The initial interaction of a pathogen with the host is a critical time for both the pathogen and the host. This early encounter with the pathogen is so vital for survival of the host that through the course of evolution invertebrates, vertebrates, and plants have developed an inheritable innate immune system to counteract pathogens. The innate immune system is one arm of the host’s defense immune system that coupled with the acquired immune system protects the mammalian hosts from an array of different pathogens. The innate immune system responds to an assortment of ligands associated with the pathogen that have been classified as pathogen associated molecular patterns (PAMPs) that require special pathogen recognition receptors (PRRs) for detection (Janeway and Medzhitov, 2002). Subsequently, the acquired immune response involves the host’s T- and B-cells that respond to signals produced by the innate immune system to direct the acquired immune response in a more pathogen-specific manner (Akira et al., 2006).

One family of PRRs is the Toll-like receptors (TLRs) of which humans have 10 members and mice have 12 (Takeuchi and Akira, 2010). The location that TLRs are assembled by the host cell reflects the type of ligand that it might encounter during the interaction with a pathogen. TLR1, TLR2, TLR4, TLR5, TLR6, and TLR11 are found on the surface of cells. TLR11 in mice recognizes uropathogenic bacteria (Zhang et al., 2004) or a profiling-like protein from the protozoan Toxoplasma gondii (Yarovinsky et al., 2005), but in humans, it is present as a pseudogene. On the other hand, TLR3, TLR7, TLR8, and TLR9 are expressed in intracellular endosomal compartments (Kawai and Akira, 2010). For the purpose of this present report, we are focusing on only the surface assembled TLRs (see Figure 1 ). In the former group, TLR1 combines with TLR2 (heterodimer) to recognize triacyl-lipopeptides, and the heterodimer TLR2/TLR6 recognizes diacyl-lipopeptides (Akira et al., 2006; Kang et al., 2009; Kurokawa et al., 2012; Nakayama et al., 2012). The TLR4 homodimer recognizes lipopolysaccharides (LPS) but may require other accessory factors, such as LPS binding protein (LBP), myeloid differentiation factor 2 (MD2), and CD14 for maximal activation (Poltorak et al., 1998; Shimazu et al., 1999; Akira et al., 2006). Unlike other TLRs assembled on the surface of the cell, a subset of TLR4-MD2-LPS complexes may be recruited into an endosomal compartment to activate an alternative signal transduction pathway for the induction of proinflammatory cytokines and type I interferons (IFNs) (Kagan et al., 2008). The homodimer TLR5 is activated by the bacterial flagellin subunit, which is the major subunit of the assembled bacterial flagella (Hayashi et al., 2001; Donnelly and Steiner, 2002; Smith et al., 2003).

Our interest has been focused on specific Gram-negative, bacterial pathogens (Yersinia pestis, Burkholderia mallei, Burkholderia pseudomallei, and Francisella tularensis) that are considered Tier 1 biological agents by the Centers for Disease Control and Prevention and are also Biosafety Level 3 (BSL3) biological agents. When the surface of these Gram-negative pathogens was examined, we saw an array of potential structures/organelles that protected or shielded the pathogen from unfriendly environments, for example, antimicrobial peptides or antibiotics or the host’s immune response. In Figure 1 we show an example of potential components on the outer surface of the Gram-negative, motile pathogen B. pseudomallei 1026b, which is the etiological agent of melioidosis, that could interact with TLRs on the surface of host cells. One of the most abundant molecules is lipopolysaccharide (LPS), which has been estimated to cover three-quarters of the bacterial surface with approximately 1-2 X 10⁶ molecules (Whitfield and Trent, 2014). Lipid A is the endotoxin part of LPS and is the agonist for the host’s TLR4. The acylated Lipid A is embedded within the outer membrane leaflet and anchors the LPS molecule to the surface of the pathogen. Other structures that may be present are extracellular polysaccharides (EPS) that include a capsule and/or exopolysaccharide molecules. Both molecules protect the bacteria from adverse environmental conditions, from the host’s immune responses, such as opsonization/phagocytosis, and from complement mediated bacterial lysis (Whitfield, 2006; Warawa et al., 2009; Whitney and Howell, 2013; Schmid et al., 2015). Another molecule present is lipoprotein (Lpp) (Braun lipoprotein), which is also one of the most abundant proteins of the bacterial cell (~1 x 10⁶ molecules per cell) (Braun and Rehn, 1969; Li et al., 2014). Most Lpps are unbound or free (2/3) and a smaller fraction (1/3) is in the periplasm embedded in the inner leaflet of the OM, with some molecules linked to the peptidoglycan (PG) layer (bound portion) (Braun and Rehn, 1969; Inouye et al., 1972). Diacylated and triacylated lipoproteins are synthesized and inserted into the membrane of bacterial cells (Wilson and Bernstein, 2016). A subset of the unlinked or free Lpp molecules, however, is exported through the OM and embedded into the outer leaflet of the OM with the carboxy-terminal end exposed to the outer surface of the bacterial cell (Cowles et al., 2011; Kovacs-Simon et al., 2011; Nakayama et al., 2012; Wilson and Bernstein, 2016; Asmar and Collet, 2018). The presence of triacylated lipoproteins was once thought to be exclusive to Gram-negative bacteria, but they have also been found in some Gram-positive bacteria, which were reported to only synthesize diacyl lipoproteins (Kovacs-Simon et al., 2011; Kurokawa et al., 2012; Nakayama et al., 2012).

Another bacterial organelle associated with some bacteria is the flagellum that is involved in motility and chemotaxis, which leads to enhanced adhesion, biofilm formation, and virulence (Josenhans and Suerbaum, 2002; Galyov et al., 2010; Duan et al., 2013; Haiko and Westerlund-Wikstrom, 2013). The flagellin molecule, which is the major subunit of flagella, can be divided into 4 domains (D0, D1, D2, and D3) (Yonekura et al., 2003; Ramos et al., 2004; Beatson et al., 2006). The amino acids in domains D0 and D1 contain the most conserved sequences within the flagellin molecule (Yonekura et al., 2003; Ramos et al., 2004). The highly variable D2 and D3 domains are arranged into folded β-sheets and protrude outward from the assembled flagellum core. The highly conserved sequences in regions D0 and D1 of flagellin presents a protein signature that are the target for TLR5 (Hayashi et al., 2001). At least one region in the N-terminal D0 region and another in the C-terminal D1 region of flagellin are required for activation of TLR5 (Smith et al., 2003; Andersen-Nissen et al., 2005). Furthermore, for full activation of TLR5, the interaction of two TLR5 polypeptides (homodimer) and two flagellin subunits are required (Yoon et al., 2012; Ivicak-Kocjan et al., 2013; Forstneric et al., 2016).

Pathogenic bacteria have evolved methods to mitigate the protective innate inflammatory response of the host. One of the best examples is presented by the Gram-negative pathogen Y. pestis, the etiological agent of the zoonotic disease plague (Perry and Fetherston, 1997). During the transmission of Y. pestis and infection of the mammalian host, the pathogen is subjected to two different temperature conditions (Hinnebusch et al., 2017). While in a flea at 23-28°C, and while in the mammalian host at 37°C. Kawahara et al. (2002) reported that Y. pestis grown at 27°C produced a mixture of tri-, tetra-, penta-, and hexa-acylated lipid A molecules. However, when Y. pestis was grown at 37°C, only the lesser acylated lipid A forms were present. When they examined the biological activity of lipid A from Y. pestis grown at 27°C it was more active on human cells in inducing the production of TNF-α than lipid A from the pathogen grown at 37°C. The temperature dependent changes in acylated forms of lipid A were also seen in two other pathogenic Yersinia strains (Y. pseudotuberculosis and Y. enterocolitica) (Rebeil et al., 2004). To test the hypothesis that Y. pestis can alter lipid A biosynthesis to overcome the host innate immune response, Montminy et al. (2006) genetically modified Y. pestis to produce a more potent LPS. In this case, the Y. pestis strain with the more potent lipid A substantially activated the innate immune system compared to the wild-type pathogen and was cleared from the host more effectively than the wild-type strain in a mouse model of plague.

Although much is known about the molecular structures of PAMPs and TLRs, it is not clear how these two components interact with each other in situ. For example, when the mammalian host is first exposed to an aerosol of the pathogen, the pathogen may enter through the nasal cavity where it may proceed to infect the host. When and how the innate immune system is activated during this initial phase of infection is still unclear, and it may be confined to localized areas of the host depending on the number of pathogenic bacteria involved. Currently, it is speculated that TLRs are activated by PAMPs that are released or escaped from the pathogen during growth or eventually released in the presence of human serum (Tesh et al., 1986; Zhang et al., 1998; Hellman et al., 2000; Kitchens et al., 2000; Vesy et al., 2000; Matsuura, 2013). In previous innate immune studies with Yersinia pestis the bacterial lipid components were purified or bacterial cells inactivated before they were examined for their ability to activate TLRs (Kawahara et al., 2002; Rebeil et al., 2004; Montminy et al., 2006; Matsuura et al., 2010). Thus, we asked if we could detect changes in TLR2 and TLR4 activation by a live pathogen, such as Y. pestis CO92, when grown at 28°C (more potent lipid A) or at 37°C (less potent lipid A) that would reflect the modification of lipid A at the two temperatures as previously reported (Kawahara et al., 2002; Rebeil et al., 2004; Montminy et al., 2006). We also asked how other live Tier 1, Gram-negative pathogens, such as Burkholderia mallei (Bm), Burkholderia pseudomallei (Bp), and Francisella tularensis (Ft) Schu S4, would interact with specific PRRs on a mammalian cell. B. mallei is the etiological agent of glanders, which is a zoonotic disease usually associated with horses, donkeys, and mules, and is found in the Middle East, Africa, and Central and South Americas (Whitlock et al., 2007; Verma et al., 2014). B. pseudomallei, which causes melioidosis is an endemic disease of humans in South-east Asia and Northern Australia that may mimic other diseases, such as tuberculosis (Cheng and Currie, 2005; Wiersinga et al., 2012). F. tularensis is a small, highly infectious, Gram-negative, intracellular pathogen that causes tularemia (Ellis et al., 2002). Ft subspecies tularensis Schu S4 (Type A) is a highly virulent strain found primarily in North America, while the live vaccine strain (LVS) of Ft is an attenuated stain of the subspecies holarctica (Type B) can be found in Europe, the Far East, and North America. Ft subspecies novicida is another closely related strain of Ft but is less virulent in animal models. The present report is the result of our studies to assess the interaction of live Gram-negative pathogens with TLR2, 4, and 5, that are expressed on the surface of host cells.

Our aim was to assess how live, Tier 1 Gram-negative bacterial cells interacted with cell surface expressed TLRs, which we believe to be representative of the early encounter of the pathogen with the host. In our initial study we investigated the hypothesis that Yp can alter the biosynthesis of lipid A depending on growth at 28 C (more reactive lipid A) or 37 C (less reactive lipid A). We could directly detect the changes in TLR2 or TLR4 activation with whole cells of Yp at the two different temperatures that likely reflected the changes in lipid A biosynthesis. The changes in lipid A biosynthesis at these two temperatures have been well documented with hexa-acyl lipid A being synthesized at 28 C (Kawahara et al., 2002; Rebeil et al., 2004; Montminy et al., 2006; Matsuura et al., 2010). At the higher temperature of 37°C associated with mammalian infection, TLR2 was primarily activated with concurrent suppression of TLR4 activation that coincided with the biosynthesis of tetra- or penta-acylated modified forms of lipid A as previously reported (Kawahara et al., 2002; Rebeil et al., 2004; Montminy et al., 2006; Matsuura et al., 2010). We were not able to observe the same temperature effects on TLR2 or TLR4 activation by Bp K96243 or Bm FMH. This observation supports a recent report that growth of Bp K96243 at different cultural conditions did not alter lipid A modification and that by MALDI-TOF lipid A spectra analysis tetra- and penta-acylated species were consistently present (Sengyee et al., 2018).

Another observation we noted using live, whole cells was the mitigation of TLR2 and TLR4 activation by the Bp 1026b capsule. The capsule of Bp and the closely related Bm is a virulence factor and contributes to resistance to serum killing, decrease in opsonization, and persistence in infected hosts (DeShazer et al., 2001; Atkins et al., 2002; Reckseidler-Zenteno et al., 2005; Warawa et al., 2009; Zhang et al., 2011; Heiss et al., 2012; Woodman et al., 2012). These anti-microbial properties of the capsule can be attributed in part to the prevention of complement C3 and antibody deposition on the pathogen (Reckseidler-Zenteno et al., 2005; Zhang et al., 2011; Woodman et al., 2012). We now showed that capsule can attenuate the early innate immune response of the host to the presence of the pathogen. This in turn could directly affect the expression of cytokines/chemokines needed for the early recruitment of innate immune cells, such as phagocytic macrophages and neutrophils, and NK cells, to limit the spread of the invading pathogen (Woodman et al., 2012; Wiersinga et al., 2018). Thus, we have revealed another important immune property of the capsule that contributes to the virulence of these Gram-negative pathogens.

Our results, however, were different from a previous report that found no significant difference in TLR4 activation by a capsule negative Bp compared to the wild-type Bp strain (Sengyee et al., 2018). One major difference between our study and that of Sengyee et al. (2018) was that they used heat-inactivated bacterial cells (80°C for 1 h) to assess TLR4 activation, while we used live Bp in our studies. High temperatures have been reported to cause changes in the outer membrane and loss of LPS of Gram-negative bacteria (Russell, 2003). In addition, heating (70°C for 15 min) depolymerized flagella of Salmonella typhimurium which would alter its interaction with TLR5 (Smith et al., 2003). We did see another difference in our studies with Bp 1026b wild-type and capsule mutant LPS extract in the activation of TLR2 or TLR4 as previously observed (West et al., 2008). They reported that their Bp LPS preparations activated TLR4 and not TLR2, and that TLR4 activation was dependent on CD14 and MD-2. In our studies we saw a higher TLR2 response than a TLR4 response to LPS from both Bp 1026b wild-type and capsule mutant LPS extracts, which was similar to the TLR response to live Bp cells. On closer examination of their report it appears that their Bp LPS (and lipid A) preparations were more highly purified than the Bp LPS extracts used in our studies. They eliminated the presence of phospholipids and other lipoproteins that could activate TLR2. Nevertheless, we did not see any apparent differences between the wild-type and capsule mutant LPS extracts by gel analysis and TLR activity.

Our studies with Francisella strains revealed the absence of TLR4 activation compared to the other Gram-negative pathogens in the current study. This was most striking with Ft Schu S4 (a type A strain) cells where even TLR2 activation was absent. Ft Schu S4 had been previously reported to show little stimulation of the innate immune response in human dendritic cell culture (Chase et al., 2009), or in a microarray analysis of infected human monocytes (Butchar et al., 2008). The lack of TLR2 and TLR4 activation by Ft Schu S4 cells may reflect in part the presence of an unusual lipid A moiety associated with Francisella spp. (Wang X. et al., 2006; Raetz et al., 2009). Approximately 90% of lipid A embedded in the outer membrane has been described as “free lipid A” which was present in Ft LVS and F. novicida (Wang X. et al., 2006). This fraction of lipid A lacks Kdo and hence is not glycosylated with core subunits, or contains O-antigen sugars. Another major difference in lipid A in Francisella is the presence of only tetraacylated chains that are 16 and 18 carbons in length compared to the variable number of acylated chains that are typically 12 to 14 carbons in length in many bacteria (Raetz et al., 2009; Okan and Kasper, 2013). In contrast to Ft Schu S4, Ft LVS (a type B strain) and F. novicida, which are not considered human pathogens, exhibited significant TLR2 activation. LPS prepared from Ft Schu S4 and Ft LVS, also activated TLR2. Because Ft Schu S4 cells were negative for activation of TLR2 and TLR4, it is likely that TLR2 activation by LPS from Ft Schu S4 is from the presence of tri-acylated lipoproteins, or other TLR2 activating polysaccharides, for example capsule, in the LPS extract (Kim et al., 1996; Cole et al., 2006; Wang et al., 2006; Cole et al., 2007; Butchar et al., 2008; Thakran et al., 2008). We also do not know how these components are organized on the surface of Ft Schu S4 and Ft LVS, however, because we see differences between the activation of TLR2 by the two different Ft species, it suggests that organization of the outer surface of the microorganisms is different. Tri-acylated lipoproteins have been generally associated with the outer membranes of Gram-negative bacterial and activate the heterodimer TLR1/TLR2, although some Gram-positive microorganisms are able to synthesize both di-and tri-acylated lipoproteins (Takeuchi et al., 2002; Jin et al., 2007; Thakran et al., 2008; Kovacs-Simon et al., 2011; Nakayama et al., 2012). On the other hand, di-acylated lipoproteins are recognized only by the TLR1/TLR6 heterodimer. LPS from Ft LVS and F. novicida had previously been reported to show no activation of TLR2 or TLR4 (Cole et al., 2006; Hajjar et al., 2006), therefore, activation of TLR2 with cells could be from the lipoproteins or other glycoproteins associated with the outer membrane of the cell (Thakran et al., 2008). Since Ft Schu S4 cells did not activate TLR2 or TLR4 it implicates the presence of another component, such as a capsule or envelope that shields the pathogen from the host’s innate immune response differently from that in the less pathogenic Francisella strains (Ft LVS or F. novicida) (Rowe and Huntley, 2015).

Our observation using live, whole cells with flagella attached to Bp K96243 or Bt E264 activated TLR5 was noteworthy. Activation of TLR5 was authenticated in a number of ways: 1. Purified FliC from St or Bp K96243 activated TLR5; 2. An anti-TLR5 mAb inhibited TLR5 activation by FliC from both these microorganisms but did not affect TLR2 or TLR4 activation; 3. The anti-TLR5 mAb also eliminated TLR5 activation by whole-cells of motile pathogenic Bp K96243 and motile nonpathogenic Bt E264; 3. Filtration of Bp K96243 and Bt E264 cells through a 0.2 um filter eliminated TLR5 activation; and 4. Deletion of fliC in both Burkholderia species abrogated TLR5 activation by both mutants. A high resolution crystal structure of Bp K96243 FliC (residues 25 to 378) has been reported (Nithichanon et al., 2015). The modeled FliC structure contained the folded-domains consisting of the N- and C-terminal regions that are adjacent to each other as found in the St FliC crystal structure. It was not clear if the consensus sequence required for TLR5 activation would be available in this FliC model.

Activation of TLR5 by Bp 1026b FliC has been previously reported (Birnie et al., 2019; Dickey et al., 2019). Whether activation of TLR5 by flagella can occur, or even the possibility of flagella attached to bacterial cells can activate TLR5 has been in doubt (Smith et al., 2003; Andersen-Nissen et al., 2005). Smith et al. (2003) showed that TLR5 activation by filamentous flagellin (flagella) was much less compared to flagellin; however, it still could activate TLR5 albeit at a lower activity. On a weight basis flagellin activated TLR5 3-5 fold more than flagella filaments (~60-fold vs 20-fold activation at ~1 ng/ml, respectively). It was also suggested that heating (70°C, 15 min) flagella filaments disassembled the structure and released flagellin subunits. It was not clear to what extent disassembly would occur. In a follow up study heat-inactivated, flagellated bacteria was used to examine TLR5 activation by bacteria from different classifications (Andersen-Nissen et al., 2005).

Our results suggest that either Burkholderia flagellin or flagella can activate TLR5. For activation of TLR5 to occur the flagellin molecule must occupy a binding site on each TLR5 monomer prior to homodimer formation and eventual activation of MyD88-dependent signaling to activate NF-κB (Yoon et al., 2012; Ivicak-Kocjan et al., 2013). Consequently, for TLR5 dimer formation to occur with a flagellum it may take one (linking both TLR5 monomers) or two flagellums with each occupying a FliC binding domain on a TLR5 monomer to initiate the dimerization and activation process. These proposed events with flagella would be a more complex process than that initiated with flagellin molecules. This could explain the difference in the ability of soluble flagellin compared to insoluble flagella to activate TLR5 (Smith et al., 2003). Also, because the sequence of Bp K96243 FliC was not identical to that of St FliC (33% identical overall), the TLR5 recognition site in Bp K96243 FliC may be more accessible than in St FliC. We did see that the Bp K96243 FliC could activate TLR5 at a higher rate than an equivalent amount of St FliC (see Figure 7 ), even after considering the differences in mass of the FliC molecules. More studies will be needed to further confirm our observation of activation of TLR5 by the Bp flagellum or by other strains. In contrast, there are motile pathogenic bacterial strains that have evolved to avoid activating the host’s TLR5, such as Helicobacter pylori and Campylobacter jejuni, that have changes in the critical 89 – 96 amino acid region of FliC needed for TLR5 activation (Ramos et al., 2004; Andersen-Nissen et al., 2005). It would be an advantage for the host to have the ability to detect either flagellin or flagellum to mitigate infection by a motile pathogen. Heating to inactivate the pathogen may well raise questions of the quality of the interaction with TLRs. Nevertheless, microorganisms have been heat-inactivated to examine their interaction with TLR 5 (or other TLRs) (Andersen-Nissen et al., 2005; West et al., 2008; West et al., 2013; Chaichana et al., 2017; Birnie et al., 2019; Dickey et al., 2019). Hence, we used two methods in the present study to examine intact flagella interaction with TLR5. Firstly, we used live, motile Bp cells to inoculate HEK293-TLR5 cells, and secondly, a filter (0.2 µm) to exclude motile bacteria (with flagella attached, see Figure 1 ) to demonstrate that TLR5 activation could be abolished. This was confirmed with two different species of Burkholderias. It cannot be ruled out that there was some free flagellin in our cell suspension, however. One exception where eliminating bacterial cells by filtration did not completely abolish TLR5 activation in the cell-free filtrate was in the study with the Bt E264 (DDI3196+pMo168+fliC) partially complemented fliC mutant ( Figure 8F ). With the complemented fliC mutant, we saw a large increase in TLR5 activation compared to that of the wild-type parent strain. After a suspension of the complemented fliC mutant was passed through a 0.2 um filter, TLR5 activation was abrogated, except at the higher cell concentrations where we saw an increase in TLR5 activity in an equivalent amount of filtrate. We believe this to be FliC present in the filtrate that was over produced by the expression plasmid pMo168+fliC and, subsequently, secreted or leaked from the complemented mutant. Further studies are needed to examine the phenotype of the complemented fliC strain more closely as regards to the presence of flagella and to examine the efficacy of filtration of other motile microorganisms to prevent TLR5 activation.

Although in the past many interactions of TLRs have been evaluated with purified agonist in the absence of the microorganism, our results suggest that early activation of TLRs may also occur when PAMPs are still associated with a live infectious agent. We believe this to be a possibility in the initial stages of infection even before the invading agent has established an active infection. It has been suggested that the agonist must be dissociated (or secreted) from the microorganism, for example during growth, before it can interact with TLRs (Tesh et al., 1986; Zhang et al., 1998; Hellman et al., 2000; Kitchens et al., 2000; Vesy et al., 2000; Matsuura, 2013). Possibly, both free and associated PAMPs can activate TLRs during the early period of infection of the host by the pathogen. On the other hand, the agonist might dissociate from the microorganism after binding to the TLR or with the participation of an unidentified enzyme/mechanism that can free the agonist from the surface of the microorganism. Without the free agonist, the TLR4 endocytic pathway (TRIF-dependent) would be excluded. Further studies are needed, for example cross-linking studies, to examine the early interaction of the microorganism with TLRs to resolve some of the possible mechanisms for TLR activation in the host.

In conclusion, it has become more apparent in the evolution of successful pathogenic bacteria that avoidance of the host’s TLR4 response is a common theme (Arpaia and Barton, 2013). Taken together the host’s surface expressed TLRs, TLR2 (TLR1/2, TLR2/6), TLR4, and TLR5, all recruit a common cytosolic signal transduction protein MyD88 that leads to the eventual activation of NF-κB and AP-1. Therefore, why have pathogens appear to commonly evolved mechanisms to suppress or avoid activation of TLR4? It may be because activation of TLR4 can lead to two host innate immune responses to the pathogen: a MyD88-dependent response and a TRIF-dependent response. The alternative TLR4 endocytic pathway or TRIF-dependent response leads to the production of anti-microbial type 1 interferons (IFNα/β) and delayed activation of NF-κB (Barton and Medzhitov, 2003; Kagan et al., 2008). Depending on the pathogen, IFNα/β appears to restrict pathogen growth, activate macrophages, and upregulate protective cytokines, although there were some reports of detrimental effects on the host (McNab et al., 2015). Studies on the early interaction of live pathogenic microorganisms with specific TLRs could augment the discovery of new therapeutic compounds for the treatment of infectious diseases.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

KA and DD designed, performed experiments, and wrote the manuscript. RB and JD performed experiments and contributed to writing the manuscript. DF performed the statistical analyses of the studies and contributed to the manuscript. DW was involved in the design of the studies and contributed to the manuscript. PW was involved in the design, funding, administration of the studies and contributed to the manuscript. All authors approved the submitted version of the manuscript.

The authors would like to thank JSTO/Defense Threat Reduction Agency for their support of USAMRIID project no. 923678.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.