The strains, plasmids and primers employed in this study are listed in Supplementary Tables 1, 2. The M. xanthus strains were cultivated in Casitone-based rich-nutrient CTT medium (Hodgkin and Kaiser, 1977) for growth assays. E. coli strains were routinely grown on Luria-Bertani (LB) agar or in LB broth. The incubation temperatures were 37°C and 30°C for E. coli cells and M. xanthus cells, respectively. When required, final concentrations of 40 μg/ml kanamycin, 100 μg/ml ampicillin, 34 μg/ml chloramphenicol, and 10 μg/ml tetracycline were added to the media.

The two groEL genes and one groES gene in the M. xanthus DK1622 genome are MXAN_RS23765 (groEL1), MXAN_RS21695 (groEL2), and MXAN_RS23760 (groES). The NCBI accession numbers for the GroEL1, GroEL2, and GroES protein sequences of M. xanthus DK1622 are WP_011554876.1, WP_011554465.1, and WP_002640434.1, respectively. The conserved domain information of GroEL and GroES was obtained from the CDD protein family (cdd239460 for GroEL and cdd238197 for GroES of M. xanthus DK1622, respectively; Marchler-Bauer et al., 2017). Based on the sequence similarity of the conserved domains, we performed RPS-BLAST (Reverse Position-Specific BLAST), which searches against the NCBI genome database (http://www.ncbi.nlm.nih.gov/genome/), and retrieved the superset of all the GroEL and GroES protein sequences in prokaryotic genomes. The E-value for the search was set to be equal to or higher than the default cutoff of 0.01.

A whole-genome-based myxobacterial phylogenetic tree was constructed online utilizing the CVTree3 program and a composition vector approach (Zuo and Hao, 2015).

Deletion of groES was carried out in M. xanthus DK1622 using the positive-negative KG cassettes (Ueki et al., 1996). Genomic DNA from DK1622 served as the template for PCR amplification of the upstream and downstream homologous arms utilizing Pfu DNA polymerase. The arms were fused at the BamHI site to create internal deletion fragments, which were cloned into SmaI-digested pBJ113, forming the deletion plasmid pBJ-groES. The deletion plasmid was transferred via electroporation into M. xanthus DK1622 cells as previously described (Kashefi and Hartzell, 1995). Individual Km-resistant colonies were selected and inoculated onto CTT agar plates supplemented with 1% galactose for a second round of screening. Deletion mutants were identified based on their galactose resistance and kanamycin sensitivity phenotypes, as well as by PCR and sequencing verification (Li et al., 2010).

The groES gene was fused to the pilA promoter (1 kb), and the fused sequence was digested with XbaI and EcoRI and ligated into the site-specific integration plasmid pSWU30, which had been digested with XbaI and EcoRI, to obtain the plasmid pSWU-groES. This plasmid was transferred via electroporation into M. xanthus DK1622 cells as previously described (Kashefi and Hartzell, 1995). Individual tetracycline-resistant colonies were selected, and the selected mutant was designated YL0308. The mutant was further employed to delete the local groES gene via electroporation with pBJ-groES. The km^r and ter^r mutants were selected, and proper integration of the fragment was confirmed by PCR and sequencing (Lobedanz and Sogaard-Andersen, 2003).

The groES gene, including the groES-groEL1 intergenic region, was fused to the promoter (500 bp) of groEL2, which was then fused upstream of groEL2. The fused fragments were digested with XbaI and EcoRI and ligated into the site-specific integration plasmid pSWU30, which was digested with XbaI and EcoRI to obtain the plasmid pSWU-P_groEL2+groES+groEL2. This plasmid was transferred via electroporation into groEL2 deletion mutant YL0302 cells. Individual colonies with tetracycline-resistance were selected, and the mutant was designated YL1101. All primers mentioned above are listed in Supplementary Table 2.

The promoter of either groEL2 or groEL1 was fused to the groES gene, and the fused fragments were digested with BamHI and EcoRI and ligated into the Myxococcus-E. coli shuttle plasmid pZJY41 (which had also been digested with BamHI and EcoRI) to obtain the plasmids pZJY-P_groEL2+groES and pZJY-P_groEL1+groES. The constructed plasmids were then transferred via electroporation into M. xanthus DK1622 cells, and the mutants were designated YL1102 and YL1103, respectively (Supplementary Table 1).

The M. xanthus groEL and groES genes were separately cloned into pET22b and over–expressed in E. coli BL21(DE3) cells. The heterogeneously expressed GroELs and GroES were purified as described (Klunker et al., 2003), with some modifications. Briefly, BL21 cells were grown at 37°C in LB medium with 100 μg/ml ampicillin to an OD600 of 0.7 and induced with 1 mM IPTG. Cells were resuspended in lysis buffer (25 mM Tris-HCl, pH 8.0, 300 mM NaCl, 2 mM dithiothreitol) and lysed via sonication. After gentle ultrasonication, the mixtures were centrifuged at 4°C for 30 min (12,000 rpm). The supernatant containing His-tagged GroES was applied to a Ni²⁺-NTA column and purified by Superdex 200 chromatography. His-tag sequences were not added to the GroEL proteins to maintain target protein activity. The supernatants containing GroEL were concentrated using tubes with a 100-kDa cut-off and were further purified by chromatography (Superdex 200).

An ATPase assay was performed as previously described (Figueiredo et al., 2004) with minor modifications. Briefly, 1 μM GroEL1 and GroEL2 were individually added to a reaction buffer containing 20 mM MOPS, pH 7.5, 100 mM KCl, 5 mM MgCl₂, and 0.5 M ammonium sulfate with or without 2 μM GroES at 37°C. The reaction was initiated by adding 2 mM ATP and stopped with 20 mM CDTA after 10 min. Liberated inorganic phosphate was quantified by the malachite green assay (Adenosinetriphosphatase assay kit from Nanjing Jiancheng Bioengineering Institute, China) after incubation at 25°C for 30 min. The absorption was measured at 640 nm. Three independent replicates were performed for each experiment.

Twenty micromolars malate dehydrogenase (MDH) was denatured in 1 M guanidinium HCl for 60 min, and the reaction system was diluted 100-fold with buffer A (50 mM Tris-HCl, pH 7.0, 0.5 M ammonium sulfate, and 2 mM ATP). Next, 1 μM GroEL1 and GroEL2 were separately added into buffer A with or without 2 μM GroES. The concentration of added chaperonin was determined by NanoDrop 2000 Spectrophotometer. The reaction mixture was incubated at room temperature for 60 min (Figueiredo et al., 2004). Fifty millimolars CDTA was added to stop the reaction. The enzymatic activity of MDH was measured every 10 min at 340 nm using buffer B, containing 50 mM Tris-HCl, pH 7.0, and 0.5 M ammonium sulfate, 1 mg/ml bovine serum albumin, 0.22 mM NADH, 0.55 mM oxaloacetate, 1 mM CDTA. For citrate synthase (CS) refolding assay, after a similar process, the enzymatic activity of CS was determined every 10 min at 412 nm using the Citrate Synthase Activity Assay Kit (Sigma-Alorich). Three independent replicates were performed for each experiment.

Proteins were purified as described above. Excess amount of GroEL1 (20 mg) or GroEL2 (20 mg) proteins was mixed with 1–4 mg His-tagged GroES in suspension buffer (25 mM Tris-HCl, pH 8.0, 300 mM NaCl, 2 mM dithiothreitol, and 2 mM ATP) for 3 h. The mixtures were eluted through nickel columns. Then the columns were washed three times using washing buffer, and the bound proteins were eluted from the columns with 500 mM imidazole and detected by SDS-PAGE.

Hetero-expression of the duplicate GroELs and GroES was performed using pET22b-GroEL1-NoHis, pET22b-GroEL2-NoHis, and pET28a-GroES-NoHis respectively (Supplementary Table 1), in E. coli BL21(DE3) cells. BL21 cells were treated as described as above. After gentle ultrasonic treatment, cells were centrifuged at 4°C for 30 min (12,000 rpm) for the native PAGE assay. Compared to normal sodium dodecyl sulfate (SDS)-PAGE, the protein buffer and gel for native PAGE contained no SDS or mercaptoethanol. Electrophoresis was performed in a mixture of ice and water to avoid the depolymerization of multimer proteins.

To investigate the in vivo refolding activities of GroEL1/2 proteins with or without the presence of GroES, we constructed the plasmids pET22b-HrcA, pBAD33-GroESL1/2, and pBAD33-GroEL1/2 (Supplementary Figure 1). The compatible plasmids pET22b-HrcA and pBAD33-GroESL1/2 (or pBAD33-GroEL1/2) were transformed into E. coli Top 10F' cells (Supplementary Table 1), and transformants were selected under the double-resistance conditions of 34 μg/ml chloramphenicol and 100 μg/ml ampicillin. When the culture OD600 reached 0.7, 100 mM IPTG was added for the induction. Then, cells were harvested, resuspended in lysis buffer containing 25 mM Tris-HCl, pH 8.0, 300 mM NaCl, 2 mM dithiothreitol, and lysed via sonication. After gentle ultrasonication, the mixtures were centrifuged at 4°C for 30 min (12,000 rpm), which divided the mixture into supernatants and sediments. The samples were tested the expression and solubility of HrcA protein by SDS-PAGE.

M. xanthus DK1622 cells and groEL mutants were inoculated at final concentration of 1 × 10⁷ cells/ml in CTT medium and cultured for 36 h to the exponential growth stage, respectively. Then 0.5 ml of each of the cultures was transferred into 50 ml fresh CTT medium for further cultivation. The cultures were harvested at periodic time points, and the RNA was extracted immediately using a bacterial RNA extraction kit (Thermo Fisher) according to the manufacturer's instructions. The purified RNA extracts were reverse-transcribed to cDNA. Quantitative real-time PCR was performed in a total reaction volume of 20 μl containing 1 μl of 250 nM primers, 10 μl of SYBR Green PCR master mix, 8.5 μl of RNase-free water, and 0.5 μl of a 10-fold-diluted cDNA template. 16S rRNA gene was used as the reference. The primers used for real-time PCR are listed in Supplementary Table 2.

Mid-log-phase cultures of M. xanthus cells were harvested as described above. After heat-shock treatment at 42°C for 30 min or 60 min, the cells were immediately serially diluted and plated on 1.5 and 0.3% CTT agar. After 6 days, colony-forming unit (CFU) numbers were calculated, using untreated cells as a control. The survival rate was calculated as CFU divided by the cell number before heat shock (5 × 10⁹; Li et al., 2010).

The difference significance was analyzed statistically by using IBM SPSS Statistics for independent-samples T-test.

We surveyed the occurrence of groEL and groES genes by RPS-BLAST searching based on sequence similarity, and retrieved a total of 5,658 groEL and 5,253 groES genes in 4,861 completely sequenced prokaryotic genomes (Supplementary Table 3). The number of groEL genes in a genome range from zero to seven, and 93.5% of these genomes contain one or more groEL genes (Table 1). Consistent with the previous report (Williams and Fares, 2010), Mycoplasma often lack the groEL genes (67 of the 105 sequenced genomes). In addition, some Spiroplasma, Ureaplasma, and unclassified bacteria have no groEL gene. In Archaea, strains from Euryarchaeota and TACK group also lack groEL. In the genomes with single groELs, the groEL gene is always neighbored with a groES gene, forming a bicistronic groESL operon. There are 884 bacterial genomes possessing two or more groEL genes (19.5% of the 4,861 sequenced genomes). Of the total 5,658 groEL genes, 4,824 were found closely downstream a groES gene in a groESL operon, while the other 834 (14.7% of groEL genes) have no neighboring groES gene. The stand-alone groEL genes are distributed in 770 genomes; thus, some genomes contain two or more stand-alone groEL genes. This is consistent with a previous survey of 669 bacterial genomes in which the majority of groEL genes co-occurred with a neighboring groES gene, while others stood alone (Lund, 2009). Notably, of the total 5,253 groES genes, 435 stand alone with no neighboring groEL gene; these genes are distributed across 419 genomes.

There are 24 sequenced myxobacterial genomes available in the GenBank database. Among them, 19 genomes contain two groEL genes, two contain three groEL genes, and three contain single groEL genes. Figure 1 is a diagrammatic sketch of the occurrence of groEL and groES genes in myxobacteria (details referred to Supplementary Table 4). All myxobacterial genomes possess at least one complete bicistronic groESL operon. Interestingly, the loss of a neighboring groES is rather a characteristic of duplicate groEL genes in the Cystobacterineae suborder of myxobacteria, while the four sequenced genomes of Sorangium and Chondromyces in the Sorangineae suborder and the marine halophilic strain Haliangium ochraceum DSM14365 contain a second groESL operon. Of the two genomes possessing three copies of groEL, the Sorangineae strain (S. cellulosum So0157-2) has two complete groESL operons and a stand-alone groEL, while the Cystobacterineae strain (C. fuscus DSM2262) has a complete groESL operon and two stand-alone groEL genes. The three myxobacteria with single groEL genes are newly identified myxobacterial strains that are phylogenetically distant from “the classical myxobacteria.” Although normally coinciding at a high taxonomic level, the occurrence of the groES and groEL genes may vary at the level of genus or even the species. For example, while one of the two groEL genes in Cystobacter violaceus Cb vi76 neighbors groES and one does not, the C. fuscus DSM 2262 strain has three groEL genes, of which one neighbors a groES gene and two stand alone. S. cellulosum So ce56 possesses two complete groESL clusters, but S. cellulosum So0157-2 contains two groESL clusters and a third stand-alone groEL gene.

M. xanthus DK1622 possesses two groEL genes that share 79% similarity of their amino acid sequences (Goldman et al., 2006; Li et al., 2010). Based on our previous studies, a single groEL1 or groEL2 gene is capable of individually supporting the growth of DK1622 cells, but these two genes function divergently in predation and development processes (Li et al., 2010; Wang et al., 2013). Another significant characteristic of the two paralogous groEL genes is their organization differences in genome. The groEL1 gene (MXAN_RS23765) lies downstream of groES, forming a complete bicistronic groESL operon, verified using RT-PCR (referred to Supplementary Figure 2), while groEL2 (MXAN_RS21695) does not neighbor with a groES gene.

We attempted to knock out the groES gene but failed, suggesting that inactivation of the groES gene is fatal to DK1622 cells. To confirm the essentiality of groES, we complemented this gene at the Mx8 attB site in the DK1622 genome using pSWU30 to form a mutant with two copies of the groES gene (strain YL0308). We further attempted to delete the original groES from YL0308, and, as expected, obtained a viable mutant (YL0309), having an ectopically expressed groES gene and the stand-alone genes groEL1 and groEL2. The YL0309 mutant had a nearly identical growth curve to the wild-type strain DK1622 (P > 0.05 for each detected time point; Figure 2). Thus, the groES gene is essential for the cell survival of M. xanthus DK1622. The groEL1 and groEL2 genes were alternatively deletable for cell survival, which suggested that GroEL2 also required GroES in order to function. M. xanthus cells were likely able to self-regulate groEL and groES gene expression to meet cellular requirements for these proteins.

To ascertain the relationship between the two GroELs and the single GroES protein of M. xanthus DK1622, we hetero-expressed these genes in E. coli cells and purified the target proteins. The in vitro binding activities of GroES and GroEL1/GroEL2 were assayed using two methods. We designed GroES with a His-tag at the C-terminus, which made the protein to be able to bind to a nickel column. The GroEL proteins had no His-tag and were unable to bind to the nickel column by themselves. When added to a GroES-containing column, either GroEL1 or GroEL2 was able to retain on the column (Figure 3A). As the concentration of the column-bound GroES increased, the recovery of GroEL proteins by the column also increased. Furthermore, GroEL is an ATPase, and binding to GroES reduces the ATPase activity of GroEL proteins (Martin et al., 1991; Viitanen et al., 1991). After the addition of GroES, the ATPase activities of the GroEL1 and GroEL2 proteins decreased to a similar extent (P < 0.001 and 0.001 < P < 0.01 for GroEL1 and GroEL2, respectively; Figure 3B). The above results confirmed the binding activities of both GroEL1 and GroEL2 to GroES in vitro. Based on the results in Figure 3B, GroEL1 showed higher in vitro ATPase activity than GroEL2 in the absence or presence of GroES, suggesting their disparate refolding activities in cells.

We assayed the in vitro refolding activities of the two GroELs with the malate dehydrogenase (MDH) protein. In the presence of GroEL1 or GroEL2, but without GroES, denatured MDH exhibited slight renaturation. When GroES was added, the MDH-folding activity of GroEL1 had a slightly higher rate than that of GroEL2 after 30 min (P < 0.05), but the final refolding activities were nearly identical (~25% re-natured MDH) for both GroEL1 and GroEL2 (Figure 3C). In addition, we also performed the refolding assay on citrate synthase (CS), which showed similar result as the MDH refolding (Supplementary Figure 3).

We estimated the in vivo polymerization of the Myxococcus GroEL1 and GroEL2 proteins in E. coli cells in the presence or absence of the GroES. As the native-PAGE shown, the alone-expressed GroEL1 and GroEL2 proteins had a molecular weight (Mw) of ~800 kDa, respectively (the lanes 3–4 and 6–7 in Figure 4A). The molecular weight is about 14-times of that of the GroEL proteins (the Mws of GroEL1 and GroEL2 are 57.9 and 58.1 kDa, respectively, and their 14-mer polymers are 810.6 and 813.4 kDa, respectively). Similarly, the only-expressed GroES proteins formed a 7-mer complex (lanes 1–2 in Figure 4B; the Mw of GroES is 10.7 kDa, and the Mws of 7-mer GroES are 74.9 kDa). Co-polymers of GroEL and GroES were also observed in E. coli cells containing co-expressed groEL1/groEL2 and groES genes of M. xanthus DK1622. The GroEL1-GroES and GroEL2-GroES complexes formed retarded bands with a similar molecular size (lanes 1–2 and 8–9 in Figure 4A; the Mws of 7-mer GroES plus 14-mer GroEL1 or 14-mer GroEL2 are 885.5 or 888.3 kDa, respectively). These results suggested that the individual GroEL1 and GroEL2 proteins both existed as 14-mer homopolymers, each of which was able to bind the 7-mer of GroES in vivo.

HrcA is a negative regulatory protein that is able to bind to the CIRCE (Controlling Inverted Repeat of Chaperone Expression) regions in front of the groESL operons and thus reduce the expression of groEL and groES genes (Wilson et al., 2005). The HrcA protein relies on GroEL to complete refolding and is a potential GroEL substrate. We assayed the in vivo refolding functions of the two GroEL proteins with or without GroES by expressing the groEL, groES and hrcA genes from M. xanthus in different combinations in E. coli cells. SDS-PAGE result indicated that in the absence of the DK1622 groES gene, HrcA primarily existed in the form of inclusion bodies in cells when co-expressed with only the groEL1 or groEL2 gene from M. xanthus (Figure 4C). However, when the DK1622 groES gene was further introduced into in the E. coli cells containing the hrcA and groEL1/groEL2 genes from M. xanthus, the HrcA protein was then mostly soluble (Figure 4D). These results suggested that both GroEL1 and GroEL2 proteins from M. xanthus were able to aid the refolding of the HrcA proteins in the presence of the M. xanthus GroES in E. coli cells.

The above results indicated that either GroEL1 or GroEL2 requires GroES in order to function in M. xanthus. The two groELs and the single groES had to coordinate their expressions for synergy. Our previous studies showed that the two groELs had a balance of their expressions in M. xanthus cells (Li et al., 2010; Wang et al., 2013). In this study, we further assayed expressions of the groES gene in the wild type strain DK1622, the groEL1/2 deletion mutants (YL0301 and YL0302) and their corresponding complementary strains (YL0901 and YL0902) by using quantitative PCR. In the wild type strain DK1622, groES gene expression increased and reached its maximum during the first 24 h of incubation in CTT medium. Then, the expression decreased and reached its lowest level after ~42–48 h of incubation (Figure 5). There was a slight increase in groES expression in the late stationary growth stage. This groES expression curve was highly consistent with that of the groEL genes in DK1622 (Li et al., 2010). Deletion of the groEL2 gene, which has no neighboring groES gene, resulted in the decreased expression of the groES gene at each time point. This result is similar to those obtained following the deletion of the groEL1 gene (Figure 5). When groEL1 or groEL2 with its own promoter was complemented ectopically at the attB site in the genome of the groEL1 deletion mutant YL0301 (YL0901 and YL0902), the mutants recovered the total groEL expression levels comparable to that in M. xanthus DK1622 (Wang et al., 2013). Interestingly, the groES expression levels in YL0901 and YL0902 were also recovered to that in the wild type strain (Figure 5). These results suggested that there was a self-regulation mechanism to balance the expressions of the groEL and groES genes for their synergetic functions in M. xanthus cells.

In M. xanthus DK1622, the groES and groEL1 genes are in an operon under the same promoter, while groEL2 stands with a different promoter (Supplementary Figure 4 demonstrates their construction in genome). To confirm the possible coordinated expression and self-regulation of isolated groEL and groES genes, we further introduced an additional groES gene into DK1622 using the shuttle plasmid pZJY41 under the groEL2 or groEL1 promoter (referred to Supplementary Figure 4), forming the mutants YL1102 and YL1103, respectively. Compared with that in the wild-type strain DK1622, the expressions of groES markedly increased in both the YL1102 and YL1103 mutants (P < 0.001) and the groEL1 promoter is stronger than the groEL2 promoter (Figure 6). Interestingly, the expressions of either the groEL1 or the groEL2 genes also increased correspondingly (P < 0.001). The groES expression levels were comparable to the total expression changes of groEL1 and groEL2 in the two mutants. The results further demonstrated that the expression levels of the groEL and groES genes appeared to be self-regulated in M. xanthus, even when distributed in different places in the genome.

Stand-alone groELs are quite common in bacteria. In myxobacteria, the second groEL gene exists alone in the Cystobacterineae suborder but forms a complete groESL operon in the Sorangineae suborder (Figure 1). Based on our results, the stand-alone groEL gene functioned in a groES-dependent manner, at least in M. xanthus DK1622. We thus sought to determine whether an artificial groESL2 operon was able to change the coordinated expression of groEL and groES genes in M. xanthus. We inserted a second copy of the groES gene in front of groEL2 in DK1622, following the native groEL2 promoter (Supplementary Figure 4) to form a complete bicistronic groESL2 operon (mutant YL1101). The co-expression of groEL2 and the inserted groES in the artificial groESL operon was verified by RT-PCR amplification (Supplementary Figure 2). However, the total expression levels of groES, as well as that of groELs, in this mutant were still similar to those of DK1622 (P > 0.05); whereas the groES and groEL1 expressions in the groEL2 deletion mutant (YL0302) were both significantly decreased (P < 0.001; Figure 7A). The co-chaperonin GroES, in combination with the chaperonin GroEL, helps bacteria resist heat-shock conditions (Moparthi et al., 2013). We further assayed the heat-shock response of the M. xanthus cells containing two complete groESL operons. The survival rate of the YL1101 mutant after heat-shock treatment showed almost no differences from that of the wild-type strain DK1622 (P > 0.05); whereas the groEL2 deletion mutant decreased the survival rate upon heat shock (P < 0.001; Figure 7B). The results demonstrated that the single groES gene was molecularly equivalent for the expressions and functions of the duplicate groEL genes in M. xanthus cells.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.