Regulator of Actin-Based Motility (RoaM) Downregulates Actin Tail Formation by Rickettsia rickettsii and Is Negatively Selected in Mammalian Cell Culture.

The etiological agent of Rocky Mountain spotted fever, Rickettsia rickettsii, is an obligately intracellular pathogen that induces the polymerization of actin filaments to propel the bacterium through the cytoplasm and spread to new host cells. Cell-to-cell spread via actin-based motility is considered a key virulence determinant for spotted fever group rickettsiae, as interruption of sca2, the gene directly responsible for actin polymerization, has been shown to reduce fever in guinea pigs. However, little is known about how, or if, motility is regulated by the bacterium itself. We isolated a hyperspreading variant of R. rickettsii Sheila Smith that produces actin tails at an increased rate. A1G_06520 (roaM [regulator of actin-based motility]) was identified as a negative regulator of actin tail formation. Disruption of RoaM significantly increased the number of actin tails compared to the wild-type strain but did not increase virulence in guinea pigs; however, overexpression of RoaM dramatically decreased the presence of actin tails and moderated fever response. Localization experiments suggest that RoaM is not secreted, while reverse transcription-quantitative PCR (RT-qPCR) data show that various levels of RoaM do not significantly affect the expression of the known rickettsial actin-regulating proteins sca2, sca4, and rickA. Taken together, the data suggest a previously unrecognized level of regulation of actin-based motility in spotted fever group rickettsiae. Although this gene is intact in many isolates of spotted fever, transitional, and ancestral group Rickettsia spp., it is often ablated in highly passaged laboratory strains. Serial passage experiments revealed strong negative selection of roaM in Vero 76 cells. IMPORTANCE The mechanism of actin-based motility of spotted fever group Rickettsia has been studied extensively, but here, we provide genetic evidence that motility is a regulated process in R. rickettsii. The findings also suggest that serial passage of rickettsial strains in cell culture may cause the bacteria to lose essential genes that are no longer conserved under natural selective pressure. These findings are likely relevant to the interpretation of studies concerning virulence determinants of rickettsiae.

INTRODUCTION

Rickettsia rickettsii is an obligately intracellular pathogen that causes Rocky Mountain spotted fever, a disease with the highest mortality of any tick-borne rickettsial disease (1, 2). Although originally recognized during the early 1900s in the Bitterroot Valley of western Montana, it has since become clear that the highest incidence in the United States occurs in the south-central states (3, 4). In the last 2 decades, case numbers have also begun to rise in the southwestern United States and Mexico (5). Numbers of spotted fever rickettsiosis caused by other members of the genus have also risen, seemingly correlated with a rising coincidence of people and ticks, as different species of ticks increase their range and numbers (6, 7).

Members of the spotted fever group of rickettsiae, including R. rickettsii, utilize the polymerization of host actin to generate long actin tails that propel the bacterium through the cytoplasm of the host cell (8, 9). Actin-based motility contributes to a bacterium’s ability to infect adjacent cells and has been studied using several facultative intracellular bacteria as models, including Listeria monocytogenes, Shigella flexneri, and Burkholderia pseudomallei (10–12). For these model pathogens, impairing actin-based motility results in smaller plaque sizes in cell culture models (11, 13, 14) and typically attenuates virulence (15). The development of plaque assays for rickettsiae enabled characterization of strains based on different plaque phenotypes (16–19). Comparison of plaque phenotypes, especially within the same species, has allowed the unraveling of genetic factors that lead to those phenotypes (20, 21). Within the spotted fever group, a reduced plaque size led to recognition that Sca2 is responsible for actin-based motility, and its ablation leads to decreased virulence in the guinea pig model (20, 22, 23). However, neither plaque size (21, 24) nor opacity (21) alone is a reliable indicator of virulence.

Mutations in roaM, a gene of unknown function, were initially identified in comparative genomics studies of the genetic differences between virulent and avirulent R. rickettsii (25, 26). The avirulent Iowa strain has an M4I amino acid change compared to the virulent Morgan and Sheila Smith strains, while the Morgan strain contained a base pair insertion resulting in an approximately 20% truncation of the protein product. Further investigations of this gene in Sheila Smith revealed a point deletion in some, but not all, R. rickettsii Sheila Smith subclones. Upon inspection, the emergence of the point mutation in roaM correlated with the emergence of large, clear plaques in later passages of R. rickettsii Sheila Smith. Annotated spotted fever group rickettsia genomes indicated frequent truncation of roaM even within a single species. Further investigation revealed a rapid loss of roaM in mammalian cell culture, suggesting that it may be beneficial in the tick vector but selected against in cell culture. Therefore, we pursued the possible connection between roaM and actin-based motility as well as stability of wild-type roaM during mammalian cell passage.

RESULTS

The genotype of roaM in Sheila Smith derived strains correlates with different plaque phenotypes.

The parental Sheila Smith small-plaque variant (SSS) carries an intact roaM and displays a small plaque phenotype. To assess the significance of roaM and its potential relation to plaque size, we generated several strains of R. rickettsii Sheila Smith with various roaM genotypes (Fig. 1A). SSS-pROAM-FL and SSS-pROAM-T constitutively produce N-terminally FLAG-tagged full-length and truncated versions of RoaM, respectively (Fig. S1). A genetic knockout of roaM (SSS-ΔroaM) was generated via plasmid-based homologous recombination (Fig. S2). Growth curves (Fig. 1B) show similar rates of growth of all R. rickettsii variants and mutants in the exponential phase between 0 and 2 days postinfection. Deletion or truncation of roaM resulted in larger plaque sizes at both 8 and 15 days postinfection compared to SSS (Fig. 1C and D). The overexpression of full-length roaM resulted in reduced plaque size. Interestingly, overexpression of truncated roaM in SSS (SSS-pROAM-T) resulted in an intermediate phenotype, suggesting a dominant or perhaps competitive effect over wild-type RoaM. The chromosomal copy of roaM for SSS-pROAM-T was PCR amplified and Sanger sequenced to rule out inadvertent recombination or truncation.

FIG 1

Differential roaM genotype in R. rickettsii Sheila Smith-derived strains results in varied plaque-size phenotypes. (A) Schematic of the genotype of critical strains used in this study. SSS, Sheila Smith wild-type small-plaque variant; SSL, Sheila Smith large-plaque variant; SSS-pROAM-FL, SSS with ectopically expressed, full-length roaM; SSS-pROAM-T, SSS with ectopically expressed truncated roaM producing a 211-amino-acid (aa) variant as found in SSL; and SSS-ΔroaM, SSS with a chromosomal deletion of roaM. (B) Growth curves for the above strains reveal no substantial difference in exponential growth rate. (C) Representative photographs of R. rickettsii plaques on six-well plates stained at either 8 or 15 days postinfection. Bar = 1 cm. (D) Pooled data from three independent experiments measuring plaque area, measuring 50 plaques per strain per experiment. Error bars represent standard deviations. Standard analysis of variance (ANOVA) followed by Tukey’s multiple comparison was performed on results. ****, P ≤ 0.0001; n.s., not significant.

Production of ectopically expressed, FLAG-tagged RoaM in R. rickettsii. A Western blot was performed for Vero cells infected with R. rickettsii expressing FLAG-tagged proteins of interest to demonstrate expression. Asterisks indicate full-length products. Degradation products are not labeled. Samples were dissolved in 2× Laemmli sample buffer, run on an SDS-PAGE gel, and transferred to PVDF. The blot was probed with anti-FLAG and anti-GAPDH as a loading control and visualized with Li-Cor secondary antibodies on a Li-Cor Odyssey CLx instrument. Arrowheads indicate nonspecific, anti-FLAG cross-reactive bands observed in Sheila Smith and Iowa strains. Download FIG S1, TIF file, 0.3 MB.

Strategy for generating SSS-ΔroaM. (A) Diagram of plasmids used to induce homologous recombination. (B) Agarose gel for PCR test for the presence of wild-type roaM and recombination using a primer that anneals outside the region of recombination (ANP0212) and a primer that anneals within either roaM (ANP0213) or the recombinant cassette (ANP0214). “Initial” indicates the initial isolation of SSS-ΔroaM, while “4×” indicates the series of four sequential plaque isolations to separate the knockout strain from the wild-type strain; the knockout strain was subsequently used in experiments. Molecular size standards are shown on the left (in kilobase pairs). Download FIG S2, TIF file, 0.4 MB.

Infection of Vero cell monolayers reveals differential patterns of spread determined by roaM genotype.

Infection at a low multiplicity of infection (MOI) of 0.1 results in distinctly small foci with limited cell-to-cell spread of rickettsiae at 24 h postinfection (hpi) for SSS (Fig. 2A). Overexpression of roaM FL exacerbates these small foci. Comparatively, rickettsiae with truncated roaM or lacking roaM appear more evenly distributed throughout the monolayer. The average number of rickettsiae per cell was determined. SSL, SSS-ΔroaM, and SSS-pROAM-T on average contained between 2 and 4 rickettsiae, while SSS and SSS-pROAM-FL contained between 4 and 5 rickettsiae, with a larger standard deviation (Fig. 2B). Images were also quantified for the ratio of infected cells to total cells, revealing that infection with SSL, SSS-ΔroaM, and SSS-pROAM-T led to more frequent cell-to-cell spread with a resultant higher percentage of infected cells at 24 hpi. Taken together, these observations indicate that full or partial deletion of roaM, or complementation with truncated roaM, induced more rapid cell-to-cell spread.

FIG 2

Low-MOI infection of Vero cells reveals differing rates of cell-to-cell spread among strains. (A) Vero cell monolayers on coverslips were infected at an MOI of 0.1, incubated at 34°C, and fixed with methanol at 24 or 48 h. An IFA was performed using anti-Na/K ATPase to visualize membrane boundaries (magenta), MAb 13-2 (66) for rickettsiae (green), and DAPI as a nuclear stain (cyan). Representative wide-field microscopy images are shown. The control with no primary antibody to rickettsiae (NPC) is a single, representative z-plane. Five randomly acquired wide-field images each from three independent infections for each strain were analyzed using Imaris 9.7 software by automated segmentation of nuclear and cell boundaries, as well as rickettsia identification. (B and C) Data from all experiments were pooled to show (B) the number of rickettsiae per ROI, with ROI representing single z-planes of individually segmented Vero cells at 24 h postinfection, and (C) the average percentage of infected cells per field. Error bars represent standard deviation. Statistics performed include standard ANOVA followed by Dunnett’s multiple-comparison test. ****, P ≤ 0.0001; **, P = 0.0014; n.s., not significant.

Truncation, deletion, and complementation with a truncated variant result in an increase in actin tail formation.

Smaller plaque size of intracellular pathogens on host-cell monolayers has previously been correlated with loss of actin tails (14, 20). To determine if the differences in spread between strains are related to the production of actin tails, Vero cells were infected with R. rickettsii strains at an MOI of 0.5, fixed at 10 hpi, and stained with phalloidin to visualize actin tail production (Fig. 3A). Quantification of actin tail production was accomplished using Imaris cell imaging software to identify actin filaments associated with rickettsiae from z-stacks obtained on a wide-field microscope and then corrected for error by normalizing to the values for Rcl sca2::Tn (Rcl is R clear, a higher-passage derivative of R), which is known not to produce long actin tails (Fig. 3B) (20). SSL, SSS-ΔroaM, SSS-pROAM-T, and Rcl produced actin tails at approximately three to four times the rate of SSS. Conversely, SSS-pROAM-FL produced almost no actin tails, although individual examples of SSS-pROAM-FL generating actin tails could be found at later time points, such as 48 hpi (Fig. 3C). This suggests that the lower rate of actin tail formation in SSS-pROAM-FL is not due to an overall inability to generate them.

FIG 3

Ablation of roaM or complementation with a truncated variant results in an increased rate of actin-tail formation. Vero 76 cell monolayers on coverslips were infected with R. rickettsii at an MOI of 0.5 and fixed with paraformaldehyde at 10 or 48 h postinfection. IFAs were performed, visualizing actin with phalloidin-Alexa Fluor 568 (magenta) and rickettsiae with MAb13-2 (green) and using DAPI as a nuclear stain (cyan). (A) Representative maximum intensity projections derived from wide-field images showing production of actin tails by different R. rickettsii strains at 10 h postinfection. The control with no primary antibody to rickettsiae (NPC) is a single, representative z-plane. (B) Three independent experiments, each with five randomly acquired z-stacks for each strain, were analyzed using Imaris 9.7 to automatically identify rickettsiae and actin filaments likely to be tails (by filtering for size and sphericity). Each experiment was corrected for error by subtracting the results for production of tails by the sca2 mutant, which is known to not generate any large actin tails (20). The resulting data matched results from a manually annotated subset. Error bars represent standard errors of the means. Standard ANOVA followed by Dunnett’s multiple-comparison test was used to determine the significance of results compared to SSS. ****, P ≤ 0.0001; ***, P = 0.0002; **, P = 0.005 to 0.01. (C) Representative maximum intensity projection derived from a wide-field acquired z-stack showing actin tail production by SSS-pROAM-FL at 48 h postinfection.

Determination of roaM translation start site indicates that the large plaque size of the R. rickettsii Iowa strain is due to a previously identified point mutation.

We investigated in detail the M4I nonsynonymous mutation in roaM in Iowa compared to Sheila Smith (26). The Iowa strain makes large, opaque plaques (27). The first four annotated amino acids in RoaM for Sheila Smith are MMSM, raising the possibility that more than one of these ATG codons could serve as the actual translation start site. To address this point, a reporter plasmid to measure promoter activity was constructed, pRrReporter (Fig. 4A). Promoter sequences up to the potential translational start site of interest were cloned into pRrReporter to drive the transcription and translation of GFPuv (green fluorescent protein with maximal fluorescence under UV light) (Fig. 4B). The guanidine residue highlighted in Fig. 4B was empirically determined by capping-RACE (rapid amplification of cDNA ends) (28) to be the transcriptional start site (Fig. S3). Measured fluorescence intensity can then be normalized to constitutively expressed mCherry, driven by the rompA promoter. Fluorescent microscopy images for the reporters in Escherichia coli and R. rickettsii (Fig. 4C) were quantified using FIJI image analysis (Fig. 4D and E). No variants of the roaM promoter were able to drive production of GFP in E. coli (Fig. 4D). In R. rickettsii, only P 3rd was able to produce detectable GFP fluorescence. This suggests that the third methionine is the actual first residue of RoaM. Subsequent numbering of amino acids here is based upon this translational start site.

FIG 4

Determination of roaM translational start site via promoter-driven fluorescence analysis. (A) Schematic of the pRrReporter plasmid. (B) Diagram of the roaM promoter in SSS and variant promoters cloned into pRrReporter to generate different reporter plasmids. (C) Representative wide-field microscopy images of reporter constructs in E. coli and R. rickettsii. Images were acquired with identical microscopy parameters, and brightness/contrast were adjusted uniformly for all images within bacterial species. (D and E) Quantification of GFP expression normalized to mCherry expression for reporter constructs in E. coli (D) and R. rickettsii (E) and the average relative intensity for identified bacteria was obtained for three independent experiments, each utilizing three fields per strain. The data were subsequently normalized to an unexpressed, putative pseudogene (PG). Brown-Forsythe and Welch ANOVA with Dunnett’s multiple-comparison tests were performed on the prenormalized data. **, P = 0.0044; *, P ≤ 0.0214; n.s. not significant. Error bars represent standard errors of the means.

Sequence alignment for determination of transcriptional start sites of rompB/sca5 and roaM. Results from Sanger sequencing of individual clones from the primer ANP0453 to the translational start site (TSS) of roaM and rompB are shown aligned to the region upstream of the translational start site of those genes. The empirically determined transcriptional start site and the translational start site are highlighted in cyan. Download FIG S3, TIF file, 0.4 MB.

Because P Iowa did not produce detectable GFP, the Iowa strain may not generate RoaM due to the nonsynonymous point mutation affecting the translational start site. To confirm this, a rabbit polyclonal antibody against RoaM was generated and used to probe Western blots of purified rickettsiae (Fig. 5A). Anti-RoaM detected a band in SSS approximating the predicted size of 41.5 kDa, and a smaller band in SSL close to the predicted size of truncated RoaM, 24.6 kDa. Preincubation of the antibody with the immunizing peptide antigen inhibited antibody binding to either the full-length or truncated RoaM, thus demonstrating specificity of the reaction. No bands were detected for SSS-ΔroaM or Iowa, indicating that Iowa does not produce significant amounts of RoaM. The transcription of roaM was also tested by amplifying PCR fragments from cDNA internal to roaM, A1G_06525, and gltA as a control (Fig. 5B). SSS-ΔroaM does not produce the fragment corresponding to roaM but does transcribe A1G_06525, likely as readthrough from the transcription of gfpuv. The Iowa strain produces transcripts for both roaM and A1G_06525, even though it does not translate roaM in any significant amount. Note that roaM and A1G_06525 are predicted to be transcribed as an operon (29), and PCR amplification from rickettsial cDNA using a primer pair with annealing sites internal to roaM and internal to A1G_06525 produce a band, confirming the transcript as polycistronic (Fig. S4).

FIG 5

R. rickettsii Iowa does not produce RoaM despite transcribing roaM. (A) Western blot of samples solubilized in 2× Laemmli sample buffer and probed with anti-RoaM (red) and MAb 13-2 (green) with or without 5 μg/mL competing peptide antigen. White stars indicate full-length and truncated RoaM. (B) Agarose gel of PCR products for gltA, roaM, and A1G_06525 from cDNA derived from SSS, SSS-ΔroaM, and Iowa strains.

roaM and A1G_06525 constitute a polycistronic mRNA. Agarose gel for a PCR product from primers designed to amplify a region starting within roaM and ending within A1G_06525 from cDNA and gDNA derived from SSS. Molecular size standards are shown on the left (in kilobase pairs). Download FIG S4, TIF file, 0.3 MB.

RoaM is localized to the bacterial cytosol and is not secreted.

To establish whether RoaM is secreted, we generated a SSS strain expressing glycogen synthase kinase 3β (GSK-3β)-tagged RoaM FL. This strategy was previously used (30) to determine that R. rickettsii RARP2 is a secreted effector. If a protein of interest is secreted and exposed to host cytosol, the GSK tag will be phosphorylated. A Western blot using total versus phospho-specific GSK-3β antibodies is used to determine the phosphorylation status of the protein. In the case of GSK-RoaM FL, we did not detect a phosphorylated band (Fig. 6A). By way of confirmation, we also utilized our FLAG-tagged RoaM strains in immunofluorescence assays (IFA). We found that the FLAG epitope from our nonsecreted control and RoaM variants could be detected within rickettsiae only when a lysozyme treatment and additional permeabilization step was added (Fig. 6B, compare upper to lower panels). The pattern of localization of RoaM appears to match that of FLAG-SS-MTase (30), a methyltransferase that is not secreted and is predicted to be localized to the rickettsial cytosol. Taken together, the results of these two experiments suggest that RoaM is not secreted.

FIG 6

RoaM is not secreted by R. rickettsii. (A) Western blots of Vero 76 cell lysates infected with R. rickettsii expressing GSK-tagged proteins. Cells were infected at an MOI of 1, and lysates were prepared at 24, 48, and 72 hpi. Samples were run on duplicate gels and probed for phosphorylated (left) and total (right) GSK-tagged protein (red). MAb 13-3 against rOmpA (green) (66) was used as a loading control. Images are representative of three independent experiments. White stars in the 48-hpi panels indicate GSK-SS-RARP2, GSK-RoaM-FL, and GSK-GFP and proteolytic fragments of each. (B) Immunofluorescence assay for R. rickettsii Iowa expressing Sheila Smith RARP2 (Iowa-pSS-RARP2) (positive control for secretion), R. rickettsii Iowa expressing RrIowa_1431 (Iowa-pSS-MTase) (negative control for secretion) (30), and SSS-pROAM-FL and SSS-pROAM-T probed with anti-FLAG to visualize localization of ectopically expressed FLAG-tagged proteins. Cells shown in upper right images were permeabilized with 0.1% Triton x-100 only; those in lower right images were additionally treated with lysozyme and a secondary 0.1% Triton X-100 permeabilization. Uninfected and Iowa-pSS-RARP2 no-primary-antibody control samples underwent lysozyme treatment and secondary permeabilization steps. Large pleomorphic structures in Iowa-pSS-RARP2 panel are endoplasmic reticulum-derived structures resulting from expression of SS-RARP2 (30).

Real-time quantitative PCR reveals no roaM-dependent reduction in expression level of genes implicated in actin-based motility or cell-to-cell spread.

To determine whether RoaM reduces the formation of actin tails by inducing transcriptional repression of genes important for actin-based motility and spread, the expression of sca2, rickA, and sca4 between strains was assessed by qRT-PCR (Fig. 7). Sca2 and RickA are involved in the polymerization of actin (20, 23, 31, 32), and thus, overexpression of either of these genes might have accounted for the increase in actin tail formation for some strains. Additionally, Sca4 mediates cell-to-cell spread by R. parkeri by interacting with vinculin to decrease the physical tension at cell-cell junctions and allow membrane protrusions into neighboring cells (33); therefore, its overexpression could also provide a partial explanation for the observed differences in actin-based motility between roaM mutants. Instead, no significant overexpression of sca2, rickA, or sca4 was observed between strains when normalized to either of two control genes, gltA and rpoD. Likewise, for SSS-pROAM-FL, no significant repression of these genes was observed. Based on these data, we conclude that it is unlikely that RoaM acts at the level of transcriptional regulation of these known actin-regulating genes. However, other unknown factors may be regulated by RoaM.

FIG 7

roaM expression has no discernible effect on the transcription of rickettsial genes known to influence motility or cell-to-cell spread. Three combined independent qRT-PCR experiments comparing the expression of sca2, sca4, and rickA normalized to gltA or rpoD for different R. rickettsii strains, with various roaM expression shown as an experimental control. Error bars represent standard deviations. Data were analyzed by two-way ANOVA followed by Dunnett’s multiple-comparison test revealing no significance for the expression of sca2, sca4, and rickA for strains compared to SSS with the exception of one comparison; the difference in sca2 expression between SSS and SSS-pROAM-FL normalized to rpoD, with a P value of 0.0183. Each qRT-PCR experiment included three technical replicates per data point and controls without cDNA, negative controls without reverse transcriptase, and positive controls with gDNA to validate the results of the experiment. Melting curves were examined to ensure fidelity of each primer set.

Examples of both full-length and truncated roaM are found bioinformatically within different rickettsial species, and roaM genotype is predictive of plaque size phenotype in other R. rickettsii strains.

Using genomic sequences available on the PATRIC bioinformatics resource center (34), we compared the state of the operon containing roaM and A1G_06525 in different isolates of several Rickettsia spp. (Fig. 8). When genomic sequences of multiple isolates are available for the same species, it is often possible to find examples where the full-length roaM is present and others where it is ablated. This suggests that roaM may be mutated during different linages and passage histories. In the case of R. heilongjiangensis, the type strain 054, which has been maintained in Vero cells since its isolation in 1982, was sequenced in 2011 (35) and contains a truncated version of roaM. By comparison, R. heilongjiangensis isolated directly from multiple individual ticks and genomically sequenced contained full-length roaM homologs (36). These observations suggest that roaM may be selected against during repeated passage in cell culture.

FIG 8

Annotated genotype of roaM in various Rickettsia spp. and isolates. Schematic of roaM and A1G_06525 operons from various Rickettsia spp. and isolates based on annotated genomes on PATRIC (34). Yellow bars indicate the approximate length of the truncated roaM mutants. Critical mutations, differences between isolates, and the translated size of RoaM are listed to the right.

To determine if the integrity of roaM could broadly predict plaque size for other isolates of R. rickettsii, we plaqued other strains in our collection on Vero cell monolayers (Fig. 9A) and measured the area of 50 plaques for each strain (Fig. 9B). The roaM/A1G_06525 operon was PCR amplified from isolated genomic DNA and Sanger sequenced (Fig. 9C). An association was observed between truncations of roaM and a larger plaque size among R. rickettsii isolates.

FIG 9

Genotype of roaM in R. rickettsii strains is correlative with plaque size. (A) Photographs of R. rickettsii plaques from different strains. Bar = 1 cm. (B) Measured plaque area for R. rickettsii strains at 8 days postinfection. Results represent a single experiment where 50 plaques were measured for each strain. Error bars represent standard deviations. One-way ANOVA (Kruskal-Wallis) followed by Dunn’s multiple-comparison test was performed. The average plaque area for SSS was significantly different (<0.0001) from that for SSL, Hauke, Hino, Rcl, and Iowa, while plaque areas for Sao Paulo and HLP 7421 were not significantly different. Hauke, Hino, Rcl, and Iowa were not significantly different from SSL. (C) Schematic of the roaM- and A1G_06525-containing operon as empirically determined by PCR amplification and Sanger sequencing. Mutations and overall amino acid length of RoaM are listed on the right.

Serial passage of R. rickettsii Sheila Smith small in Vero cells results in the spontaneous generation of roaM mutants.

To determine if we could recapitulate the loss of roaM in cell culture, SSS was serially passaged every 3 to 4 days in Vero cells over a 5-week period. Three separate lineages were maintained. Large, clear plaques were first detected by passage 6. These plaque variants were readily apparent by the eighth passage and became more abundant over subsequent passages (Fig. 10A and B). For each lineage, a large clear plaque was isolated from passage 10 and subcloned twice. Sanger sequencing of the roaM region from DNA amplified from each of these plaques revealed that each lineage developed a novel mutation in roaM (Fig. 10C). The ease with which we were able to generate phenotypically large, clear plaques and associated mutations in roaM suggests that there is strong negative selection of roaM when R. rickettsii is cultured in Vero cells.

FIG 10

Serial passage of SSS in Vero cells leads to roaM mutations and an emergent large-plaque phenotype. Three separate linages were maintained. (A) Examples of plaques at passages 4, 8, and 12 from lineage B. Bar = 1 cm. (B) Ratios of small to large, clear plaques for lineages A, B, and C over successive passages. Each data point represents the average of two replicate wells, and error bars represent standard deviations. (C) Schematic of operon containing roaM and A1G_06525, which were PCR amplified from a single large, clear isolated plaque from passage 10 of each lineage and subsequently Sanger sequenced. Mutations and overall amino acid length of RoaM are listed on the right.

Overexpression of full-length RoaM delays onset and duration of fever in guinea pigs.

To determine the impact of various roaM genotypes on virulence, groups of three female Hartley guinea pigs were each inoculated intradermally with 100 PFU of the given strains, and temperature was monitored daily via a subdermal transponder (Fig. 11). Guinea pigs infected with most strains developed an appreciable increase in fever between days 3 and 4, except for those with SSS-pROAM-FL, for which the increase in fever began between days 6 and 7. The duration of fever for SSS-pROAM-FL was 6 days compared to 7 to 9 days for the other strains. The delayed onset and shorter duration of fever for guinea pigs infected with SSS-pROAM-FL suggests that the observed reduction in actin tail formation for this strain in vitro may account for a decrease in overall virulence.

FIG 11

Overexpression of RoaM delays onset and duration of fever in guinea pigs. Three guinea pigs per strain were challenged intradermally with 100 PFU of R. rickettsii. Temperature was monitored by intradermal transponder. Values are means and standard deviations. Guinea pigs became febrile between days 3 and 4 for most strains and exhibited subsidence of fever by day 11. Animals infected with the overexpressing SSS-pROAM-FL strain showed a delayed onset and shorter duration of fever.

DISCUSSION

Here, we describe RoaM, a previously unknown regulator of actin dynamics in Rickettsia spp. RoaM was first identified in genomic comparisons as one of the few genes differing between the avirulent Iowa strain and virulent Sheila Smith, Morgan, and R strains of R. rickettsii (25, 26). Subsequent bioinformatic analysis of roaM from different spotted fever group rickettsiae revealed a mix of full-length and truncated versions of the gene, often from within the same rickettsial species. Where passage history was available, it appeared that minimally passaged strains were more likely to possess a full-length copy. Genetic disruption of roaM or expression of full-length and truncated versions indicated that RoaM is implicated in plaque size and actin-based motility. RoaM was not secreted from the rickettsiae and thus does not appear to directly interact with actin. Transcriptional analysis showed no effect of RoaM on expression levels of known rickettsial actin-interacting genes, thus suggesting that additional, unknown, regulatory mechanisms control actin-based motility in rickettsiae.

The disruption of roaM in R. rickettsii Sheila Smith resulted in larger plaque sizes and increased frequency of actin tail formation but no increase in virulence in a guinea pig model of infection. However, overexpression of roaM resulted in even smaller plaques than in the wild-type SSS and decreased frequency of actin tail formation. In the case of roaM overexpression, we observed a modest reduction in fever response and duration of symptoms. Although plaque size can be a useful phenotype and may be indicative of changes in actin-based motility, it does not necessarily equate with virulence. For example, in the Iowa strain, small- and large-plaque variants have been isolated, yet neither is virulent in guinea pigs (24), presumably due to the multiple mutations observed in comparison to virulent strains (26). Similarly, both clear (lytic) and opaque (nonlytic) plaque phenotypes have been isolated from both the virulent R strain and avirulent Iowa strains of R. rickettsii, yet neither plaque phenotype differs in virulence from the parental strain (21).

As obligately intracellular pathogens, Rickettsia spp. possess a relatively minimized genome and have been hypothesized to be in a perpetual state of genome reduction due largely to their dependence upon the host for many nutrients and growth factors that they would otherwise need to synthesize (37). In the case of agents of vector-borne diseases, complex interactions with the host and vector likely impose positive or negative selective pressure on different genes (38). Extended laboratory passage of facultative intracellular bacterial pathogens (39–43) is well known to frequently lead to a loss of virulence factors due to the absence of host-imposed selective pressures. A more complex dynamic may exist in vector-borne diseases. The interesting possibility arises that certain rickettsial genes may be essential in either the vector or the host but detrimental in the other and, thus, a balance must be achieved. Wild-type roaM seems to be found in fresh isolates of R. rickettsii but appears to be strongly selected against during cell culture passage. The rapid loss of roaM during laboratory passage was recapitulated experimentally in Vero cells. This suggests that there are rickettsial genes essential in the tick reservoir that are unnecessary or even deleterious in mammalian hosts. The strong negative selective pressure against RoaM in Vero cells may explain our failure to complement the SSS-ΔroaM mutant with full-length roaM despite repeated efforts.

Other rickettsial genes, such as that encoding a major surface antigen, rOmpA, appear to have been lost from the Iowa strain of R. rickettsii (24–26). The Iowa strain is unique in that it has been passed 271 times through embryonated chicken eggs. It was originally described as mildly virulent and then as having increased virulence and then became avirulent (44). The original isolate is no longer available, and most of these studies were done before plaque purification of rickettsiae was developed. Later, a targeted knockout of rompA revealed that rompA is not essential for virulence in a guinea pig model. It was speculated that the gene might be necessary for maintenance in of R. rickettsii in ticks (45); however, the recognition that rompA is disrupted in Rickettsia peacockii (46, 47), a commensal in ticks, suggests that may not be the case.

It is also possible that there are rickettsial genes necessary for survival or virulence in mammalian cells that are superfluous in ticks. Certainly, virulent strains of R. rickettsii can be detrimental to the tick vector. The death of Dermacentor andersoni infected with R. rickettsii was noted by Burgdorfer in 1963 (48), and a more thorough study by Niebylski et al. (49) showed a very high mortality rate for infected ticks and introduced the effects of temperature on the survival of infected ticks. Labruna et al. similarly found that infection of Amblyomma aureolatum with the sympatric strain R. rickettsii Taiaçu also resulted in mortality and reduced oviposition of infected ticks (50). Still other studies using different tick-Rickettsia sp. combinations have found minimal or no apparent effects of rickettsial infection on the arthropod vector (51–54). Although many of the differences in these studies may be accounted for by the differences between tick species, infection with a rickettsia lacking a mechanism to check its activation of actin-based motility could be responsible for additional cellular damage to the vector, leading to decreased survival. Although not confirmed experimentally, it is conceivable that extended passage through ticks could select against rickettsial genes essential in mammals but not required in ticks. Such a mechanism could explain the existence of rickettsial species such as R. peacockii, which is transmitted transovarially in ticks yet for which there is no evidence for transmission to mammals and which has not been propagated in mammalian cell culture (55, 56). Multiple genes thought to contribute to the virulence of R. rickettsii, including rarP2, rompA, sca1, rickA, dsbA, the protease II gene, and a putative phosphoethanolamine transferase gene (47), are deleted in R. peacockii.

The effects on virulence of R. rickettsii after serial passage in different hosts or vectors may have been presaged by Price in the 1950s (57). Virulent strains passaged by spleen transfer through cottontail rabbits, a target of Haemaphysalis leporispalustris ticks resulted in decreased virulence. These data were recapitulated for both dogs and opossums (57). However, these strains were able to maintain their virulence if passed through cottontail rabbits from one animal to another via Dermacentor andersoni nymphs. Perhaps this series of experiments was an early indicator that there are important selective pressures inherent in the environment within the tick. It is possible that the difference in selective pressures might be leveraged as a tool to explore loss or gain of function under various culture conditions (58). Serial passage in different arthropod and mammalian cell types or whole animals may reveal mutations in genes important for managing the relationship between bacterium, host, and vector. There have been several alternative methods for the culture of Rickettsia spp. explored using tick and mosquito cell lines (56, 59), and these methods might provide greater genomic stability or a different menu of selective pressures for gene loss than mammalian cell culture.

The observation that disruption of RoaM can increase actin tail formation and cell-to-cell motility has implications for the interpretation of studies in actin-based motility that rely on measurement and characterization of actin tails. An allele strongly downregulating tails in some situations could also exist, and because the stimulus for turning actin motility on or off is not yet known, this could result in the appearance of being immotile, as R. peacockii appears to be despite possessing sca2. Clearly, roaM presents a confounding factor that will need to be resolved for future actin-based motility studies in Rickettsia species. An additional consequence is that R. rickettsii roaM mutants produce large, robust plaques, which may lead to unintentional selection during plaque isolation procedures. The unknowing usage of hyperspreading mutants may also have consequences for the survival and fecundity of the arthropod vector.

If rickettsiae containing wild-type roaM are able to regulate their cell-to-cell spread, it would not be surprising to find that the reason for this is to more efficiently utilize the resources in a particular host cell until those resources become limiting. At such a time, an unknown signal may cause Rickettsia to activate its motility. RoaM is cytosolic, and it has no effect on the transcription of known rickettsial genes related to motility, although a broader screen will be necessary to rule out a possible role as a transcriptional regulator. The functional mechanism of RoaM may otherwise be based on protein-protein interactions or signaling. Some clues may be inferred from ectopic expression of RoaM T resulting in increased motility. RoaM T could achieve this phenotypic effect by multimerizing incorrectly with wild-type RoaM. Alternatively, if RoaM and A1G_06525 interact, as might be inferred from their synteny, overexpression of RoaM T may titrate A1G_065525 away from its normal function, where it would interact with wild-type RoaM. Recent work with Rickettsia parkeri has shown that actin tails disengage when in a protrusion before entry into an adjacent cell (33), and here may be an instance where RoaM might negatively regulate actin tails. Both Phyre2 and DELTA-BLAST suggest a small degree of structural similarity to the myosin heavy chain (60, 61). It is tempting to speculate whether RoaM might have a similar structural role and perhaps be involved in mechanosensing, similar to myosins (62). Future work will necessarily be driven by the need to reveal the mechanism by which RoaM regulates actin-based motility and is likely to identify additional factors controlling actin-based motility.

MATERIALS AND METHODS

Cell lines and rickettsiae.

Vero76 cells (ATCC CCL-81) were grown in RPMI medium plus 5% heat-inactivated fetal bovine serum (FBS) at 37°C in a 5% CO2 atmosphere. Rickettsia rickettsii Sheila Smith (CP000848.1) organisms were grown at 34°C in Vero 76 cells in M199 medium plus 2% heat-inactivated FBS in a 5% CO2 atmosphere. For purification of rickettsiae, infected Vero cells were lysed by Dounce homogenization followed by centrifugation through a 30% Renografin pad. Rickettsiae were washed twice in 250 mM sucrose and stored in brain heart infusion broth (BHI) at −80°C. Each rickettsial stock was titrated for PFU by plaque assay. For plaque assays, Vero cells were grown to monolayers in 6-well plates in RPMI 1640 medium with 5% FBS as previously described (19, 20). The medium was removed, and 100 μl of R. rickettsii-containing BHI dilutions was added directly to each well. Plates were incubated for 30 min at 34°C and 5% CO2. Vero cells and R. rickettsii dilutions were overlaid with 5 mL of M199 medium with 5% FBS and molten 0.5% low-melting-point agarose and were grown for 5 days at 34°C and 5% CO2. Wells with visible plaques were stained with 0.5 mL thiazolyl blue tetrazolium bromide (MTT; 0.3 mg/mL in K36 buffer) overnight, PFU were counted the next day, and PFU/mL were calculated. PFU were used to calculate multiplicities of infection (MOI).

Strain histories of R. rickettsii from the RML culture collection used here are listed in Table 1.

TABLE 1

Strain histories of R. rickettsii in the Rocky Mountain Laboratories collection used here

Strain	Passage historya	Clone	roaM b
Sheila Smith (SS) (1961 seed)	5EP/4VC	1	FL
SSS	5EP/4VC/1MP/11VC	1	FL
SSL	5EP/4VC/1MP/14VC	2	T
Rclc	10EP/3VC/1 MP /12VC	1	T
Rcl #9-7 Sca2 KO	10EP/3VC/1 MP /17VC	#9-7	T
Iowa	271EP/1GP/9VC/1MP/14VC	1	T
Sao Paulo	8EP/4VC/1MP/6VC	1	FL
HLP 7421	7EP/6VC/1MP/5VC	2	FL
Hauke	?/1TC/7VC	1	T
Hino	?/5EP/8VC	1	T

Number of passages in embryonated eggs (EP), Vero cells (VC), BALB/c mice (MP), guinea pig (GP), and tissue culture (TC). ?, unknown number of passages.

Genotype of roaM. FL, full length; T, truncated (truncation varies with strain; see the text).

R clear (Rcl) is a higher-passage derivative of R. It is the strain transposon mutagenized to generate the Sca2 knockout (20).

PCR amplification and plasmid and genomic DNA isolation.

PCR products were amplified from target DNA using primers ordered from Integrated DNA Technologies (Table S3) with Q5 high-fidelity DNA polymerase (New England Biolabs [NEB], M0491), isolated using a GeneJet PCR purification kit (Thermo Scientific, K0702), and ligated into vector backbones using T4 DNA ligase (NEB, M0202), all per manufacturers’ instructions. Generated plasmids were transformed into NEB stable competent E. coli (NEB, C3040H) and were prepared using GeneJet plasmid miniprep or maxiprep kits (Thermo Scientific, K0503 and K0492). Isolation of rickettsial genomic DNA was accomplished by processing individual stock aliquots with the DNeasy blood and tissue kit (Qiagen, 69504). Sanger sequencing to confirm constructs and PCR products derived from genomic DNA were performed by ACGT Inc.

Strains. Download Table S1, DOCX file, 0.02 MB.

Plasmids. Download Table S2, DOCX file, 0.01 MB.

Primers. Download Table S3, DOCX file, 0.02 MB.

Construction of R. rickettsii strains.

Plasmids pROAM-FL and pROAM-T were generated from pRAMF2-N and transformed into the small-clone variant of Sheila Smith (SSS) to constitutively overexpress full-length and truncated RoaM (Fig. S1). A genetic knockout of roaM was generated via recombination with a plasmid-based cassette expressing GFP and rifampin resistance. The strategy employed was based on the FRAEM method employed for Chlamydia (63). Briefly, a cassette consisting of constitutively expressed GFP and rifampin resistance genes flanked by two regions of homology (2,039 bp upstream and 1,433 bp downstream) flanking the knockout target, roaM, were cloned into pRKO (Fig. S2A). This fragment was then cloned into a backbone that constitutively expresses mCherry. After transformation and passaging, homologous recombination was selected for by picking GFP-positive, mCherry-negative plaques. This strategy was, however, unsuccessful in generating knockouts of other genes, including a putative pseudogene and genes where knockouts have already been generated via transposon mutagenesis.

Growth curves for R. rickettsii strains.

For each strain, five 25-cm2 flasks of confluent Vero cells were infected at an MOI of 0.25 and rocked at room temperature for 30 min. Eight milliliters of M199 with 2% FBS was then added, and flasks were incubated at 34°C. To harvest, medium was removed, and cells were scraped into 1 mL of BHI. Specimens were added to a microcentrifuge tube containing presterilized glass beads, and host cells were disrupted by using a Mini Beadbeater (BioSpec Products) for 10 s. Samples were divided and stored at −80°C in an isopropanol cooler until quantitation of PFU.

Plaque size measurements.

To determine average plaque size for various R. rickettsii strains, Vero cells in six-well plates were infected with rickettsiae diluted in BHI such that approximately 10 to 40 plaques per well were obtained. Plates were stained with MTT (0.3 mg/mL in K36 buffer) at 8 or 15 days postinfection. Plaques were imaged on a Bio-Rad ChemiDoc MP system with transillumination. The diameter of plaques was measured with a digital caliper (Mitutoyo, 500-171-30) at a consistent horizontal angle, from top to bottom, left to right, to avoid bias.

SDS-PAGE and Western blotting.

Rickettsia-infected Vero 76 cell lysates were generated by adding 100 μL 2× Laemmli buffer (64) per well of a 24-well plate. Samples were inactivated by heating at 99°C for 10 min. Samples were run on an SDS-PAGE gel and transferred to a polyvinylidene difluoride (PVDF) membrane at 100 V for 1 h and 15 min in Towbin transfer buffer (65) containing 20% methanol. Membranes were dried overnight, briefly rehydrated in methanol, rinsed with Tris-buffered saline (TBS), and blocked for 1 h in Odyssey blocking buffer (Li-Cor, 927-50000) at room temperature. Primary antibodies were diluted in blocking buffer with 0.2% Tween 20, while secondary antibodies IRDye 680RD goat anti-rabbit immunoglobulin and IRDye 800CW goat anti-mouse immunoglobulin (Li-Cor, 926-68071 and 926-32210) were diluted 1:10,000 in blocking buffer with 0.2% Tween 20 and 0.02% SDS. Blots were then visualized on an Odyssey CLx system and analyzed in Image Studio 5.2 (Li-Cor).

Secretion of RoaM was assayed utilizing GSK-tagged proteins as described previously (30). Vero 76 cells were grown in a 24-well plate and infected with GSK-tagged protein expressing strains of R. rickettsii at an MOI of 1 for 24, 48, and 72 h before harvesting in Laemmli sample buffer as described above. Western blots were probed with anti-phospho-GSK-3β (1:1,000) (Cell Signaling Technology, 5558) to detect phosphorylation of the GSK tag or anti-GSK-3β tag (1:1,000) (Cell Signaling Technology, 9325) as a control for production of the GSK-tagged proteins. Monoclonal antibody (MAb) 13-3 (1:1,000) was used as a loading control for rickettsiae.

Generation of RoaM polyclonal antibodies.

Polyclonal antibodies to RoaM were generated commercially (GenScript) by immunizing two New Zealand White rabbits with a three-booster series of 0.2 mg keyhole limpet hemocyanin (KLH)-conjugated synthetic peptide derived from RoaM (NREILTHEEIFKC) predicted to be antigenic by a proprietary algorithm (GenScript). A C-terminal cysteine was added to aid in conjugation to the KLH. Sera were collected and antigen affinity purified.

Immunofluorescence assays and microscopy.

Vero cell monolayers were grown on round 12-mm no. 1.5 coverslips (Glaswarenfabrik Karl Hecht GmbH & Co. KG) in 24-well plates using RPMI medium with 5% FBS. Monolayers were infected at the desired MOI as specified in the figure legends and synchronized by centrifuging at 750 × g at room temperature for 5 min followed by incubation at 34°C until the desired time point for analysis. Cultures were rinsed with 500 μL Hanks’ balanced salt solution with Ca2+ and Mg2+ (Gibco, 14025076) and then fixed with either 1 mL 4% paraformaldehyde in PBS or 1 mL of methanol for 20 min. After permeabilization, blocking, and probing with specific antibodies, coverslips were mounted on slides using ProLong Diamond antifade mounting medium (Thermo Fisher, P36961). All immunofluorescence experiments incorporated controls to which no primary antibody was added in order to control for potential nonspecific binding. Imaging was performed on a Nikon Ti2 Eclipse wide-field microscope equipped with a Nikon DS-Qi2 camera, Lumencor SOLA light engine, and Semrock filters DAPI-3060A, GFP-4050B, and LED-TRITC-A, using a Plan Apo λ 60× oil immersion objective with a numerical aperture (NA) of 1.4. Acquisition of random fields was accomplished using Nikon NIS-Elements Advanced Research version 5.11.02, autofocusing on the DAPI (4′,6-diamidino-2-phenylindole) channel.

A modified permeabilization strategy with FLAG-tagged constructs was also performed to assess localization of RoaM (30). Vero 76 cells were grown on coverslips in a 24-well plate and infected at an MOI of 1 with FLAG-tagged protein expressing R. rickettsii strains. After fixation at 48 h postinfection with 4% paraformaldehyde in PBS, all samples were permeabilized for 10 min with 250 μL Triton X-100 in PBS at room temperature. A subset of samples were then incubated with 1 mg/mL lysozyme in PBS (Sigma, L6876) for 10 min at room temperature, followed by an additional 10 min treatment with Triton X-100 in PBS at room temperature. Coverslips were probed with anti-FLAG (Sigma, F1804) (1:500 in PBS) for 1 h at room temperature, followed by 3 washes with 500 μL PBS and then goat anti-mouse IgG H&L (heavy plus light chain) conjugated to Alexa Fluor 568 (Abcam, ab175701) (1:500) for 1 h at room temperature. DAPI (1 μg/mL in PBS) was applied to each well for 5 min at room temperature as a counterstain. Each well was washed 3 times with 500 μL/well PBS before mounting in ProLong Diamond.

Analysis of R. rickettsii foci of infection.

Vero 76 monolayers on glass coverslips in 24-well plates were infected at an MOI of 0.1. Cultures were rinsed with 500 μl of HBSS and then fixed and permeabilized with 1 mL/well methanol for 20 min at room temperature at 24 and 48 hpi. Specimens were then blocked with 250 μl 1% bovine serum albumin (BSA) for 1 h at room temperature, probed with 1 μg/mL MAb 13-2 (rOmpB) (66) and anti-Na/K ATPase (Abcam, ab76020; 1:500) in PBS, and incubated at 4°C overnight. Coverslips were washed 3 times with 500 μl of PBS, labeled with goat anti-mouse IgG H&L-Alexa Fluor 488 (1:500 in PBS) (Abcam, ab150117) and goat anti-rabbit IgG H&L-Alexa Fluor 568 (1:500 in PBS) (Abcam, ab175695) for 1 h at room temperature, and then counterstained with DAPI (1 μg/mL) for 5 min at room temperature. Coverslips were washed 3 times with 500 μL PBS before mounting in ProLong Diamond. Five random wide-field images were acquired for each strain and processed in Imaris 9.7.2. Individual cells were segmented based on the Na/K ATPase staining of the cell membrane and DAPI staining of the nucleus using the Imaris cell module to generate regions of interest. Rickettsiae were automatically identified and attributed to their respective regions of interest (ROI) to assess the number of rickettsiae and the percentage of infected cells per strain.

Quantification of actin tails.

Vero 76 monolayers on glass coverslips were infected at an MOI of 0.5, incubated at 34°C in Vero 76 cells in M199 medium plus 2% heat-inactivated FBS in a 5% CO2 atmosphere. Cultures were fixed at 10 hpi in 4% paraformaldehyde for 20 min at room temperature, permeabilized with 0.1% Triton X-100 for 15 min at room temperature, and then blocked with 1% BSA in PBS for 1 h. Coverslips were probed with 1 μg/mL MAb 13-2 in PBS overnight at room temperature. Coverslips were then washed 3 times with PBS and probed with goat anti-mouse IgG H&L-Alexa Fluor 488 1:400 in PBS (Abcam, ab150117) and Alexa Fluor 568 phalloidin (1:400 in PBS) (Thermo Fisher, A12380) for 3 h at room temperature. Coverslips were counterstained with 1 μg/mL DAPI for 5 min at room temperature. Coverslips were washed 3 times with 500 μL PBS before mounting in ProLong Diamond. Five random, wide-field 10-μm z-stacks were acquired in 0.27-μm steps for each strain. Actin tails were quantified using Imaris 9.7.2 by identifying phalloidin-stained actin filtered by mass and sphericity and then identifying rickettsiae that were located within 1.06 μm of the identified filaments. The 1.06-μm threshold for association between software-defined rickettsiae and actin filaments to define rickettsiae as producing tails was empirically determined by comparison to manually scored data subsets. Data were corrected for misidentification of actin filaments as tails in close proximity to Rickettsia by using R. rickettsii Rcl sca2::Tn, which does not produce actin tails (20). Manual scoring of three z-stacks for each strain from one experiment gave data equivalent to those obtained with the automated methodology.

roaM promoter analysis.

R. rickettsii strains carrying GFP reporter constructs were used to infect Vero 76 monolayers on coverslips at an MOI of 0.25 and fixed with 4% paraformaldehyde at 24 hpi. Rickettsia-infected cells were then permeabilized with 0.1% Triton X-100 for 5 min and counterstained with 1 μg/mL DAPI for 5 min. Coverslips were then washed 3 times with PBS before mounting in ProLong Diamond. before mounting. Overnight cultures of E. coli strains carrying reporter constructs were diluted 1:100 in LB and incubated for 3 h at 37°C. Samples were then diluted 1:50 in LB, and 5 μL of each sample was allowed to dry onto a coverslip before mounting in ProLong Diamond antifade mounting medium. For each experiment, three images were taken for each strain and processed with the FIJI distribution of ImageJ2 (67). ROI were generated by defining a mask based on the TRITC (tetramethyl rhodamine isocyanate) channel by applying a Gaussian blur with a sigma of 2, rolling ball background subtraction set to 50 pixels (rickettsiae only), and thresholding using the Otsu algorithm (68). The resulting ROI were then applied to the original GFP and TRITC channels to obtain intensity values for each individual ROI with an area between 0.5 and 10 pixels.

Polyclonal antibodies to RoaM were generated commercially (GenScript) by immunizing two New Zealand White rabbits with a three-booster series of 0.2 mg of a KLH-conjugated synthetic peptide derived from RoaM (NREILTHEEIFKC) predicted to be antigenic by a proprietary algorithm (GenScript). A C-terminal cysteine was added to aid in conjugation to the KLH. Sera were collected and antigen affinity purified.

R. rickettsii RNA isolation and RT-qPCR.

Vero cells were infected with R. rickettsii at an MOI of 0.1 and incubated for 48 h at 34°C. Medium was poured off, and cells were scraped into 1 mL PBS and centrifuged for 5 min at 3,000 × g in a microcentrifuge tube. The supernatant was removed, and the cell pellet was dissolved in 1 mL TRIzol LS reagent (Invitrogen, 10296028) preheated to 99°C. RNA was extracted by using Direct-zol RNA Miniprep Plus (Zymo Research, R2070) following the manufacturer’s protocols, including the on-column DNase I treatment. Each RNA sample was treated with 2 U of RNase-free DNase I (New England Biolabs, M0303) for 15 min at 37°C and then further purified using the RNA Clean & Concentrator-5 kit (Zymo Research, R1013) as per the manufacturer’s protocol, including the on-column DNase I treatment. Purified RNA was assessed for quality with a NanoDrop 2000 spectrophotometer (Thermo Fisher). cDNA was generated using the SuperScript IV first-strand synthesis system (Invitrogen, 18091050) with 200 ng of RNA per sample, amplified with random hexamers, and RNase H was added as per the manufacturer’s protocol. Controls run in the absence of reverse transcriptase were generated simultaneously. For qPCR, 20-μL reaction mixtures were assembled using SYBR green PCR master mix (Applied Biosystems, 4309155) with a 25 nM concentration of each primer (Table S3) and 2 μL cDNA. The program used was as follows: 10 min incubation at 95°C, followed by 40 cycles of 95°C for 15 s, 56°C for 30 s, and 60°C for 30 s. Amplification and melting curves for qPCR were performed on a Roche LightCycler 480, and cycle threshold (C) values were calculated with the LightCycler 480 software using the Absolute Quantification 2nd Derivative Maximum function. Genomic DNA (gDNA) positive controls and negative controls, including reactions in the absence of reverse transcriptase, and controls without cDNA were performed, and appropriate criteria for controls were met. C values were used to calculate ΔΔC fold effect relative to the control genes gltA and rpoD (69).

Determination of transcriptional start sites.

To determine the transcriptional start sites of roaM and sca5/ompB, the capping-RACE method was utilized (28, 70). Briefly, RNA was isolated from infected cells as above, and a 5′ cap structure was added to approximately 5 μg of RNA using vaccinia capping enzyme (NEB, M2080S) per the manufacturer’s instructions. RNA was reisolated from this reaction mixture using an RNA Clean and Concentrator-5 kit (Zymo Research, R1013). Approximately 100 ng of capped RNA was incubated with 2 pmol gene-specific primer for either roaM or ompB at 65°C for 5 min and chilled on ice for 2 min. Then, 0.5 mM concentrations of deoxynucleoside triphosphates (dNTPs), 5 mM dithiothreitol (DTT), 100 U SuperScript IV, and 1× SuperScript IV buffer were added, and the reaction mixture was incubated at 50°C for 60 min. Template switching oligomer (TSO; 1 μM; ANP0451), an additional 100 U SuperScript IV, 1× SuperScript IV buffer, and 5 mM DTT were added. Next, 2 mM MnCl2, 2 mM MgCl2, and 0.1% BSA (NEB, B9000S) were included to increase the efficiency of the addition of cytosine residues by the reverse transcriptase upon interaction with the cap structure (71), and the reaction mixture was incubated at 42°C for 90 min. Two units of E. coli RNase H was added, and the reaction mixture was incubated for 20 min at 37°C to remove RNA. The reaction mixture was then used for nested PCR with Taq DNA polymerase (NEB, M0273) using primers ANP0452 and ANP0453 with gene specific primers. PCR products were then cloned into pGEM-T Easy (Promega, A1360) for sequencing.

Serial passage experiment.

Plaque-purified R. rickettsii SSS was used to infect Vero 76 cells in three T25 flasks to initiate three different lineages. Each flask was harvested on Tuesdays and Fridays (alternating 3- and 4-day incubations) by scraping the host cell monolayer into 1 mL of BHI and disrupted for 10 s with glass beads. One hundred microliters of supernatant was used to inoculate the next T25 flask of Vero 76 cells, and the remainder was frozen. Even-numbered passages were plaqued at several dilutions and stained with MTT after 10 days of incubation at 34°C. At passage 10, a single large plaque from each lineage was plaque cloned twice and used for DNA isolation. The roaM region was amplified (primers ANP0212 and ANP0276) and Sanger sequenced.

Guinea pig fever curves.

Female Hartley guinea pigs (300 to 500 g; 5 to 8 weeks of age) (Charles River Laboratories) were obtained and housed in compliance with a protocol (2017-064.2) approved by the Rocky Mountain Laboratories Animal Care and Use committee. To monitor temperatures, transponders (Bio Medic Data Systems, Inc., Seaford, DE) were implanted subcutaneously. One hundred PFU was delivered via intradermal injection, and temperature was recorded daily for 16 days postinfection. At 28 days postinfection, sera were collected for antibody titration and animals were euthanized.