Previous characterization of S. mutans ΔyidC1 and ΔyidC2 deletion mutants revealed severe impairment of growth of the ΔyidC2 strain upon exposure to acid, osmotic, and oxidative stress (Hasona et al., 2005; Hasona et al., 2007; Funes et al., 2009). However, the ΔyidC2 mutant reproducibly resolved acid-sensitivity by acquiring a point mutation, W138R, in the TM2 domain of YidC1 (Mishra et al., 2019). Unlike deletion of yidC2, deletion of yidC1 does not manifest in sensitivity to environmental stress except under conditions of zinc-excess. Since, yidC1 ^W138R expression conferred tolerance to acid-stress in a ΔyidC2 background, we hypothesized that this strain might show altered sensitivity to Zn-excess. We compared growth of S. mutans wild-type (WT), ΔyidC1, and ΔyidC2 strains, as well as strains expressing yidC1 ^W138R or yidC1 ^W138A in a double ΔyidC1/yidC2 background, under non-stress, acid stress, and Zn-excess conditions (Figure 2). As expected, deletion of yidC2 resulted in impaired growth and pronounced acid sensitivity. Growth of the ΔyidC2 strain under non-stress conditions was restored by expression of yidC1 ^W138R , and growth at pH 5 was restored by expression of either yidC1 ^W138R or yidC1 ^W138A . Deletion of yidC1 impacted S. mutans growth in THYE containing 2.5 mM ZnCl₂ far more than deletion of yidC2, and zinc tolerance was restored by expression of yidC1 ^W138R but not by yidC1 ^W138A . These results highlight the importance of S. mutans YidC1 residue 138 under several biologically relevant environmental conditions and demonstrate that expression of yidC1 ^W138R phenocopies expression of yidC2 when either is present as the sole functional yidC.

Often, genes that contribute to the same biological pathways will manifest similar phenotypes upon deletion. To test the degree to which a S. mutans Δcls mutant shares known ΔyidC2- or ΔyidC1-associated properties, WT, Δcls, ΔyidC1, and ΔyidC2 strains were compared for their ability to tolerate low pH, high salt, and Zn-excess conditions (Figure 3A). In addition to impaired growth at pH 5, the Δcls mutant demonstrated pronounced sensitivity to 3% NaCl, another ΔyidC2-associated phenotype. The Δcls strain also acquired moderate sensitivity to 2.5 mM ZnCl₂, but resembled ΔyidC2 more than ΔyidC1 in this regard (Figure 3A).

To further evaluate whether YidC2 and CL act in concert to protect against environmental stress, we hypothesized that yidC2 and cls would be transcriptionally upregulated by identical stressors. To test this question, promoter-reporter gene fusions of yidC1, yidC2, and cls were constructed. The genetic context of yidC1 is well conserved among Gram-positive bacteria with yidC1 in an operon with the gene encoding the protein subunit of ribonuclease P (rnpA). In S. mutans, 17 bp of rnpA, including the stop codon, overlaps with yidC1. The start codon of the downstream gene, jag, is located 12 bp from the stop codon of yidC1. Bacterial promoter prediction tools (BPROM) identified two distinct −10 and −35 sites upstream of yidC1 (Solovyev and Salamov, 2011). One set of −10 and −35 boxes is located 256 and 276 bp upstream of yidC1 and within the rnpA gene, while a second set is located 48 and 67 bp upstream of rnpA. No transcription termination sites were identified between rnpA, yidC1, and jag suggesting these genes comprise an operon. Based on these predictions, the 470 bp promoter region upstream of the yidC1 start site was used to construct a yidC1 promoter-lacZ fusion construct. Unlike yidC1, the genomic context of yidC2 is not conserved. In S. mutans, yidC2 is present 91 bp downstream of an acylphosphatase encoding gene, acp, and ∼400 bp upstream of the translation elongation factor, greA, which is oriented opposite to yidC2. The genetic context of yidC2 suggests it is likely transcribed from its own promoter. A−10 (TAAAAT) and a −35 (ATTGAGT) box were predicted 40 and 64 bp upstream of the yidC2 start site, respectively, and a 283 bp segment of the yidC2 promoter was used to make its lacZ promoter-reporter construct. Previous work from the Quivey lab identified the cls transcriptional start site 5’ to the upstream gene, wapA (MacGilvray et al., 2012). A 300 bp segment upstream of wapA was used to make the cls promoter-lacZ reporter construct. All three promoter-reporter gene fusions were integrated in a single copy at the phnA-mtlA2 locus within the S. mutans UA159 chromosome.

To determine the impact of varying environmental stressors on yidC1, yidC2, and cls expression, β-galactosidase activities were measured in response to acid stress (pH 5), osmotic stress (3% NaCl), Zn-excess (2.5 mM ZnCl₂), and cell envelop stress (250 μg mL^-1 bacitracin) conditions and compared with exponential growth-phase expression for each respective promoter (Figure 3B). Acid stress and bacitracin-induced envelope stress led to ∼2-fold induction of yidC2 and cls, while yidC1 promoter activity remained unchanged under all environmental stress conditions tested. Osmotic stress and Zn-excess had no measurable effect on the activity of any of the promoters. These results are consistent with the hypothesis that YidC2 and CL likely function together to enable acid and envelope stress tolerance. PI and PS biosynthetic pathways have not been characterized in S. mutans or other streptococci. Therefore, mutant strains are not available to explore potential phenotypic overlap between ΔyidC1 and ΔyidC2 strains and mutants lacking anionic phospholipids other than CL.

Due to challenges in adequate expression and purification of recombinant YidC1 and YidC2 proteins in sufficient quantity for lipid-protein interaction studies, we used coarse grain molecular dynamics (CGMD) simulations to study the impact of membrane lipids on S. mutans YidC paralogs in our current study. CL, PI, and PS together comprised ∼90% of the total phospholipid content of cytoplasmic membrane samples prepared from S. mutans cells harvested from stationary phase cultures (Morales-Aparicio et al., 2020). CL, PI, and PS are anionic phospholipids with negatively charged polar head groups (Supplementary Figure S3). We modeled CL, PI, and PS at a 3:1:1 molar ratio mimicking the S. mutans lipid bilayer. To start, monomeric YidC1 or YidC2 simulations containing the CL-rich tri-lipid mixture were analyzed to identify potential interactions between monomeric forms of YidC1 and YidC2 with the membrane lipids. A potential monomeric organization of YidC1 and YidC2 was evaluated first based on crystallographic data of various YidCs from B. halodurans (Kumazaki et al., 2014a; Kumazaki et al., 2014c), E. coli (Kumazaki et al., 2014b), and Thermotoga maritima (Xin et al., 2018; Nass et al., 2022). Here, and for simulations described below, single systems were built using the CHARMM-GUI (Jo et al., 2008; Lee et al., 2016) Martini maker (Qi et al., 2015; Hsu et al., 2017) and were energy minimized and equilibrated in the Groningen Machine for Chemical Simulation (GROMACS) (Abraham et al., 2015). After equilibration, systems were cloned into 5 replicates (Knapp et al., 2018) to ensure equivalent starting energy profiles, and production was carried out for a total of 10 μs per simulation. The average RMSD of Cα-atoms provides insight into system stability for a given protein throughout the simulation. All systems that we assessed were observed to quickly reach equilibrium within ∼2 Å Cα RMSD compared to the final snapshot of respective equilibration steps (Supplementary Figure S4). PyLipID was then used to calculate lipid residency times by residue (Song et al., 2022). Lastly, we calculated the RMSF values for each amino acid residue during the course of the simulation of YidC1 and YidC2 in the 3CL:PI:PS mixture to understand environment-dependent protein dynamics (Supplementary Figure S5). RMSF values represent the extent of fluctuation for each residue around their average position. Interestingly, none of the YidC1 or YidC2 residues were significantly occupied (>3 μs) by CL in the simulated CL-rich bilayer (Supplementary Figure S6). Average RMSF values of residues comprising the cytoplasmic loops as well as the TM helices were less than 2 Å in both YidC1 and YidC2 monomers suggesting relatively inflexible structures. Collectively, our results support the notion that monomeric YidC1 and YidC2 do not represent the functional forms of the insertases in vivo. This observation is predicated upon the expectation that insertases exist in vivo as oligomers or in heteromeric complexes. Since CL interaction is likely related to function, and significant CL interaction is observed only for dimer simulations (discussed below), we conclude that insertase monomers are likely not functionally relevant units in a CL-rich environment.

We next tested whether the assembly of YidC1 or YidC2 as dimers within the simulated bilayer would change the potential lipid-interacting interfaces of the proteins. A dimeric YidC structure has been proposed in the literature based on bioinformatic analysis of an ancestral YidC-like protein as the progenitor of SecY (Lewis and Hegde, 2021), as well as experimental data on E. coli YidC and other Oxa1 family proteins, EMC3 and GET1 (Oliver and Paetzel, 2008; Ravaud et al., 2008; McDowell et al., 2020; O'Donnell et al., 2020; Pleiner et al., 2020). Recently a preprint supporting a functional dimeric pore-like structure of E. coli YidC, based on fluorescence correlation spectroscopy, Blue-Native PAGE, reconstitution of purified protein into vesicles, and single molecule microscopy, was published (Knyazev et al., 2023).

To model how dimeric YidC1 or YidC2 interact with membrane phospholipids, simulation systems containing the CL-rich tri-lipid mixtures were built around constructed homodimers of WT YidC1 and WT YidC2. Notably, the TM4 helices of S. mutans both YidC1 and YidC2 align with TM5 of E. coli YidC, which was proposed by Lewis and Hegde to be present at the dimer interface (Lewis and Hegde, 2021). Thus, dimer structures were built by cloning and placing a 180-degree rotated monomer respective to the original monomer with TM4-TM4 distance measured to be 7–8 Å using PyMol. Dimeric simulations were energy-minimized and equilibrated identically to monomeric simulations, and production was carried out for 10 μs prior to trajectory analysis. As with monomeric simulations, dimeric YidC1 and YidC2 displayed average RMSD values less than ∼2 Å for Cα-atoms when fit to the final equilibration frame of respective equilibration runs (Supplementary Figure S4). High CL densities were observed around multiple residues for both YidC1 and YidC2 dimeric structures (Figures 4A–C). Although the number of CL-occupying residues was comparable for YidC1 and YidC2, 38 and 32, respectively, CL occupied distinctly different locations within each dimer. Residues within TM1, TM4 and TM5 helices were strongly occupied by CL in dimeric YidC1 in comparison to multiple residues within TM2, TM3, and TM4 helices in dimeric YidC2. Notably, CL significantly occupied the TM4-TM4 dimer interface in both YidC1 and YidC2. In our simulations, cytoplasmic loop 1 residues at the junctions of TM1 and TM2 were significantly occupied by CL in the YidC1 dimer, but those sites were unoccupied in the YidC2 dimer. A lack of CL occupancy at the lipid-water interface in the TM1 and TM2 helices of YidC2 suggests a more flexible C1 loop in YidC2 compared to YidC1, which is supported by RMSF analysis (Supplementary Figure S5). In addition, RMSF analyses suggest structural plasticity in dimeric structures for both YidC1 and YidC2 when compared to their monomeric forms (Figures 4D, E). The coincident observations of increased dynamics and enhanced CL interaction as dimers is intriguing.

When assembled as dimeric forms in the CL-rich environment, differences in RMSF values between YidC1 and YidC2 were quite substantial. This information is illustrated by a YidC2 minus YidC1 heat map (Figure 4F). The YidC2 dimer appears more dynamic by approximately 1 Å compared to YidC1 for each aligned residue pair. Although a 1 Å RMSF difference in magnitude at a given individual position may not seem high, statistically significant differences (p<0.05) were observed for nearly every homologous residue. Collectively, these results reveal a level of structural flexibility for the YidC2 dimer in a CL-rich environment that is not as strongly shared by the YidC1 dimer. In addition, both YidC1 and YidC2 dimers acquired substantially increased CL occupancies compared to their respective simulated monomers, and the patterns of CL occupancy varied between the paralogs. A dimer-promoting role of CL was previously reported for SecYEG (Corey et al., 2018); therefore, it is not surprising that YidC1 and YidC2 both formed stable dimers in the simulated CL-rich environment. None of the YidC1 dimer residues showed significant PI or PS occupancy in the CL-rich tri-lipid mixture (Supplementary Figure S7). In contrast, the YidC2 dimer showed significant residency times with all three anionic phospholipids in the simulated bilayer (Supplementary Figure S7).

To understand how YidC1^W138R might impact lipid interaction and protein dynamics, we also simulated this mutated protein in a CL-rich environment for 10 μs per replicate. Structure files were prepared by introducing the point mutation to the WT, membrane-aligned, YidC1 monomer or dimer model prior to system generation in CHARMM-GUI. When evaluated in either monomeric or dimeric forms in the simulated CL-rich bilayer, YidC1^W138R became stable in <1 µs (Supplementary Figure S4). RMSF analysis and residue-wise comparison with the WT YidC1 dimer did not identify significant differences suggesting that the W138R point mutation did confer an overall dynamic structure like that observed for the YidC2 dimer. Both YidC1 and YidC1^W138R dimers were less dynamic than the YidC2 dimer in the CL-rich environment (Supplementary Figure S5). Nevertheless, the W138R point mutation within TM2 of YidC1 resulted in a loss of CL interactions with multiple residues within TM1, and TM1-C1 loop and TM2-C1 loop junctions (Figure 5A). Lipid occupancy maps further showed gain of CL interactions with residues within TM2 and TM3, a feature characteristic of the YidC2 dimer (Figures 5A, B). Because it is negatively charged, CL would be expected to interact better with positively charged Arg, compared to hydrophobic Trp. This was evidenced by the increased CL residency time at R138 of the mutant YidC1 dimer compared to W138 of the wild-type YidC1 dimer. Notably, the lipid occupancy map of the YidC1^W138R dimer acquired multiple similarities to that of the YidC2 dimer except at the C-terminal tail, which more closely resembled the YidC1 dimer (Figure 5A). Neither PI nor PS showed any significant occupancy with the YidC1^W138R dimer (Supplementary Figure S8). Collectively, evaluation of lipid interactions with the varying YidC monomers and dimers reveals the significance of CL in the YidC1^W138R acquisition of YidC2-like features, namely a gain in ability to support acid tolerance and a loss in tolerance to Zn-excess. Similar to monomeric YidC1 and YidC2, the YidC1^W138R monomer likewise did not demonstrate significant lipid occupancies (Supplementary Figure S6), further supporting a dimeric structure as the functionally relevant form of YidC insertase paralogs.

The acquisition of YidC2-associated behaviors by the YidC1^W138R mutant prompted us to also engineer substitution mutations at position 152 of YidC2. These included S152W, S152A, and S152R mutations, with each gene variant produced as the sole yidC paralog in a ΔyidC1/yidC2 background. First, we assayed each strain for growth under non-stress conditions (Figure 6). As expected, the ΔyidC2 strain demonstrated an obvious growth defect compared to the ΔyidC1 strain. None of the amino acid substitutions at YidC2 position 152 conferred a severe growth defect of S. mutans in liquid culture, although growth of the YidC2^S152A variant was somewhat more impacted than the YidC2^S152R and YidC2^S152W strains. Similar results were observed in the efficiency of plating assays. Importantly, when S. mutans growth was evaluated at pH 5, the YidC2^S152A and YidC2^S152W strains were both acid sensitive, as was the ΔyidC2 deletion mutant (Figure 6). This indicates that neither YidC1, nor YidC2^S152A or YidC2^S152W variants, can efficiently insert those substrates necessary for acid tolerance in S. mutans. In stark contrast, the YidC2^S152R strain tolerated acidic pH even better than the ΔyidC1 strain that produces unmodified YidC2. That is, YidC2-mediated acid tolerance was improved by the YidC2^S152R substitution. Again, results were recapitulated in efficiency of plating assays. Collectively, these results demonstrate the critical importance of residues 138 and 152 to YidC1 and YidC2s’ respective abilities to support acid tolerance, or not. Taken together, our results reveal that a Trp at this position is associated with acid sensitivity, while a Ser or Arg at this position enables the insertase to support acid tolerance as is observed for unmodified YidC2, YidC2^S152R, and YidC1^W138R.

Lastly, S. mutans was grown in the presence of excess Zn (Figure 6). As expected, growth of the ΔyidC1 mutant was severely impaired in liquid culture and on agar plates. The YidC2^S152W strain also acquired the severe ΔyidC1-associated zinc sensitivity phenotype. In other words, the YidC2^S152W variant became YidC1-like. In addition, the YidC2^S152A strain also became intolerant to excess zinc, though more so in liquid culture than on agar plates. Again, the result of the YidC2^S152R mutation was opposite to that of the other amino acid substitutions tested. In this case, zinc tolerance was notably improved in liquid culture. When tested on agar plates, the YidC2^S152W mutation was as severe as eliminating yidC1, but less impact was observed for YidC2^S152R followed by YidC2^S152A in efficiency of plating assays. Lack of complete concordance of results in liquid culture compared to those on agar plates could stem from differences in insertion efficiency or functionality of relevant substrates under conditions of varying aeration, or as yet unknown changes in lipid composition in the presence of excess zinc.

Because the yidC2 ^S152W -expressing strain phenotypically resembled the strain expressing yidC1 only, we wanted to understand the impact of the S152W mutation on phospholipid interactions at the level of individual residues in comparison to dimeric unmodified YidC1 and YidC2. As with dimeric YidC1^W138R simulations, YidC2^S152W/R structure files were prepared by introducing the point mutation to the WT, membrane-aligned, YidC2 dimer model prior to system generation and production MD in the simulated CL/PI/PS tri-lipid bilayer. RMSD values equilibrated to <2.5 Å compared to the final snapshot of equilibration, indicating system stability (Supplementary Figure S4). After PyLipID analysis, the CL occupancy profile of YidC2^S152W dimers clearly showed decreased intensity at all positions, including S152, compared to YidC1 or YidC2 dimers (Figures 7A, B; Supplementary Figure S9). The most apparent CL occupancy observed for the YidC2^S152W dimer was at residues within the cytoplasmic C1 loop close to the TM1 interface. This would be expected to confer rigidity to the C1 loop region similar to that observed for the YidC1 dimer. PI occupancy for the YidC2^S152W dimer followed a pattern generally similar to that of CL occupancy; however, PI occupancy was gained for residues within the C-terminal tail of the YidC2^S152W mutant dimer (Supplementary Figure S9). Similar to results observed for CL occupancy, the YidC2^S152W mutation also resulted in an overall loss of PS occupancy at numerous residues within the YidC2^S152W dimer. This distinguished it from the unmodified YidC2 dimer with regard to phospholipid residency (Supplementary Figure S7). Comparison of RMSF values of YidC2^S152W residues with those of the YidC2 dimer showed a ∼0.5–1 Å decrease in the RMSF values for approximately 2/3 of YidC2^S152W residues (Figure 7C, left). Although these decreases in RMSF values in YidC2^S152W were small compared to those in unmodified YidC2, they were statistically significant as determined by t-test indicating the S152W mutation resulted in a global decrease in YidC2 protein dynamics (Figure 7C, left, Supplementary Figure S5).

Phenotypic characterization of the panel of YidC2 point mutants showed that substitution of Arg for Ser at position 152 favored growth under all conditions tested. Similar to every YidC variant tested, YidC2^S152R quickly equilibrated within the CL-rich membrane, and the CL occupancy profile of the YidC2^S152R dimer was similar to that observed for the WT YidC2 dimer (Supplementary Figures S7, S9). However, S152R showed a lower CL residency time compared to S152. Residue-based RMSF analysis suggests an overall decrease in the dynamics of YidC2^S152R in a CL-rich environment compared to the results for unmodified YidC2 (Figure 7C, right). Interestingly, YidC2^S152W RMSF values compared to those of WT dimeric YidC1 are not statistically significant, which is taken as further supportive evidence that YidC2^S152W phenocopies WT YidC1 (Figure 7D, left). In contrast, RMSF values of the YidC2^S152R dimer suggest increased dynamics compared to unmodified dimeric YidC1 (Figure 7D, right). That is to say, the YidC2^S152W dynamic profile was more similar to that of YidC1 than to that of WT YidC2, and the YidC2^S152R dynamic profile was statistically significantly different from either WT YidC1 or WT YidC2. The latter observation was somewhat surprising since in other cases we were able to relate dynamic behavior observed in CGMD experiments to improved YidC2 functionality in vivo. An important caveat, however, is that our simulations were carried out in the absence of a confirmed substrate. Respective substrates of bacterial translocation systems, of which there are potentially hundreds, are largely unknown. Even in the most widely studied organism E. coli, only a handful of membrane-localized substrates are verified and utilized in model experimental systems.

Bacterial strains and plasmids used in this study are listed in Supplementary Table S1. All of the S. mutans mutants used in this study were derived from S. mutans strain UA159 (Ajdic et al., 2002), and were routinely grown in Todd-Hewitt broth (BBL, Becton Dickinson) supplemented with 0.3% yeast extract (THYE) at 37 °C in 5% CO₂/95% air atmosphere (v/v). Spectinomycin (1 mg ml^-1), kanamycin (1 mg ml^-1), or erythromycin (10 μg ml^-1) were added, where appropriate, for the growth and selection of S. mutans strains. E. coli strain C2987 (NEB) was used for genetic manipulations and was routinely grown in Luria–Bertani (LB) medium (10 g l^-1 tryptone, 5 g l^-1 yeast extract and 5 g l⁻¹ NaCl) at 37°C with vigorous shaking or on LB agar (1.5%) with appropriate selection. Ampicillin (100 μg ml⁻¹), or erythromycin (250 μg ml⁻¹) were used for the growth and selection of E. coli transformants.

Site-directed mutagenesis of yidC1 and yidC2. A standard PCR-based mutagenesis technique was used for site-directed mutagenesis (Zheng et al., 2004). pSM15 and pSM20 (Mishra and Brady, 2021) were used as templates for the mutagenesis of yidC1 and yidC2 genes, respectively. Primers used for mutagenesis are listed in Supplementary Table S2.

S. mutans strains were engineered to harbor the yidC1 or yidC2 variant of interest expressed under the regulatory control of the respective native promoter. Next, the residual yidC1 or yidC2 gene was deleted by allelic replacement with an antibiotic resistance cassette. Replacement cassettes consisting of the wild-type or mutant yidC1 or yidC2 gene were fused to an antibiotic resistance marker gene followed by the 3′-flanking regions of yidC1 or yidC2, respectively. To make the cassettes, PCR amplification products of the three regions were assembled by the NEB HiFi DNA assembly kit following the manufacturer’s instructions. The assembled products served as templates for PCR to generate gene the complete replacement cassettes using end primers. These were then used for transformation of S. mutans strain UA159. Once transformants were confirmed to harbor the desired yidC1 or yidC2 variant replacements, they were transformed with an allelic replacement construct that comprised the PCR-amplified 5′- and 3′-flanking regions of the residual yidC joined to an antibiotic resistance encoding gene. All mutations were confirmed by PCR and sequencing of the relevant genomic loci. Production of YidC1 or YidC2 and their derivatives was confirmed by Western blot using anti-C-terminal tail-specific rabbit antibodies (Mishra and Brady, 2021). Primers used for construction of deletion mutants are listed in Supplementary Table S2.

Promoter-reporter gene fusions were constructed for yidC1, yidC2, and cls. For the yidC1 promoter, 470 bps of DNA located directly upstream of the yidC1 start codon, which included the promoter and partial sequence of rnpA was amplified using primers listed in Supplementary Table S2. For the yidC2 and cls promoter-reporter gene fusions, 289 and 300 bp segments upstream of yidC2 and wapA start codons were amplified, respectively. PCR products corresponding to promoters were subcloned between BamHI/SacI digested pMZ plasmid (Chen et al., 2002; Liu et al., 2009), which contains a promoterless lacZ gene derived from Streptococcus salivarius for integration within S. mutans genomic DNA by a double crossover event at mtlA-phnA locus. Promoter-lacZ fusion containing plasmids were transformed in S. mutans UA159 competent cells and the integrity of the gene fusions was verified by PCR amplification of the relevant locus followed by sequencing.

Exponential phase cultures (OD₆₀₀ = 0.2) of promoter-reporter fusion strains grown in THYE were pelleted and resuspended in THYE pH 7.0 (control), THYE pH 5 (acid stress), THYE +3% NaCl (osmotic stress), THYE +2.5 mM ZnCl2 (Zn-excess), THYE +250 μg mL^-1 bacitracin (cell envelope stress) for 45 min. Pellets were harvested and washed with Z-buffer ((60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, freshly-added 50 mM β-mercaptoethanol), and β-galactosidase activity was determined and expressed as described previously (Seaton et al., 2015).

Overnight cultures grown in THYE broth with appropriate selection agents were diluted 1:50 in THYE and grown to mid-exponential phase (OD₆₀₀ = 0.2–0.4). Following 1:100 dilution of mid-log phase cultures, bacterial growth was measured using a Bioscreen C instrument as described (Mishra and Brady, 2021). Wells of the honeycomb plates used in the Bioscreen C instrument were covered with ∼50 μl of heavy mineral oil to achieve anaerobic growth under non-stress and acid stress conditions, except under conditions of Zn-excess in which mineral oil was not used because zinc toxicity is only observed under aerobic conditions. For efficiency of plating assays, mid-exponential phase cultures were adjusted to OD₆₀₀ = 0.2. Serial ten-fold dilutions were made in THYE, and 4 μL aliquots of each dilution were spotted on THYE agar (pH 7.0, pH 5.0, 3% NaCl, or 2.5 mM ZnCl₂). Plates were incubated at 37 °C in air with 5% CO₂ for 2 days before documentation.

Three-dimensional models of the monomeric YidC1 and YidC2 were built using the I-TASSER homology modeling webserver using B. halodurans YidC2 (PDB ID: 3WO6) and E. coli YidC (PDB ID: 3WVF) as templates (Kumazaki et al., 2014a; Kumazaki et al., 2014b; Yang et al., 2015; Yang and Zhang, 2015) (Figure 1B). Alpha Fold predicted models were similar to those derived by I-TASSER except that YidC1 and YidC2 C-terminal tail structures were not predicted with high confidence by Alpha-Fold. Head-to-head YidC1 or YidC2 dimer files were built by cloning and placing the 180-degree rotated monomer from the original monomer. TM3-4 of each monomer was designated as the interface with a minimum backbone-backbone distance of 7–8 Å at the closest point, measured via Pymol. Monomer and dimer structure files were aligned in a mock membrane using the PPM 2.0 webserver (Lomize et al., 2012). Point mutations were introduced into each WT insertase membrane aligned model structure using the PyMol mutagenesis wizard in order to build mutant systems. Groningen Machine for Chemical Simulation (GROMACS) (Abraham et al., 2015) system files of box size 200 nm, 310.15 K, and with 50 mM NaCl, and 1 bar pressure were built using the CHARMM-GUI (Jo et al., 2008; Lee et al., 2016) Martini maker (Qi et al., 2015; Hsu et al., 2017). Single systems of each protein, monomer or dimer, were built into lipid mixtures containing either 60 mol% CL or PE, with 20 mol% PI and 20 mol% PS. Lipid acyl chains were 1-palmitoyl-2-oleoyl (PO) in all cases. A schematic diagram of phospholipid molecules employed in this study is shown in Supplementary Figure S1. Each system was energy minimized and equilibrated according to default CHARMM-GUI mdp settings with the Martini2.2 forcefield (Monticelli et al., 2008; Periole et al., 2009; d et al., 2013), using the Berendensen Barostat (Berendsen et al., 1984) v-rescale thermostat, and the reaction-field method of electrostatics (Hünenburger and Gunsteren, 1998). After equilibration, each system was replicated for production to ensure equivalent starting energy profiles for a total of 5 replicates per system (Knapp et al., 2018). Production was carried out for 10 µs with identical parameters with the exception of the Parrinello-Rahman Barostat (Parrinello and Rahman, 1981) in place of the Berendensen Barostat. Production mdp files were altered only to influence simulation time length. Computational time was provided by the Texas Tech High Performance Computing Center (HPCC). Protein Cα root-mean-square deviation (RMSD) analysis was performed with the gmx rms tool and calculated as average Cα RMSD as a function of simulation time compared to the final structural snapshot from equilibration trajectories per simulation. RMSD is presented as mean ± s.d. from the 5 replicate systems. Average root-mean-square fluctuations (RMSF) per residue were calculated using the gmx rmsf tool. RMSF data is presented as mean ± s.d. of each residue throughout simulation across 5 replicates. Two-tailed t-tests from RMSF data were calculated between per replicate residue RMSF values across datasets. Lipid residency times were calculated using PyLipID (Song et al., 2022) with an interaction start distance of 0.55 nm and an end distance of 1.0 nm. Data are presented as the mean of 5 replicates.