits the alternative pathway of complement and is a method of immune evasion for pathogenic organisms. Whether B. burgdorferi BB0405 also binds Factor H is unknown at this time. BB0838 also may play an important role in the infectious life cycle of B. burgdorferi. When global transposon mutagenesis of B. burgdorferi was performed by Norris and colleagues, no transposon insertions were identified within the bb0838 gene [109, 110], suggesting it is an essential protein.

The computational and bioinformatics studies presented have identified novel OMPs from B. burgdorferi (BB0838 and BB0406) and confirmed that BB0405 also is a B. burgdorferi OMP. Furthermore, BB0405 and BB0406 were shown to have pore forming properties, suggesting they may play a role in allowing Borrelia spp. to sample and respond to environmental changes. In addition to a better catalog of candidate OMPs from various borrelial species, the computational framework utilized here could also help to identify new vaccine candidates for future studies. As it relates to vaccine-development, the overall clustering observed in Fig. 5 is provocative and points out specific candidate OMPs that could be targeted specifically for future Lyme disease or relapsing fever vaccine studies. Most important, however, is the fact that it may now be possible to identify potential vaccine candidates that could target both of these important human diseases with a single or multi-subunit vaccine consisting of OMPs shared by both groups. Apart from this applied aspect of vaccine development, basic mechanisms of molecular pathogenesis for this wide array of spirochetes could also be revealed in the various OMP protein clusters, which could imply functional specialization of specific OMPs that have evolved to support specific activities by Lyme-disease or relapsing-fever spirochetes. Further studies will be required to delineate among the proteins identified as to which are actual OMPs and which may be false positives. Finally, while the focus of our study was limited to Borrelia spp., it seems self-evident that a similar strategy could be used to identify OMPs from bacteria other than spirochetes to identify new vaccine targets for many different human diseases.

OM, outer membrane; OMP, outer membrane protein; PK, proteinase K