Early infection-induced natural antibody response

There remains to this day a great gap in understanding as to the role of B cells and their products—antibodies and cytokines—in mediating the protective response to Francisella tularensis, a Gram-negative coccobacillus belonging to the group of facultative intracellular bacterial pathogens. We previously have demonstrated that Francisella interacts directly with peritoneal B-1a cells. Here, we demonstrate that, as early as 12 h postinfection, germ-free mice infected with Francisella tularensis produce infection-induced antibody clones reacting with Francisella tularensis proteins having orthologs or analogs in eukaryotic cells. Production of some individual clones was limited in time and was influenced by virulence of the Francisella strain used. The phylogenetically stabilized defense mechanism can utilize these early infection-induced antibodies both to recognize components of the invading pathogens and to eliminate molecular residues of infection-damaged self cells.

There remains to this day a great gap in understanding as to the role of B cells and their productsantibodies and cytokines-in mediating the protective response to Francisella tularensis, a Gramnegative coccobacillus belonging to the group of facultative intracellular bacterial pathogens. We previously have demonstrated that Francisella interacts directly with peritoneal B-1a cells. Here, we demonstrate that, as early as 12 h postinfection, germ-free mice infected with Francisella tularensis produce infection-induced antibody clones reacting with Francisella tularensis proteins having orthologs or analogs in eukaryotic cells. Production of some individual clones was limited in time and was influenced by virulence of the Francisella strain used. The phylogenetically stabilized defense mechanism can utilize these early infection-induced antibodies both to recognize components of the invading pathogens and to eliminate molecular residues of infection-damaged self cells.
Francisella tularensis is a Gram-negative coccobacillus belonging to the group of facultative intracellular bacterial pathogens proliferating in phagocytic cells, including neutrophils. Several subspecies of F. tularensis cause serious tularemia infection in both animals and humans. A safe and effective vaccine continues to be sought to protect against respiratory infection with hypervirulent F. tularensis subsp. tularensis 1 . One of the reasons we do not yet have a suitable vaccine relates to the considerable gaps in our knowledge of the body's defense mechanisms against this dangerous pathogen.
The immune mechanisms contributing to the expression of protective immune response against F. tularensis remain a matter of debate. The theories and views as to the roles of individual cell types and subtypes in generating protective immune response are as diverse as are the infection models used to reveal their function. Originally, the T cell-mediated immune processes based on IFN-γ-dependent activation of macrophages were regarded as constituting the decisive factor in host defense against tularemia [2][3][4][5] . Although we have intensively studied macrophages as the primary hosts for Francisella replication [6][7][8][9][10] , the evaluation of their role in protection remains to this day dependent on the experimental approach taken and virulence of the Francisella strains used 11,12 . The contribution of polymorphonuclear leukocytes to the limitation and/or dissemination of infection around the body is similarly ambiguous [13][14][15] . The currently prevailing opinion is that neutrophils create detrimental effects by participating in tissue destruction following F. tularensis infection and thus contribute to pathogenic processes 16 . The primary interaction of dendritic cells with F. tularensis, which, as in other phagocytic cells, escapes from the phagosome and proliferates within the dendritic cell cytosol, ends with cytokine-producing cells that have not been sufficiently stimulated and therefore have limited capacity to influence the protective immune response 17,18 . As in the case of macrophages, however, the signaling pathways of dendritic cells are differentially regulated by virulent and attenuated strains of Francisella 19 .
There remains to this day, a great gap in our knowledge as to the role of B cells in mediating protective response to F. tularensis infection. The basic function of B lymphocytes within the humoral branch of the adaptive immune system is to generate a spectrum of specific antibodies. According to the prevailing notion, however, the specific antibodies are not a decisive component of protective immunity. Early studies on lymphocyte response to the vaccine strain of F. tularensis demonstrated response of both T and B cell populations 32 . Gradually, it was Results F. tularensis elicits early robust total serum antibody response. Our initial finding of early antibody production determined by ELISA in SPF mice after F. tularensis infection allowed for two interpretations. This reflected either a reaction of cross-reacting pre-existing antibody clones or the rapid natural response of innate B cell subsets. For this reason, we used the GF mice for further experiments and we compared their antibody response with the response of SPF mice. GF animals have no pre-existing contacts with bacteria whereas SPF mice have a natural, pre-existing microbiota, including many non-Francisella species. To determine the kinetics of early antibody response to infection, SPF and GF Balb/c mice were infected with F. tularensis subsp. holarctica FSC200. To ensure direct interaction of antibody-producing cells with bacteria, we used intraperitoneal infection, where direct interaction of individual B cell subtypes is possible without prior processing of the immunogen by antigen-presenting cells. Moreover, the peritoneal cavity provides a suitable model for the study of B cells because of the presence there of a unique peritoneal cavity-resident B cell subset known as B1a cells, in addition to B1b cells and conventional B2 cells. After 12, 24, and 48 h of colonization, total serum antibody levels were measured in each group of mice (Fig. 1). The antibody isotypes composition of Balb/c SPF and GF murine sera obtained from naïve mice without any bacterial colonization were very low and therefore are not shown in this figure. The basal individual antibody isotypes levels for GF mice were significantly lower in comparison with the corresponding isotypes levels for SPF mice. A lower level of circulating IgG1 and IgG2b was observed in normal uninfected GF mice when compared with uninfected SPF mice ( Supplementary Fig. S1). A higher level of light kappa and lambda chains in analyzed SPF sera, however, are probably associated with higher content of antibodies in the SPF sera in both normal and infected mice ( Supplementary Fig. S2).
The infection of GF as well as SPF mice with F. tularensis confirmed that infection induced the early antibody response as soon as 12 h post infection in both mice strains (Fig. 1). The infection with F. tularensis significantly increased the levels of all antibody isotypes in GF sera and in sera of SPF mice with the exceptions of IgD and IgE. The IgG1 and IgG2b isotypes dominated in sera of SPF mice, while IgM, IgA, and IgG3 predominated in GF mice. The kinetics of individual antibody isotypes 48 h post infection suggest the infection induces an immediate humoral stress response that is dominantly produced by early-induced natural antibodies.
Early robust IgM response occurred only after the infection of GF mice with F. tularensis subsp. holarctica FSC200 at all monitored time points. The IgM antibodies constitute the major component of the natural antibodies and should also be the first class of antibodies produced during early antibody response to infection. Our results demonstrate that this corresponds only to the response of GF mice, as the dominant isotype of early infection-induced antibodies of SPF mice are of IgG2b isotype. Assessed globally, the serum IgM and IgG antibodies were barely detectable in both GF and SPF mice before Francisella infection. The early humoral response of GF mice to infection is more intense, however, than is that of SPF mice 12 h post infection (Fig. 2).
The increased level of infection-induced IgM antibody isotype in GF mice at 12 h postinfection and other time intervals in comparison with the levels of IgM isotype in sera of infected SPF mice may be caused by there    The immunoglobulin subclasses of serum samples were determined using a Quantibody mouse immunoglobulin isotype array while following the manufacturer's instructions. The antibody isotypes compositions of Balb/c SPF and GF murine sera obtained from naïve mice without any bacterial colonization were very low and are not shown in these graphs. The basal individual antibody isotypes levels for GF mice were significantly lower in comparison with the corresponding isotypes levels for SPF mice (data not shown). The Quantibody mouse immunoglobulin isotype arrays were performed in biological triplicates (individual sera) for all time intervals and were independently repeated at least three times. Reference isotypes are presented. Values are expressed as mean ± standard deviation (SD) and analyzed for significance using Student's two-tailed t-test. Differences were considered statistically significant at **p < 0.01 and *p < 0.05 and are denoted by asterisks. www.nature.com/scientificreports/ prepared. Other gels were electroblotted onto polyvinylidene difluoride membrane. Immunoblotting was subsequently performed with mice tularemic sera from SPF and GF mice. Biological triplicates of SPF and GF sera were prepared at selected time points and immunoblotting experiments were performed twice with each serum. Normal murine SPF and GF serum was prepared from blood of Balb/c mice without exposure to any F. tularensis strain. Coomassie-stained 2-DE gels were then prepared to excise the selected spots that were digested and the reactivities on bacterial extracts were determined. The mass spectrometry approach successfully assessed in total 87 different F. tularensis proteins that reacted with the sera obtained from mice at 12, 24, and 48 h post infection (Fig. 3). A similarity analysis demonstrated that the proteins reacted according to the model (GF versus SPF) that was used and the time intervals when the sera were obtained after infection. The dominant immunoreactive protein spots of F. tularensis FSC200 detected on at least three immunoblots are shown on the reference 2-D protein map (Fig. 4) and the proteins are summarized in Supplementary Table S1. A total 50 immunoreactive Francisella proteins were unique targets of GF murine sera with no appearance of interaction with SPF sera isolated during the intervals of innate immune response to F. tularensis infection (see Fig. 3 and Supplementary Tab. S1). A number of immunoreactive proteins identified in this study already have been described using different infection models, including human tularemia at the stage of adaptive immune response (see references for Supplementary Table S1). Some studies were conducted to identify and characterize the virulence factors that F. tularensis uses to infect a wide variety of hosts, induce disease, and evade immune defense mechanisms. Some of them are listed as immunoreactive proteins. Sera isolated from GF as well as SPF mice during very early time points (i.e., 12, 24, and 48 h) postinfection reacted also with several virulence factors already identified. In this respect, the robust reactions to Francisella virulence factors were mainly observed using GF murine serum obtained 12 h after intraperitoneal inoculation by bacteria.
The early Francisella-induced antibodies reacted also with SodB (FTS_1747), KatG (FTS_1471), DnaK (FTS_1167), and GroES (FTS_1671), which are proteins that already have been reported to be released or secreted www.nature.com/scientificreports/ by F. tularensis but the mechanisms for their secretion remain unclear. In addition, we identified 6 immunoreactive proteins annotated with signal peptide according to SignalP 5.0 software, namely KatG (FTS_1471), OmpA (FTS_1295), GTP-dependent nucleic acid-binding protein EngD (FTS_0935), Omp protein of unknown function (FTS_0008), glycerophosphoryl diester phosphodiesterase (FTS_1476), and hypothetical protein (FTS_0815). All the mentioned proteins are listed along with their characteristics in Supplementary Table S1. All identified proteins reacting with the sera from Francisella-infected GF and SPF mice were sorted into Clusters of Orthologous Groups (COG) functional categories (Fig. 5). Two dominant COGs relate to biogenesis at the stage of protein post-translational modification and chaperon-accompanied protein translocation. Sixty-three of the 87 identified antibody clones reacted with bacterial proteins that have orthologs or analogs in eukaryotic cells. The remaining 24 prokaryotic targets are characteristic for F. tularensis, of which 7 are hypothetical proteins. The identities of those proteins were determined according to the UniProtKB database. Among the  Table S1 (designated letters and numbers of protein spots were used only for internal purposes and do not necessarily imply the same identity of the labeled proteins). Polyclonal peroxidase-conjugated goat anti-mouse IgG antibody was used for secondary antibody detection. All experiments were performed in biological triplicates (individual sera) for all time intervals and were independently repeated at least three times. www.nature.com/scientificreports/ eukaryotic analogs or orthologs, mitochondrial and cytosolic proteins were found to be the dominant targets of early Francisella-induced antibodies (Fig. 6). The reactivities of all SPF and GF murine sera obtained at the selected time points are described in Supplementary Table S1. Of 50 unique GF antibody specificities, the serum of GF mice contains 12 with no eukaryotic orthologs (Table S2). These antibody specificities have protein targets that, according to the Clusters of Orthologous Groups categorization, are involved on the one hand in amino   www.nature.com/scientificreports/ acid metabolism and transport, protein intracellular trafficking and secretion, and membrane biogenesis, and on the other hand in cell cycle control and replication and repair (see Fig. 6).

Francisella-induced early unspecific antibody response. The sera from Balb/c mice infected with
F. tularensis subs. holarctica FSC200 obtained in early 12 h interval after infection reacted with different sets of F. tularensis subsp. holarctica FSC200 and F. tularensis LVS proteins used in the form of whole-cell lysates for immunoblotting. The intensity of reaction corresponded to virulence of the bacterial strain used for infection. The early humoral response of mice to F. tularensis subs. holarctica FSC200 is more intense in terms of both the quantity of antibodies in serum and the spectrum of recognized targets when compared with LVS counterparts.
In parallel with immunoblotting of sera obtained from GF and SPF mice infected by F. tularensis against F. tularensis whole-cell lysate proteins we used as targets also the whole-cell lysates of Klebsiella pneumonia (clinical isolate) and Pseudomonas aeruginosa CCM 3955. Both bacteria are conditionally pathogenic and can incidentally colonize the airways. The reaction of early Francisella-induced antibodies on these targets was weaker than was the response against F. tularensis whole-cell lysate, but it was still very significant ( Supplementary Fig. S4). The reactions of protein targets of both alternative microbes with the sera obtained from F. tularensis-infected GF mice were more intense than was the response of sera obtained from SPF mice. The GF control sera failed to react with Klebsiella pneumoniae and Pseudomonas aeruginosa proteins. The control sera obtained from SPF mice reacted, but only slightly, with the proteins of both bacteria. The target proteins for the sera from Francisella-infected mice were not identified because we did not search for them. If the sera from both GF and SPF Francisella-infected mice reacted with proteins of unrelated bacteria we tested also the response of these sera with the whole-cell lysate of Balb/c peritoneal cells. In this case, too, we have demonstrated the interaction of early infection-induced antibody clones obtained from GF and SPF mice with protein targets of whole-cell lysate of peritoneal cells. The responses of both GF and SPF mice appear to be similar in intensity, but they respond differently in terms of their protein targets ( Supplementary Fig. S5).

Discussion
The immune response to natural infection with F. tularensis as well as the response to immunization of experimental animals or vaccination of humans with attenuated strains induced strong cell-mediated immunity. Originally, the delayed-type skin test or the in vitro lymphocyte stimulation test demonstrated the expression of cell-mediated immune response at the end of the first week after natural infection or vaccination 20,54 . Similarly, the utilization of whole-blood culture DNA synthesis response and IL-2 and IFN-gamma secretion authenticated the second week as marking the time interval needed for the Th1 type of T-cell-mediated immunity expression after vaccination of human volunteers with a live vaccine strain 55 . In parallel, it was demonstrated that a substantial antibody response of human patients occurred within the second week of infection 54 . Studies on the host genetic background of resistance to experimental Francisella infections [56][57][58] and the participation of neutrophils, NK cells, or B cells in the resistance to primary tularemia infection 13,33,59,60 drew attention to early innate immunity. This innate (natural, genetic, or inherent) immunity (or resistance) has been defined as "the ability of an organism to resist the effect of an ecological or physiological agent in the absence of any acquired (reactive) immunity" 61 . In other words, innate immune responses occur because cells are able rapidly to react to invading microorganisms without the requirement for an antigen-specific adaptation. Innate immune responses in the context of Francisella models operate until the fifth day post infection 6 . During this time period, the unspecific early protective immunity can be demonstrated in murine models between the second and third days post infection with vaccine strain of F. tularensis 33,62 ; human volunteers respond to LVS vaccination by activation of γδ T cells and NK cells 24 h post vaccination 63 . Transcriptome analysis has demonstrated the induction of proinflammatory and innate immunity-related genes at early post-vaccination time points, meaning ≤ 48 h 64 .
Although we have been investigating immunity against F. tularensis infection for many decades our knowledge as to the role of natural humoral mechanisms remains relatively rudimentary. We have identified, rather serendipitously, several antibody clones isolated 12 h post infection from sera of SPF mice immunized with LVS that reacted with F. tularensis proteins and were absent in the sera of control mice. Such an early antibody response could be either a response of memory B cells resulting from previous interactions with microorganisms or an induction of T-cell-independent early natural antibody production induced by F. tularensis strain FSC200 infection. We decided to resolve this dilemma using GF mice that should not have had any interaction with bacteria during their ontogeny, and not even with bacteria that constitute the gut microbiome of SPF mice.
According to earlier studies, the antibody production of GF mice in response to corpuscular antigens, such as from sheep red blood cells 65,66 or hamster erythrocytes 67 , was comparable to that seen in conventional mice. Similarly, only limited differences were observed between conventional and GF mice in response to hydrophilic polysaccharide dinitrophenyl-lysine-Ficoll or dinitrophenylated bovine serum albumin 66 . In all these early studies, the humoral immune response or competence was monitored in the intervals corresponding to induction of acquired immune responses. Antibody production during early time intervals corresponding to induction of innate immune responses was not in focus, despite the fact that the natural antibodies have been shown-beyond immediate protection against infection-to play a substantial regulatory role in the actions of the defense system and homeostasis 68 .
Our study revealed the early humoral innate immune response of GF as well as SPF mice to the intracellular bacterium F. tularensis at least from the 12th hour after intraperitoneal inoculation with the bacteria. The majority of antibody clones reacted with bacterial proteins that have orthologs or analogs in eukaryotic cells. The dominant targets were cytosolic and mitochondrial proteins. Some of the clones reacted also directly with eukaryotic proteins separated by SDS PAGE of naïve murine peritoneal cells lysate. Considering that natural antibodies (autoantibodies) have been shown to recognize evolutionarily fixed epitopes shared between foreign www.nature.com/scientificreports/ and self-antigens 69 , the results of our study correspond to the generally accepted target profile of natural antibodies. The isotypes composition and their relative proportions in the serum of GF mice also correspond to the general characteristics of natural antibodies. The kinetics of antibody production suggest a temporary nature of humoral innate response, at least in the sense that some antibody clones (not the B cell clones) disappear within the next 12 h. Such kinetics suggest that different (sub)types of B cells and plasma cells might participate in the early natural humoral immune response to infection. Part of the clones persists, however, and these specificities can be identified even in the acquired immune response phase. Antibody specificities that were detected in both the innate immunity phase and the acquired immunity phase may result from the B cell responses regulation, selection of the B cell repertoires, and/or the regulation of B cell development, all of which are processes modulated by early produced natural antibodies 70 . An anamnestic response of memory B cells derived from an innate B cell pool by infection may ensure selected specific and long-term acquired humoral responses. This comparison of the GF and SPF mice antibody response intensities leads to the conclusion that previous interaction with bacteria instructed the immune system of SPF mice and that the uniform humoral response of the naïve innate immune system of GF mice is more limited, and especially at IgM isotype level. Moreover, data in the literature proves that the exogenous antigenic stimulation has changed the relative proportion of cells secreting IgM isotype antibody on the one hand and cells secreting IgG and IgA antibody isotypes on the other. This means that exogenous antigenic stimulation has a great effect on the background IgM-, IgG-, and IgA-secreting cells, and, most likely, it also changes the repertoire of antigenic specificities of the antibodies produced 71 .
It has been demonstrated that antibodies-including natural antibodies-can influence the immune processes through diverse mechanisms, among these being opsonization of microbes, complement activation, formation of circulating antigen-antibody complexes and their trapping in lymphoid organs, antibody-dependent cellmediated bactericidal activity, signaling through Fc receptors, and modulation of the inflammatory response. Moreover, natural antibodies may prevent pathogen dissemination to target organs and influence the immunogenicity through enhanced antigen-trapping in secondary lymphoid organs 72,73 . All the activities of natural antibodies mentioned above can modulate the induction or regulation of acquired immune responses 73 .
Among the early induced natural antibody clones were clones reacting with ClpB, SodB, KatG, Hfq, IglA, IglB, and IglC proteins, which already had been demonstrated to be Francisella virulence factors responsible for intracellular survival and stress response 74 . In addition to the KatG and SodB proteins that are secreted, antibody clones reacting with the other secreted proteins DnaK and GroES also have been identified. It is noteworthy that antibody clones reacting with these virulence factors encompassing both unique bacterial proteins and proteins having eukaryotic orthologs or analogs, with the exception of SodB, were identified only at time point 12 h postinfection. All these antibody specificities may potentially modulate the early stages of interaction between Francisella and the host cell. In this context, we would mention that KL Elkins et al. have disclosed in several studies an early T-cell-independent and B-cell-dependent specific or eventually nonspecific protective response that can be demonstrated between the second and third day after primary sublethal infection of mice 33,62,75,76 . Due to the fact that, according to a notion prevailing since the late 1990s, the antibodies are produced up to about the fifth or seventh day after vaccination or infection, early resistance to Francisella and/or Listeria has not been attributed to the involvement of antibodies. According to the results presented here, however, the early production of infection-induced natural antibody clones reacting with F. tularensis proteins may be the critical partner of the early cellular responses controlling the presence of bacteria in the body. A search for direct proof of antibody involvement in the induction of this early anti-infective resistance should be a matter for further experiments.

Conclusion and perspectives
We conclude that both GF and SF mice respond to F. tularensis strain FSC200 intraperitoneal infection by early IgM, IgG3, and IgA antibody production that can be detected in serum as early as 12 h after infection. Some antibody specificities production has a temporary character. The spectrum of 87 antibody specificities, consisting of a group of bacterial proteins, of which 63 have orthologs or analogs in eukaryotic cell proteins, along with the kinetics of their production, lead to the conclusion that this is a production of natural antibodies, probably by B-1a cells induced by F. tularensis intraperitoneal infection. Preliminary testing of cell subtypes infection in murine peritoneum proved B-1a cells to be the dominant target of Francisella invasion (data not shown). It had already been demonstrated that body cavity B-1 cells are activated by infection and can be considered-unlike lymph tissues B-1 cells-as responder cells 70,77,78 . Our results, however, do not correspond to those obtained in a model of influenza infection where no specific IgM antibodies in serum were detected at early intervals after intranasal infection 77,78 . The reason for this difference can be seen in the different cell types, and including B cell subtypes, that encounter infectious agents as first responders on the nasal mucosa and in the lung in comparison with those which are first responders in peritoneum. Differences in pathogen type (i.e., virus versus bacterium) also can play a substantial role.
Such an early production of antibodies raised several questions. First, where is the origin of this antibody production? Is it directly in the peritoneum or does it occur after translocation of B cells to lymph organs? Further, which B cell subset(s) really produce(s) these antibodies and which receptor(s) and signaling pathways are involved in this phenomenon? What is the primary target of the antibodies thus produced? Are they primarily autoantibodies or are they already targeted, due to the phylogenetic relationship between eukaryotes and prokaryotes, as well as phylogenetic memory, to components of an invading microorganism? Finally, how do these early produced antibodies participate in the Francisella-host cell interactions and to what extent can they influence the induction of protective immune response? The positive effects of B-1a cells and of antibodies produced early by B-1a cells already have been reported [79][80][81] .
The phylogenetically stabilized defense mechanism can utilize these early infection-induced antibodies both to recognize the components of the invading pathogens and to eliminate the pathogen-damaged own cells. Such www.nature.com/scientificreports/ early defense steps can thus be characterized as a booster of opsonophagocytosis (or neutralization) of pathogens that is one of the key events in the induction of acquired immunity mechanisms that already are targeted upon eliminating the pathogen 82 (Table 1). Nevertheless, there remain substantial gaps in our understanding of cooperation between the early humoral and cellular branches of the innate and acquired immune systems in expression of protective responses to F. tularensis infection that can lead us to the design of an effective and safe vaccine.

Material and methods
Bacteria. Francisella 83 was prepared according to the manufacturer's instructions, pH was adjusted with HCl to 6.8 (unless otherwise stated), and the medium was sterile-filtered rather than autoclaved. Following incubation, bacterial colonies were lifted from the plates and resuspended in saline to reach optical density 1.00 (corresponding to bacterial 5 × 10 9 CFU/ml). The actual number of bacteria in the suspension utilized for the experiments was determined by serial dilutions and the number of colony-forming units (CFU) was calculated. Klebsiella pneumonia and Pseudomonas aeruginosa CCM 3955 were cultivated under standard cultivation procedures using blood agar plates.

Animals.
Female BALB/c-SPF mice (SPF) were purchased from Velaz (Unetice, Czech Republic). Female germ-free (GF) BALB/c (hereafter GF mice) mice were kindly provided by the Institute of Microbiology of the Czech Academy of Sciences (Department of Gnotobiology, Novy Hradek, Czech Republic). The GF mice were housed under sterile conditions in Trexler-type plastic isolators with free access to autoclaved tap water and 50 kGy-irradiated sterile pellet diet Altromin 1414 (Altromin, Lage, Germany). Before experiments GF mice were transferred into sterile indiviually ventilated cages (IVC, Tecniplast, Italy). The SPF mice were housed under specific-pathogen-free conditions and GF mice were housed in micro-isolator cages under germ-free conditions. The mice were placed into sterile boxes with air-conditioning and stabilized temperature of 22 ± 2 °C. Light mode was 12 h of light and 12 h darkness 45   www.nature.com/scientificreports/ The retrieved suspension was centrifuged at 400×g for 5 min to concentrate any cellular components. Peritoneal lavage cells were then lysed with RIPA buffer (Thermo Fisher Scientific, USA) to obtain extracted proteins. The extraction was done according to the manufacture's procedure. The protein suspension was subsequently precipitated using trichloroacetic acid/sodium deoxycholate (TCA-DOC) with acetone wash. Briefly, the protein suspension was mixed with 1/10 of the volume of 0.15% sodium deoxycholate (DOC) and incubated on ice for 15 min. The sample was then mixed with 1/10 of the original volume of 72% trichloroacetic acid (TCA) and the suspension was again incubated on ice for 15 min. Subsequently, the mixture was microfuged at 12,000×g, at 4 °C for 15 min. Supernatant was discarded and 1 ml of acetone chilled to − 20 °C was added. All steps of TCA-DOC precipitation were repeated twice. The pellet was air-dried for 5 min, subsequently resuspended in a minimal volume with 2-D sample buffer, then subjected to 2-D gel analysis.
Infection of mice and collection of murine sera. The SPF and GF mice were intraperitoneally inoculated with an F. tularensis FSC200 dose of 10 2 CFU in 0.2 ml total volume of saline. At three time intervals postchallenge (12 h, 24 h, and 48 h), infected mice were killed by cervical dislocation, blood was collected, and serum was prepared according to a common procedure. Sterility of serum was determined by plating on McLeod agar plates for 72 h. Obtained serum was pooled and tested for specific antibody titers using enzyme-linked immunosorbent assay (ELISA). Serum aliquots were stored in volume 500 μl sera at − 80 °C. Normal murine serum was prepared from blood of Balb/c mice without exposure to any F. tularensis strain. www.nature.com/scientificreports/ Mass spectrometry profiling. In-gel digestion. Selected protein spots were excised from representative colloidal blue-stained gels and subjected to in-gel digestion. Briefly, gel pieces were covered with 100 mM Tris/HCl (pH 8.5) in 50% acetonitrile (ACN) at 30 °C for 20 min. Next, 50 mM ammonium bicarbonate (pH 7.8) in 5% ACN was added to each gel piece. Subsequently, gel pieces were vacuum-dried, covered with 0.1 μg of trypsin in 50 mM ammonium bicarbonate (pH 7.8) in 5% ACN, then gently shaken at 37 °C overnight. Protein digests were then extracted using 2% trifluoroacetic acid and 98% ACN. The final extracts were reduced using a vacuum dryer to approximately 15 μl and frozen at − 80 °C. The resulting peptides were mixed with 5 mg/ml of α-cyano-4-hydroxycinnamic acid in 50% acetonitrile and 0.1% trifluoroacetic acid. Peptide samples with matrix solution were spotted onto a MALDI plate.

Serum analysis procedures.
Protein identification and characterization. The protein identification and characterization analysis of the murine sera have been done according 45,84 . Briefly, the mass spectra were recorded in the positive reflectron mode on a 4800 MALDI-TOF/TOF mass spectrometer (AB Sciex, Foster City, CA, USA) and operated in its delayed extraction mode. Internal calibration of mass spectra was conducted utilizing the tryptic autolytic peptides. Acquired data were evaluated using GPS Explorer Software v.3.6 (AB Sciex, Australia) (https ://sciex .com/ produ cts/softw are), which integrates the Mascot search algorithm against the Francisella tularensis subsp. holarctica OSU18 genome database. The following parameters were used for protein identification: 100 ppm error, optional oxidation of methionine, fixed carbamidomethylation of cysteine residues, and one possible missed cleavage site allowed. Proteins were considered to be identified with confidence when a minimum of two peptide sequences per protein were identified and the GPS protein score confidence interval was equal to 100%. Trypsin was selected as the proteolytic enzyme. The identities of all proteins were determined according to the UniProtKB database 85 (https ://www.unipr ot.org).

Statistical analysis.
Each immunoblot experiment was conducted at least in triplicate while using three biological replicates. Pearson correlation coefficient was calculated using GraphPad Prism 5.01 software (Graph-Pad Software, San Diego, CA, USA) (https ://www.graph pad.com/scien tific -softw are/prism /). All statistical analyses were also performed using GraphPad Prism 5.01 software (GraphPad Software, San Diego, CA, USA). In case of Quantibody immunoglobulin isotype array, four technical replicates were measured (according to the manufacturer's instructions). Reference antibody isotypes and light kappa and lambda chains are presented as well as reference immunoblots. Values are expressed as mean ± standard deviation (SD) and analyzed for significance using Student's two-tailed t-test. Differences were considered statistically significant at **p < 0.01 and *p < 0.05, unless otherwise stated, and are denoted by asterisks.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).