RcsD, an inner membrane protein, has an HPt domain at its C-terminus (Takeda et al., 2001; Wall et al., 2018). The rcsD gene in Y. pestis has undergone a frameshift after codon 642 due to a single nucleotide deletion (Figure 1A). However, rcsD_pe is still functional as deletion of the N-terminal region (1–1846 bp, ΔrcsD_N-term) of rcsD_pe reduced biofilm formation and Congo red (CR) pigmentation (Figure 1B and C; Sun et al., 2008), while replacement of rcsD_pe with rcsD_pstb, and rcsB deletion significantly increased biofilm formation (Figure 1B; Sun et al., 2008). To further characterize the differences between rcsD_pe and rcsD_pstb, biofilm formation was determined in Y. pestis wild type and rcsD N-terminal deletion strains carrying plasmid-encoded copies of each gene, respectively (Figure 1B and C). Consistent with our previous report (Sun et al., 2008), overexpression of rcsD_pstb significantly increased CR pigmentation and biofilm formation in vitro (Figure 1B and C). By contrast, overexpression of rcsD_pe decreased the CR phenotype and formation of biofilm (Figure 1B and C), indicating that rcsD_pe might play a different role to rcsD_pstb in the modulation of Rcs signalling.

Next, we constructed two mutants, one lacking the entire rcsD_pe gene, and the other lacking a C-terminal fragment (ΔrcsD and rcsD_C-term) (Figure 1D). In contrast to the rcsD N-terminal deletion mutant (ΔrcsD_N-term), these two strains showed comparable phenotypes with increased CR adsorption and in vitro biofilm formation (Figure 1E and F). The opposing effects on biofilm formation in these mutants suggest that the C-terminal region may contribute to the inhibition of biofilm formation. To test this, we expressed rcsD_C-term, rcsD_N-term, and rcsD_pe in a rcsD deletion mutant. Overexpression of rcsD_C-term and rcsD_pe but not rcsD_N-term greatly decreased CR pigmentation and biofilm formation (Figure 1E and F). Collectively, these results demonstrate that RcsD_pstb positively regulates, while the frameshifted rcsD_pe negatively regulates, biofilm formation, and that this activity depends on the C-terminus.

The C-terminal HPt domain of rcsD is phosphorylated by RcsC and subsequently transfers a phosphate group to the response regulator RcsB in vitro (Takeda et al., 2001). Given that frameshifted rcsD_pe is functional, and expression of the C-terminus of rcsD is sufficient to repress biofilm formation, we hypothesized that a small protein containing the HPt domain is expressed by rcsD_pe. To identify the putative open reading frame (ORF) of this latter protein, we analysed candidate translational initiation sites in the C-terminus of rcsD_pe by searching for AUG, GUG, and UUG codons, which account for approximately 80%, 12%, and 8% of start codons in bacterial genomes (Srivastava et al., 2016). We found two putative ORFs (encoding genes composing 462 and 573 nucleotides, respectively, Figure 2—figure supplement 1A) with UUG start codons, denoted as UUG^–462 and UUG^–573. Ectopic expression of these two putative ORFs in Y. pestis KIM6+ showed comparable biofilm formation to expression of full-length rcsD_pe (Figure 2—figure supplement 1B). Mutation of these two start codons alone (UUG^–462→UUA, UUG^–573→CUU) or together did not alter the function of rcsD_pe (Figure 2—figure supplement 1C), indicating they are not the start sites promoting expression of the functional protein.

AUU, CUG, and AUC are occasionally used as start codons in bacteria (Cao and Slavoff, 2020; D’Lima et al., 2017). We identified a putative ORF, encoding a 103-residue protein, initiated by an AUU codon, which also had a predicted RBS binding site (Figure 2A). Mutation of this codon or the predicted RBS region in rcsD_pe abolished its function, while replacement of AUU with a strong AUG start codon significantly enhanced its function (Figure 2B and C). Furthermore, ectopic expression of the 103-residue HPt domain with an AUU start codon, but not with a GGU codon, using a modified pBAD vector in the Y. pestis KIM6+ wild type strain, or rcsD deletion mutant, strongly repressed biofilm formation and CR pigmentation (Figure 2—figure supplement 1D), indicating that AUU is a functional start codon initiating translation of RcsD-Hpt.

Next, we introduced 3xFlag and 6xHis tags to rcsD_pe and rcsD_pstb and analysed their expression by western blotting. A band in the tagged RcsD_pstb strain was observed at the expected size (Figure 2D, Lane3), but no visible band was detected in the tagged rcsD_pe strain (Figure 2D, Lane2). To enhance detection sensitivity, the rcsD_pe lysates were enriched before western blotting. An approximately 15 kDa band, corresponding to the predicted protein containing the HPt domain, was detected in both rcsD_pe, rcsD_C-term, and rcsD_pstb expressing strains (Figure 2D and Figure 2—figure supplement 1E, see Figure 2—source data 1, Figure 2—figure supplement 1—source data 1 for details), but not detected in the putative start codon mutated rcsD_pe expressing strain (Figure 2—figure supplement 1E, see Figure 2—figure supplement 1—source data 1 for details). In addition, full-length RcsD was detected in rcsD_pe but not in rcsD_C-term and rcsD_pe-stop (a stop codon was introduced upstream of the frameshift site) expressing strains (Figure 2D and Figure 2—figure supplement 1F, see Figure 2—source data 1, Figure 2—figure supplement 1—source data 2 for details), indicating that intact RcsD is weakly expressed in wild type Y. pestis through translational recoding (Rodnina et al., 2020). This is in accordance with the translational lacZ reporter system, which detected 1% readthrough once frameshifted (Figure 2—figure supplement 1G). Consistent with these results, Y. pestis with rcsD_pe-stop showed similar biofilm and CR phenotype as the rcsD_N-term deletion strain (Figure 2C and data not shown). Taken together, these data suggest that a 103-amino acid protein, designated as RcsD-Hpt, is expressed by rcsD_pe and functions as a negative regulator of biofilm formation in Y. pestis.

RcsD is a phosphorelay protein that can transfer phosphate from the conserved His residue in its HPt domain to a conserved Asp in the receiver domain of RcsB and can also dephosphorylate this site (Ancona et al., 2015; Takeda et al., 2001). An H844A mutation in the Y. pestis wild type or rcsD_N-term deletion strain showed a comparable phenotype to a full rcsD deletion strain (Figure 2B and C and Figure 1E), while plasmid-borne expression of the mutated version of rcsD_pe or rcsD-hpt in Y. pestis wild type and its rcsD mutants displayed a similar phenotype to an empty vector control (Figure 2—figure supplement 2A). In addition, plasmid-borne expression of the mutated version of rcsD_pstb in the Y. pestis rcsD_N-term deletion strain showed significantly less biofilm than that of rcsD_pstb (Figure 2—figure supplement 2B), indicating H844 might be important for the dephosphorylation ability of RcsD. Next, we modelled the structure of RcsD-Hpt with AlphaFold2 (Fowler and Williamson, 2022; Figure 2E), which revealed a similar structure to HptB, an HPt orphan protein in P. aeruginosa (Figure 2F; Chen et al., 2020). Like HptB, the predicted structure of RcsD-Hpt shows an elongated bundle of four helices α2, α3, α4, and α5, covered by the short N-terminal α1 helix. The imidazole side chain of the conserved active-site histidine residue His50 (His844 in RcsD_pstb) is located near the middle of helix α3 and protrudes from the bundle where it is exposed, as His57 is in HptB. Taken together, RcsD-Hpt may function as a classical HPt orphan protein, and the conserved His residue is crucial for its function.

RcsF and IgaA, which regulate environmental stress sensing (Cho et al., 2014; Guo and Sun, 2017; Wall et al., 2018), transfer signals to RcsD through interaction of IgaA with its periplasmic domain (Wall et al., 2020). Given that RcsD-Hpt does not encode a functional periplasmic domain, we hypothesized that the roles of RcsF and IgaA are dispensable in the Y. pestis Rcs signalling system. We constructed rcsF and igaA deletion mutants in the Y. pestis wild type and rcsD_pstb substitution strains, respectively. In the Y. pestis rcsD_pstb substitution strain, deletion of rcsF increased CR adsorption and biofilm formation (Figure 3A and B), while deletion of igaA completely abolished biofilm formation and CR binding (Figure 3C and D). Furthermore, overexpression of RcsF but not its C125S mutant (as Escherichia coli RcsF cysteine mutants are inactive Rogov et al., 2011), decreased biofilm formation and CR adsorption (Figure 3A and B), while expression of IgaA but not its C413S mutant (the conserved cysteine is essential for the function of IgaA in Salmonella enterica Pucciarelli et al., 2017) complemented the phenotype of the igaA deletion strain (Figure 3C and D). These results indicate that when receiving signals transduced from RcsF and IgaA, RcsD_pstb mediates decreased phosphorylation of RcsB, and thus Rcs signalling is switched off. Alternately, deletion or ectopic expression of rcsF or igaA does not regulate biofilm formation and CR pigmentation in wild type Y. pestis KIM6+ (Figure 3A–D), suggesting a decoupling of the requirement for RcsF and IgaA and the Rcs system after the transition from rcsD_pstb to rcsD_pe.

RcsC, a bifunctional histidine kinase and phosphatase, phosphorylates and dephosphorylates RcsD, which subsequently, via RcsB, activates or represses expression of its target genes (Guo and Sun, 2017; Wall et al., 2018). Deletion of RcsC resulted in increased biofilm formation and CR pigmentation (Sun et al., 2008; Figure 3E), indicating that RcsC is involved in Rcs signalling in Y. pestis. In addition, high expression of RcsD-Hpt in the rcsC deletion mutant did not affect biofilm formation and pigmentation (Figure 3E), indicating that RcsC is required for RcsD-Hpt-mediated biofilm repression in Y. pestis. His489 and Asp885 in RcsC are the predicted autophosphorylation and subsequent transfer receipt sites, respectively (Clarke et al., 2002; Latasa et al., 2012). A D885A mutation in RcsC abolished its function in an rcsD_pstb substitute or wild type strain (Figure 3F–H), indicating that Asp885 was crucial for phosphate transfer in both conditions. An H489A mutation in RcsC did not alter its function in the wild type strain (Figure 3F and H), indicating that His489 is not important for the phosphorylation of RcsD-Hpt. Surprisingly, mutation of His489 resulted in decreased biofilm formation and CR adsorption in the rcsD_pstb substitution strain (Figure 3G and H), indicating that His489 might function as a phosphate reservoir involved in dephosphorylation of RcsD. RcsC T903A constitutively activates Rcs in S. enterica (García-Calderón et al., 2005; Latasa et al., 2012). The corresponding T913A mutation in RcsC strongly decreased biofilm formation and CR pigmentation in the RcsD_pstb substitution strain (Figure 3F–H), indicating that an RcsC T913A mutant stimulates Rcs in a similar manner to S. enterica. The same mutation in RcsC resulted in enhanced biofilm formation in Y. pestis, indicating that an RcsC T913A mutant has an impaired ability to phosphorylate RcsD-Hpt. Taken together, these data indicate that RcsC remains a crucial component of the Rcs phosphorelay system in Y. pestis.

RcsB is a phosphoacceptor in the Rcs system and contains a conserved Asp site in its receiver domain which can be phosphorylated by RcsD (Fredericks et al., 2006; Takeda et al., 2001). Phosphorylated RcsB regulates expression of biofilm-related genes (hmsT, hmsD, hmsP, and hmsHFRS) in Y. pestis (Fang et al., 2015; Guo et al., 2015; Sun et al., 2012). Expression of rcsD_pe or rcsD_pstb only conferred altered biofilm formation in the presence of RcsB (Figure 4A), indicating that RcsD_pe and RcsD_pstb might differentially modulate the phosphorylation of RcsB. To test this hypothesis, we detected the phosphorylation of RcsB using Phos-tag SDS-PAGE gels and western blotting (Madec et al., 2014). As expected, replacement of rcsD_pe by rcsD_pstb in Y. pestis resulted in decreased phosphorylation of RcsB, while mutation of the conserved Asp residue (D56Q) in RcsB abolished phosphorylation (Figure 4B, see Figure 4—source data 1 for details). Consistent with the phosphorylation status of RcsB, the transcription and expression of HmsT were differentially regulated by RcsD_pe and RcsD_pstb (Figure 4C and D, see Figure 4—source data 2 for details).

Loss of function in rcsA is a crucial step for Y. pestis to establish flea-borne transmission (Sun et al., 2008; Sun et al., 2014). We therefore speculated that mutation of rcsD might also play a role in the adaptation of Y. pestis to the flea. We therefore infected the Oriental rat flea, Xenopsylla cheopis, with Y. pestis wild type (KIM6+), the rcsD_pe::rcsD_pstb and the rcsD_N-term deletion strain. Bacterial burdens in infected fleas at 0, 7, and 28 day (s) post infection were not significantly different between any strain combinations (Figure 5A). In addition, we did not observe significant differences in flea blockage, despite different in vitro capacities to form biofilms in these strains (Figure 5B). Taken together, these results suggest that the frameshift mutation in rcsD does not alter the infection, persistence, and blockage-forming capacity of Y. pestis in fleas.

We fortuitously observed that a Y. pestis rcsD_pstb strain, but not the wild type, displayed a pgm- phenotype after 4 weeks of flea infection. The pigmentation phenotype (pgm+) is defined by Y. pestis absorption of exogenous hemin or CR to form pigmented colonies. A spontaneous deletion of the 102 kb pgm locus imposed by the instability of two flanking IS elements results in a pgm- phenotype (Fetherston et al., 1992; Tong et al., 2005). Interestingly, an accumulation in pgm locus mutations was previously reported for strains exhibiting enhanced biofilm formation (Fetherston et al., 1992; Kirillina et al., 2004; Silva-Rohwer et al., 2021). Prior to the infectious blood meal, no colonies exhibited a pgm- phenotype when the wild type and rcsD_pstb strains were plated on CR agar plates. Twenty-eight days post infection, we observed the pgm- phenotype in bacteria isolated from all fleas infected with the rcsD_pe::rcsD_pstb strain. The mean percentage of isolates displaying the pgm- phenotype was 16.3%, ranging from 0.1% to 47.9%, whereas only one out of ten fleas infected with the wild type strain displayed this phenotype with a frequency of ~1.0% (Figure 5C). We further analysed the pgm mutation rate of Y. pestis in liquid medium. To mimic the environment in the flea, Y. pestis were grown in liquid medium for several days, where they remained in stationary phase before reinoculation. Consistent with the flea infection data, Y. pestis encoding rcsD_pstb displayed a significantly higher pgm- mutation rate relative to the wild type strain (Figure 5D). Strains with deletion of rcsD-Hpt or mutated rcsD_pe also exhibited increased pgm mutation rates (Figure 5D), indicating that RcsD-Hpt is important for stability of the pgm locus. This data agrees with previous reports of strains showing increased pgm locus loss concomitant with enhanced biofilm production levels (Fetherston et al., 1992; Kirillina et al., 2004; Silva-Rohwer et al., 2021). PCR analysis confirmed that the pgm- phenotype was caused by the spontaneous deletion of the pgm locus (data not shown). Taken together, these results suggest that the frameshift mutation in rcsD promoted stable maintenance of the pgm locus of Y. pestis KIM6+ in infected fleas.

Rcs has been reported to modulate virulence in other bacteria (Li et al., 2015; Wall et al., 2018; Wang et al., 2012). KIM6+ is an avirulent derivative of the fully virulent KIM strain, which was cured of the pCD1 plasmid. To investigate the role of Rcs on pathogen virulence in a mammalian host, we took advantage of the Y. pestis biovar Microtus strain 201 (Zhang et al., 2014), a human-avirulent but rodent-virulent strain, isolated from a natural reservoir, the Brandt’s vole (Microtus brandti) (Song et al., 2004). Mutation of rcsB or rcsD in the Y. pestis biovar Microtus strain 201 showed a similar CR absorption and in vitro biofilm phenotype as the Y. pestis KIM6+ strain (data not shown). To our surprise, deletion of rcsB or replacement of rcsD_pe by rcsD_pstb in this strain did not significantly affect their virulence when mice were subcutaneously infected with different doses of bacteria (Figure 6A–C).

Rcs has been reported to modulate the expression of many genes in response to environmental stress. To characterize the genes regulated by the Rcs system in Y. pestis, we performed RNA-seq on total RNA isolated from Y. pestis biovar Microtus strain 201, rcsB deletion and rcsD substitution strains cultured at 26°C and 37°C. A total of 139 genes (43 upregulated and 96 downregulated) and 49 genes (18 upregulated and 31 downregulated) were significantly differentially expressed (|logFC|≥1, p≤0.05) in the rcsB deletion and rcsD substitution strain, respectively, when compared with the wild type strain at 26°C (Figure 6D, Figure 6—figure supplement 1A–1B , and Supplementary file 4). At mammalian temperature (37°C), 37 genes (26 upregulated and 11 downregulated) and 177 genes (97 upregulated and 80 downregulated) were significantly differentially expressed (Figure 6D, Figure 6—figure supplement 1C–1D and Supplementary file 4). Several differentially expressed genes (DEGs) identified by RNA-seq were verified by quantitative real-time PCR (qRT-PCR), indicating comparable patterns of expression (Figure 6E and Supplementary file 4). Substitution of rcsD_pstb had an opposing effect on gene expression to deletion of rcsB (Figure 6F and Supplementary file 4), indicating that the frameshift mutation in rcsD lessens the regulatory function of Rcs. Furthermore, Rcs positively regulated genes such as those encoding the type 6 secretion system, and those related to biosynthesis of yersiniabactin and ABC transporter genes (Figure 6F). These observations suggest that Rcs might play an important role in the environmental fitness and virulence of Y. pestis, and thus could be required for the flea-mammalian host transmission cycle in the wild.

To investigate the role of the rcsD mutation in the evolution of Y. pestis, we analysed the evolutionary changes that occurred during the divergence of Y. pestis from Y. pseudotuberculosis (Figure 7 and Supplementary file 5). Mutation of rcsD is present in all Y. pestis that harbour five genetic changes (pde3’, ymt, rcsA, pde2, and ureD) required for flea colonization (Figure 7), except for the 0.PE7 branch (Figure 7). An IS element in the ompC gene is one of two IS elements driving instability of the pgm locus (Figure 7 and Supplementary file 5; Fetherston et al., 1992; Tong et al., 2005). The IS element in ompC was present in most modern Y. pestis branches, but not in RT5, an ancient Y. pestis strain isolated from a bubonic plague patient during the Bronze Age (Spyrou et al., 2018). This indicates that emergence of the rcsD mutation is likely not due to loss of the pgm locus enabled by IS elements. Although multiple rcsD mutations are present across the phylogeny (Figure 7 and Supplementary file 5), the HPt encoding region is present in all sequenced Y. pestis isolates, indicating an important role of RcsD-Hpt in refining stable blockage-mediated flea-borne transmission of Y. pestis.

This study utilized Y. pestis strain KIM6+, which is derived from the sequenced strain KIM strain (Deng et al., 2002), but is cured of the pCD1/pYV plasmid required for mammalian virulence, and is competent for flea blockage (Hinnebusch et al., 1996) and biofilm formation (Darby et al., 2002; Sun et al., 2008). Studies in mice utilized the biovar Microtus strain 201 which was avirulent to human but fully virulent to mice (Zhang et al., 2014).

Deletion of rcsC, rcsF, and igaA was achieved using a one-step method to integrate PCR products into the chromosome with pKD46, as previously described (Datsenko and Wanner, 2000). Double mutants were made by the sequential application of pKD46-mediated deletion. CRISPR-Cas12a-assisted recombineering were used to introduce point mutations, deletions, insertions, and gene replacements in this study (Yan et al., 2017). All strains were verified by PCR, DNA sequencing, and plasmid complementation.

For construction of the hmsT::lacZ reporter, 350 bp of hmsT upstream sequence, together with the first seven codons of the ORF, were amplified by PCR using KIM6+ chromosomal DNA as the template. The DNA fragments were digested with HindIII and BamHI restriction enzymes and cloned into pGD926 (Ditta et al., 1985; Vieille and Elmerich, 1990), generating plasmids pYC593 and pYC287. Plasmids expressing rcsC (D885A), rcsC (T913A), rcsD (H844A), rcsD::rcsD-3xflag, and rcsD-3xflag::rcsD-3xflag-his6 were generated by overlapping PCR as described previously (Li et al., 2015).

For inducible rcsD expression, the gene was cloned downstream of the arabinose-inducible promoter of plasmid pBAD/Myc-His (Invitrogen). The plasmid used for determining readthrough was generated by cloning a partial sequence of rcsD (159 bp for rcsD_pstb and 158 bp for rcsD_pe) containing the 8T (frameshifted region) into pMal-lacZ, generating plasmids pMal-rcsD_pstb-lacZ and pMal-rcsD_pe-lacZ. An additional stop codon was introduced into the 158 bp rcsD_pe sequence of pMal-rcsD_pe-lacZ, generating plasmid pMal-rcsD_pe-stop-lacZ.

All strains and plasmids used in this study are shown in Supplementary file 1 and oligonucleotides used in this study are shown in Supplementary file 3.

Microtiter plate biofilm assays were performed as previously described (Sun et al., 2012). Briefly, bacteria were cultured overnight in LB broth supplemented with 4 mmol CaCl₂ and 4 mmol MgCl₂. Cultures were subsequently diluted into 96-well plates and incubated with shaking for 24 hr at 26°C. The wells were washed, and the adherent biofilm was stained with crystal violet, solubilized with 80% ethanol and 20% acetone, and measured by A₆₀₀. Results are from three independent experiments with three technical replicates per experiment.

β-Galactosidase activities were measured as previously described (Guo et al., 2015; Sun et al., 2012). Briefly, overnight cultures of Y. pestis harbouring lacZ reporters were diluted to an OD₆₀₀ of 0.05 and grown in LB broth at room temperature to an OD₆₀₀ of 1.5. The active β-galactosidase, encoded by the lacZ reporter gene in Y. pestis strains, can cleave o-nitrophenyl-β-D-galactopyranoside (ONPG) substrate to a bright yellow product. The cells were lysed and ONPG solution were added. After incubation at 37°C, the reaction was stopped by adding 1 M Na₂CO₃, then absorbance was measured at 420 nm. Results were normalized against cell density and incubation time, and shown in Miller units (Miller, 1972). At least two independent experiments with technical triplicates were performed.

qRT-PCR was carried out as previously described (Sun et al., 2011). Briefly, cells were first grown in LB broth overnight before diluting to an OD₆₀₀ of 0.05 in LB and incubating at room temperature to an OD₆₀₀ of 0.8. Total RNA was isolated using the Rneasy Mini Kit (Qiagen). Residual DNA was removed by treatment with rDNase I (Ambion) and confirmed by PCR. cDNA was synthesized from the RNA and used for quantitative PCR on an Applied Biosystems unit (Quant Studio 5). The quantity of mRNA was normalized relative to the reference gene 16sRNA (YP_r1). The relative mRNA expression levels in each strain were normalized to the wild type samples. Primers and probe sets used in this study are listed in Supplementary file 3. Results from three independent experiments performed in technical triplicate were analyzed by one-way analysis of variance (ANOVA) with Bonferroni’s test.

Western blotting was performed as previously described (Ren et al., 2017). For detection of enriched RcsD_pe, Y. pestis strains expressing rcsD fused with 3xFlag-His6 were grown at 26°C to stationary phase. Cells were harvested by centrifugation and disrupted by sonication. The protein was enriched by Ni-nitrilotriacetic acid His resin for western blot analysis. For detection of enriched RcsD and RcsD-Hpt in Figure 2—figure supplement 1E and F, Y. pestis strains expressing rcsD fused with 3xFlag were grown at 26°C to stationary phase. Cells were harvested by centrifugation and disrupted by sonication. The protein was enriched by flag resin for western blot analysis. For detection of HmsT, Y. pestis strains harbouring a plasmid expressing HmsT with 3xFlag were grown at 26°C to stationary phase. Cells were harvested by centrifugation and disrupted by sonication. Approximately 20 ng proteins were loaded for detection of HmsT. The proteins were separated on 10% SDS-PAGE gels transferred to PVDF membranes (Millipore), analyzed by immunoblotting with an anti-Flag antibody produced by Invitrogen (Catalog number: MA1-91878-HRP; RRID: AB_2537626), and detected with ECL Western Detection Reagents (Bio-Rad). Resulting bands were quantitated by densitometry using NIH ImageJ (Gallo-Oller et al., 2018).

For detection of protein phosphorylation, acrylamide gel was mixed with 25 μM Phos-tag acrylamide (AAL-107, Wako) and 25 μM MnCl₂ (Madec et al., 2014). Cells grown to an OD₆₀₀ of 0.8 were centrifuged and pellets were resuspended in PBS containing Protease Inhibitor Cocktail (Roche) and then lysed with a sonicator. Samples were quickly loaded onto gels containing 25 µM Phos-tag acrylamide and 25 µM MnCl₂ and run. After a 10 min wash with WB transfer buffer supplied with 1 mM EDTA, followed by a 10 min wash with transfer buffer without EDTA. Proteins were transferred to a PVDF membrane and blotted using an anti-Flag antibody produced by Invitrogen (Catalog number: MA1-91878-HRP; RRID: AB_2537626).

Total RNA was extracted using the RNeasy Kit (Qiagen). RNA-seq and expression quantification were performed by Genewiz. Gene expression levels were further normalized using the fragments per kilobase of transcript per million mapped reads method to eliminate the influence of different gene lengths and sequencing depth (Wang et al., 2009). The edgeR package was used to identify DEGs across samples with fold changes ≥2 and a false discovery rate-adjusted p (p-value)≤0.05 (Anders and Huber, 2010; Robinson et al., 2010). DEGs were then subjected to an enrichment analysis of GO function and KEGG pathways (Harris et al., 2004; Kanehisa and Goto, 2000).

Animals were handled in strict accordance with the Guidelines for the Welfare and Ethics of Laboratory Animals of China and all the animal experiments were approved by the Institutional Animal Care Committee of Military Medical Sciences. Bacterial cultures at 37°C were washed twice with PBS (pH 7.2) and then subjected to serial 10-fold dilutions with PBS. Dilutions were plated onto brain heart infusion (BHI) agar plates to calculate the numbers of colony-forming units (CFU). For each strain, different doses of bacterial suspension were inoculated subcutaneously at the inguinal region of 10 female BALB/c mice (aged 6–8 weeks), which were obtained from Charles River Laboratories (Beijing, China). Survival was monitored at regular intervals, and a survival curve was generated with GraphPad Prism 5.0. p-Values were determined using the log-rank (Mantel–Cox) test and the Gehan–Breslow–Wilcoxon test; p<0.01 was considered statistically significant.

Flea infections and blockage analysis were carried out as previously described (Silva-Rohwer et al., 2021; Sun et al., 2014). Y. pestis strains were grown overnight in 3 mL HIB at 26°C with shaking, then diluted into 100 mL HIB to cultivate at 37°C without shaking. The following day harvested cells were suspended in sterile PBS, and optical density at absorbance of 600 nm was determined. A commercial preparation of heparinized mouse blood (BioIVT, New York) was inoculated with Y. pestis to a final concentration of CFU/mL of ~5 × 10⁸ to 1 × 10⁹. X. cheopis fleas were allowed to feed on the infected blood through a mouse skin membrane. Studies with mice were performed in strict accordance with the U.S. National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals (National Research Council Committee for the Update of the Guide for the Care and Use of Laboratory Animals et al., 2011) and as approved by the Washington State University Institutional Animal Care and Use Committee.

Mice: A CD-1 mouse breeding colony originally sourced from Envigo (https://www.envigo.com/model/hsd-icr-cd-1) is maintained at WSU. Males and females are used. Neonates between the ages of 2–6 days are used for feeding fleas for breeding and maintenance of infected fleas.

Fleas: X. cheopis fleas are maintained at WSU since 2010 in Dr. Vadyvaloo’s lab. These fleas were originally sourced from Dr Joseph Hinnebusch’s lab at the NIH. Males and females were used for experiments.

For in vivo pigmentation phenotype detection, 10 fleas infected with Y. pestis wild type or the rcsD_pe::rcsD_pstb strain were collected at the end of the infection period (T=28 days). Fleas were individually triturated in sterile PBS, and the fractions were serially diluted and cultivated on a CR plate. Two days later, CFU enumeration relevant to the pigmentation phenotype was performed. For in vitro pigmentation assays, fresh Y. pestis KIM6+ strains were inoculated into BHI medium and grown at 26°C for 24 hr. Cultures were diluted 1:100 into fresh BHI medium every 3 days. The cultures at 0, 1, 4, and 7 days post inoculation were plated on CR plates for analysis. The loss of the pgm locus was confirmed by PCR using two sets of primers described in Supplementary file 3. One set of primers targets the hmsS gene, which can obtain PCR products from the wild type strain but not the pgm locus deletion mutant. Another set of primers target the upstream and downstream regions of the pgm locus, and can be used to obtain PCR products from the pgm locus deletion mutant but not the wild type strain.

Figure legends detail the quantification and statistical analyses methods. We conducted the statistical analyses by GraphPad Prism.