PlzA is a bifunctional c-di-GMP biosensor that promotes tick and mammalian host-adaptation of Borrelia burgdorferi

In this study, we examined the relationship between c-di-GMP and its only known effector protein, PlzA, in Borrelia burgdorferi during the arthropod and mammalian phases of the enzootic cycle. Using a B. burgdorferi strain expressing a plzA point mutant (plzA-R145D) unable to bind c-di-GMP, we confirmed that the protective function of PlzA in ticks is c-di-GMP-dependent. Unlike ΔplzA spirochetes, which are severely attenuated in mice, the plzA-R145D strain was fully infectious, firmly establishing that PlzA serves a c-di-GMP-independent function in mammals. Contrary to prior reports, loss of PlzA did not affect expression of RpoS or RpoS-dependent genes, which are essential for transmission, mammalian host-adaptation and murine infection. To ascertain the nature of PlzA’s c-di-GMP-independent function(s), we employed infection models using (i) host-adapted mutant spirochetes for needle inoculation of immunocompetent mice and (ii) infection of scid mice with in vitro-grown organisms. Both approaches substantially restored ΔplzA infectivity, suggesting that PlzA enables B. burgdorferi to overcome an early bottleneck to infection. Furthermore, using a Borrelia strain expressing a heterologous, constitutively active diguanylate cyclase, we demonstrate that ‘ectopic’ production of c-di-GMP in mammals abrogates spirochete virulence and interferes with RpoS function at the post-translational level in a PlzA-dependent manner. Structural modeling and SAXS analysis of liganded- and unliganded-PlzA revealed marked conformational changes that underlie its biphasic functionality. This structural plasticity likely enables PlzA to serve as a c-di-GMP biosensor that in its respective liganded and unliganded states promote vector- and host-adaptation by the Lyme disease spirochete.


Introduction
Lyme disease, a multisystem disorder characterized by cutaneous, neurologic, cardiac, and rheumatologic manifestations [1,2], is caused by the spirochete Borrelia burgdorferi [3]. With 35,000 confirmed cases reported to the Centers for Disease Control and Prevention annually [4,5], Lyme disease is the most prevalent arthropod-borne infection in the United States [6]. Based on insurance claims data, Kugeler et al. [5] estimated that during 2010 to 2018 the incidence of Lyme disease was �476,000 cases per year.
In nature, B. burgdorferi cycles between a hard tick vector and a vertebrate reservoir host, usually small rodents [3,[7][8][9]; the generalist feeding behavior of Ixodes spp. is responsible for transmission of B. burgdorferi to humans, an incidental host [10,11]. Because transmission of B. burgdorferi is transstadial, larvae must acquire spirochetes by feeding on an infected reservoir host. Following acquisition, spirochetes enter a quiescent state within the midguts of flat nymphs [3,12]. The subsequent nymphal blood meal stimulates a replicative burst during which spirochetes replicate exponentially, traverse the midgut epithelium, migrate through the hemocoel to the salivary glands, and, following penetration of salivary acini, access the next host via the salivary stream [7,13,14]. To maintain its complex bi-phasic life cycle, B. burgdorferi must coordinate the expression of colonization factors and protective surface molecules and adjust its physiologic state to contend with vastly different environmental threats and nutrient profiles encountered in mammals and arthropods [3,15,16].

Protection of spirochetes by PlzA during tick feeding is c-di-GMPdependent
Prior studies by us [36,39] and others [38,41,43,49,57] demonstrating that both the Hk1/Rrp1 TCS and PlzA are required for survival of B. burgdorferi during the blood meal strongly suggest that PlzA functions as a c-di-GMP biosensor in feeding ticks. To confirm this experimentally, we took advantage of a finding by Mallory et al. [51] demonstrating that recombinant PlzA containing an arginine to aspartic acid substitution at residue 145 (R145D), the first position of the RxxxR motif, completely abolished c-di-GMP binding. At the outset, we confirmed this finding by comparing the ability of recombinant wild-type and PlzA-R145D His-tagged proteins to bind a fluorescent c-di-GMP analog ( To evaluate the consequence of the R145D substitution on PlzA function in ticks, we inserted the mutated allele into the native plzA locus of BbP1473, a wild-type (wt) strain B31 A3-68 isolate (S1 Table). The resulting mutant, designated plzA-R145D, was tested in parallel with isogenic wt, ΔplzA and plzA complemented (plzAcomp) strains in larvae infected via immersion and then fed to repletion on naïve C3H/HeJ mice as previously described [39,42]. In all comparisons, the plzA-R145D mutant was indistinguishable from its ΔplzA counterpart. By semi-solid phase plating, no viable plzA-R145D or ΔplzA spirochetes were recovered from replete larvae (Fig 1A). Immunofluorescence assay (IFA) of larval midguts revealed only sparse spirochete remnants for both of the PlzA mutant strains (Fig 1B). By qPCR, we detected 1-log 10 lower burdens for the plzA-R145D and ΔplzA mutants in replete larvae compared to the wt and plzAcomp strains (Fig 1C). Based on prior studies [39,41,42,45,54] and the IFA results herein, we attribute the decreased spirochete burdens for both mutants to their destruction during tick feeding. Collectively, these data establish unequivocally that the protection afforded by PlzA during the blood meal is c-di-GMP-dependent.

Loss of PlzA markedly impairs spirochete infectivity in a c-di-GMPindependent manner
Unlike Δhk1 and Δrrp1 strain, which display wt infectivity in mice [39,41,42], ΔplzA organisms are highly attenuated [49,57]. Thus, while the phenotypes for all three mutants are highly similar, if not identical, in ticks (Fig 1), their phenotypes in mice are dichotomous. Importantly, these data suggest that PlzA function in mammals is c-di-GMP-independent. To garner support for this notion, we compared infectivity of plzA-R145D and plzA strains in C3H/HeJ mice in parallel with wt and plzAcomp strains. Consistent with prior studies [49,57], infectivity of the ΔplzA mutant was markedly impaired compared to wt at two-weeks post-infection (Table 1). Of the ten mice infected with the ΔplzA strain, only two seroconverted (S2 Fig). Though not every tissue from the two ΔplzA-infected mice was culture positive (Table 1), in these animals, the mutant clearly disseminated from the site of inoculation. In contrast, all mice infected with the plzA-R145D mutant seroconverted and nearly all tissues were culturepositive at two-weeks post needle-inoculation (Table 1 and S2 Fig), thereby confirming that PlzA function in the mammal does not require binding of c-di-GMP. Complementation restored infectivity of the ΔplzA strain to wt levels ( Table 1).
As noted earlier, attenuation of PlzA-deficient strains has been attributed to abnormal growth and/or motility in vitro in standard BSK medium [49,57]. However, in side-by-side growth curves with wt, Δrrp1, ΔplzA and plzAcomp strains, the virulent plzA-R145D mutant exhibited a more pronounced growth defect than the attenuated ΔplzA null mutant (p = 0.025) (S3 Fig). Moreover, we saw no significant difference in growth between Δrrp1 spirochetes, which are fully virulent ([41,42] and below), and the ΔplzA mutant (S3 Fig). Thus, while the ΔplzA strain grows more slowly than its wt and plzAcomp counterparts in vitro (p<0.05), this phenotype does not explain its markedly reduced virulence in mice. By darkfield microscopy, we also compared the motility of wt, Δrrp1 and ΔplzA strains in BSK-II medium. As previously reported [43], Rrp1-deficient spirochetes display a faster run speed and significantly decreased flexing compared to wt, essentially locking them in "run" mode (S1 and S2 Movies). In contrast, the swimming behaviors of wt and ΔplzA (S1 and S3 Movies, respectively) were indistinguishable, as previously noted [49,57].

Attenuation of ΔplzA spirochetes is RpoS-independent
The reduced virulence of PlzA-deficient spirochetes has been attributed to loss of BosR expression with subsequent ablation of the RpoN/RpoS pathway [52]. In our hands, however, wt, ΔplzA and plzAcomp strains express comparable levels of RpoS and OspC, the prototypical RpoS-dependent gene product, in vitro following temperature-shift (Fig 2A). As a control, lysates were immunoblotted for GlpD, a known c-di-GMP-dependent downstream target of

PlzA overcomes an RpoS-independent immune bottleneck during early infection
To examine whether host-adaptation prior to needle-inoculation enables spirochetes to overcome the early infection defect caused by loss of PlzA, wt, ΔplzA, plzA-R145D and plzAcomp strains were cultivated in DMCs and then immediately used to inoculate (1 × 10 4 spirochetes) naïve C3H/HeJ mice. Analysis of whole cell lysates confirmed that all strains had properly hostadapted prior to inoculation (S4A Fig). In contrast to mice inoculated with in vitro-cultivated ΔplzA organisms (Table 1), all mice needle-inoculated with DMC-cultivated ΔplzA organisms seroconverted (Table 2 and S4B Fig) and were culture positive at three or more tissue sites at two weeks post-infection ( Table 2). As expected, plzA-R145D and plzAcomp strains displayed infectivity comparable to that of the wt parent (Table 2). To determine whether adaptive immune pressure contributes to the attenuation of the ΔplzA mutant, we assessed infectivity of wt, ΔplzA and plzAcomp strains (1 × 10 4 spirochetes) in immunodeficient Prkdc scid (scid) mice four-weeks post-inoculation. As shown in Table 3, the ΔplzA strain infected scid mice at near wt levels. Collectively, these data suggest that PlzA promotes the expression of one or more virulence-related gene products required to evade host adaptive immunity.

Constitutive production of c-di-GMP ablates spirochete virulence
Kostick et al. [43] previously reported that overexpression of Rrp1 in a wild-type B. burgdorferi background had no effect on motility or chemotaxis in vitro but substantially attenuated virulence in mice infected by needle-inoculation. These results suggest that c-di-GMP is Total mice infected 5 10/10 2/10 5/5 5/5 1 Serology is based on immunoreactivity of serum from individual mice against whole cell lysates of wild-type B. burgdorferi strain B31 cultivated at 37˚C in vitro.
Immunoblot data for individual mice are presented in S2 Fig.  2 Wild-type and ΔplzA strains were compared to plzA-R145D and plzAcomp strains in separate experiments (5 mice per strain, per experiment). 3 Data represent culture positivity for the designated tissues collected from C3H/HeJ mice two weeks after inoculation with 1 × 10 4 of wild-type (wt), ΔplzA, plzA-R145D or plzAcomp strains cultivated in vitro. 4 Total number of culture-positive tissues from all mice in the designated group.
deleterious to spirochetes during the mammalian host phase of the enzootic cycle. However, because the authors were unable to measure c-di-GMP levels in either the wild-type or Rrp1-overexpressing strains, it is unclear how the levels of this secondary messenger compare in in the two strains. Thus, to further investigate the effect of c-di-GMP on infectivity in the mammal, we took advantage of studies by Ryjenkov et al. [37] demonstrating that Slr1143, a diguanylate cyclase from an oxygenic phototroph Synechocystis sp., constitutively synthesizes c-di-GMP. As described in Methods, we generated a cp26-based suicide-vector containing a Borrelia-codonoptimized, hemagglutinin (HA)-tagged slr1143 construct expressed under the constitutive borrelial flaB promoter ( Fig 3A) and transformed it into a strain B31 Δrrp1 mutant. Constitutive expression of HA-tagged Slr1143 in this strain (designated cDGC) was confirmed by immunoblot using anti-HA antibodies ( Fig 3B). cDGC grown at 37˚C to late-logarithmic phase harbored concentrations of c-di-GMP, measured by LS-MS/MS, only slightly (3-4-fold) greater than those in the wt; as expected, no c-di-GMP was detected in the Δrrp1 control ( Fig 3C). During in vitro growth, constitutive synthesis of c-di-GMP by Slr1143 functionally complemented loss of Rrp1 based on restoration of GlpD expression (Fig 3B), wt growth kinetics (Fig 3D), and normal motility (S4 Movie). Lastly, we examined the survival of the cDGC strain in ticks (Fig 4). As expected [39,[41][42][43]45], and similar to the ΔplzA and plzA-R145D mutant phenotypes (Fig 1), the Δrrp1 mutant did not survive the blood meal ( Fig 4A and 4B), though qPCR detected

PLOS PATHOGENS
c-di-GMP as a biosensor for the B. burgdorferi enzootic cycle residual spirochete DNA ( Fig 4C). Importantly, Slr1143 functionally complemented the Δrrp1 mutation, restoring the mutant's ability to survive during the larval blood meal (Fig 4). We next used the cDGC strain to assess how continued synthesis of c-di-GMP affects the ability of Lyme disease spirochetes to establish infection, disseminate and persist in mice. Naïve C3H/HeJ mice were needle-inoculated (1 × 10 4 spirochetes) with wt, Δrrp1 and cDGC strains and infection was assessed by tissue culturing at two weeks post-inoculation. In contrast to mice infected with the wt and Δrrp1 strains, all of which were culture positive at each site tested, none of the mice inoculated with the cDGC strain yielded positive tissue cultures or seroconverted (S2 Table).

Production of c-di-GMP antagonizes the RpoN/RpoS pathway through PlzA
We postulated that antagonism of RpoS-dependent gene regulation by c-di-GMP might explain the avirulence of the cDGC strain. To investigate this possibility, we cultivated wild-  Fig 5A). Notably, we also detected GlpD in the cDGC strain (Fig 5A), indicating that c-di-GMP is able to promote glp gene expression when RpoS-mediated repression of glp gene transcription [20,36,55,62] is antagonized. Surprisingly, by both immunoblot (Fig 5A) and qRT-PCR (Fig 5B), rpoS/RpoS levels in all three strains were equivalent in DMCs, confirming c-di-GMP interferes with RpoS function at the post-translational level. Of note, constitutive synthesis of c-di-GMP had no effect on OspC expression in vitro, suggesting this phenomenon is unique to host-adaptation (S5 Fig).
To determine whether the interference of c-di-GMP on RpoS function in DMCs is mediated via PlzA, we inserted the constitutive-expressed/-active diguanylate cyclase (P flaB -slr1143-HA) cassette into the Δplz background, generating ΔplzA+cDGC, and evaluated the protein profiles of wt, ΔplzA, plzA-R145D, ΔplzA+cDGC and cDGC strains cultivated in DMCs by SDS-PAGE and immunoblot (Fig 6). Whereas the presence of c-di-GMP (cDGC) interfered with both positive and negative aspects of RpoS-dependent gene regulation in DMCs, the absence of PlzA (ΔplzA+cDGC) restored normal RpoS function in vivo. Consistent with results   PlzA has a flexible bipartite domain structure that potentially explains its differential function Galperin and Chou [64] recently categorized bipartite 'xPilZ-PilZ' proteins from diverse bacteria into nine classes based on their N-terminal non-canonical PilZ-related domains. PlzA belongs to the 'PilZN3-PilZ' class based on the presence of a newly described N-terminal PilZN3 domain, which is predicted to form a six-stranded PilZ-like β-barrel. To date, crystal structures have been reported for only three xPilZ-PilZ proteins in c-di-GMP-liganded and -unliganded states (S6A Fig): (i) MrkH, a PilZN2-PilZ DNA binding transcriptional factor that promotes biofilm formation in Klebsiella pneumoniae [65,66]; (ii) FlgZ, a PilZN-PilZ protein,

PLOS PATHOGENS
c-di-GMP as a biosensor for the B. burgdorferi enzootic cycle which functions as a flagellar brake in Pseudomonas putida [67]; and (iii) PlzD, a PilZNR-PilZ protein, which regulates virulence and motility in Vibrio cholerae [68,69]. In all three, binding of c-di-GMP by the C-terminal PilZ domain induces a large rotational change that brings their N-and C-terminal β-barrels into proximity, with c-di-GMP intercalated at xPilZ-PilZ domain interface. Root-mean-square deviation (RMSD) values for liganded and unliganded MrkH, Shapes of the Kratky plots ( Fig 7A) revealed that liganded-PlzA is well-ordered, while the unliganded protein is unfolded and/or highly flexible, confirming a major structural rearrangement upon binding of c-di-GMP. We next used trRosetta [70], I-TASSER [71,72] and SWISS-MODEL [73] to generate structural models for liganded-PlzA and then screened twenty of the resulting models against our experimental SAXS data (S4 Table). One model, generated using trRosetta, produced the lowest χ 2 value (7.14) which, after SREFLEX refinement [74], improved to 6.66 (S7C Fig). Of note, refinement did not require structural changes within the N-or C-terminal domains, just reorientation of the β-barrels. Next, the refined model was docked with c-di-GMP using HADDOCK [75] and fit into the SAXS envelope. As shown in Figs 6D and S7D, the RxxxR c-di-GMP binding motif is located within the extended interdomain linker, while the (D/ N)hSxxG motif is located within the C-terminal PilZ β-barrel. Helices α1 and α2 in the N-terminal PilZN3 domain are positioned in proximity to its unique C-terminal α-helix (CT-α) and RxxxR motif. Electrostatics analysis of liganded-PlzA indicates three positively charged surface regions; one contains the c-di-GMP binding site, while the other two are located in grooves within the N-and C-terminal domains (S7E Fig, dashed lines) and could be available for interactions with DNA or other proteins. Due to its high ambiguity score, an envelope for unliganded PlzA could not be generated from the SAXS data, further confirming that c-di-GMP locks both domains of PlzA into a static, condensed conformation. and PilZ β-barrels are colored in salmon and yellow, respectively. The α helices in the PilZN3 and PilZ domains are highlighted in light green and cyan, respectively, while the C-terminal α-helix in PilZN3 is shown in purple. c-di-GMP binding residues in RXXXR and DXSXXG motifs are highlighted in dark green. c-di-GMP (red) was docked into the model using HADDOCK [75]. https://doi.org/10.1371/journal.ppat.1009725.g007

PLOS PATHOGENS
c-di-GMP as a biosensor for the B. burgdorferi enzootic cycle Note added in proof: Singh et al. [123] recently reported a 1.6 Å high resolution crystal structure for Borrelia (Borreliella) burgdorferi PlzA in complex with two molecules of c-di-GMP (PDB ID: 7MIE). The structural model and domain organization for PlzA presented herein are highly similar to the solved structure reported by Singh et al.

Discussion
In many bacteria, c-di-GMP inhibits motility [76] and, by a variety of allosteric mechanisms, regulates the transition from a motile planktonic to a biofilm-associated, communal lifestyle [77][78][79][80][81]. In B. burgdorferi, c-di-GMP production is tied to environmental sensing by the Hk1/ Rrp1 TCS in response to unknown biochemical cues generated during tick feeding [38][39][40][41][42][43][44]. Spirochetes unable to produce c-di-GMP cannot withstand the onslaught of nutritional and/or biophysical stressors unleashed by the blood meal and, consequently, are destroyed in the midguts of both larvae and nymphs [39,41-43]. Since the signals driving Hk1/Rrp1 activation emanate from the blood meal [38-43], production of and signaling by c-di-GMP is restricted to the tick phase of the enzootic cycle. Indeed, not only is c-di-GMP not required during the reservoir phase [39,41,42], herein, we present evidence that it is inimical, at least in part, because it antagonizes RpoS function(s) required for the establishment and maintenance of mammalian infection by B. burgdorferi. A conundrum the spirochete faces is that it must produce c-di-GMP during two 'opposing' tick phases-acquisition, the RpoS-OFF state when spirochetes colonize the vector, and transmission, the RpoS-ON state when spirochetes regain motility and exit the midgut [20,34,35,42,55,82]. B. burgdorferi appears to have partially resolved this dilemma by integrating the adaptive changes mediated by c-di-GMP-liganded PlzA into the ON and OFF states of the RpoS genetic program in ticks along with divergent motility responses during the acquisition and transmission blood meals. Moreover, in PlzA, the spirochete has evolved a novel c-di-GMP effector protein [64] that serves as a biosensor for the presence or absence of c-di-GMP to promote, respectively, vector-or mammalian hostadaptation.
In vitro and in vivo studies of PlzA function yield widely disparate results. In vitro, the consequences of PlzA deficiency are negligible (i.e., minor growth and motility defects), whereas the in vivo effects related to the loss of PlzA are dramatic [38,44,49,57]. As shown here and elsewhere [49,57], ΔplzA spirochetes are destroyed in ticks and, in mice, they exhibit marked attenuation of infectivity; importantly, neither phenotype can readily be related to the modest motility defect observed in vitro by others [49,57]. As noted by Novak et al. [44], B. burgdorferi mutants defective in motility (e.g., cheA2 and pdeA) not only survive but replicate exponentially within feeding ticks [50,83]. Rather, the bulk of available evidence indicates that lysis of Δhk1 and Δrrp1 spirochetes during the blood meal is due to decreased expression of permeases for the uptake and utilization of alternative carbon sources (e.g., glycerol, N-acetylglucosamine (NAG) and chitobiose) with consequent inability to support both energy generation and cell envelope biogenesis [16,41,42,45,54]. Lysis of ΔplzA and plzA-R145D strains under these same conditions implies that these transcriptional effects of c-di-GMP are, at least in part, PlzAdependent. Other investigators previously have raised the possibility that PlzA function in mammals might be c-di-GMP-independent [44,52,53,57]. We confirmed this by demonstrating that the virulence of the plzA-R145D strain is comparable to wild-type. In other bacteria, the regulatory effect of PilZ domain proteins on the flagellar motor requires c-di-GMP [84][85][86]; results with the plzA-R145D strain also argue that PlzA's role in borrelial virulence is unrelated to motility. While non-canonical PilZ domain proteins (e.g., Vibrio cholerae PlzB and Xanthamonas campestris Xcc6021) that are unable to bind c-di-GMP have been linked to virulence [64,68,87], to our knowledge, PlzA is the first example of a c-di-GMP biosensor with dual functionality. Because the effector functions of c-di-GMP-liganded and -unliganded PlzA, respectively, are strictly segregated to the tick and mammalian stages of the enzootic cycle (Fig 8), it seems almost certain that they reflect discrete interaction partners and downstream effector mechanisms [49,53,57].
In independent studies, Sze et al. [45] and He et al. [52] reported that inactivation of rrp1 and plzA, respectively, resulted in reduced levels of BosR in vitro, leading to the conclusion that reduced transcription of rpoS is responsible for the virulence defect of the ΔplzA mutant. Our data are not in accord with this reasoning, as we observed no effect on rpoS/RpoS levels either in vitro or in DMC-cultivated ΔplzA spirochetes. Moreover, we detected comparable levels of rpoS/RpoS in wt, Δrrp1 and cDGC strains in DMCs ([42] and herein). Collectively, these results indirectly confirm that the levels of RpoS are not adversely affected by loss of either liganded-or unliganded-PlzA and that, consequently, one must look elsewhere to understand how PlzA interfaces with RpoS. They also strongly infer that the early infection defect of ΔplzA spirochetes is both RpoS-and c-di-GMP-independent and, by extrapolation, that PlzA promotes mammalian host-adaptation via a mechanism(s) that is extrinsic to the RpoN/RpoS pathway. Although this RpoS-/c-di-GMP-independent function of PlzA is cryptic at present, one can glean insights into its general features from three facets of the ΔplzA murine phenotype uncovered herein: (i) The ΔplzA defect is overcome by a small subpopulation of organisms; in 2 of 10 C3H/HeJ mice inoculated with the ΔplzA strain cultivated in vitro, the mutant spirochetes survived the inoculation, disseminated and persisted in metastatic sites. Consistent with these findings, Pitzer et al. [49] found that they could increase the proportion of mice with disseminated infection by increasing the ΔplzA inoculum. (ii) The ΔplzA defect can be bypassed to a substantial degree if the mutant is host-adapted in DMCs prior to inoculation. (iii) The ΔplzA defect involves, at least in part, evasion of adaptive immunity since ΔplzA infectivity is substantially greater in scid mice. A body of evidence indicates that, once inoculated, spirochetes must overcome stochastic bottlenecks created by host barriers to cause systemic infection [88,89]. Unliganded-PlzA appears to participate in one or more regulatory pathways that increase the probability that an infecting population will contain a sufficient number of organisms with the 'appropriate' transcriptional profile to surmount these bottlenecks. Fig 8 presents our working model for how gene regulation by PlzA, c-di-GMP and RpoS interdigitates to sustain B. burgdorferi within its dual-host lifecycle. In mammals, when c-di-GMP signaling is normally OFF, one sees RpoS-mediated repression of tick-phase genes [36, 55,56,62,90]. When this 'gatekeeper' function of RpoS was first noted [55,61], predating the discovery of c-di-GMP signaling in B. burgdorferi [37-39,41,43,46], we postulated that RpoS-mediated repression was induced by mammalian host-specific signals [55,61]. However, two subsequent, closely related lines of evidence recently led us to propose an alternative hypothesis, namely, that antagonism of RpoS within feeding ticks is alleviated in the mammal by the cessation of c-di-GMP synthesis [36,42,82]: (i) RpoS-mediated repression of tick-phase genes does not occur concurrently with the RpoS-ON state in feeding nymphs and (ii) in engorged nymphs, ΔrpoS organisms express significantly higher levels of RpoS-repressed genes than wt [42,82]. Herein, we assessed this counterintuitive idea by engineering a B. burgdorferi strain (cDGC) that constitutively synthesizes c-di-GMP at levels comparable to those produced by wt strains. The results were striking and, once again, underscore the dichotomy between in vitro and in vivo gene expression by B. burgdorferi. In vitro, 'ectopic' expression of c-di-GMP had no discernible effect on the RpoN/RpoS pathway, while in vivo we saw a dramatic PlzA-dependent dampening of RpoS activity that affected both RpoS-upregulated and RpoS-repressed gene products. Our finding that this antagonism occurred without diminution

PLOS PATHOGENS
c-di-GMP as a biosensor for the B. burgdorferi enzootic cycle of either rpoS transcript or protein, suggesting that liganded-PlzA directly or indirectly interacts with RNA polymerase-RpoS holoenzyme (RNAP-RpoS). In conjunction with these new insights, one can envision how liganded-PlzA could act as a 'brake' for RpoS-dependent gene regulation in ticks. During acquisition, transcription of rpoS by the Rrp2/BosR/RpoN complex is quickly shut OFF in the midgut in response to unknown signals [16,20]. Conceivably, to expedite the switch from an RpoS-ON to -OFF, liganded-PlzA also could interfere with transcription by residual RNAP-RpoS holoenzyme allosterically, while BBD18, which is RpoSrepressed and c-di-GMP-induced [36,42], would target free RpoS for proteolytic degradation [91][92][93][94]. During transmission, partial restriction of RNAP-RpoS holoenzyme by liganded-PlzA could enable spirochetes to sustain expression of RpoS-repressed tick phase genes (i.e., ospA, glps) while at the same time allowing reduced expression of RpoS-upregulated genes required for transmission (i.e., bba64) or early infection (i.e., ospC, dbpBA) [36, 82,95]. Once within the mammal, loss of c-di-GMP signaling, presumably hastened by the spirochete's two phosphodiesterases, PdeA and PdeB [44,48,50], removes impedance of the RpoS gatekeeper and enables unliganded-PlzA to assume its alternative, c-di-GMP-independent function(s) either directly or indirectly via an as yet unidentified interactive partner.
The crystal structures of MrkH [66], FlgZ [67], and PlzD [96] in their liganded and unliganded states show that c-di-GMP binding causes large spatial rotations in the N-and C-terminal domains and reorganization with their inter-domain linkers, with minimal structural changes in the individual domains. c-di-GMP-induced conformational changes in PlzA [51,57], confirmed herein by SAXS, also likely underlie this novel effector protein's dual functionality in ticks and mice. In its liganded state, PlzA forms a stable, compact protein, while the apo-protein is partially unfolded and/or significantly flexible; our analysis further suggests that apo-PlzA requires an interaction partner to promote structural stabilization and functionality. The extended length of the PlzA linker region likely contributes to even greater flexibility between its PilZN3 and PilZ domains [51]. Reorientation of the three α-helices of the PilZN3 domain in liganded PlzA would create a new interface for interactions distinct from those of the unliganded protein. Expression data for the glp operon indicate that liganded-PlzA directly or indirectly promotes expression of this RpoS-repressed locus by RNAP-RpoD (σ 70 ) [42,53]; this could occur by two possible mechanisms (Fig 8). One, analogous to MrkH, is by DNA binding as a transcription factor [66,97]. As with MrkH [66,97], the surface electrostatics of our PlzA model indicate large positively charged regions, including major grooves in N-and C-terminal β-barrels. The other, as proposed by Zhang et al. [53], is by allosteric interaction with RNAP-RpoD. The possibility that PlzA can act as a positive or negative regulator for RNAP-RpoD and RNAP-RpoS holoenzymes, respectively, presents a unifying and, therefore, appealing mechanism for PlzA's transcriptional effects.
Efficient migration between the vector and reservoir hosts is essential for perpetuation of B. burgdorferi in nature [3,[7][8][9]; once within a reservoir or incidental human host, motility becomes critical for dissemination and tissue invasion [98][99][100]. In other bacteria, there is an inverse relationship between intracellular c-di-GMP concentrations and motility, with low levels promoting motility and high levels stimulating adherence and biofilm formation [76,101]. In Gram-negatives, YcgR orthologs acts as a 'clutch' to slow flagellar rotation in response to cdi-GMP by directly interacting with the flagellar motor at the rotor-stator interface [76,86,102]. Results presented herein show that B. burgdorferi deviates from this motility control paradigm. Comparison of the swimming behaviors of wt, Δrrp1 and cDGC strains in vitro revealed that elevated c-di-GMP exerts a potent inhibitory effect on motility; however, as shown here and elsewhere [49,57], inhibition by c-di-GMP appears to be largely independent of PlzA given that neither ΔplzA nor plzA-R145D exhibited discernably aberrant motility in BSK-II. Along similar lines, using a ΔpdeAΔplzA double mutant, Pitzer et al. [49] demonstrated that elevated c-di-GMP regulates motility in B. burgdorferi by a PlzA-independent mechanism. Thus, B. burgdorferi appears to contain an as yet unidentified c-di-GMP-responsive regulator of motility that does not contain a recognizable PilZ domain. The broader implication is that the spirochete has, at least in part, separated c-di-GMP control of motility from PlzA-mediated gene expression which is needed for survival during the acquisition and transmission blood meals. Previously, we reported that spirochete transmission during the nymphal blood meal is biphasic, occurring initially via a non-motile, replicative process termed 'adherence-mediated migration' followed by an invasive, motile phase [13]. In support of this theory, we demonstrated that the contents of engorged midguts inhibit B. burgdorferi motility, now attributable to luminal tick factor(s) that stimulates synthesis of c-di-GMP via activation of the Hk1 sensor. Presumably, in the small number of spirochetes that reach the midgut basement membrane, these stimulatory cues diminish, allowing c-di-GMP levels to drop low enough to restore motility. Spirochetes lacking RpoS survive within the nymphal midgut, indicating indirectly that Hk1/Rrp1 pathway is active, but remain confined within the luminal space [82]. Thus, the ON/OFF state of the RpoN/RpoS pathway likely determines whether spirochetes colonize or penetrate the midgut epithelium in concert with precise, spatiotemporal regulation of c-di-GMP production [42,82].

Ethics statement
All experiments involving animals conducted at UConn Health were performed in accordance with The Guide for the Care and Use of Laboratory Animals (8th Edition) [103] using protocols reviewed and approved by the UConn Health Institutional Animal Care and Use Committee [Animal Welfare Assurance (AWA) number A347-01].

Routine DNA manipulation and cloning
Oligonucleotide primers used these studies are described in S3

Construction of B. burgdorferi plzA site-directed mutant
To distinguish between the c-di-GMP-dependent and -independent effector functions for PlzA, an arginine to aspartic acid point mutation was introduced at residue 145 (R145D) by site-directed mutagenesis [51]. Briefly, an AgeI site was introduced into the pbb0733-Easy suicide vector [49] in the plzA-bb0734 intergenic region using the QuikChange II site-directed mutagenesis kit (Agilent Technologies Inc., Santa Clara, CA) and primers BB0733AgeImut-F and BB0733AgeImut-R. A P flgB -GentR cassette [111] was inserted into the AgeI site of pbb0733-Easy in the same orientation as plzA. A single point mutation (R145D) was introduced in the plzA coding region using primers BB0733-R145D-F and BB0733-R145D-R. The resulting suicide vector (pbb0733R145DGenta-Easy) was linearized by digestion with NotI, electroporated into B. burgdorferi strain B31 A3-68 Δbbe02 as previously described [112]. Transformants were screened for the gentamycin-resistance cassette by PCR using PlessGent-F and PlessGent-R. The R145D point mutation was confirmed by sequencing. A single transformant containing the same plasmid profile as the parent was selected for further analysis and designated plzA-R145D.

Generation of B. burgdorferi strain expressing a constitutively active diguanylate cyclase
A Δrrp1 mutant was generated by transforming B31 5A18 NP1 [113] with pΔBB0419 [38]. A single streptomycin-resistant clone containing the same plasmid profile as the parent was selected for further analysis. A codon-optimized version of slr1143, encoding a constitutively active diguanylate cyclase from Synechocystis sp.
[37] and C-terminal hemagglutinin (HA) tag, was synthesized by Genewiz (South Plainfield, NJ). slr1143opt-HA was fused to the B. burgdorferi flaB promoter via overlapping PCR using the primers listed in S4 Table. The PflaB-slr1143opt-HA cassette was cloned into the AatII site of EcAG265 [106], a modified version of the cp26-based E. coli-B. burgdorferi shuttle vector pMC2498 [60] in which the promoterless gfp cassette has been replaced with an AatII site. The resulting plasmid (EcAG284) was confirmed by sequencing and then electroporated into Δrrp1 as previously described [112]. Transformants were screened for slr1143opt-HA by PCR using primers 5' bb0733 ORF and 3' bb0733 ORF. A single gentamycin-resistant clone containing the same plasmid profile as the parent was selected and designated cDGC. To generate a ΔplzA strain that constitutively synthesizes c-di-GMP, the cassette encoding PflaB-slr1143opt-HA followed by GentR was subcloned into pUC19 containing~1200 bp framing the rrp1 coding region, so that PflaB-slr1143opt-HA + GentR replaced rrp1 (pEcAG391). This plasmid was transformed into ΔplzA and a single gentamycin-resistant clone containing the same plasmid profile as the parent and designated ΔplzA+cDGC.

Murine infection studies
Five to eight-week-old female C3H/HeJ or NOD.Cg-Prkdc scid /J (scid; Jackson Laboratories, Bar Harbor, ME) mice were needle-inoculated with 1 × 10 4 organisms cultivated in vitro or from freshly harvested DMCs. At designated time points, animals were sacrificed, and blood and tissues collected for serology and culturing in BSK-II containing Borrelia antibiotic cocktail (0.05 mg/ml sulfamethoxazole, 0.02 mg/ml phosphomycin, 0.05 mg/ml rifampicin, 0.01 mg/ml trimethoprim and 2.5 μg/ml amphotericin B). Cultures were examined weekly by darkfield microscopy for up to 4 weeks. Seroconversion was determined by immunoblotting B. burgdorferi whole cell lysates with infected mouse serum, diluted 1:1000, followed by incubation with HRP-conjugated secondary antibody (Southern Biotechnology Associates, Birmingham, AL) diluted 1:20,000 and detection using SuperSignal West Pico chemiluminescence substrate (Pierce, Rockford, IL).

Tick experiments
Pathogen-free Ixodes scapularis larvae were purchased from Oklahoma State University Tick Rearing Facility (Stillwater, OK). Immersion feeding of naïve larvae was performed as previously described [90] using the method established by Policastro et al. [114]. Pools of 10 replete larvae per strain were processed for DNA extraction and semi-solid phase plating. A separate pool of 10 larvae was processed for immunofluorescence using FITC-conjugated anti-Borrelia antibody (Kirkegaard and Perry Laboratories, Gaithersburg, MD) as previously described [90]. Replete larvae used to assess borrelial gene expression were infected by whole body infestation of needle-inoculate mice as previously described [90]. To assess gene expression during transmission, infected I. scapularis nymphs were fed for at least 72 h on naïve C3H/HeJ mice as previously described [90]. Fed larvae (~100-150 per pool) and nymphs (~20-25 per pool) were crushed into TRIzol reagent (Invitrogen) and stored at -80˚C until RNA was extracted.

qRT-PCR
Total RNA was isolated from triplicate pools of replete larvae, unfed and fed nymphs, and DMC-cultivated organisms) as previously described [90]. Following DNase treatment, RNA was converted to cDNA using SuperScript III (Life Technologies), including a negative control with no reverse transcriptase. cDNAs were assayed in quadruplicate using SsoAdvanced Universal SYBR (rpoS) or Probe (flaB) Mix (Bio-Rad) using the primers described in S3 Table. Transcript copy numbers were calculated using the iCycler post-run analysis software based on internal standard curves and then normalized against copies of flaB as previously described [90].

Measurement of c-di-GMP levels
Cultures (50 mls) of wt B31 5A18 NP1, Δrrp1, and cDGC were grown to late-logarithmic phase and cells were harvested for ethanol extraction as previously described [42]. Extracted supernatants were filtered through a 0.22 μm syringe-filter, concentrated via SpeedVac, and resuspended to a final volume of 0.1 ml in HPLC grade water. c-di-GMP was detected by ultraperformance liquid chromatography (UPLC) in tandem with mass spectrometry (MS) using an Acquity UPLC system coupled to an Acquity TQD mass spectrometer (Waters Corporation, Milford, MA). The separation of c-di-GMP was achieved using a High Strength Silica (HSS) reversed-phase UPLC column. Briefly, the eluent system was composed of 0.1% formic acid in water (pH 2.9) (eluent A) and 0.1% formic acid in acetonitrile (eluent B). 98% eluent A was held for 0.5 min followed by a gradient to 100% eluent B in 4 min, held for 0.5 min, then switched back to 98% eluent A at a flow rate of 0.4 ml/min. An Acquity HSS T3 column (2.1 by 100 mm; 1.8 μm particle size; Waters) was used with a sample injection volume of 10 ul. The column and autosampler were maintained at 35˚C and 20˚C, respectively. Detection of cdi-GMP was performed in electrospray ionization (ESI) negative-ion mode using the multiple-reaction monitoring mode. For ESI-MS/MS analysis, the following ion transition, cone voltage (CV), and collision energy (CE) were used: c-di-GMP m/z 689.1 (precursor ion) and 150.0 (product ion); CV, 66 V; and CE, 56 eV. The ESI capillary voltage was 3 kV, the source temperature was set at 150˚C, and the desolvation temperature was set at 400˚C. The flow rate of the desolvation gas (N2) was set at 650 liters/h. The Waters IntelliStart software was utilized for analyte signal optimization. Statistical analysis for obtaining calibration and quantification results for c-di-GMP. was performed using Waters QuanLynx, which is included in the Mas-sLynx software v.4.2. The concentration of c-di-GMP was calculated by interpolation of the observed analyte peak area with the corresponding calibration curve. Concentrations were determined as c-di-GMP (nmol/μg) of wet cell pellet weight.

Small angle X-ray scattering (SAXS) data acquisition and analysis, structural modeling, and c-di-GMP docking
Prior to SAXS, recombinant PlzA (135, 67 and 33 μM) was incubated with or without 270 μM of c-di-GMP in 500 mM NaCl and 20 mM Na 2 HPO 4 (pH 7.4) . SAXS data were acquired on the Bio-SAXS beamline BL4-2 at the Stanford Synchrotron Research Laboratory using a Rayonix MX225-HE CCD detector. All scattering data (to a maximum q of 0.5 Å −1 ) were collected at a wavelength of 1.3 Å for ten consecutive 2-second exposures. Results from the buffer alone, with or without c-di-GMP, were subtracted from the liganded and unliganded scattering, respectively. Data were analyzed using the ATSAS package [115]. Kratky plots and radii of gyration (R g ), extrapolated from the Guinier region of the Guinier plot, were computed using PRIMUS [116]. Scattering curves for liganded-PlzA were scaled and merged in PRIMUS primarily using the low q data for PlzA at 33 μM and the high q range data for PlzA at 67 μM. Unliganded-PlzA was analyzed using data set collected from PlzA at 67 μM only. P(r) functions were calculated using GNOM and ambiguity scores by AMBIMETER [117]. Ab initio shape determination was performed using DAMMIN [118] followed by DAMAVER [119]. Twenty three-dimensional models of PlzA were predicted using trRosetta [70], I-TASSER [71,72] and Swiss-Model [73] (S4 Table). Theoretical scattering curves were computed from different structural models and compared to the experimental scattering curves using FoXS [120]. The best-fitted model was refined by normal mode analysis from SAXS data using SRE-FLEX [74]. Coordinates for c-di-GMP were extracted from the crystal structure of Vibrio cholerae VCA0042/PlzD (PDB 2RDE) [96]. HADDOCK v2.2 [75] was used to dock c-di-GMP into the SREFLEX refined PlzA model. In the docking protocol, PlzA residues R145, R149, D182, A184 and G187 were designated as active residues to apply distance restraints. Superimposition of the PlzA model into the SAXS envelope structure was performed by SUPCOMB [121]. PyMOL Molecular Graphics System v2.3.2 (Schrödinger, LLC, New York, NY) was used for structure visualization, calculation of RMSD and surface electrostatics, and image rendering.

Statistical analysis
Growth curves were compared using the CGGC permutation test [122], with 1000 permutations. All other pairwise comparisons were evaluated by unpaired Student's t-test with twotailed p values and a 95% confidence interval using Prism v8.4.3 (GraphPad Software, San Diego, CA).
Supporting information S1