Neisseria gonorrhoeae MS11 and derivative strains were grown on GCB agar plates with Kellogg’s supplements or in GCBL medium with 0.042% sodium bicarbonate and Kellogg’s supplements (“cGCBL”; Kellogg et al., 1963; Morse and Bartenstein, 1974) at 37°C.

Plasmids for this study (Table 1) were generated by PCR amplification of N. gonorrhoeae MS11 chromosomal DNA with primers listed in Table 2, followed by restriction digest with listed enzymes (Table 2). Purified, digested inserts and vectors were ligated overnight with T4 DNA ligase. Ligations were transformed in TAM1 E. coli (Active Motif).

To construct pAKK128 and pAKK129, primers iga-end-out and lacZ937-R were used to PCR-amplify the parAB promoter region and ~1 kb of the 5′ region of lacZ from MMC545 (wild-type SLs) and MMC546 (SL_ΔBC) chromosomal DNA. The PCR products were digested with ClaI (upstream of the promoter region) and XhoI (within lacZ), resulting in ~0.9-kb fragments that contained the parAB promoter with either the wild-type stem-loops or mutant stem-loops fused to the first 839 bp of lacZ. pMR115 + 1, which contains the full lacZ gene fused to a different promoter, was digested with ClaI and XhoI. The PCR products were ligated into the digested plasmid and transformed into E. coli TAM1, generating pAKK128 (wild-type SLs-lacZ) and pAKK129 (SL_ΔBC -lacZ). The constructs were confirmed by sequencing.

Plasmid pMMC25 was made using site-directed mutagenesis to alter the −10 promoter element of NcngR_093 from TACGCT to GACGGA: two fragments were amplified from the MS11 chromosome using primers (1) nc093_sdmF1 + nc093R1 and (2) nc093F1 + nc093_sdmR1. Base pair changes are shown in bold (Table 2). Fragments were purified and then used in equal parts as the template for a secondary PCR with primers nc093F1 + nc093R1. pIDN1 vector and purified secondary PCR product were digested with SacI/XhoI and ligated together with T4 DNA ligase before transformation into TAM1 E. coli.

Gonococcal strains were generated by spot transformation on GCB agar plates (Callaghan and Dillard, 2019). All strains are derived from MS11. Table 3 specifies transformations for this study as (transforming DNA) x (parent strain).

Gonococcal transformations with pMMC38 were re-streaked for screening on GCB agar plates containing 2 μg/ml chloramphenicol (Cm2). The fastest-growing colonies from Cm2 plates were re-streaked to GCB + Cm10 plates, from which single colonies were isolated for PCR screening and sequence confirmation.

Synthetic DNA gene blocks were used to transform GC directly by spot transformation and introduce new constructs by homologous recombination at the iga/trpB complementation locus. Gonococci transformed by gene blocks were re-streaked onto GCB + 40 μg/ml X-gal agar plates. For MMC545 and MMC564, white colonies were chosen for PCR screening and sequence confirmation. For MMC546, blue colonies were chosen.

Construction of BACTH constructs is described in the “BACTH assays” section, below.

RNA isolation and qRT-PCR were performed as described in (Ramsey et al., 2015), using SYBR green reagents. When comparing MS11 and KH655, RNA isolation was performed using TRIzol and the Zymo Direct-zol RNA Miniprep kit. DNase and cDNA preparation were unchanged. Quantitation was achieved by the ΔΔC_T method or with standard curves from MS11 genomic DNA, and Student’s t tests determined significance following previous studies (Applied Biosystems, 1997; Yuan et al., 2006). Primers are listed in Table 2.

Western blots were performed on PVDF membranes against the FLAG epitope, with the exception of Supplementary Figure S4 (described below). After protein transfer, membranes were blocked with 5% milk in 1X TBS + 0.1% Tween 20 (TBST). M2 Mouse α-FLAG primary antibody (Sigma Aldrich) was used at a concentration of 1:20,000 in TBST. Goat α-mouse secondary antibody (Santa Cruz Biotechnology) was also diluted 1:20,000 for use. Samples containing 6 μg protein were loaded per lane unless otherwise noted. Protein amounts were determined using the Bradford assay (Bio-Rad). All blots were visualized using the LI-COR Odyssey® Fc imaging system. For subcellular fractionation samples, α-CAT (Sigma) was used at 1:14,000 and α-SecY (Genscript) at 1:5,000. Horseradish peroxidase-conjugated secondary antibody mouse α-rabbit (Santa Cruz Biotechnology) was used at 1:20,000 dilution.

The western blot for the subcellular fractionation experiment shown in Supplementary Figure S4 used 4 μg protein per lane, and was transferred onto a nitrocellulose membrane. Blocking and primary antibodies were performed as above. LtgA was detected using 1:5,000 mouse monoclonal α-LtgA (final concentration ~0.17 μg/ml) primary antibody. 800CW goat α-mouse secondary antibody was used to detect ParA-FLAG3, ParB-FLAG3, and LtgA, 680RD goat α-rabbit secondary antibody was used to detect SecY and CAT.

A non-piliated variant of N. gonorrhoeae strain MMC545 was grown from freezer stocks on GCB plates overnight. Colonies were swabbed into cGCBL to start 3 ml cultures at OD₅₄₀ = 0.25, and cultures were grown to mid-log phase (2 h). Cultures were diluted back to OD₅₄₀ = 0.3 with cGCBL and aliquoted into the Biolog Phenotype Microarrays (PMs) with pipetting to resuspend the desiccated compounds of interest. We tested PMs 5, 8, 9, 10, 12, 13, 15, 16 (Biolog, #12141, 12,183, 12,161, 12,212, 12,213, 12,215, 12,216, respectively). We performed in vivo β-galactosidase assays by incubating these plates in the Biotek Synergy HT plate reader for 12 h at 37°C with agitation. OD₄₉₂, OD₅₄₀, and OD₆₆₀ reads were taken every 30 min. According to Tang et al., normalized β-galactosidase activity was calculated as O D 630 i n d i g o O D 492 c e l l d e n s i t y = a × O D 492 − O D 630 b × O D 630 − O D 492 , where a = 0.762, the correction factor for cell density and b = 0.267, the correction factor for indigo. To calculate the correction factor a, OD₄₉₂ and OD₆₃₀ were measured for non-piliated MMC545 gonococcal cultures during 16.5 h growth in a blank Biolog plate (six wells, n = 204 data points). Plotting OD₆₃₀ as a function of OD₄₉₂ yielded a linear relationship with R = 0.977, and the slope of the linear line of best fit is the correction factor a (Tang et al., 2013).

GCB agar plates of piliated MMC545 were grown 16–20 h at 37°C, 5% CO₂, then swabbed into 2–4 ml cGCBL. Dilutions of 10⁻⁴–10⁻⁵ (80 μl) were spread plated on GCB + 40 μg ml⁻¹ X-gal plates. Atop the spread culture, a 0.25-inch disk (Hardy Diagnostics) was placed and saturated with 10 μl of the compound of interest. Colony color was assessed after 36–48 h of incubation at 37°C with 5% CO₂ and colony color was visually assessed 36–48 h later.

GGI genes were PCR amplified from MS11 chromosomal DNA using primers specified in Table 2. PCR products and vectors were restriction enzyme digested (specified in Table 2, “Enzyme” column) and ligated. BACTH vectors are specified in Table 1. Final plasmids were confirmed by DNA sequencing. Plasmids of interest were co-transformed into E. coli BTH101 and plated on LB agar plates with 0.5 mM IPTG, 40 μg/ml Xgal, and appropriate antibiotic selection (Table 1, antibiotic selection needed for both co-transformed plasmids). Plates were incubated 40–48 h at 30°C before assessing blue colony color. Antibiotics were used at the following concentrations: 100 μg/ml ampicillin, 50 μg/ml kanamycin. For β-galactosidase assays using BACTH constructs, cells were grown overnight at 30°C in LB with appropriate antibiotics and 0.5 mM IPTG and β-galactosidase activities were measured as described by Miller (1972).

Neisseria gonorrhoeae assays were performed according to Ramsey et al. (2015). Briefly, N. gonorrhoeae was grown overnight on GCB plates and swabbed into cGCBL at an OD₅₄₀ ~ 0.25. After 3 h of aerated growth by rotation, 0.5 ml samples were collected for protein quantification. Cultures were chilled for 20 min on ice, then cells were collected from 2 ml samples by centrifugation, resuspended in 400 μl Z buffer (Miller, 1972) containing 0.002% SDS (Ramsey et al., 2015), aliquoted into 96 well plates, and exposed to ONPG at final concentration of 0.92 mg/ml. Protein concentration was assessed by Bradford assay and substituted for optical density to calculate the output in Miller units (Miller, 1972). Absorbance measurements were taken using a BioTek Synergy HT plate reader. For E. coli carrying pAKK128 or pAKK129, overnight cultures were diluted to an OD₆₀₀ of 0.25 in LB with 25 μg/ml chloramphenicol and grown at 37°C for 3 h with rotation. The OD₆₀₀ of the cultures was measured. The cultures were placed on ice for 20 min, and then 1 ml was pelleted and the cells resuspended in 1 ml of Z buffer. A 10 μl volume of the cell suspension was mixed with 990 μl of Z buffer, then 40 μl of chloroform was added, and the samples were vortexed. Samples were incubated at 28°C for 5 min. Three 100 μl aliquots of each sample were placed in a flat bottom 96-well plate. A volume of 30 μl of ONPG (4 mg/ml) was added to each well, and the OD₄₂₀ and OD₅₅₀ were measured every 5 min. β-gal units were calculated using the Miller equation.

Isolation of soluble and total membrane fractions was performed according to Ramsey et al. (2014) with the following modifications: at least four 3 ml cultures of each strain were grown for each fractionation experiment. Washed cell pellets were resuspended in 0.5 ml 0.01 M Tris–HCl (pH 7.0) before sonication. Samples were sonicated for a total of 50-, 10-s intervals with ≥30 s on ice between each pulse. Ultracentrifugation was performed at 65,000 rpm in a Beckman TLA110 rotor for 1.5 h.

Outer membrane samples were also prepared as described in (Ramsey et al., 2014), although for this study cells were harvested from six gonococcal cultures, 4 ml each, in cGCBL grown from OD₅₄₀ = 0.25 for 3 h. Cells were collected by centrifugation at 10,000 rpm for 10 min at 4°C and washed once with cold PBS before proceeding.

Investigations of the expression of the gonococcal T4SS have begun to reveal a complex regulatory network, with transcriptional, translational, and post-translational mechanisms all at play (Pachulec et al., 2014; Ramsey et al., 2014, 2015; Callaghan et al., 2021). Quantitative transcript data indicate that for both the traH operon (traH, traG, and atlA) and the parA operon (containing parA and parB), transcripts are readily detected in vitro. However, proteins encoded on the traH operon are difficult to detect and attempts to visualize ParA and ParB expression have yet to be reported (Ramsey et al., 2015). The expression of TraH and TraG was shown to be controlled by an RNA switch, and we report here that parA uses a similar switch. We discovered a putative pair of stem-loops in the 5′ UTR of parA, by manual curation of intergenic GGI regions. The stem-loop proximal to the promoter (“SL2”) occludes the translational start site (TSS) and Shine-Dalgarno sequence of the parA operon mRNA (Figures 1A,B). We were unable to identify an energetically favorable alternate secondary structure that releases any part of the ribosome binding site (RBS) in the parAB 5′UTR secondary structure.

To determine the necessity of the stem-loop structure for regulation, we deleted the inner portion of the stem-loop sequence. By removing the inside “leg” of each stem (legs B and C, creating “SL_ΔBC”) the formation of the secondary structure becomes much less favorable, increasing the Gibbs free energy (ΔG) of the structure from −25.3 to −4.68 kcal/mol (Figure 1B). This deletion also removes the bases that pair with the TSS “AUG” in the wild-type structure, leaving it more easily accessible. We cloned the wild-type and SL_ΔBC 5′UTR _parA constructs into E. coli plasmids, making translational fusions with a lacZ reporter. The fusions were made such that the lacZ start codon and subsequent coding sequence replaced those of parA. The wild-type 5′UTR resulted in low levels of LacZ activity, whereas the SL_ΔBC mutant gave approximately 10-fold increased levels (Figure 1C). These data indicate that the stem-loop structures are functional in gene regulation and can perform such regulation in the absence of gonococcal-specific factors.

We introduced the SL_ΔBC mutation into N. gonorrhoeae and examined effects on transcription and translation. We measured relative transcript abundance using qRT-PCR to test for transcriptional effects, looking for direct effects on parA or possible indirect effects on other T4SS genes. The SL_ΔBC deletion did not significantly alter transcript levels for any of the four tested genes, one gene from each of the four GGI operons necessary for secretion (operon 1: traD, operon 2: traK, operon 3: traH, terminal operon: parA; Figure 2A). This result suggests that the secondary structure is not a determinant of transcription activity nor mRNA stability for parA, nor does it affect transcript levels for genes in other T4SS operons.

Since SL2 would occlude the ribosome binding site and start codon of the parA mRNA, we next asked whether the stem-loops control protein expression. To detect the partitioning proteins by western blot, we added a FLAG3 epitope tag (three repeat copies of the FLAG epitope tag in tandem) to the C-terminus of either ParA or ParB by making genetic changes at the native loci. The epitope-tagged constructs were introduced into both wild-type gonococci (MS11) and the stem-loop deletion strain. In the wild-type background, ParA-FLAG3 was undetectable by western blot, and ParB-FLAG3 was very faintly visible. However, in the stem-loop mutant strains, expression of both proteins was greatly increased and easily visualized via western blotting against the FLAG epitope (Figure 2B). The control of ParB translation by a switch regulating ParA expression is possible because the start codon of parB overlaps the stop codon of parA, making it likely that parA and parB are translationally coupled like many of the gonococcal T4SS genes (Hamilton et al., 2005). We conclude that the stem-loops in the parA 5′UTR control protein expression from the parA-parB mRNA, revealing a putative riboswitch mechanism of control.

For screening and semi-quantitative assessment of protein expression in N. gonorrhoeae, we introduced stem-loop – LacZ reporter constructs into the iga/trpB complementation locus on the gonococcal chromosome. We fused the lacZ gene to either the wild-type stem-loops (MMC545) or the stem-loop deletion sequence SL_ΔBC (MMC546) such that the lacZ translational start site is the native parA start site, normally occluded by the wild-type stem-loop structure. This construct was placed under the control of the opaB promoter, which is constitutively active in gonococci. β-galactosidase assays with these strains confirmed that the wild-type stem-loops expressed little LacZ protein whereas the stem-loop deletion mutant allows ample LacZ expression, increasing LacZ expression approximately 400-fold (Figure 3A).

Since no alternate structure for the 5′-UTR was identified, and the translation start site for ParA lies entirely on SL2 leg D, we questioned whether SL1 was playing a role in stem-loop-mediated regulation. To probe the utility of SL 1 in this system, we created a LacZ reporter strain with five base pair changes in SL1 leg A (SL_Amut5), predicted to make folding of stem-loop 1 less favorable (ΔG_SL1-WT = −8.3 kcal/mol, ΔG_SL1-Amut5 = −2.8 kcal/mol; Figure 3B). Surprisingly, these mutations abolished β-galactosidase activity to undetectable levels, below wild-type levels (Figure 3A), indicating a role for SL1 in structure and/or stability of the RNA secondary structure.

Sequence predictions of the mRNA containing SL_Amut5 indicate a propensity for SL2 to elongate in the absence of strong SL1 folding (Figure 3C). At its native locus, a portion of SL1 leg B is able to pair with the beginning of the parA gene, creating six new base pairings and extending SL2. In the lacZ reporter constructs, SL2 is also predicted to elongate by pairing bases of SL1 leg B with the beginning of the lacZ, forming seven new base pairings in a slightly different configuration (Figure 3C). The predicted secondary structures of SL2-parA and SL2-lacZ are very similar, with ΔG = −18.5 and −19.1, respectively. The elongated SL2 structure has a more favorable free energy of folding (predicted ΔG_SL2 decreases by 4.7 kcal/mol in the lacZ constructs, 2.8 kcal/mol in the parA constructs when it adopts the elongated conformation), which could explain the decreased protein output from the SL_Amut5 construct.

Thus, it seems plausible that SL1 contributes to the formation of the wild-type SL2, and prevents the extension of SL2 into a longer and more stable structure. The wild-type stem-loop structure allows for a limited amount of protein expression – far lower than we observe in the complete disruption of these structures, but still detectable by β-galactosidase assay (Figure 3A). However, the formation of a structure with an even tighter occlusion of the ribosome binding site, as we observe in the absence of proper SL1 folding, introduces the possibility that binding of an unknown element of SL1 leg A could be a mechanism to completely abolish protein expression of ParAB. These stem-loop mutation results suggest a protein regulation system that can be finely tuned, both increasing and decreasing translation as the cell responds to environmental stimuli.

If the 5′UTR sequence is a switch, what are its biologically relevant activators? We saw two potential avenues for RNA switch activation. Firstly, ligand binding could induce conformational changes that make the RBS more accessible to the ribosome. Alternatively, but not exclusively, an sRNA could interact with the stem-loops to alter their structure and allow translation initiation.

The specific roles or mechanisms of partitioning activity have not been extensively explored in the gonococcal T4SS. Although we have built hypotheses around findings in other systems, there is ample variation in how these proteins function (Atmakuri et al., 2007; Lutkenhaus, 2012; Gruber et al., 2016). We decided to begin characterizing these proteins by looking at sequence homology, interaction partners, and localization.

Interestingly, the ParA and ParB encoded on the GGI may not be cognate partners. ParA contains a conserved domain from the P-loop NTPase superfamily of proteins, which are found abundantly in protein and DNA localization roles (Pfam accession cl38936; Marchler-Bauer et al., 2017; El-Gebali et al., 2019; Supplementary Figure S3). ParAB cognate pairs are canonically found adjacently encoded, which is indeed the case for the GGI (Bignell and Thomas, 2001; Hamilton et al., 2005; Pachulec et al., 2014; Figure 1A). On the other hand, the gonococcal ParB contains a conserved domain from the ParB family protein of the Pseudomonas fluorescens Pf-5 genetic island-1 (PFGI_1) class of integrating conjugative elements (Pfam superfamily cl26723, family TIGR03764; Marchler-Bauer et al., 2017; El-Gebali et al., 2019; Supplementary Figure S3). The founding members of this protein family are not encoded in immediate proximity to a ParA partner (Klockgether et al., 2004; Paulsen et al., 2005), and of the 41 protein architectures in the CDART database, only five have an adjacent P-loop NTPase (Pfam cl38936) domain (Geer et al., 2002). Consistent with this finding, neither nucleotide alignment search nor translated nucleotide alignment searches using the basic local alignment search tool (BLAST) identified homology of the N. gonorrhoeae parAB gene region to sequence from any organism outside of the Neisseriaceae family in which both parA and parB homologues were present, although parA and parB are individually homologous to many genes within their respective families (Altschul et al., 1990). Thus, while each gonococcal protein is likely to fit the role for one half of a partitioning protein pair, it is unclear whether these two proteins work together as a cognate pair, nor whether the GGI-encoded parA and parB were evolutionarily acquired as a unit.

We used a Bacterial Two-Hybrid (BACTH) system to test for direct interactions between ParA and ParB with the other predicted cytoplasmic and transmembrane proteins of the gonococcal T4SS. This system uses two fragments, T18 and T25, of the catalytic domain of Bortedella pertussis adenylate cyclase, fused to the N- or C-terminal end of two proteins of interest. If an interaction between the proteins of interest brings T18 and T25 into sufficient proximity, functional complementation results in cAMP synthesis inducing transcriptional activation of the lactose operon (Karimova et al., 2001; Battesti and Bouveret, 2012). Using this system functional complementation can therefore be detected on agar plates with X-gal or by β-galactosidase assay. We tagged ParA and ParB with either T18 or T25 at both the N- and C-termini. These constructs were tested for interactions with the N- and C-termini of the other cytoplasmic and transmembrane gonococcal T4SS proteins: TraI, TraC, TraB, TraD, TraE, TraG, and TraL. Transmembrane proteins were tagged at the N- or C-terminus based on predicted topology, such that the tag will be cytosolic (Koch et al., 2020). This large screen identified only two definite interactions for each ParA and ParB: each protein gave a positive interaction result with itself and with TraI, the T4SS relaxase. Only one combination of fusion proteins indicated an interaction between ParA and ParB directly: ParA-T25 interacted with T18-ParB, but none of the other combinations gave a positive result (Figures 4A–C).

The plasmid partitioning proteins of F-plasmid, SopA and SopB, constitute a Walker-type ATPase (SopA) and DNA-binding partner (SopB; Watanabe et al., 1992; Schumacher and Funnell, 2005). We used the BACTH system to test for interactions between F-plasmid SopAB and TraI, looking to gain information on where the gonococcal system parallels or differs from better-characterized T4SSs. Additionally, we used this system to ask whether our gonococcal proteins of interest were able to interact with their counterparts in the F-plasmid system.

We created both N- and C-terminal fusions of SopA, SopB, and TraI from F-plasmid with the T18 and T25 fragments and tested them for interactions amongst themselves and with elements of the putative gonococcal relaxosome, as well as the cytoplasmic ATPase TraC (a homologue of VirB4, the most conserved element across T4SSs; Alvarez-Martinez and Christie, 2009; Guglielmini et al., 2013; Koch et al., 2020). For clarity, F-plasmid proteins will be specified by “F” (e.g., TraI_F) and Neisseria proteins by “N” (e.g., TraI_N) for these constructs. Apart from the expected dimerizations for the SopA_F and SopB_F proteins and the expected SopA_F/SopB_F interaction (Bartosik et al., 2014), we observed a weak TraI_F dimerization and weak SopA_F/TraI_F interactions (Figure 4D). For unknown reasons co-expression of a plasmid expressing TraC_F and a plasmid expressing ParB, TraI and in particular ParA homologues led to a decreased cell number in overnight cultures.

Interactions between F-plasmid partners helped confirm the utility of our approach and identified a parallel relaxase-partitioning protein interaction. Several mixed interactions have been reported between proteins of the F-plasmid and gonococcal systems previously, however none testing elements of the putative gonococcal relaxosome (Koch et al., 2020). The following cross-system interactions were observed, however in none of these cases were the proteins seen interacting in all possible N- and C-terminal or T18- and T25-terminal configurations: TraC_F-ParA_N (3 of 8 potential interactions) and TraC_F-ParB_N (4 of 8 potential interactions), SopA_F-ParB_N (2 of 4 potential interactions), SopB_F-TraC_N (2 of 4 potential interactions) and TraI_F-TraC_N (2 of 4 potential interactions; Figures 4E,F).

To better understand where the partitioning proteins act to facilitate DNA secretion, subcellular fractionation of FLAG3-tagged ParA and ParB was used to separate soluble from membrane-associated proteins. Strains used for the fractionation studies had the stem-loop deletion in the native site parA 5′UTR to overexpress ParAB and allow visualization on western blots. These strains also had the chloramphenicol acetyltransferase gene cat expressed at the aspC/lctP complementation site, to be used as a cytosolic protein control (Ramsey et al., 2014). Based on sequence predictions, we expected both proteins to be entirely cytosolic (Bernsel et al., 2009). However, western blotting against the FLAG epitope revealed that ParA fractionated exclusively with the membrane fraction of culture lysates. Furthermore, ParB is present in both the soluble and membrane fractions, with the membrane fraction having greater ParB signal than the soluble fraction (Figure 5). Isolation of outer membrane proteins from the total membrane fraction revealed no ParA or ParB in the outer membrane, indicating that both proteins associate with the inner membrane (Supplementary Figure S4).