The evolutionary timeline for Mycoplasmataceae was calculated by TimeTree in timeline mode using the website http://www.timetree.org/. Divergence times for M. pneumoniae and Mycoplasma hyopneumoniae and for M. hyopneumoniae and M. bovis were located by using node time mode on the same server.

Enolase sequence alignment was performed by using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) with the default settings. There were two sources for mycoplasma enolase sequences. First, partial sequences were from the complete annotated genome sequences (Supplementary Dataset S5). The rests were from the non-redundant protein sequences database. For the same species, only one strain was used. The incomplete sequences and controversial sequences were discarded. Finally seventy-three mycoplasma enolase sequences were collected. To maintain consistency with the structural data, nonmycoplasma enolases were selected preferentially when their structure was known. Considering that mycoplasmas were the main body of the alignment, only representative species were selected from among spiroplasmas, archaeas, eukaryotes, and cell-walled bacterias.

Genes coding Mb Eno (NCBI accession number: WP_013456550) and Mp Eno (NCBI accession number: WP_010874963) were optimized according to E. coli codon usage and synthesized by GenScript Biotech Corp. (Nanjing). Then, the two genes were inserted into the pET21a vector and transformed into the BL21(DE3) E. coli strain. Protein expression and purification were carried out according to the previous methods (Chen et al., 2019).

Mb Eno and Mp Eno were concentrated and diluted to 6 and 12 mg/ml for crystal screening. The crystals were screened by using the sitting drop vapor diffusion method at 291 K. Mp Eno crystals were grown in 0.2 M sodium citrate tribasic dihydrate, 0.1 M Tris-hydrochloride and 30% v/v polyethylene glycol 400 at pH 8.5 after 7 days. Mb Eno crystals were grown in 0.14 M calcium chloride dihydrate, 0.07 M sodium acetate trihydrate, 14% v/v 2-propanol and 30% glycerol at pH 4.6 after 1 week. The Mp Eno and Mb Eno crystals were diffracted at 1.8 and 1.7 Å, respectively, on beamline BL19U1 at the Shanghai Synchrotron Radiation Facility (SSRF). X-ray diffraction data were merged, integrated and scaled by using HKL3000 software. The structures of Mp Eno and Mb Eno were solved by molecular replacement using the enolase structure of Mhp Eno (Protein Data Bank [PDB] ID: 3j6i) as a reference model with Phaser in the CCP4 program suite (Hough and Wilson, 2018). REFMAC5 (Kovalevskiy et al., 2018) was used for initial restrained rigid-body refinement. COOT (Emsley and Cowtan, 2004) was used for manual model building. Further refinement was performed with Phenix (DiMaio et al., 2013). Finally, the stereochemical quality of the final models was further evaluated with the program PROCHECK. Structural analysis was performed by using the CCP4 program and PyMOL software. Structural similarity assessment and the generation of heat maps were carried out with the Dali server (http://ekhidna2.biocenter.helsinki.fi/dali/). The oligomerization interfaces were analysed by PISA server (https://www.ebi.ac.uk/pdbe/pisa/).

Mb Eno and Mp Eno were diluted with PBS to a concentration of 0.01 mg/ml. After glow-discharge treatment, 5 μl of each sample was added to the front side of the copper grid. After standing for 60 s, the copper grid was dried with filter paper. Five microliters of 2% uranyl acetate was immediately added to the copper grid for staining and then dried with filter paper. The staining step was repeated once, after which the grid was dried naturally. The samples were observed with an FEI Tecnai Spirit electron microscope.

The enolase sequences used to construct the evolutionary tree were the same as it used in sequence alignment. The evolutionary tree was generated by MEGA 10 (Kumar et al., 2018). Maximum likelihood was used as the statistical method. After protein sequence alignment was carried out, the amino acid substitution models were searched using the aligned sequences. LG mode was used for the substitution model. The gamma distribution was used to calculate rates among sites. Partial deletion was chosen for gap/missing data treatment. All other parameters were set to the default values. To construct the 16S rRNA tree, GTR + G + I mode was used for the substitution model (Kumar et al., 2018). The neighbor-joining tree method was used for statistical test.

Two hundred eighty-one complete mycoplasma genomes (sixty-three species) were downloaded form GeneBank (Supplementary Dataset S5). Genomes without annotations, genomes with mistaken annotations, and genomes with incomplete data were discarded. HGTree database (http://hgtree.snu.ac.kr/) and DarkHorse HGT Candidate Resource (http://darkhorse.ucsd.edu/) were used for the detection of lateral gene transfer and the identification of phylogenetically atypical proteins on a genome-wide basis (Podell and Gaasterland, 2007; Jeong et al., 2016). The ACLAME database (http://aclame.ulb.ac.be), ICEberg 2.0 (http://db-mml.sjtu.edu.cn/ICEberg/) and PHASTER (http://phaster.ca/) were applied to search for integrating conjugative elements, mobile genetic elements, cis-mobilizable elements and phage and prophage sequences (Leplae et al., 2010; Arndt et al., 2016; Liu et al., 2019).

SPR was performed at 25°C with a CM5 sensor chip in a BIAcore X100-Plus (GE Healthcare). Purified swine PLG and FN (Sigma) were separately diluted to 50 μg/ml and immobilized on the chip with a resonance unit (RU) of approximately 2000 by using an amine coupling kit (Biacore AB). Binding kinetics were analyzed by using HBS-EP buffer consisting of 10 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 0.05% (v/v) P20 (a surfactant) (Biacore AB). Analytes (Mp Eno and Mb Eno) at a series of increasing concentrations (0–4,000 nmol/L) were used in the experiment. The dissociation phase was monitored for 1,000 s by allowing the proteins to flow over the chip at a flow rate of 30 μl/min. Kinetic model equation K D   =     k d k a (K _D is equilibrium dissociation constant, k _a is association rate constant, k _d is dissociation rate constant) was used to calculate affinity values (Day et al., 2002).

Far-WB analysis was carried out according to a previous procedure (Chen et al., 2019). In brief, approximately 20 μg of Mhp Eno, Mp Eno and Mb Eno was transferred to a PVDF membrane. BSA was used instead of mycoplasma enolases as a negative control. After blocking with skimmed milk, the membrane was incubated with 5 μg/ml plasminogen or fibronectin (Roche). Then anti-plasminogen or anti-fibronectin antibody (Abcam; 1 μg/ml) was used to interact with the membrane as the primary antibody. Finally, horseradish peroxidase (HRP)-conjugated anti-IgG (Boster; 1:5,000 dilution) and electrochemiluminescence kits (Thermo Scientific) were used to detect proteins in the membrane.

Enolases from different species in a nonredundant protein sequence database were aligned, and mycoplasma enolases did not show good sequence similarity with each other. Takeing M. pneumoniae enolase (Mp Eno) and Mycoplasma genitalium enolase (Mg Eno) as example, the similarity between them was only 77.85% (Figure 1, Supplementary Figure S1 and Supplementary Dataset S1). According to the percent idendity matrix of mammalian enolases, Streptococcus enolase and Enterobacteriaceae enolase, the normal enolase similarity between two species with close evolutionary relationships normal is over 90% (Supplementary Figures S2–S4 and Supplementary Datasets S2–S4). For example, the similarity between enolases from Streptococcus pneumoniae and Streptococcus pyogenes is 93%. Even fly enolase has a percent identity of 78.24% with lobster enolase, although flies and lobsters are only in the same phylum, Arthropoda. Even the similarity between enolases from Streptococcus pneumoniae and Enterococcus hirae is 83.06%, and these bacteria are members of the same Lactobacillales order but belong to different families (Figure 1, Supplementary Figure S1 and Supplementary Dataset S1). Overall, the identity between any two mycoplasma enolases was seldom higher than 90%. The lower similarities between Mp Eno or Mg Eno and other mycoplasma enolases were more obvious (Figure 1, Supplementary Figure S1 and Supplementary Dataset S1).

In our previous studies, the S3/H1, H6/S6, H7/H8, and H13 regions were found to be unique features of Mhp Eno in terms of both sequence and structure (Chen et al., 2019). Here, we confirmed that all mycoplasma enolases have a sequence pattern that obviously differs from those of other (archaea, cell-walled bacteria and eukaryotic) enolases, especially in the above-mentioned S3/H1, H6/S6, and H7/H8 regions and H5/H6 region (Figure 2). This sequence pattern also differs among enolases from different Mycoplasma species. In the S2/S3 loop, four obviously different sequences were found in total of seventy-six mycoplasma-related enolases (three spiroplasma enolases). The S3/H1 region (containing inserted TKYE) was found to be a specific feature of Mhp Eno in our previous studies. However, we found that fifty-eight mycoplasma enolases had this region, but the enolases of S. eriocheiris, M. pneumoniae and M. genitalium lacked this region, showing the same sequence pattern contained in enolases from cell-walled bacteria and eukaryotes (Figure 2). Fifty-three mycoplasma enolases contained a deletion between site 150 and site 160 in the H4/S4 loop region; however, only twenty-three mycoplasma enolases had an insertion in this region (Figure 2). Other minor differences in the H3/H4 H5/H6, H6/S6, H7/H8 and H11/S10 regions were found among the different mycoplasma enolases (Figure 2). Compared to mycoplasma enolases, enolases from other organisms in the same phylum or domain share a relatively conserved sequence pattern (Figure 2). These results indicate that mycoplasma enolases adopted their own sequence patterns but that these patterns are not strictly conserved.

The enolases of most prokaryotes were found to be octamers. However, dimeric enolases have also been observed in some species (Kühnel and Luisi, 2001; Franklin et al., 2015). In our studies, enolase from M. pneumoniae was found to be a dimer, and that from M. bovis was shown to be an octamer. These results were confirmed by multiple methods. First, denatured Mp Eno and Mb Eno showed the same monomeric molecular weight of 50 kDa on an SDS-PAGE gel (Figures 3A,B). For native PAGE analysis, octameric Mhp Eno was used as a standard molecule. The Mp Eno band ran at approximately 100 kDa, which is between 90 and 130 kDa. The Mb Eno band was above 130 kDa, which is near the band corresponding to Mhp Eno (Figure 3C). This means that soluble Mp Eno and Mb Eno were dimers and octamers, respectively. Second, during size-exclusion chromatography, Mp Eno eluted at approximately 14 ml, corresponding to a molecular weight of 100 kDa, which indicates the dimeric isoform. Mb Eno eluted at approximately 11 ml, corresponding to a molecular weight of 400 kDa, which means that Mb Eno is an octamer. Third, negative stain electron microscopy of the proteins was used to directly image the two molecules. According to the EM images, Mb Eno appeared as a flower-like ring composed of eight small subunits, indicating an octamer. However, Mp Eno was too small to observe its appearance (data not shown) (Figure 3D). Finally, in the resolved structures, Mp Eno is clearly a dimer, and Mb Eno is an octamer (Figure 4). The contacting interfaces for the oligomerization of Mp Eno and Mb Eno were examined. The detail information was listed in Table 1. There is only a dimeric interface that holds the two monomers together in Mp Eno. Mp Eno dimeric interface has 22 hydrogen bonds and buries a total of 1717 Å² surface area. There are 54 amino acids from one molecule and 53 amino acids from another molecule to form the Mb Eno interface. For Mb Eno, however, there are two types of oligomerization interfaces: dimeric interface and octameric interface. The octameric interface fastens the neighboring heart-like dimers to from a ring-like octamer. The dimeric interface of Mb Eno was composed of 49 residues from one monomer and 51 residues from another monomer. There are 34 hydrogen bonds, 12 salt bridges and lots of Van der Waals forces to form an interface area of about 1,856.4 Å². The octameric interface of Mb Eno is on the opposite side of that of the dimer, forming a surface area of about 1,279.9 Å². A total of 67 residues from the two neighboring enolase monomers were involved in this interface. There are 21 hydrogen bonds and 2 salt bridges contributing to the formation of octameric interface (Table 1). In our studies, the oligomerizations of Mp Eno and Mb Eno seem to be stable in different buffers (Tris−HCl buffer and PBS) and different concentrations (>6000 μg/ml and 10 μg/ml). Phenomenon of coexist of monomers, dimers, and octamers in solution found in enolase from Trichomonas vaginalis was not detected in Mp Eno and Mb Eno (Mirasol-Meléndez et al., 2018). All these results clearly demonstrate that Mb Eno exists as an octamer and that Mp Eno exists as a dimer. In most cases, enolases from two evolutionarily related species share the same oligomerization state. For example, enolases from both Streptococcus pneumoniae (Ehinger et al., 2004) and Streptococcus suis (Lu et al., 2012) are octamers. Therefore, it is somewhat unusual that different isoforms are found in mycoplasma enolases.

The structures of enolase Mp Eno and Mb Eno crystals were determined at 1.8 Å and 1.7 Å respectively (Table 2). As enolase is a relatively conserved enzyme, Mb Eno and Mp Eno have overall structures very similar to those of other enolases (Figures 4A,B). Dimeric Mp Eno is composed of two enolase monomers with a heart-like or butterfly like shape, similar to the shapes of other dimeric enolases (Figure 4C). Octameric Mb Eno consists of four dimers of enolase monomers, forming a disc-like shape with a center tunnel (Figure 4D). Both Mb and Mp enolase monomers are composed of an N-terminal cap domain and a C-terminal TIM barrel domain. The C-terminal domain of Mb Eno and Mp Eno shows a topology of β2α2βα2(βα)5, which differs from the traditional β2α2(βα)6 topology of other enolases (Figures 4A,B).

A detailed structural comparison between mycoplasma enolase and other determined enolase structures in the PDB database was carried out. First, the structural superimposition of enolases from different species was performed (Supplementary Figure S5). The overall folds of all enolases overlapped well. As indicated by the sequence alignment, notable differences were still found in the S3/H1, H6/S6, H7/H8, and H13 regions (Figure 5 and Supplementary Figure S5). Insertions in these regions were found in mycoplasma enolases. This difference manifests as differences in the length of the loop or the insertion or deletion of a helix. However, amino acid mutations were not reflected in the structures. For example, although obviously different amino acid arrangements were observed in the H5/H6 region (Figure 2), no structural differences were detected (Supplementary Figure S5). Thermal stabilities of different enolase structures were also checked (Figure 5). It is clear that S3/H1 loop and H6/S6 loop regions studied here are seems to be stable, while most of them have a low or moderate b-factor value. S6/H7 loop regions or Helix 7 regions have relative high b-factors in their overall structures. However, the values of Helix 7 are still acceptable in mycoplasma enolase structures (Figure 5).

To quantify the structural similarity among enolases from different species, “all-to-all” structural comparison was carried out (Supplementary Figure S6). According to the heat map, species with close evolutionary relationships clearly have high structural similarity and are clustered together (Supplementary Figure S6). However, similar to the low identity among mycoplasma enolase sequences, structures of mycoplasma enolase also did not show good similarity with each other. Both Mb Eno and Mp Eno showed the best structural similarity with E. hirae enolase with a Z-score of 61.7. Furthermore, both Mb Eno and Mp Eno showed the second-best similarity with enolase from B. subtilis, with a Z-score of 61.4. The Z-score for Mp Eno and Mhp Eno was only 58.5, and that for Mp Eno and Mhp Eno was only 60.9. Mhp Eno is the most structurally similar to the enolases of B. subtilis, S. aureus and E. hirae, with Z-scores of 62.7, 62.2 and 61.5, respectively. The Z-score for Mb Eno and Mhp Eno was only 61.3 (Supplementary Table S1). These analytical results indicate that the structures of mycoplasma enolases also differ from one another.

However, unlike other enolases, both Mb Eno and Mp Eno have an extra helix, helix 7 (H7), which was first found in Mhp Eno and shown to be a species-specific loop in Mhp Eno (Figure 5). To determine whether all mycoplasma enolases have helix 7, we checked the alignment and found that the amino acids in H7 are relatively conserved among almost all mycoplasma enolases (Figure 2, Figure 5 and Figure 6A). Helix 7 is composed of a motif with the sequence of F-K/G-K-L/F-K-X-A-I. Within this motif, the first amino acid, F, and third amino acid, K, are strictly conserved, and the other sites are occasionally substituted by other amino acids. However, all other nonmycoplasma enolases lack this motif (Figure 6). This was further confirmed by structural analysis (Figures 6B,C). Therefore, we confirmed that helix H7 is characteristic to mycoplasma enolases in terms of both sequence and structure.

The long loop after helices H7 and before H8 also only exists in Mycoplasma species (Figure 2). However, the length of the H7/H8 loop differed among the three mycoplasma enolases examined (Supplementary Figure S5B and Figure 6). Mb Eno had the longest H7/H8 loop, which was longer than that of Mhp Eno by 3 additional amino acids. Approximately 85% of the Mycoplasma species within Mycoplasmataceae share this long H7/H8 loop sequence.

To better understand the evolutionary relationship among different enolase proteins, we constructed an evolutionary tree by using enolase protein sequences from different species. We simultaneously constructed a tree by using 16S rRNA sequences from the corresponding species to determine the evolutionary relationship among these specific species (Figure 7). To conduct a parallel comparison, an evolutionary tree was also built using conserved EF-TU proteins (Ludwig et al., 1993) from the corresponding species. It is easily found that the three different evolutionary trees exhibited good reconciliation. Mycoplasmas were grouped into three branches: the pneumoniae group, the hominis group and the spiroplasma group. The evolutionary relationship evidenced by the enolase tree was consistent with that calculated by 16S rRNA tree but was different from the traditional classification. For example, M. mycoides, M. leachii, M. capricolum and several other species are clustered with spiroplasmas other than mycoplasmas. However, as shown by the evolutionary trees, enolase proteins from different kingdoms or phyla have much closer relationships than the overall evolutionary relationships of the species themselves. In other words, the genetic distances between the kingdoms Archaea, Bacteria and Eukarya are much longer than the corresponding distances between their enolase proteins. However, the distances among different mycoplasma enolases were much longer than those in the EF-TU and 16S rRNA trees (Figure 7). For example, the distances between M. hyopneumoniae and Mycoplasma ovipneumoniae were 0.25, 0.1 and 0.04 in the enolase, EF-TU and 16S rRNA trees, respectively. The distances between M. pneumoniae and M. genitalium were 0.288, 0.031 and 0.017 in the enolase, EF-TU and 16S rRNA trees, respectively. These results indicate that enolases from different Mycoplasma species exhibit longer evolutionary distances than the specific species themselves (Figure 7). These results indicate that enolases are conserved among different kingdoms but divergent in Mycoplasmataceae.

Horizontal gene transfer is always a factor of disturbance in evolutionary analyses. However, according to our analysis, it is very unlikely that horizontal gene transfer occured in Mycoplasma enolase genes. First, we checked 281 public genomes of mycoplasmas and found that only one enolase gene could be identified in each genome. Second, mycoplasma enolase genes are not included in the list of phylogenetically atypical proteins, which are potential candidates for horizontal gene transfer, as calculated by DarkHorse HGT Candidate Resource and HGTree. Third, none of the integrated conjugative elements, mobile genetic elements, cis-mobilizable elements, or phage and prophage sequences were detected in mycoplasma enolase genes by using ICEberg 2.0, ACLAME or PHASTER. Therefore, we can conclude that mycoplasma enolase genes were generated by parent-to-progeny inheritance rather than horizontal gene transfer.

To determine whether divergence in enolase exists in all the other clades, enolases from Enterobacteriaceae, Streptococcus and Mammalia were randomly selected and used to construct evolutionary trees (Supplementary Figure S7). Enolases from the same class among these Enterobacteriaceae, Streptococcus and Mammalia species were closely clustered in the trees. In fact, the distances from one enolase to another in the same class were too short to distinguish. This means that enolases in these species are well conserved or did not undergo much independent evolution (Supplementary Figure S7). However, in contrast, the branches of different species in the mycoplasma enolase tree are long enough to be well distinguished (Figure 7A). These results indicate that enolase divergence is specific to Mycoplasmataceae. However, we have not determined whether this finding applies to every class. Currently, this divergence has been found in only Mycoplasmataceae.

Furthermore, we constructed separate evolutionary trees for the S3/H1, H6/H8 (including the H6/S6 loop and H7/H8 loop) and H4/S4 regions (Figure 8). According to the S3/H1 loop tree, the original evolutionary relationship was clearly disrupted. Mycoplasma ovis and Synechococcus elongatus had the closest relationship. E. coli and C. aurantiacus were mixed in the Mollicutes clade. Campylobacter jejuni showed a closer relationship with eukaryotes than with other species (Figure 8A). The same findings were observed in the H4/S4 loop tree. The H4/S4 loop of E. coli enolase had a much closer relationship with those of eukaryotic enolases. Cell-walled bacterial clades were distributed in branches of Mollicutes (Figure 8B). These results indicate that the S3/H1 and H4/S4 regions of mycoplasma enolases are hot spots for evolution. However, the H6/H8 region tree and species evolutionary tree were consistent. In the H6/H8 region tree, all Mollicutes species are clustered (Figure 8C). Cell-walled bacteria were also clustered. This indicates that the evolution of this region is quite consistent with the evolution of the corresponding specific species. This is also consistent with the previously observed phenomenon that helix 7 and the H7/H8 loop are almost absolutely conserved in mycoplasmas. This means that the H6/H8 region including helix 7 of mycoplasma enolases, may have evolved from a common ancestor. This further confirms the species-specific nature of mycoplasma enolases.

Different isoform statuses, structural features and sequence patterns have been observed in different mycoplasma enolases. To determine whether they have different functions, we checked their affinity to PLG and FN. Mb Eno, Mp Eno and Mhp Eno are all reported to be important adhesins and can interact with PLG and FN. SPR is widely used for bacteria moonlight enzymes to detect their binding affinities with PLG and FN (Matta et al., 2010; Yu et al., 2018). Therefore, we performed SPR to determine the affinity between PLG/FN and the three enolases. According to the results, the affinities of the three enolases to PLG were different. Mhp Eno had the best value of 62.5 nM. The K _D of Mb Eno was 95.23 nM. The affinity of Mp Eno to PLG was about 300 nM. These results indicate that all three enolases could interact with PLG and showed slight differences from each other (Table 3 and Supplementary Figure S8). For FN, this difference was more significant. Affinity of Mp Eno to FN was too low to be detected. The FN affinity of Mb Eno and Mhp Eno were 485.8 and 74.08 nM respectively, with 10-fold difference (Table 3 and Supplementary Figures S8E–G). These results were also confirmed by far-WB (Supplementary Figure S8) and indicate that mycoplasma enolases have similar binding functions, but their binding affinities differ.