Protein network analysis indicates that Ebola virus, Neisseria meningitidis and Trypanosoma brucei trigger common host defense response pathways

Background: Sepsis is the most severe final state of infection. Might pathogens belonging to different kingdoms of life and pathogen classes trigger analogous responses pathways in the host? Is there a common denominator or strategy to prevent it? Here we use interactomics and comparative evolutionary analysis of three serious infectious pathogens, spanning and parasites: that target the blood during their infectious life cycle. Results: We analyze 2797 unique human proteins targeted by the pathogens. The comparison resulted in specific and shared protein-protein interactions (PPIs) with each pathogen, derived from orthology searches of experimentally validated PPIs. Furthermore, we narrow down the data set of effected genes of the human (defense) immune response and describe human proteins predicted to participate in at least two of the compared host-pathogen networks. Only four proteins were common to all three host-pathogen interactomes. This interaction clade of four proteins, common to all three host-pathogen networks, suggests a central cluster of host response formed by elongation factor 1-alpha 1 (EF-1-alpha-1), the SWI/SNF complex subunit SMARCC2 (matrix-associated actin-dependent regulator of chromatin subfamily C), the dolichyl-diphosphooligosaccharide--protein glycosyltransferase subunit 1, and the tubulin beta-5 chain. These hub proteins are functionally connected, suggesting according to the collected data including experimental information on VP24 from Ebola virus that these, as well as some joined interaction partners, may orchestrate a common host response to infection.

This interaction clade of four proteins, common to all three host-pathogen networks, suggests a central cluster of host response formed by elongation factor 1-alpha 1 (EF-1alpha-1), the SWI/SNF complex subunit SMARCC2 (matrix-associated actin-dependent regulator of chromatin subfamily C), the dolichyl-diphosphooligosaccharide--protein glycosyltransferase subunit 1, and the tubulin beta-5 chain. These hub proteins are functionally connected, suggesting according to the collected data including experimental information on VP24 from Ebola virus that these, as well as some joined interaction partners, may orchestrate a common host response to infection.

Background
About 10 13 human cells coexist with 10 14 bacterial, fungal, and protozoan cells, i.e. thousands of microbial species. Microbes create the human flora, limited to areas of the body such as skin, mouth, large intestine, and vagina. We live in close intimacy with microbes. However, only some pathogens (including microbes, fungi or viruses) may cause symptomatic infections -or death-in humans, one example being meningococcal bacteria (Rouphael and Stephens 2012;van Deuren, et al. 2000). How can this happen? When is the eco-system broken? Often immunocompromised individuals get infected, highlighting the fundamental role of the immune system (Alberts 2002).
In contrast, dedicated pathogens (including virus, bacteria or even eukaryotic organisms) do not require an immunocompromised or injured host. Pathogen physiology allows them to cross cellular barriers and trigger specific responses in the invaded organism that may contribute to pathogen survival and multiplication, as happens for Ebolavirus and Trypanosoma sp (Goeijenbier, et al. 2014;Moole, et al. 2015;Ponte-Sucre 2016;Rewar and Mirdha 2014;Stijlemans, et al. 2016). Thus, healthy people are "invaded" by dedicated pathogens (Alberts 2002). Cohabitation between populations at risk and their animals -as pathogen reservoirs -, is common in disease transmission (Welburn, et al. 2015).
Ecological and evolutionary hints, together with the disease dynamics, are fundamental for understanding and analysing the temporal, organizational and spatial interactions, even within a single pathogen. That is, hours to months, cellular to ecosystem levels, and local to pandemic spread of diseases (Welburn, et al. 2015). Additionally, some pathogens circulate between individuals of a single species, while others flow among multiple hosts, need arthropod vectors, or subsist in reservoirs. May pathogens belonging to different kingdoms of life and pathogen classes trigger analogous responses pathways in the host?
Appreciating that protein-protein interactions and conservation testify these different levels and scales of interactions paving the route to infection for the pathogen, as well as providing the backbone of defenses for the host, we herein compare for the first time the conserved and specific protein interactions of three serious pathogens [Filovirus ebolavirus (ZE for Zaire ebolavirus, Bundibugyo ebolavirus), Neisseria meningitidis ( NM for N. meningitidis) and Trypanosoma brucei (Tb for T. gambiense and rhodesiense, T. brucei sp.)], all sharing a blood-stage. The following detailed bioinformatics analysis of all available data highlights specific and conserved interactions and discusses implications for conserved immune mechanisms activation, and infection biology.

Conserved host and pathogen-specific pathways
Binding of pathogen proteins to specific host targets may affect specific host functions, including defense mechanisms. However, in general, they do not disrupt completely host networks, as the interactions mainly occur due to biophysical properties of the involved proteins. Herein we analyzed host-pathogen interactions comparing a virus (Zaire ebolavirus, Bundibugyo ebolavirus, ZE), a prokaryotic bacterium (Neisseria meningitidis, NM) and an eukaryotic parasite (Trypanosoma brucei gambiense and rhodesiense. Tb) that all three can cause severe infection including sepsis. By predicting physical proteinprotein interactions (PPIs) between host and each pathogen proteins, we described common pathways existing on the three host-pathogen interaction networks.
Detailed information on disease characteristics for ZE, NM and Tb are listed and explained in Additional file 1 and Additional file 2. The focus on this information is on general responses occuring during the infection processes comparing all three pathogens.
Thorough analysis of the summarized data points out functional similarities and differences occurring during infection processes. Therefore, individual host-pathogen PPIs could be established based on these functional similarities and differences. This information -and a literature-driven pathway analysis-, was a key step to address the next step of our investigation based on interactome data and comparisons between the three host-pathogen interactions.
Thus, to translate the observed characteristics into interactome data and describe the differences between ZE, NM and Tb, we performed a systematic analysis using bioinformatical techniques. We mapped physical protein-protein interactions (PPIs) between host and each pathogen, using orthology relations and host-pathogen interaction databases. Additionally, a stringent mapping pipeline, that refines interaction predictions by considering known domain-domain interactions (see material and methods) was used.
This approach produces high-quality interactome data with a low false-discovery rate (Itzhaki, et al. 2006;Zhou, et al. 2013).
To characterize the function of proteins in the above mentioned interactions we used the GO annotation. Results are depicted in Fig. 1 and 2. Fig. 1 displays the number of genes involved in the interactome for each pathogen and its counterpart in humans. Although each protein-protein interaction networks express unique characteristics, the overall profile of gene type involved in the interaction is similar not depending on to which animal kingdom the pathogen belongs. That is, most host proteins involved in the interaction are expressed either in the plasma membrane, or relate to DNA or RNA functions, or are mitochondrial proteins. Furthermore, with increasing complexity of the invading pathogen, an increasing number of proteins seem to be involved in all compartments or functions.
On the other hand, most pathogen proteins participating in the interaction are unknown proteins.

ZE.
The results indicate that the overall profile of protein type contributing to the interaction is similar and does not depend to which kingdom the pathogen belongs thus suggesting a "meta language" involved in host-pathogen interaction.
Due to this overlapping network of interacting proteins, belonging to pathogen and host that participate in the interactome for each pathogen and its human counterpart, we analyzed (see Additional file 3) the role of the described PPIs in host defense. The resulting subnetwork suggests the presence of common mechanisms to fight these three pathogens as depicted in Fig. 3 and summarized in Table 1  Redundancy in PPIs is permeating and fundamental for survival during such hostpathogen battles (Phizicky and Fields 1995): either to impair or to promote host protection, relying on innate and cellular defense mechanisms (Dyer, et al. 2007).
Domain analysis reveals that exposed protein domains with high structural and physicalchemical interaction affinity are central, conserved (Rual, et al. 2005) and selected by evolution in this long term battle (Aloy, et al. 2003;Littler and Hubbard 2005;Panchenko, et al. 2005;Teichmann 2002;Valdar and Thornton 2001). Host-pathogen interaction details and individual protein function including host defense-related signaling pathways and compartments are given in Table 1. This table categorizes unique (U) and conserved (C) interactions. Specific host defenses elicited by shared protection mechanism against all three pathogens include interferon, antigen-antibody interaction, the HLA system, shared cytokines and chemokines and proteasome associated proteins. Regulation of this shared host defense network is achieved by transcription factors via cell recognition associated proteins. Moreover, the ZE host defense interaction system centers on two human proteins, one at the nuclear level (related to proliferation) and one with multiple locations at the cytoplasm, Golgi apparatus, and plasma membrane: the CD antigen CD317, or stromal bone marrow antigen-2, a lipid raft associated protein (see also Additional file 5).
Pathogen interacting proteins re-occur in all three host-pathogen networks In the next step of analysis, all proteins that participate in the interactome, represented in all three pathogen-human interaction networks were analyzed. A total of 2797 unique human proteins were targeted by the pathogens. Interesting, data related to shared interaction nodes are detailed in Additional file 6 and summarized in Table 1, stressing again host-defense-activation-system-related proteins. Fig. 4 summarizes unique, shared, and common interactions for all three pathogen-human PPIs.
Pathogens tend to interact with hubs in the host network, in order to hijack them for their own profit (Yang, et al. 2011). However, the host network remains/stays robust, as a catastrophic failure of host cell metabolism would hinder pathogen survival and proliferation as well as kill the host (Crua Asensio, et al. 2017). The eukaryotic cells herein represented by trypanosomes achieve the highest number of specific interactions (1414), followed by Neisseria (753) and only then by 57 ZE-specific interactions. Pathway classifications were collected by different methods using Netpath (Kandasamy, et al. 2010), BioCarta (Nishimura 2001), KEGG (Kanehisa, et al. 2008) and Reactome (Vastrik, et al. 2007) database classification. The obtained results contributed to further dissect pathways involved in host-pathogen interaction: For example and regarding the trypanosome interactome, the Netpath database perceives EGFR, TSLP and androgen receptor pathways, while the databank reactome recognizes host defense mechanisms, immune signalling, cell senescence, and response to stress as top categories involved in the host-pathogen interaction. According to the BioCarta database, further signalling relies mainly on Thrombospondin-1 (TSp-1) induced apoptosis, Extracellular-signal regulated kinases/FYN proto-oncogene, src family tyrosine kinase/Tyrosine kinase (Erk/Fyn/Tyr) activation and EGF signalling. These results are further supported by KEGG which identifies NOD-like receptor signalling, antigen processing, and trypanosomal defense pathways, as central for the host-pathogen interactome. In fact, the innate DB (database) classifies the innate response against trypanosomes as relying on IL2 and IL12 as well as osteopontin and atypical NF-kB mediated processes (Additional file 7).
Although this is a common network for all three pathogens, the host defense details of course vary for each disease. Thus, interactions of host proteins specifically shared only between the smallest pathogens NM and ZE include mRNA stability, gene expression control, and mRNA splicing (reactome classification). Interestingly Netpath reveals again androgen receptor pathways as important, and that the TNFalpha response pathway is involved in the host response to these two pathogens.
Interestingly, and according to the reactome database, ZE and Tb, pathogens at the two ends of complexity and size, share specific defense systems centering on pathways for eukaryotic translation, elongation, life cycle, and infection. This involves ribosomes and PID-BardI signalling, and is modulated by Rb and p53 proteins according to KEGG. Netpath on the other hand shows a shared TSLP pathway response and again the TNFalpha response.
In conclusion, our data suggest that conserved host-pathogen interactions always include host defense and differentiation/maturation pathways. Additionally, the two bigger interactomes (NM, Tb) also involve reprogramming of host-cell senescence and apoptosis; the two smaller pathogens both influence mRNA, and Tb and ZE impact protein translation in their favor and induce specific host defense pathways.

Central interactome proteins shared by all pathogens and humans, implications
The previous results illustrate that out of the 2797 unique targeted proteins, only four [the elongation factor 1-alpha 1 (EF1a), the SWI/SNF complex subunit SMARCC2, the dolichyl-diphosphooligosaccharide--protein glycosyltransferase subunit 1, and the tubulin beta-5 chain] were common to all three host-pathogen interactomes. These top four hub proteins share a common trait: they are exploited by pathogens to gain entry to the host and survive inside it. As illustrated in Fig. 5, (colour codes differentiating all three pathogens), these four central proteins are strongly and well connected proteins. It is known that such hub proteins are often well conserved (Brown and Jurisica 2007;Fraser, et al. 2002) and hence present good pathogen targets (Durmus Tekir, et al. 2012).
Based on the results presented herein we would like to suggest that these host proteins may represent potential targets for the design of infection disruption tools, as due to their central role, their manipulation would probably stop all three infections. For example, EF1a activity is related to functional changes in the host cell so that infection can better exploit cell ongoing protein synthesis. The SWI/SNF-related matrix-associated actindependent regulator and the tubulin beta-5 chain may target hubs of the cytoskeleton related to infection and their modulation may block all three infections. The Dolichyldiphosphooligosaccharide--protein glycosyltransferase subunit 1 might provide an Achilles-heel on glycosyl and lipid metabolism, which any of the three infective agents exploits. Recent data for NM underline that tubulin interactions influence tubulin polymerization, an essential step for pathogen phagosome engulfment or escape (Tala, et al. 2014). Moreover, Table 2 examins in detail the conserved four human proteins and their interactions. It gives some more detail about the function and processes they are involved in as well as available direct evidence that these proteins are interacting with any of the pathogens. For Ebola virus this has been verified regarding VP24 and all four human interactions (Garcia-Dorival et al., 2014;Pichlmair et al., 2012).

Discussion
Proteins often interact with other protein molecules either from the same organism or bridging host organism and an invading microbe. The latter interaction has direct implications for host and pathogen, as it may either impair defenses (pathogens) or help to defend the microbe (host) by activating innate or cellular or systemic defense (immune) mechanisms. This all highlights the importance to understand the human-pathogen interactome, to provide insights into the pathophysiological mechanisms of infection potentially common to many infectious diseases. Herein we analyzed the interactome between the human host and a eukaryote, a bacterium, and a virus. The integrated results are interpreted around two key questions: are there conserved host and pathogen-specific pathways fundamental for defense (immune system) activation? if so, which interactive pathways are shared between all three or just two kingdoms of life? For the discussion we focus on host-defense responses unique to a particular pathogen and those shared by the different pathogens.

A common defense network for blood borne pathogens
To analyze if there is a common defense-mechanism against bloodborne pathogens, starting from conserved pathways and PPIs, we will elaborate on the herein obtained results.

ZE interacts with the complement cascade
Cell entry mechanisms for ZE have not been well defined (Falasca, et al. 2015). However, cholesterol-enriched lipid raft microdomains, or plasma membrane microdomains, seem to represent the main gateway used by this virus (entry and exit) (Bavari, et al. 2002;Falasca, et al. 2015). On the other hand, by stimulating antibody production and promoting the appearance of virus-antibody complexes, that bind to monocytes and macrophage Fc-receptors, ZE enhance its infectivity (Furuyama, et al. 2016;Takada, et al. 2007;Takada, et al. 2003a). Virus-antibody complexes also activate the complement cascade and facilitate virus entry into cells. Still, neither mechanism completely explains the non mediated Fc-receptors or complement cascade promoted increased infectivity described in ZE infected primate kidney cells (Ito, et al. 2001;Takada, et al. 2003b;Watanabe, et al. 2000). ZE stimulated antibody production may enhance infectivity due to its binding to specific antigen epitopes located at two or more molecules of monomeric IgG antibodies. This process permits complement cascade component C1 to bind to the antibodies Fc portion. The complex constituted by virus, antibodies, and C1, then binds C1q ligands at the immune cell surface, promoting either virus binding to ZE-specific receptors or endocytosis into target cells by intracellular signaling via C1q ligands (Takada, et al. 2003a).
Once internalized, a cytopathogenic effect of ZE causes annihilation of infected cells.
Monocytes/macrophages and dendritic cells (DCs) are ZE early replication sites and exert key roles in the dissemination of the virus from the spleen and lymph nodes to other tissues; and fibroblasts, hepatocytes, adrenal cells and epithelial cells can be infected by the virus (Falasca, et al. 2015). Further amplifying mechanisms of the initial cytopathogenic effect result in the destruction/impairment of vital body functions. Two proteins are involved in the ZE-human host defense interactome as herein described.
PCNA (see Table 1), a proliferating cell nuclear antigen (Cyclin), essential for replication and involved in DNA repair, epigenetics and chromatin remodeling (Moldovan, et al. 2007).
ZE evades Type-I IFNs response (IFN-α and IFN-β), thus eliminating the protective role of IFN-α. Additionally, multifunctional molecules like VP35 and VP24 -associated with viral particle maturation and immune activity suppression-, as well as inhibition of the tetherin activity seem to participate in the evasion process mechanisms.
Interestingly, the bone marrow stromal antigen 2 of the human-ZE interactome (see Table   1), belongs to the IFN-dependent antiviral response path. Its expression restricts viral factors once the virus or its viral components are detected by immune host cells.
Thus, upon virus recognition, a cascade of controlled interactions occur between signaling molecules, up to the upregulation of interferon-stimulated genes expression that further activate neighbor cells, and increase the expression of viral restriction factors, among them tetherin (Douglas, et al. 2010;Kuhl and Pohlmann 2012;Le Tortorec, et al. 2011). In conclusion, one of the two ZE proteins involved in host defense interaction mechanisms guarantees proliferation, while the second one restricts viral spreading. The implications of this balance should be further analyzed.
Additionally, ZE infection induces a massive cytokine/chemokine production by peripheral blood mononuclear cells (PBMC) or monocytes/macrophages. This occurs even after the first hour of ZE exposure, prior to virus gene expression. Shed glycoprotein molecules activate non-infected DCs and macrophages, inducing the secretion of pro-and antiinflammatory cytokines via TLR4 stimulation.
Furthermore, downstream effects of antigen-presenting cell dysfunction affect adaptive immunity. A massive natural killer (NK) cell loss in the peripheral blood impact the capacity of infected cell clearance and may be partially responsible for the unbalanced maturation signals for DCs. Thus, lymphoid depletion and necrosis occur in spleen, thymus and lymph nodes of dying patients, with a drastic decrease in the number of circulating T lymphocytes in succumbing people but not in survivors. Interestingly, despite the large loss of lymphocytes at work, no signs of lymphocyte virus infection is detected, suggesting a bystander mechanism of apoptosis, probably a non-apoptotic form of cell death or necrotic cell death (Douglas, et al. 2010). At the end, the overwhelming immune response disturbs the balance and promotes the precipitation of the survival of both, the host and the pathogen.
NM and Tb activate common pathways in the host Diseases caused by both NM and Tb have a blood stage and both, accidentally or not, may end invading the brain and causing either meningitis, in the case of NM, or a second stage disease in the case of Tb. Interestingly, the respective interactome, between each pathogen and the human defense mechanisms, overlap in high proportion (see Table 1). A plethora of molecules related to antigen presentation, as well as to cytokine and chemokine secretion and antibody production are involved in the respective interactome, suggesting that both pathogens activate in a similar way host defense and consequently, the immune system. In fact, conserved inflammatory molecules: Interleukin-2 receptor subunit beta, TNF receptor-associated factor 6 and Interleukin-18-binding protein, participate in the interactome, trigger processes such as thrombo-inflammation in different human host areas in a pathogen-specific way (Table 1).
LTBP4, the latent-transforming growth factor beta (TGF-β)-binding-protein-4, participate in the human-NM interactome, but not in the human-Tb interactome. LTBP4 is an extracellular matrix (ECM) structural component and TGF-β signaling regulator (Robertson, et al. 2015). Herein we comment on two different host-pathogen interacting mechanisms, present in both host-pathogen networks, fundamental for disease blood stage before invading the central nervous system: antigenic variation and complement activation.
Variations in blood parasitemia in cases of Tb are linked to host immune responses, but also to variation on the surface antigenic molecules expressed by the parasites (Stijlemans, et al. 2016). The glycoprotein coat of bloodstream parasites or Variant Surface Glycoprotein -VSG-is crucial for evading host antibody activity, and the complement cascade (Donelson, et al. 1998;Horn and McCulloch 2010;Pays 2006;Vanhamme, et al. 2004). Thus, to avoid these defense mechanisms, a high-frequency antigenic variation of the surface VSG occurs. Parasites are "hidden" from the immune system and therefore survive while in the bloodstream. Upon invading mammalian host bloodstream, the surface VSG triggers a humoral immune response. To cope with it, a small proportion of trypanosomes express a "new VSG", survive, and impair the Common to NM and Tb diseases is also the depletion of the complement cascade function.
In trypanosoma infections by Tb gambiense and Tb rhodesiense, reduction in complement activity is caused, at least partially, by massive activation triggered by the parasites themselves (Vincendeau and Bouteille 2006). This happens also in the cattle disease produced by T. congolense, where both the classical (Nielsen and Sheppard 1977) and alternative pathways (Tabel 1982) are activated, and by Tb brucei, where the classical pathway activation by VSG occurs in an antibody-independent fashion (Musoke and Barbet 1977). The complement cascade seems thus to be unable to prevent the establishment of parasites in the bloodstream (Devine, et al. 1986;Namangala 2011). Of note, Devine et al. (Devine, et al. 1986), described that Tb gambiense activates the alternative complement pathway in human serum, without activating a lytic activity. The cascade does not continue beyond the establishment of complement receptor 3 (C3) convertase on trypanosome surface.
On the other hand, human infection by NM begins by the colonization of the mucosal epithelium by the bacteria. NM adhere to the apical surface of mucosal epithelial cells in the cervix, conjunctiva, fallopian tubes, nasopharynx, rectum and urethra, by means of type IV pilin. The Opa proteins mediate intimate adherence and internalization, and adhesins and invasins, like LOS, mediate the infection of epithelial cell subgroups; the pilus and the porins interact cooperatively with CR3 on cervical epithelium (Coureuil, et al. 2013). Each mucosa might have special adhesins and molecules involved in adhesion and internalization that will not be described herein.

Conserved host and pathogen-specific pathways
Next we searched for a common denominator that could lead us through the results depicted in Fig. 3, to accomplish the potential existence of common, as well as pathogenspecific -unique-interaction mechanisms to guide the host-pathogen communication.
Detailed conserved and unique PPIs occurring in different compartments are listed in Table 1.
In Fig. 3 host proteins that interact with proteins from the three pathogens are organized according to their location (green nodes), in six different cellular compartments.
Cytoplasmic proteins include the host factor p53 (Tp53, to modulate apoptosis and inflammation for a suitable infection environment) and transforming growth factor beta-1- By this interaction EF1a is incorporated in HIV-1 virion membranes, and cleaved by the viral protease. This interaction impairs translation in vitro, a result consistent with a previously proposed model in which inhibition of translation by accumulation of Gag serves to release viral RNA from polysomes, permitting the RNA to be packaged into nascent virions (Cimarelli and Luban 1999;Ejiri 2002). Additionally, excretory-secretory products fundamental for the pathogenesis of Giardia intestinalis seem to be relevant for host parasite interactions, among them, EF1a and a 58 kDa protein, identified as arginine deiminase. EF1a localizes to the nuclear region in trophozoites but relocalizes to the cytoplasm during host cell interaction. Recombinant EF1a is recognized by giardiasis patients serum suggesting the importance of the released EF-1a during Giardia infection (Skarin, et al. 2011).

Implications for Infection Biology
The results highlight the fundamental role of "recognition" as an initial step and "effector" activation as a final step within the chain of events that would eventually lead to the specific immune (defense) system activation. In fact, the Venn Diagram describes that trypanosomes, being eukaryotes, have the highest number of specific interactions with the host, followed by NM and being the ZE-specific interactions very low. A high proportion of overlap exist between Neisseria-and trypanosome-human interactions, and these conserved PPIs focus on host proteins related to the: immune system signaling, response to stress and reprogramming of host cell senescence. On the contrary, overlaping interactions between ZE and NM focus on mRNA stability and splicing while interaction overlap between ZE and Tb concerns translation and elongation, ribosome function and life cycle, meaning focus on protein translation.
Most interestingly, the interactome analysis singles out the clade of only four interacting proteins present in the host and interacting with all three pathogens. The aforementioned clade seems to participate in assuring the proper decoding of the genomic message to produce cellular proteins and relocate them at the appropriate place. Three of these four central fundamental hub proteins clearly relate with cytoskeleton organization and remodeling, being EF1α undoubtedly related to protein translation. However, EF1α proteins, also modulate cytoskeleton activities, exhibit chaperone-like activity, and are key proteins for cell proliferation and cell death in human tumors (Abbas, et al. 2015). In fact, EF1α assists replication of many RNA viruses, such as the respiratory syncytialvirus (RSV). In this case, down-regulation of EF1α restricts the expression of viral genomic RNA and the release of infectious virus (Wei, et al. 2014); in parasites, EF1α has also been implicated in pathogenesis (Nandan and Reiner 2005) and host cell invasion (Matsubayashi, et al. 2013).
This theme suggests organizational complexity along the evolution from virus, to bacteria, to metazoans, and offers multiple avenues for regulatory control, all with the aim of producing the appropriate cellular response to external stimuli. The pathogens activate a highly conserved, concerted immune (defense) response to preserve the integrity of the host. Whether or not this is the case for the herein analyzed pathogens has to be demonstrated.

Probing the revealed protein-protein interactions of the central backbone of host defense
To validate these findings, potential experiments may be suggested with focus on the viral pathogen EBOV: In HIV viruses, the SWI/SNF complex subunit SMARCC2 (matrix-associated actin-dependent regulator of chromatin subfamily C) is a key factor in chromatin remodeling, being its innate function to suppress transcription. Thus, knocking down the HIV-1 associated SMARCC2 (Baf53) gene, increase viral gene expression from transiently and chronically infected cells (Van Duyne, et al. 2011). A similar experimental approach, and the potential functional regulation exerted by phosphorylation would shed light in EBOV infections.
On the other hand, oligosaccharyltransferase complexes seem to be fundamental host factors for the propagation and replication of dengue virus. By the use of a genome-wide clustered, regularly interspaced, short palindromic repeat (CRISPR) screen, (Lin, et al. 2017) it has been demonstrated that the oligosaccharyltransferase complexes may be functioning as scaffolds, but also may act on nonstructural dengue virus proteins to mediate efficient synthesis, folding, and/or recruitment of nonstructural proteins to specific sites in the endoplasmic reticulum (Lin, et al. 2017). Similar experiments would share light to the mechanisms involved in the conserved host-pathogen PPIs herein described.
In addition to the central interaction clade, most host cell compartments are targeted by the pathogens, the overlap between host-pathogen interactions exposed best when comparing NM and Tb. Shared interacting host genes relate to antigen-presenting and proteasome processing of antigens, as well as tissue antigens, immunoglobulins and cytokines (IL2, IL18, TNF, TGF), and their receptors. Additionally, NM plasma membrane proteins have multiple localization niches indicating the reuse of function in these niches.
This data highlights potential therapeutic targets for the design of agents to impair the infection produced by these pathogens.
What is clear is that as the sculptor of organisms is evolution, phenotypes favoring survival retain and contain information useful for this aim and permit evolution and survival to continue, and be preserved. We have defined for three pathogens, belonging to three kingdoms of life basic PPIs that seem to be central for pathogen interactions with the human host in general. The network may orchestrate an ancient common host response to infection, is validated for all four human interactions with Ebola virus protein VP24 (Garcia-Dorival et al., 2014;Pichlmair et al., 2012) and should guide infection therapy design in a more rational and efficient way.

Conclusions
We compared human host-pathogen interactions from three kingdoms of life and described conserved host responses triggered independently of the kingdom of life origin of the pathogen, virus, prokaryotic or eukaryotic pathogens. Though we span a huge evolutionary distance between the three pathogens studied regarding protein-interaction networks with the host, our results suggest and confirm by all available protein-protein interaction data and analysis of conservation and orthology an unexpected interaction clade of cytoskeletal and ribosomal proteins common to all three networks.
All three pathogens attack a similar (or the same) host mechanism that we may call a "non-identified Achilles heel of the host", central to the affected response. In combination with some joined interaction partners this mechanism may orchestrate a common defense response to all three diseases, a fundamental but general response, validated interactions are available for VP24 from ZE for all four conserved human proteins.
These findings should prompt a number of further studies to show further conservation of this host defense backbone for other pathogens by bioinformatical analysis as well as direct experimental tests of the predicted general and specific protein interactions.

Inferring protein-protein interactions based on interologs
The aim, design and setting of the study was hence to seek direct information on the proteins and protein interactions in three pathogens from different kingdoms to reveal potential conserved host interactions as well as specific or only partly shared protein interactions. The key method to obtain for such comparisons sufficient data is the inference of protein-protein interactions based on interologs. Our specific approach is given in the following.
The experimentally determined host-pathogen PPIs from PHISTO database (Durmus Tekir, et al. 2013) and PPIs present in the DIP (Database of Interacting Proteins) database (Salwinski, et al. 2004) were used as a template to reconstruct interologs (Yu, et al. 2004) based PPIs networks of human-ZE, human-Tb and human-NM. The stand-alone InParanoid program version 4.1 was used to assign proteins from the three pathogens to orthology groups. It uses the reciprocal best blast hit strategy to identify the clusters of orthologs and paralogs (Sonnhammer and Ostlund 2015). Only the seed ortholog pair whose Bootstrap values was 100% were considered to increase prediction confidence.

Pruning PPIs with domain-domain interactions
The amino acid sequences of identified non-redundant DIP based interologs host-pathogen PPIs were extracted and domains were assigned to the sequences using Pfam version 29.0 (Finn, et al. 2016). The information of interacting protein domains was prepared using the three databases Domine (Yellaboina, et al. 2011), DIMA 3.0 (Luo, et al. 2011) and IDDI database (Kim, et al. 2012). They consider the information of the Pfam families suggested to interact based on structural information present in the protein complexes available in Protein Data Bank (PDB) (Berman, et al. 2002). This list was used to parse the DIP based interologs PPIs using customized Perl scripts and only those interolog interactions were retained that consist of at least one true domain-domain interaction pair as mentioned in the used databases.

Host-pathogen PPI networks
The host-pathogen PPIs predicted from both the PHISO and domain-domain interactions supported DIP based interactions were merged and redundancies were removed.
Cytoscape version 2.8.1 (Shannon, et al. 2003) was used to visualize the all three hostpathogen PPI networks. The network hubs were determined using the cyto-Hubba plugin (Chin, et al. 2014). The networks were further curated by extensive manual effort and the connectors of proteins related to human immune system were retrieved to generate three subnetworks.

Gene-Ontology (GO) annotation
The protein sequences of pathogen interacting host proteins were extracted from all three-established host-pathogen PPIs and resulting proteins were functionally annotated by the BLAST2GO software suite version 2.4.1 (Conesa, et al. 2005).
Fisher's exact test was used to identify significantly over-represented GO terms.

Pathway over-representation analysis
The pathway over-representation analysis of conserved human proteins that were targeted by at least two pathogens was performed using the pathway analysis tool at InnateDB database (Breuer, et al. 2013). The p-value was calculated using hypergeometric tests, and Benjamini-Hochberg adjustment was used for multiple test correction.

Subcellular localization
The subcellular localization of immune related human proteins was determined with extended version of KnowPredsite (Lin, et al. 2009) available at UniLoc server (bioapp.iis.sinica.edu.tw/UniLoc/), a knowledge-based classifier for protein subcellular localization. The prediction server KnowPredsite II is capable of predicting single and multiple localization sites for both the eukaryotic and prokaryotic proteins.
Abbreviations PPI, protein-protein interaction; PPIs, protein-protein  Declarations Ethics approval and consent to participate or to publish: Not applicable, this manuscript does not include human participants, human data or human tissue.

Availability of data and material:
The datasets generated and analysed during the current study are all available in the manuscript and its supplementary materials

Competing interests:
The authors declare that they have no competing interests neither financial nor nonfinancial.

Funding:
This work was supported by the Deutsche Forschungsgemeinschaft (GRK 2157 3D tissue infect; project number 270563345) for funding. APS acknowledges the support of the Alexander von Humboldt Foundation for return visits to Würzburg University and Missionärztlisches Institut, Würzburg.

Author contributions:
TD was responsible for project design, data analysis, manuscript writing and correction; SG was responsible for data collection, bioinformatic data analysis, manuscript writing and correction; EB was responsible for data collection and manuscript correction; APS was responsible for literature collection, data analysis, manuscript writing and correction. Color code: Inteferon associated proteins, Antigen, antibody and HLA associated proteins, Cytokine and chemokine associated proteins, Proteasome associated proteins, Transcription factors associated proteins, Cell recognition associated proteins. Note that EBOV-human immune system interaction concentrates in two proteins, one at the nuclear level (proliferation) and one that can be found in the cytoplasm, Golgi apparatus and plasma membrane. On the other hand, human -immune related-proteins involved in the interaction with T. brucei and N. meninigiditis overlap in high proportion. In the table the following codes are used: U for unique, and C for conserved interactions; the numbers correspond to the number of pathogenic proteins in corresponding category interacting with them. Information about the function of these proteins can be found in the Additional file 3. \      Supplementary Files