757 MotB sequences from above were aligned using Clustal Omega (Larkin et al., 2007) with five iterative re-alignments. The phylogeny was then estimated with Quicktree using Kimura translation for pairwise distance and calculating bootstraps with 100 iterations (Howe et al., 2002). The phylogeny was midpoint-rooted in FigTree (Rambaut, 2014) for display purposes, and should be treated cautiously as with any single-protein estimate of a deep phylogeny (Pallen and Matzke, 2006; Abby and Rocha, 2012; Shih and Matzke, 2013; Prangishvili et al., 2017; Ishida et al., 2019). For the purpose of surveying sequence diversity across MotBs, the tree is adequate, and arbitrary re-rooting will not affect estimation of ASRs at nodes.

Thirteen nodes were selected at assessed divergent points between clades based on known sodium swimmers and their proximity and distance from the arbitrary root. Sequences (Supplementary Figure 2A) and conservation via sequence logo (Crooks et al., 2004) are shown in Supplementary Figure 2.

Ancestral protein sequences were reconstructed using the empirical Bayes method implemented in PAML (Yang, 2007). MotB sequences were truncated to focus on the TM region between residues 23–65 on MotB. The posterior probability distribution at each site for each ancestral node was also calculated using PAML.

The bacterial strains and plasmids used in this study are shown in full in Supplementary Table 1. The primary strains used in this work are RP6894 (Block et al., 1989) and RP3087 (Blair et al., 1991). All the E. coli strains were cultured in LB broth and LB agar [1% (w/v) Bacto tryptone, 0.5% (w/v) Bacto yeast extract, 0.5% (w/v) NaCl, and 2% (w/v) Bacto agar for solid media] at 37°C. According to the selective antibiotic resistance pattern of the plasmids, chloramphenicol (CAM), ampicillin (AMP) or kanamycin (KAN) were added to a final concentration of 25 μg/mL, 50 μg/mL, and 25 μg/mL, respectively.

We engineered the ancestral reconstructed sequences (ASR) in a predesigned form of genetic structure that included the first 22 residues of E. coli MotB, followed by 42 residues (23–65) of ancestral sequence and a final 243 residues (66–308) of the MotB chassis (ASR = 1–22 MotB; 23–65 ASR; 66–308 MotB). In summary, this used the E. coli N-terminal domain, the ASR for the TM and plug domains (TM: residues 28–49, plug: 52–65) and then the E. coli PGD and C-terminal domain (PGD: 196–225). MotB-ASRs were constructed by PCR amplification using the chimeric plasmid pSHU1234 (Nishino et al., 2015) containing PomA and PotB (a hybrid of PomB and MotB) as the cloning vector. Thirteen forward ultramer primers were designed with ∼200 nucleotide overhangs that contained a NdeI restriction site, the predicted ancestral sequences and an annealing sequence specific to MotB on PotB in pSHU1234. A common reverse primer was designed to amplify from the end of PotB with a PstI restriction site. IDT synthesized all the primers. The list of all primers is provided in Supplementary Table 2. Phusion high-fidelity (HF) DNA polymerase (NEB) was used for the PCR amplification of ASR sequences. Each reaction contained 10 μL of 5X Phusion HF buffer, 200 μM dNTPs, 0.1 μM of each primer, 10 ng of plasmid template, 3% of DMSO, 1 U Phusion HF DNA polymerase, sterile Milli-Q water to final volume of 50 μL. The reactions were started by heating to 98°C for 30 s. The PCR reactions were then subjected to 35 cycles of 98°C for 10 s, 58°C for 30 s, and 72°C for 30 s, with a final 5 min extension step at 72°C.

The amplified MotB-ASR sequences were then separated in 1% (w/v) agarose gel and purified using the Qiagen gel purification kit. Purified MotB-ASR sequences and pSHU1234 were digested with NdeI and PstI before ligation with T4 ligase (NEB) using a 1:3 vector to insert molar ratio. All the resulting MotB-ASR plasmids contain a PomA from the vector backbone and an ancestral-contemporary MotB hybrid (Supplementary Table 1).

Finally, all the MotB-ASR plasmids were transformed into stator deleted (RP6894) and MotB deleted (RP3087) E. coli strains following the chemical transformation protocol from NEB. All engineered and cloning of MotB-ASR sequences were confirmed by colony PCR and Sanger sequencing at The Ramaciotti Centre for Genomics (UNSW, Australia) using ASR sequence-specific primers (Supplementary Table 2).

We constructed pMotB via deleting MotA from the MotAB plasmid pDB108 (A gift from David F Blair Lab) and constructed pPomA and pPotB by deleting PotB and PomA, respectively, from PomAPotB plasmid pSHU1234 (Nishino et al., 2015). Site-directed, Ligase-Independent Mutagenesis (SLIM) was used (Chiu et al., 2008) with specific primers (Supplementary Table 2). We also constructed a A225D point mutant of A. aeolicus MotA from the wild type A. aeolicus MotA plasmid (pNT7) (Takekawa et al., 2015) using QuikChange Lightning Site-Directed Mutagenesis with the recommended protocol from Agilent. The primers used for this point mutation were provided in the Supplementary Table 2.

We performed swim plate motility assays according to the previous protocol (Ishida et al., 2019) with slight modifications to check whether ancestral MotB proteins could restore motility. LB swim plates [0.25% (w/v) Bacto agar] were used to screen bacteria with swimming properties. Minimal medium swim plates [10 mM of KHPO₄, 0.1% (v/v) of glycerol, 0.1 mM of Thr, 0.1 mM of Leu, 0.1 mM of His, 0.1 mM of Met, 0.1 mM of Ser, 1 mM of MgSO₄, 1 mM of (NH₂)₂SO₄, 1 μg/mL of thiamine, and 85 mM of NaCl or 85 mM of KCl, respectively, 0.25% (w/v) agar, 0.02% (w/v) of arabinose, pH 7.0] were used to determine the dependency of coupling ions (Na⁺ or H⁺) as the medium contained only the desired ion and the pure forms of nutrients that did not contain any other ionic contaminants. Antibiotics (25 μg/mL CAM or 50 μg/mL AMP or 25 μg/mL KAN) were added according to the plasmids used. Additionally, minimal swim plates with different Na⁺ concentrations in combination with 100 μM phenamil (Sigma-Aldrich) were used to assess the of Na⁺ dependency on swimming properties.

Swim plates were inoculated with a single 1 day old colony with a sterile toothpick and incubated at 30°C for 14 h (for LB swim plates) or 16 h (minimal swim plates) to allow proper development of a swimming ring. Swimming zones were first visually checked, imaged with the ChemiDoc MP Imaging System (Bio-Rad), and swimming diameters measured using ImageJ software (Version 1.52).

The effect of different Na⁺ concentrations on the rotation speed of MotB-ASRs and sodium/proton controls were determined by tethered cell assays in the presence of a range of Na⁺ concentrations following a previous protocol (Nishiyama and Kojima, 2012) with some modifications. E. coli RP3087 containing MotB-ASR plasmids were inoculated into TB broth [1% (w/v) Bacto tryptone, 0.5% (w/v) NaCl] containing 0.02% (w/v) arabinose and 25 μg/ml chloramphenicol and were grown overnight (17 h) with 180 rpm at 30°C. The overnight cultures were sub-cultured with a 50-fold dilution into fresh TB broth and incubated for 5 h with 180 rpm at 30°C to get more motile cells. At OD₆₀₀ ∼0.80, the flagella of the cells were sheared by passing the culture multiple times (∼35) through a 26G needle syringe. After shearing the flagella, the cells were washed three times with motility buffer [10 mM potassium-phosphate, 10 mM lactic acid, 100 mM NaCl, and 0.1 mM EDTA, pH 7.0]. Cells were then attached on glass slides pre-treated with an anti-E. coli flagellin antibody with a 1:10 dilution (Nishiyama and Kojima, 2012) and washed sequentially with a Na⁺ concentration gradient containing motility buffer. The rotational speed of the cells was observed using phase-contrast microscopy (Nikon) and was recorded at 20 frames per second (FPS) through the 40X objective with a camera (Chameleon3 CM3, Point Grey Research). Rotational motion of the cells was analyzed using Lab view 2019 software (National Instruments) and rotational speeds were calculated from 20 individual cells.

Growth assays were executed to exclude possible growth impact of MotB-ASRs. Growth of E. coli strain RP3087 transformed with all the MotB-ASRs, pSHU1234, and pMotB plasmids were monitored as described (Santiveri et al., 2020) with slight modifications. Briefly, overnight cultures of the test strains with OD₆₀₀ 1.0 were diluted 1:100 in 96-well plates (Corning) with fresh LB broth containing different concentrations of NaCl, 0.02% (w/v) arabinose and 25 μg/ml chloramphenicol and incubated at 37°C. The OD₆₀₀ were measured in a microplate reader (FLUOstar OPTIMA, BMB LABTECH) every hour for 8 h with a brief shaking interval before each measurement. The experiment was performed in a triplicate.

To determine the correlation of mutations at each respective site of MotB ancestral sequences, we selected 19 species of bacteria (Supplementary Table 3) where the ion selectivity correlation with residue was previously measured (Ishida et al., 2019). Each species in the 19-species subset was classified as Na⁺ or H⁺ powered, we then added our functional MotB-ASRs as a further 13 H⁺ powered units and conducted Fisher’s Exact Test analysis (in R) based on a binary response variable. We were interested in whether a binary response might be correlated with a binary amino acid predictor, so we filtered columns in the alignment to exclude columns where the two most common amino acids added up to <85% of the observed residues. We ran Fisher’s exact test to correlate each amino acid or the mutational pair with the ionic power source.

We selected nodes in our phylogeny (Figure 1) for ancestral reconstruction (ASR) by a mixture of early and contemporary nodes to increase the probability of synthesizing a motile gene. Among the selected nodes, ASR981 had descendants that were exclusively betaproteobacteria, ASR1459 had descendants that were exclusively alphaproteobacteria, ASR908, ASR1170, ASR1239, ASR1246, ASR1457, and ASR1501 had descendants that were exclusively gammaproteobacteria, and the remaining, mostly older nodes ASR758, ASR759, ASR760, ASR765, and ASR1024 had descendants that were a mixture of gamma, beta and alphaproteobacterial, with ASR1024 also including all four of the hydrogenophilalias.

The composition of the 13 nodes is shown as the proportion of species represented in the tips downstream from that node (Figure 2). All the nodes were distributed to a wide range of bacterial genera among which Pseudomonas, Xanthomonas, Burkholderia, Yersinia, Paraburkholderia, Cupriavidus, Stenotrophomonas, Marinobacter, Xenorhabdus, and Caballeronia were prevalent. We calculated the top three genera by frequency of occurrence in the tips descendant from the selected nodes. Among all the 13 MotB-ASRs, all the descendants of ASR908 were Halomonas, all of ASR1246 were Pseudomonas, all of ASR1457 belonged to Marinobacter, and remaining nodes had descendants from mixed genera (Figure 2). With regard to species that could be expected to be sodium swimmers, in our original collection of 2187 MotB homologs we had 72 Vibrio species and 13 Shewanella, species. Following our network-based selection of refined sequences (Supplementary Figure 1), seven Vibrio strains remained in our phylogeny, and ASR1459 was the nearest parental-node to these seven strains.

We calculated MotB-ASRs using maximum likelihood methods (Yang, 2007) for each of the nodes above and engineered the TM and plug domains of these ASRs into our existing MotB chassis (Supplementary Figure 2). Motility, initially evaluated with swim plates, showed that all MotB-ASR proteins were functional and could restore motility in the presence of contemporary MotA (Figure 3A and Supplementary Figure 3). However, without MotA, in the presence of existing PomA, none of the MotB-ASRs were functional (Figure 3B). This is in contrast with the controls where PotB (chimeric B-unit) was functional together with PomA and non-functional with MotA (Supplementary Figure 4).

To determine the ion dependency of motile MotB-ASRs, we measured motility both in the presence and absence of Na⁺ on minimal swim plates. The minimal swim plate assay results showed that all the ASRs were able to swim in both the presence and absence of Na⁺ (Figures 4A,B). We further checked the swimming ability of all ASRs and H⁺/Na⁺ swimming controls (MotAMotB/PomAPotB) in the presence of the Na⁺ blocker phenamil. The results showed no inhibition of motility for all MotB-ASRs and the control H⁺ swimmer, but showed complete inhibition of motility for the control Na⁺ swimmer (PomAPotB) (Figure 4C). However, the swimming diameter of all MotB-ASRs and proton and sodium controls (MotAMotB and PomAPotB), was greater at higher Na⁺ concentrations (Supplementary Figure 5). This occurred despite the observed growth rates for MotB-ASRs and controls being equivalent at all sodium concentrations (Supplementary Figure 6). To verify the ionic energy source, we characterized stator energization via tethered cell assays.

The rotational speed of all 13 MotB-ASRs and the control strains (Na⁺ swimmer and H⁺ swimmer) were measured in the presence of different Na⁺ concentrations (0 mM, 5 mM, 21.25 mM, 42.5 mM, and 85 mM) using the tethered cell assay. Rotation speed of all tested MotB-ASRs showed no dependence on Na⁺ concentration similar to the H⁺ control swimmer (MotAMotB) (Figure 5). However, the Na⁺ control swimmer (PomAPotB) showed sodium dependence, with a gradual increase in swimming speed starting from 0 Hz to 2.8 ± 1.25 Hz maximum speed in the presence of 0 mM and 85 mM NaCl, respectively (Figure 5).

Among all 13 MotB-ASRs, MotB-ASR981 showed the highest swimming speed of 4.01 ± 0.92 Hz and MotB-ASR908 showed the lowest swimming speed of 2.71 ± 0.51 Hz. The remaining MotB-ASRs showed swimming speeds between 3.07 ± 0.55 Hz and 3.73 ± 0.92 Hz in all tested concentrations of Na⁺ (Supplementary Figure 7).

We co-transformed all the 13 MotB-ASRs with an ancient wild type (WT) Aquifex aeolicus (aa) MotA (MotA^aaWT) into a stator deleted E. coli (RP6894) to evaluate compatibility between A. aeolicus and MotB-ASR stators using the swim plate motility assay. Four (MotB-ASR760, MotB-ASR765, MotB-ASR908, and MotB-ASR981) out of 13 MotB-ASRs were compatible with MotA^aaWT. All the compatible MotB-ASRs showed Na⁺ independent motility and swam both in the presence and absence of Na⁺ (presence of K⁺) containing plates (Figures 6A,B). However, the same MotA^aaWT did not show any compatibility with contemporary E. coli MotB (MotB^E) (Supplementary Figure 8A). On the other hand, the point mutant (A225D) of A. aeolicus MotA (MotA^aa225D) was compatible with A. aeolicus-E. coli chimeric MotB (MotB^AE) and produced a much larger swimming zone than MotA^aaWT and the identical MotB^AE (Supplementary Figure 8B). In contrast, no MotB-ASRs were compatible with the point mutant of MotA^aa225D and all were non-motile (Supplementary Figure 8C).

Tethered cell assay with various concentration of Na⁺ (0 mM, 5 mM, 21.25 mM, 42.5 mM, and 85 mM) was performed to confirm the ionic energy source of the four motile MotB-ASRs (MotB-ASR760, MotB-ASR765, MotB-ASR908, and MotB-ASR981) together with MotA^aaWT co-transformed cells. We also measured the rotation speed of MotA^aa225D and MotB^AE co-transformed cell in the same condition as a Na⁺ dependent control (Takekawa et al., 2015). The rotation speeds of all the cells with MotB-ASRs and MotA^aaWT were stable and showed no dependence on external sodium concentration, while in contrast the rotation of MotA^aa225D and MotB^AE showed Na⁺ dependence (Figure 7). MotA^aaWTMotB-ASR908 rotated more slowly than MotA^aaWTMotB^AE with an average swimming speed of 1.59 ± 0.60 Hz but all other MotB-ASRs rotated more quickly, with MotB-ASR765 the fastest swimmer with the average swimming speed of 3.39 ± 0.79 Hz (Figure 7).

We assembled a sequence alignment including all 13 MotB-ASRs (Supplementary Figure 2A) and 19 additional bacterial species (Supplementary Table 3) with known sodium or proton motility (Ishida et al., 2019). We examined if the addition of our 13 proton-motile strains affected significance scores for correlation between specific residues and phenotype by Fisher’s Exact Test. For the hypothesized correlation between residue 30 F or Y (Figure 1) and proton or sodium motility, where the null hypothesis of no correlation was previously strongly rejected based on data from 19 living strains (p = 0.0034), the addition of 13 additional proton motile ASR sequences weakened, but did not completely eliminate, statistical significance at the p < 0.05 level (p = 0.014). However, residue 43 remained strongly significantly correlated with ionic power source (p = 0.00041 with living sequences; p = 0.000028 with ASRs added, Supplementary Table 4).