Autotransporter-Mediated Display of Complement Receptor Ligands by Gram-Negative Bacteria Increases Antibody Responses and Limits Disease Severity

The targeting of immunogens/vaccines to specific immune cells is a promising approach for amplifying immune responses in the absence of exogenous adjuvants. However, the targeting approaches reported thus far require novel, labor-intensive reagents for each vaccine and have primarily been shown as proof-of-concept with isolated proteins and/or inactivated bacteria. We have engineered a plasmid-based, complement receptor-targeting platform that is readily applicable to live forms of multiple gram-negative bacteria, including, but not limited to, Escherichia coli, Klebsiella pneumoniae, and Francisella tularensis. Using F. tularensis as a model, we find that targeted bacteria show increased binding and uptake by macrophages, which coincides with increased p38 and p65 phosphorylation. Mice vaccinated with targeted bacteria produce higher titers of specific antibody that recognizes a greater diversity of bacterial antigens. Following challenge with homologous or heterologous isolates, these mice exhibited less weight loss and/or accelerated weight recovery as compared to counterparts vaccinated with non-targeted immunogens. Collectively, these findings provide proof-of-concept for plasmid-based, complement receptor-targeting of live gram-negative bacteria.


Introduction
From the discovery of a new pathogen, it can be many years before a vaccine that is specific to the organism is developed. Current methods involve the identification of virulence factors and pathogen-specific antigens (Ags) to compose subunit vaccines. Alternatively, inactivated whole cell vaccines are often non-stimulatory and require the addition of adjuvant for production of an effective immune response [1]. While adjuvants are effective at increasing innate immune activation, several drawbacks (non-specific immune cell activation, potential side effects, such as

Conception and Applicability of Autotransporter (AT)-Complement Receptor Ligand (CRL) Fusions to Multiple Gram-Negative Bacteria
Previously, we used an opsonizing Ab that was specific for the surface of Ft (mAb α-Ft LPS, IgG2a) to target Ft LVS vaccines to FcRs [17,18]. While this approach enhanced vaccine efficacy, it relied upon having pre-existing pathogen-specific reagents. Here, we sought to develop a rapid and broadly-applicable method of targeting intact bacteria to immune cells and pursued bacterial surface-expression of CR ligands (CRLs) (Figure 1a). We constructed our targeting platform on the backbone of the Yersinia enterocolitica (Ye) OMP YadA (Yersinia Adhesion A). YadA is an autotransporter with a N-terminal signal sequence, a collagen-binding domain (CBD), helical stalk, and a C-terminal β-barrel that anchors the protein into the outer membrane (OM) (Figure 1b). In Ye, mature YadA is a heat-resistant 132 kDa trimer, in which the stalk and CBD are surface-exposed [33][34][35]. We expressed YadA on a broad-host range plasmid in Ec and Ft and observed by western blot~44 kDa and~132 kDa α-YadA reactive bands consistent with the detection of monomers and trimers (Figure 1c). In Ec, 50% of YadA appeared to be trimeric, while, in Ft, the trimeric form accounted for~15% of YadA (not shown). In Ft, α-YadA reactive bands partitioned into the sarkosyl-insoluble phase consistent with that of an OMP ( Figure S1a) and YadA-expressing Ft displayed robust collagen-binding (Figure 1d).
We next replaced the collagen-binding domain in YadA with a FLAG tag and the murine complement protein C3d (a ligand for both CR2 and CR3 [24,36,37]) and termed this fusion YFC (YadA-FLAG-C3d). Similar to unmodified YadA, YFC partitioned as expected for an OMP ( Figure S1b,c). We also engineered a fusion harboring the minimal CR2-binding fragment of C3d, p28 [38], in order to yield YadA-FLAG-p28 (YFP) (Figure 1b). We assessed YFC and YFP in three different γ-proteobacteria: Ec, Kp, and Ft. Lysates of recombinant bacteria were probed with α-C3d (Figure 1e) or α-FLAG (Figure 1f). The expression profiles of the AT-CRLs within Ec and Kp are similar with both monomers Pathogens 2020, 9, 375 4 of 21 and trimers apparent. In the more distally related Ft i) YFC was more readily detected than YFP and ii) YFC monomers were more apparent than trimers-similar to the case with unmodified YadA. With enhanced detection, YFC-trimers are apparent in Ft, albeit at lower levels than in Ec ( Figure S2). Together, these results suggest that YadA-CRL fusions and heterologously-expressed YadA trimerize to a roughly similar extent within a given bacterium. Next, we tested the surface-accessibility of our fusions by using intact bacteria to immunoprecipitate (IP) Ab specific for FLAG, C3, or the cytoplasmic Ft protein, IglC [39], with the latter serving as a negative control (Figure 1h). YFP supported robust Ab IP by Ec and Kp, whereas YFC appeared similarly effective in Ft (Figure 1g). These observations suggested that Ec:YFP, Kp:YFP, or Ft:YFC could be tested as vaccines in models of Ec, Kp, or Ft infection, respectively. As C3d interacts with two immune receptors (CR2 and CR3), we reasoned that analysis of YFC provided the best opportunity to detect the impact of CR-targeting in vivo. Accordingly, we pursued analysis of Ft:YFC and tularemia models of infection.
Pathogens 2020, 9, x FOR PEER REVIEW 4 of 20 than YFP and ii) YFC monomers were more apparent than trimers-similar to the case with unmodified YadA. With enhanced detection, YFC-trimers are apparent in Ft, albeit at lower levels than in Ec ( Figure S2). Together, these results suggest that YadA-CRL fusions and heterologouslyexpressed YadA trimerize to a roughly similar extent within a given bacterium. Next, we tested the surface-accessibility of our fusions by using intact bacteria to immunoprecipitate (IP) Ab specific for FLAG, C3, or the cytoplasmic Ft protein, IglC [39], with the latter serving as a negative control ( Figure  1h). YFP supported robust Ab IP by Ec and Kp, whereas YFC appeared similarly effective in Ft ( Figure  1g). These observations suggested that Ec:YFP, Kp:YFP, or Ft:YFC could be tested as vaccines in models of Ec, Kp, or Ft infection, respectively. As C3d interacts with two immune receptors (CR2 and CR3), we reasoned that analysis of YFC provided the best opportunity to detect the impact of CRtargeting in vivo. Accordingly, we pursued analysis of Ft:YFC and tularemia models of infection.

Bacterial Expression of YFC Enhances Binding and Uptake by Murine Macrophages
Having determined biochemically that YFC was surface-exposed, we next sought evidence that the murine C3d moiety was functional in promoting binding to CR+ cells. Ft is an intracellular bacterium that can enter many cells (including MΦs) through multiple receptors (including CR3), depending on which opsonins are present [41,42]. We reasoned that Ft genetically decorated with functional C3d should more avidly associate with murine MΦs than Ft lacking C3d. To this end, we incubated Ft:YFC or  impressions were confirmed by quantifying MΦ-bound and total (bound plus unbound) fluorescence over a range of multiplicity of infection (MOI) with a plate-reading fluorometer ( Figure  2a,c). In these assays, we observed that significantly higher percentages of Ft:YFC bacteria remain cell-associated when compared to the empty vector controls. Additional MΦ-binding experiments with un-stained Ft were analyzed by western blot for the bacterial and cellular proteins FopA and βactin (Figure 2b,d). These results further confirmed the notion that the C3d moiety in YFC functions to enhance binding and apparent phagocytosis.  * t test p < 0.05. Results are combined from 4 independent experiments; means shown with standard deviation (SD). Un-stained Ft (-and YFC) were incubated with RAW cells at MOI = 0 ("no Ft") or 500 for 2 h at 4 • C (b) or at 37 • C (d). Following washes to remove un-bound bacteria, cell-associated material was probed by western blot with Ab specific for the Ft protein FopA and the cellular protein β-actin.

YFC-Bearing Bacteria Induce Elevated p38 and p65 Phosphorylation in Murine Macrophages
We next sought to determine whether YFC-expressing bacteria elicit distinct cellular responses. Multiple groups have reported that MΦs interacting with wildtype, unopsonized Ft produce a transient spike in MAPK phosphorylation, which wanes to baseline as NF-κB activity subsequently peaks and resolves [44][45][46][47]. In recent studies of C3-opsinized Ft and CR3 engagement, Dai et al. found that, shortly after phagocytosis, murine MΦs respond to C3-opsonized Ft with increased MAPK phosphorylation [43]. Accordingly, we predicted that YFC-expressing bacteria would also induce altered patterns of p38 and NF-κB (p65) phosphorylation. We incubated Ft:and Ft:YFC in serum-free media with RAW 264.7 cells, which were harvested at short intervals for western blot analysis ( Figure 3a). Shortly following Ft infection, the P-p38 levels decreased while the P-p65 levels increased with no consistent differences between control and YFC-expressing Ft. However, at the later time points, cells that were infected by Ft:YFC contained levels of P-p38 and P-p65 significantly higher than in Ft:infected cells ( Figure 3).
phosphorylation [43]. Accordingly, we predicted that YFC-expressing bacteria would also induce altered patterns of p38 and NF-κB (p65) phosphorylation. We incubated Ft:-and Ft:YFC in serumfree media with RAW 264.7 cells, which were harvested at short intervals for western blot analysis ( Figure 3a). Shortly following Ft infection, the P-p38 levels decreased while the P-p65 levels increased with no consistent differences between control and YFC-expressing Ft. However, at the later time points, cells that were infected by Ft:YFC contained levels of P-p38 and P-p65 significantly higher than in Ft:-infected cells ( Figure 3).

Infection of Mice with YFC-Bacteria Transiently Limits Weight Gain and Induces Higher Titers of Ab
As the C3d portion of YFC appeared to be functional in cell-based assays, we next sought to determine whether YFC-bearing bacteria would provoke altered responses in animals. In addition to CR3 on MΦs, YFC-expressing bacteria in vivo could potentially engage CR3+ dendritic cells (DCs-potent Ag-presenting cells), CR2+ FDCs, and/or CR2+ B cells. Interaction with the latter cells by C3d-linked moieties has been shown to significantly enhance Ab responses. Accordingly, we first sought to determine whether YFC-expressing bacteria are safe in vivo and if they induced altered Ab responses. To this end, we vaccinated mice intranasally (i.n.) with one of two low doses (~50 and 200 CFU) of live Ft LVS bearing either the empty vector (-) or the YFC-expression plasmid. Mice that were vaccinated with the higher dose of Ft:YFC exhibited slight weight loss on d~5-9, whereas the remaining groups steadily gained weight following vaccination; all animals appeared to be healthy and survived (Figure 4a). By ELISA, we determined the d 21 serum Ab titers against wildtype Ft LVS lysates. We observed significantly increased titers of total Ig against Ft, which included total IgG and the IgG2c isotype, among mice that were vaccinated with Ft:YFC ( Figure 4b). IgM trended towards responses. To this end, we vaccinated mice intranasally (i.n.) with one of two low doses (~50 and ~ 200 CFU) of live Ft LVS bearing either the empty vector (-) or the YFC-expression plasmid. Mice that were vaccinated with the higher dose of Ft:YFC exhibited slight weight loss on d ~ 5-9, whereas the remaining groups steadily gained weight following vaccination; all animals appeared to be healthy and survived (Figure 4a). By ELISA, we determined the d 21 serum Ab titers against wildtype Ft LVS lysates. We observed significantly increased titers of total Ig against Ft, which included total IgG and the IgG2c isotype, among mice that were vaccinated with Ft:YFC ( Figure 4b). IgM trended towards increased levels in Ft:YFC vaccinated mice (p = 0.16); IgG1 and IgA were not uniformly detected in either vaccine group (not shown).

Vaccination with YFC-Bacteria Alters Ag Recognition and Limits Morbidity Following Challenge
Having established that Ft:YFC was well tolerated and induced elevated Ab responses, we next vaccinated mice i.n. with a single dose (~200 CFU) of Ft:and Ft:YFC to further characterize vaccine-induced responses. Among Ft:YFC immunized mice, we again observed a slight decrease in weight following vaccination ( Figure 5a) and significant increases in Ft-specific serum Ig, IgG, and IgG2c titers (Figure 5b). We used live wildtype Ft to IP serum Ab and detected bound Ig heavy chain by western blot to determine whether the Ft:YFC-induced sera also contained more opsonizing Ab. Sera derived from Ft:YFC vaccinated mice contains more bacterial surface-binding Ab than does sera from control immunized mice. We also used these sera pools to probe fractionated Ft  Figure 5c. For these assays, the blots were probed with equal titers of the two sera and, as expected, the majority of bacterial Ags were equivalently recognized ( Figure 5d). However, a small sub-set (~3-4) of A-phase proteins between~50-100 kDa were more robustly recognized by the Ft:YFC immune sera. When sera from individual animals were similarly used at equal titer, we again observed elevated reactivity among Ft:YFC-immunized mice for these 3-4 A-phase Ags (Figure 5e). Collectively, Ft:YFC immunized mice produce higher titers of serum Ab that contains more opsonizing Ig and recognizes a broader repertoire of bacterial Ags. Next, these mice were challenged i.n. with one of two doses (12k or 48k CFU) of wildtype Ft LVS. These doses were known to be lethal for naïve controls (LD 50 = 1250 CFU), but expected to be sublethal for LVS immunized mice. Indeed, the PBS controls succumb to challenge by d 9 and all immunized mice ultimately survived. However, differences were apparent in the magnitude and duration of challenge-induced morbidity-as indicated by weight loss. At both challenge doses, Ft:YFC immunized mice lost less weight than their Ft:counterparts and the differences remained significant for 4-7 d (Figure 5f). When the days below baseline weight were calculated, these were also significantly lower for Ft:YFC immunized mice. Among the 12k CFU challenge animals, the Ft:YFC immunized mice were below baseline weight for 2.3 +/-1.5 d following challenge as compared to 6.2 +/-0.8 d for the Ft:vaccinated group (p = 0.001). For the 48k challenge recipients, the corresponding numbers were 5.0 +/-0.8 d for the Ft:YFC vaccinated mice and 13.8 +/-5.1 d for the control Ft:mice (p = 0.01). Finally, we compared post-vaccination (PV) and post-challenge (PC) serum Ab titers among these groups and made an unexpected observation. Ft:YFC vaccinated mice, which had higher PV titers and less challenge-induced morbidity, appeared to have a muted Ab response to challenge. For Ft:immunized mice, we observed 16-and 20-fold increases (PC/PV) in mean serum Ig titers for the 12k and 48k doses, respectively (Figure 5g). For Ft:YFC immunized mice, the fold changes were 8 and 17.

Vaccination with YFC-Bacteria Accelerates Recovery Following Challenge with a Heterologous Isolate
In the preceding work, one isolate of Ft LVS was the basis of the vaccine strains and the challenge agent. As virulence can vary among isolates, we sought to determine whether our Ft:YFC strain could similarly protect against challenge with Rocky Mountain Laboratory (RML) Ft LVS, which is~50x more virulent than ATCC LVS [48]. We immunized four groups of mice with our Ft:or Ft:YFC vaccine strains. The Ft:YFC immunized mice again displayed transient weight loss following vaccination (not shown) and significantly increased the serum Ab titers PV (not shown). The vaccinated mice were then challenged i.n. with 10 2 , 10 3 , 10 4 , or 10 5 CFU of RML Ft LVS. PBS control mice were challenged i.n. with 10 2 , 10 3 , or 10 4 CFU of RML Ft LVS. Survival among PBS controls was 100%, 37.5%, and 0%, respectively; vaccinated mice ultimately survived all challenge doses ( Figure S4). We again noted significant morbidity differences between Ft:and Ft:YFC immunized mice (Figure 6a), albeit with some distinctions. In contrast to the homologous challenge results, here the magnitude of peak weight loss did not differ between Ft:and Ft:YFC immunized mice. However, the latter did recover weight more quickly (Figure 6a) in ordeer to yield a trend towards fewer days underweight. Among the 10 5 CFU challenge animals Ft:YFC immunized mice were below baseline weight for 8.8 +/-2.9 d after challenge when compared to 12.6 +/-6.0 d for the Ft:vaccinated group (p = 0.13). Last, we compared PV and PC serum Ab titers and again observed that mice immunized with Ft:YFC required much higher challenge doses to provoke a significant Ab-recall response (Figure 6b). magnitude of peak weight loss did not differ between Ft:-and Ft:YFC immunized mice. However, the latter did recover weight more quickly (Figure 6a) in ordeer to yield a trend towards fewer days underweight. Among the 10 5 CFU challenge animals Ft:YFC immunized mice were below baseline weight for 8.8 +/-2.9 d after challenge when compared to 12.6 +/-6.0 d for the Ft:-vaccinated group (p = 0.13). Last, we compared PV and PC serum Ab titers and again observed that mice immunized with Ft:YFC required much higher challenge doses to provoke a significant Ab-recall response (Figure 6b).

Discussion
We sought to simplify vaccine development through generation of a plasmid-encoded AT-CRL fusion. The transformation of the platform into gram-negative bacteria results in the expression of a CRL (C3d or p28) on the bacterial surface; consequently, targeting them to CRs on immune cells. Engagement of CR3 on MΦs induces phagocytosis, signaling, and entrance into peripheral lymph nodes [49]. CR2 engagement on B cells and FDCs is also an important aspect of forming an adaptive immune response-inducing prolonged Ag display in follicles, uptake by and activation of B cells, and increased Ab responses [50].
In this work, we have modified trimeric AT YadA of Ye to include a CRL, yielding constructs YFP and YFC (Figure 1). The expression and trimerization of these constructs in Ec, Kp, and Ft varied. Reduced trimeric-sized bands that were observed in Ft:YFC compared to Ec:YFC and Kp:YFC are likely not indicative of a lack of trimer formation in vivo since detectable surface exposure is observed, but rather might be due to the reduced stability of the protein when boiled in reducing sample buffer and resolved via SDS-PAGE. However, reduced surface accessibility of the constructs in some strains might be due to lack of proper chaperone complexes and/or interference by host proteases preventing proper localization, trimerization, and/or transport of passenger domains through the β-barrel. In future work, we will aim to improve the broad utility of our constructs by co-expressing chaperones, such as Ye Skp, which aids in the heterologous expression of YadA [51].
Utilizing Ft as a model, we have shown that YFC expression in bacteria increases association with CR3-expressing macrophages (Figure 2). Previous studies support that this increased interaction is likely due to CRL engagement of CR3 on these cells [41,43], although we have not formerly defined the receptor specificity here. The engagement of TLRs on the surface of these cells is likely also an important aspect of CR3 binding and induction of phagocytosis. TLRs, specifically TLR2 in the case of Ft, initiate inflammatory signaling and inside-out signaling through phosphatidylinositol-3-kinase [43, [52][53][54]. This inside-out signaling results in a conformational change of CR3 from closed to open/active, allowing for CRL engagement [55,56]. In addition to increased binding, we have also observed an increased phosphorylation of p38 and p65 in MΦs at 60 min. post-incubation with Ft:YFC (Figure 3). This largely supports work by Dai et al., who showed similar increases in inflammatory signaling when C3-opsonized Ft SchuS4 were incubated with murine MΦs; contrasted by their similar experiments in human macrophages, which resulted in the inhibition of this inflammatory signaling [43].
The increased signaling we observe could also be due to intracellular detection of C3d [57]. The pathway by which this detection occurs is not fully understood. Some studies have shown that the detection of cytoplasmic C3b by mitochondrial antiviral-signaling protein (MAVS) induces NF-κB pathways, which is consistent with the increased p65 phosphorylation that we observed [58][59][60][61]. Additionally, other groups have found that the intracellular C3b protein induces the formation of autophagosomes and increases the degradation of the accompanying pathogen [58,62], a phenomenon that might also contribute to intracellular pathogen spread via trogocytosis [63] and has been suggested as a virulence mechanism for Ft [64][65][66]. In contrast, Ft has also been shown to evade autophagy host defenses [67]. Further experiments are required in order to elucidate the exact mechanism at play within our system; we propose that assessing MΦ health as well as Ft:YFC viability and/or replication in MΦs over time will contribute to our understanding.
We administered Ft:YFC and Ft:strains to C57BL/6 mice i.n. in order to assess the effect of YFC on the immune response in vivo. On d 21 PV, we observed increased pathogen-specific Ab titers as well as increased diversity in proteins recognized compared to the Ft:control (Figures 4  and 5). This promising result agrees with previous works that have shown similar titer increases when fusing C3d to a subunit vaccine; however, our approach involves Ab development against a wide array of bacterial Ags rather than a solitary peptide. Ft:YFC vaccinated mice also exhibited protection against Ft LVS challenge-induced weight loss (Figure 5f), which suggested that the increased immune response generated translates to an effective response against subsequent pathogen exposure. We assessed whether this protection would translate to protection against challenge with a heterologous, more virulent isolate of Ft, the RML strain [48]. Again, we observed that Ft:YFC vaccinated mice exhibited a faster recovery from challenge-induced weight loss (Figure 6a). In order to assess the effectiveness of our CR-targeted vaccine strain at preventing lethality, we would likely need to use a type A strain of Ft, such as SchuS4, for a highly stringent challenge. Moreover, additional assays would be required to reveal the relative contributions of the humoral and cellular responses to Ft:YFC.
An unexpected observation came from the comparisons among the PV and PC Ab titer determinations. Specifically, the finding that Ft:YFC immunized mice, which had higher PV titers, appeared to have a muted Ab response to challenge (Figures 5g and 6b). This, along with protection from challenge-induced weight loss, supports the notion that higher Ab titers prior to challenge is associated with decreased disease severity and a diminished need for a prominent recall response at low challenge doses; a phenomenon that has been observed in humans [68][69][70][71][72]. Subjects with higher pre-existing Ab titers against flaviviruses [68], Shigella [70,72], and influenza [69,71] experience reduced disease upon respective infection and reduced Ab recall response. Therefore, we believe that our results are likely translatable to other diseases as well as human models of infection.
Our advancement in Ft vaccine development is important, as there is currently no FDA-approved vaccine and human-virulent strains of Ft that have been developed as biological weapons [73]. Furthermore, we have also successfully expressed AT-CRLs in Ec and Kp (Figure 1e-g). Preliminary results support surface exposure of our platform in these other gram-negative bacteria and offer exciting applicability to pathogens that are increasingly becoming antibiotic resistant [74]. It would be interesting to test the efficacy of targeted Ec:YFP, Ec:YFC, and Kp:YFP strains in cell-based and animal studies in Ec and Kp models of infection. Based on differences between gram-negative and gram-positive cell walls, AT-CRL would not be applicable to gram-positive bacteria. However, we could envision expanding this targeting approach to vaccine development against additional gram-negative pathogens, as the AT-CRL approach does not require pre-existing pathogen-specific Abs, identification of protective antigens, or addition of adjuvants in order to enhance immune responses. Finally, the potential broad-applicability of our platform also offers an interesting opportunity for combination with pre-existing targeting strategies, such as FcR targeting, to further advance vaccine development.

Bacteria
Escherichia coli (Ec) One Shot TM TOP10 Chemically Competent (Invitrogen, Carlsbad, CA, USA) cells were used as the host strain for plasmid construction. Ec were cultured in Luria-Bertani (LB) broth or on LB agar plates with appropriate antibiotics at either 26 • C or 37 • C where indicated. Electro-competent Francisella tularensis (Ft) LVS NR646 (BEI Resources, Manassas, VA, USA) and Klebsiella pneumoniae (Kp) ATCC 13883 (provided by Robert K. Ernst, Ph.D.) were produced as follows: overnight cultures were harvested (prior to reaching an OD at 600 nm of 0.6) via centrifugation at 4000 g for 15 min. at 4 • C. Cells were placed on ice and all solutions were chilled. Cells were re-suspended in 0.5 M sucrose, spun, and again re-suspended in a smaller volume of 0.5 M sucrose until a final volume of 500 µL was reached. The cells were then aliquoted and utilized for electroporation or frozen for later use. Ft strains were cultured at 37 • C in Brain Heart Infusion (BHI) media (pH 6.8) with appropriate antibiotics or on BHI agar plates that were supplemented with 2.5% L-cysteine hydrochloride, 5% hemoglobin, and appropriate antibiotics. Kp were cultured in LB broth or on LB agar plates with appropriate antibiotics at 26 • C.

Plasmids
The pCR 2.1-Topo vector and TOPO TA Cloning Kit (Invitrogen) were used for direct cloning of PCR products. The broader-host-range shuttle vectors pF and pF2, which encode kanamycin resistance, were used to express YadA, YFC, and YFP in Ec, Kp, and Ft. The pF/F2 plasmids include a M13R priming site, a Ft groEL promoter (truncated in pF2) driving expression of inserts (see below), and a reverse priming site for pF-Reverse. Bacteria that were transformed with empty pF/pF2 vectors served as controls and they are indicated as "-" in figures.
YadA. Full length yadA was PCR amplified from Y. enterolitica DNA with primers Ye YadA 5EcoRI and Ye YadA 3Stop_PmeI and then cloned into pCR 2.1-TOPO to generate pCR 2.1-TOPO:YadA. Table S1 provides primer sequences. The YadA EcoR1/PmeI fragment was digested from pCR 2.1-TOPO:YadA, gel-purified, and then ligated into EcoRI/Sma digested pF or pF2 yielding pF:YadA and pF2:YadA (designated as "YadA" in figures).
pF:YadA SS-SβB. Two fragments of yadA were PCR amplified from Y. enterolitica DNA and separately cloned separately into pCR 2.1-TOPO prior to DNA sequence analysis. The first fragment, encoding the YadA signal sequence (YadA SS), was amplified while using primers Ye YadA 5EcoRI and 3' YadA SS_SacI. The second fragment, encoding the YadA stalk and β-barrel (YadA SβB), was amplified using primers 5' SacI_FLAG_YadA C term and Ye YadA 3Stop_PmeI. 5' SacI_FLAG_YadA C term encodes an engineered, in-frame FLAG tag and Ye YadA 3Stop_PmeI contains an additional in-frame stop codon. Following digestion of pCR 2.1-TOPO:YadA SβB with EcoRI and PmeI, YadA SβB was gel purified and then ligated to EcoRI/EcoRV digested pF, resulting in plasmid pF:YadA SβB. YadA SS was liberated from pCR 2.1-TOPO:YadA SS with EcoRI and SacI, gel purified, and then ligated to similarly digested pF:YadA SβB. Salient features of the resulting plasmid (pF:YadA SS-SβB) include an open-reading frame containing the YadA SS, a SacI site, a FLAG-tag and YadA SβB followed by 2 stop codons. The SacI site in pF:YadA SS-SβB subsequently served as the recipient site for SacI-flanked inserts encoding complement receptor ligands (below).
YFC. DNA encoding murine C3d was PCR amplified from TOPO: C3d rvs BglII BSSB while using 5' C3d w/linker AgeI and 3' C3d w/linker XmaI_SalI and cloned into pCR 2.1-TOPO prior to DNA sequence analysis. The resulting plasmid was subject to site directed mutagenesis (QuikChange Site-Directed Mutagenesis Kit, Agilent Technologies, Santa Clara, CA, USA) with 5' Mut. Destroy SacI in C3d and 3' Mut. Destroy SacI in C3d in order to eliminate the C3d-internal SacI site. The resulting mutated C3d DNA was PCR amplified with 5' SacI_C3d and 3' C3d_SacI and cloned into pCR 2.1-TOPO yielding pCR 2.1-TOPO:C3d-NoIntSacI. Following the digestion of pCR 2.1-TOPO:C3d-NoIntSacI with SacI, the C3d-NoIntSac fragment was gel purified and then ligated to SacI digested pF:YadA SS-SβB (above). A clone with the C3d fragment in the same orientation as YadA was identified and termed YFC.
YFP. DNA encoding murine p28, flanked by SacI sites, was generated via synthetic overlap PCR using primers P28 FWD SacI (for YadA) and P28 RVS SacI (for YadA) and it was cloned into pCR 2.1-TOPO yielding pCR 2.1-TOPO:SacI-p28-SacI. pCR 2.1-TOPO:SacI-p28-SacI was digested with SacI and the p28 fragment was then gel purified and ligated to SacI-digested pF:YadA SS-SβB (above). A clone with the p28 fragment in the same orientation as YadA was identified via DNA sequencing and it was termed YFP.

Bacterial Fractionation
Whole cell preps of bacteria were incubated at 2.5 × 10 7 cells/µL in 1% Tx114 (containing protease inhibitor (Sigma-Aldrich, P2714), lysozyme (Fisher scientific, Waltham, MA, USA, BP535), and benzonase nuclease (Sigma-Aldrich, E8263) in PBS) for 30 min. at room temperature. The samples were then placed on ice for 30 min. and subsequently centrifuged at 13,000 g for 30 min. at 4 • C and the Tx114 soluble (TxS) phase was separated from the Tx114 insoluble (TxI) phase. TxI was resuspended in PBS and both TxI and TxS were again centrifuged; residual TxS was removed from TxI and TxS was moved to a fresh tube to remove any residual TxI. TxS was then incubated at 31 • C for~5 min. to precipitate Tx114, and the samples were centrifuged at room temperature for 10 min. at 10,000 g in order to separate aqueous (A) and detergent (D) soluble fractions. A and D were both washed three times. D was re-suspended in 1% SDS to a final concentration of 10 8 /µL. TxI was also further processed and it was re-suspended in 0.2% sarkosyl and then incubated at room temperature for 30 min. The TxI samples were then centrifuged at 13,000 g for 30 min. at 4 • C and the sarkosyl soluble (SS) phase was removed from the sarkosyl insoluble (SI) phase. Extraction from SI was repeated via 0.2% sarkosyl incubation and centrifugation. SI was then re-suspended in PBS and both SS and SI were again centrifuged to remove residual SS from SI and residual SI from SS. SS was transferred to a fresh tube. SI was re-suspended in 1% SDS to a concentration of 10 8 /µL. All of the samples were loaded at equivalents to 10 8 bacteria per lane for SDS-PAGE.

Collagen Binding ELISA
Flat bottom Costar 96-well assay plates (Corning #9018) were coated with 1µg of type I collagen (rat tail, Sigma-Aldrich, C3867) in 100 µL of 100 mM bicarbonate/carbonate coating buffer (pH 9.6) per well, incubating overnight at 4 • C. The wells were washed three times with PBS 0.05% Tween 20 while using the BioTek ELx50 plate washer and then blocked for 2 h at room temperature with 2% heat inactivated (HI) FBS in DPBS. Ft strains were serially diluted in blocking solution (2% HI FBS in DPBS) and then incubated on the collagen-coated plate for 1 h at room temperature. The plates were then washed and Ft was detected using 1:2500 Mouse anti-Ft LPS (clone FB11, Abcam, Cambridge, UK) followed by 1:5000 Goat anti-Mouse IgG(H+L) HRP (Southern Biotech, Birmingham, AL, USA); each diluted in blocking buffer and then incubated for 1 h at room temperature. Tetramethylbenzidine (TMB, Sigma-Aldrich) substrate was added, development was stopped using 2 M H 2 SO 4 , and absorbance was measured at 450 nm while using the BioRad iMark microplate reader.

Antibody Immunoprecipitation
25 µL of bacteria in culture media (at 2.5 × 10 7 CFU/µL) were mixed with 25 µL of the appropriate antibody in culture media and then incubated for 1 h at room temperature. Mouse α-FLAG, biotinylated (Sigma-Aldrich, F9291) was used at a final concentration of 6 µg/µL. Rat α-complement C3 (Novus NB200-540) was used at a final concentration of 12 µg/µL. Mouse α-IglC (clone 1G7 hybridoma supernatant) [75] was used at a 1:45 dilution. The samples were centrifuged at 5000 g for 10 min., washed with fresh culture media, and re-suspended to 10 7 /µL in TPIG before prep for western blotting.

Bacteria Staining and Quantification of RAW 264.7 Cell Binding
For microplate-based quantification: RAW 264.7 cells were seeded on Greiner CELLSTAR black-wall 96-well plates, aiming for confluency following overnight incubation. BHI-grown Ft were stained by incubating the bacteria with 5 µM of SYTO BC Green Fluorescent Nucleic Acid Stain (Invitrogen, S34855) in DPBS at 10 7 Ft/µL, Ft were then washed and re-suspended in SAM-free DMEM. Ft strains were serially diluted in SAM-free DMEM, starting at 10 8 /well (~MOI 1000) in 100 µL, and they were incubated with RAW cells for 2 h at either 4 • C or 37 • C. The wells were washed with SAM-free DMEM before adding DPBS and measuring fluorescence with the PerkinElmer VICTOR Nivo Microplate Reader. Output was plotted as a percentage of the input Ft signal. For Microscopy: RAW cells were seeded on MatTek 24-well No. 1.5 Coverslip plates and they were incubated with stained (as above) Ft strains at MOI 100 in 1 mL SAM-free DMEM for 2 h at either 4 • C or 37 • C. Wells were washed and imaging was performed at 40x with the Zeiss Axio Observer. For RAW cell incubation with unstained Ft strains: RAW cells were seeded on Corning Costar 24-well plates, aiming for confluency overnight, and they were incubated with 5 × 10 7 Ft/well (~MOI 500) in 500 µL SAM-free DMEM for 2 h at either 4 • C or 37 • C. The wells were washed with SAM-free DMEM and adherent cells were harvested into 150 µL of 0.5% Tx100 and combined with 150 µL Laemmli sample buffer in preparation for western blotting.

RAW 264.7 Cell Signaling Experiments
RAW 264.7 cells were seeded on Corning Costar 24-well plates, aiming for confluency overnight and 5 × 10 7 Ft/well (~MOI 500) in SAM-free DMEM were added. The plates were centrifuged at 500 g for 1 min. to synchronize bacterial association with RAW cells and they were incubated at 37 • C for 5, 15, 30, or 60 min. The wells were washed and all cells were collected with 150 µL of 0.5% Tx100 and were added to 150 µL of Laemmli sample buffer in preparation for western blotting for various cell signaling proteins.

Mice and Immunizations
The Institutional Animal Care and Use Committees of Albany Medical Collage approved (ACUP # 17-09003) the work described in this manuscript. Six-week old male and female C57BL/6 mice (Taconic) were anesthetized via intraperitoneal injection of a cocktail of ketamine (5 mg/kg) (Vedco Inc, St. Joseph, MO, USA) and xylazine (4 mg/kg) (Akorn Inc, Lake Forest, IL, USA) and checked for loss of toe pinch reflex. The anesthetized mice were vaccinated intranasally (i.n.) with a P20 micropipette on d 0 with 20 µL of PBS, Ft:-, or Ft:YFC drop-wise into a single nare using PBS as a vehicle for Ft strains [17,18]. Prior to administration, the Ft doses were spotted on agar plates, which were incubated at 37 • C for enumeration of actual CFU counts.

Antibody Titer Determination
Sera were prepared from PV and PC blood collections by centrifuging clotted blood at 2000 g for 8 min. and was then analyzed for Ft-specific titers via ELISA. Flat bottom Costar 96-well assay plates (Corning #9018) were coated with 100 µL of Ft LVS whole cell lysate derived from 5 × 10 5 Ft/well in coating buffer (100 mM bicarbonate/carbonate coating buffer pH 9.6) overnight at 4 • C. Wells were washed three times with PBS 0.05% Tween 20 using the BioTek ELx50 plate washer and were then blocked with 2.5% heat inactivated (HI) FBS in DPBS. The plates were again washed and incubated with serial dilutions of immune sera in blocking solution (2% HI FBS in DPBS) for 1 h at room temperature. Ft-specific Ab were detected with 1:5000 of the respective goat anti-mouse HRP-conjugated secondary, incubated in blocking buffer for 1 h at room temperature. The plates were again washed, tetramethylbenzidine (TMB, Sigma-Aldrich) substrate was added, development was stopped using 2 M H 2 SO 4 , and absorbance was measured at 450 nm using BioRad iMark microplate reader. Titers were determined as the Ec50s of the nonlinear regression analysis of log transformed values.

Statistics
GraphPad Prism 5 or Microsoft Excel calculated the statistically significant differences. Differences in fluorescent signal measured by the PerkinElmer VICTOR Nivo Microplate Reader, differences in quantitated densitometry for signaling experiments, differences in mouse weights, and differences in days below baseline weight were all calculated via two-tailed unpaired t-test when