The type strain, TDA-040725-5^T (= CCTCC M 207170^T = DSM 21848^T) of S. eriocheiris (Wang et al., 2011) was cultured under aerobic conditions at 30°C to a late stage of exponential phase. To compare the effects of growth conditions on chemotactic activity, cells were grown in a tissue culture flask or in an anaerobic bag (AnaeroPouch kit provided by Mitsubishi gas chemical, Tokyo, Japan), for aerobic and anaerobic conditions, respectively.

The genome sequence of S. eriocheiris was completed by high-throughput sequencing via pyrosequencing. A total of 218,524 sequences with an average length of 246 bp were obtained for the S. eriocheiris genome, accounting for 39-fold genome coverages. The sequences were assembled using Newbler software of the 454 suite package, producing 14 contigs ranging from 1.2 to 406 kb. Multiplex PCR was performed to determine the relationship of contigs, and closure of the gaps was performed by sequencing the PCR products. Phred, Phrap, and Consed software package (http://www.phrap.org/phredphrapconsed.html) was used for final assembly and edits. Regions with poor sequencing quality and homopolymers were resequenced. The final consensus accuracy was 99.9982% for the S. eriocheiris genome.

Putative protein coding sequences (CDSs) were identified by Glimmer3 (Delcher et al., 1999) and ZCURVE 1.0 (Guo et al., 2003). Functional annotation of CDSs was performed through BLASTP searches against GenBank's non-redundant (nr) protein database, followed by manual inspection. Transfer RNA genes were predicted by tRNAScan-SE (v1.23) (Lowe and Eddy, 1997). The metabolic pathways were constructed using KEGG database (Kanehisa et al., 2004). Lipoprotein was predicted by LipoP 1.0 (Juncker et al., 2003) and searching PROSITE motif PS51257. Protein domain prediction and COG assignment were performed by RPS-BLAST using NCBI CDD library. All against all BLASTP was applied to determine the paralogs. Two sequences in a pair are paralogs if the remaining High-scoring Segment Pairs (HSPs) cover at least 80% of length of the shorter protein and if the identity is greater or equal to 50%.

The TCSs were predicted using both homology search and domain analysis. The homology search was performed against the Prokaryotic 2-Component System (P2CS) database (Barakat et al., 2011) through BlastP with the E < 1E-5. The computational domain analysis was performed using the method described in Song et al. (2012). Briefly, all the histidine kinases (HKs) related domains (PF00512, PF07568, PF07730, PF07536, PF06580, PF01627, PF02895, PF05384, and PF10090) and the receiver domain (PF00072) of response regulators (RRs) were used to recognize putative proteins of TCSs.

Unique genes were verified by a TBLASTN search using protein sequences of each strain against the other species' genome sequence. Concatenated protein sequences of 183 orthologous proteins of Spiroplasma species were first aligned using MUSCLE (Edgar, 2004), then the conserved alignment blocks were extracted by the Gblocks program (Guindon and Gascuel, 2003). The phylogenetic tree was built using the maximum likelihood method implemented in PHYML (Guindon and Gascuel, 2003).

The complete genome sequence and annotations of S. eriocheiris were deposited in GenBank under accession numbers CP001973.

Cultured cells were centrifuged at 11,000 × g, for 10 min at room temperature (RT) and the pellet was resuspended with a PSF buffer (75 mM sodium phosphate, containing 5% (w/v) sorbitol and 0.1% (w/v) fructose (pH 7.3) containing 0.6% methylcellulose (MC) to the original cell density. For the capillary assay, tips of a mechanical pipet were prepared containing 10 μl of 1% agar with various chemicals to be tested. The tips were soaked in 100 μl cell suspension for 30 min at 30°C. The relative cell number in the tips was estimated by ATP concentration measured by an ATP bioluminescence assay kit HS II (Sigma Aldrich, St. Louis, MO USA).

An agar drop assay was carried out in a chamber, which was assembled by one glass slide, one coverslip and two strips of double-sided tape (Islam et al., 2014). A drop of 1% agar (7 μl) was placed at the center of a glass slide, and a coverslip was attached to the glass slide using double-sided tape. Various possible attractants were selected (Daniels et al., 1980; Daniels and Longland, 1984) and added to the agar drop as necessary. Cell suspension containing 0.6% MC was injected into the chamber. The chamber was sealed by nail polish, and kept for 40 min at RT. The cell distribution at a distance in 4 mm from the agar drop was video recorded with phase-contrast microscopy (BX50, Olympus, Tokyo, Japan). The videos were analyzed by Image J ver.1.37v (http://rsb.info.nih.gov/ij/) as previously reported (Hiratsuka et al., 2005; Nakane and Miyata, 2009, 2012). The cell numbers in a video field 45.5 μm wide and 60.9 μm long were counted manually and averaged for three fields.

For tethered cell experiments, cultured cells suspended in PBS (75 mM sodium phosphate (pH 7.3) and 68 mM NaCl) containing 0.1% fructose adjusted to be equivalent to the original cell density were injected into a flow chamber and observed by phase-contrast microscopy. The liquid in the chamber was replaced by PBS containing 20 μM methionine. Cells attached to glass were video recorded and analyzed.

The cultured cells suspended in PBS to the original density were placed onto a carbon-coated EM grid, and incubated for 10 min at RT. The solution on the EM grid was removed and stained by ammonium molybdate for 1 min. To observe the internal structure, spiroplasma cells were suspended in PBS at 10-fold density of the culture, put on a grid, and incubated for 10 min at RT. The medium was replaced by 3 μl Triton X-100 solution (0.05–0.3% Triton X-100, 1 mg/ml DNase, and 5 mM MgCl₂ in PBS) and incubated at RT for 30 s. Triton X-100 solution was removed, and the grid was stained for 1 min with 1% ammonium molybdate (vol/vol) and air-dried. The samples were observed by a transmission EM (JEM1010, JEOL, Akishima, Japan) at 80 kV. EM images were captured by a FastScan-F214 (T) charge-coupled-device (CCD) camera (TVIPS, Gauting, Germany). For the isolated filament structure, a 6 μl suspension derived from 10 ml culture was applied to an EM grid. Filaments were picked by e2helixboxer.py (http://blake.bcm.edu/emanwiki/EMAN2/Programs/e2helixboxer). Image averaging was done by EMAN2, version 2.1 (http://blake.bcm.edu/EMAN2/).

Spiroplasma cells were collected, and washed three times with PBS/G (PBS containing 10 mM glucose), and suspended in PBS, and then 0.05–0.8% Triton X-100, 1 mg/ml DNase, and 5 mM MgCl₂ were added as the final concentrations. After 20 min incubation on ice, the Triton X-100-insoluble fraction was collected by centrifugation at 19,000 × g for 40 min at 4°C and suspended in PBS. The suspension was analyzed by SDS-PAGE and peptide mass fingerprinting (PMF) using autoflex speed (Bruker Daltonics, Billerica, MA, USA) as reported previously (Nakane et al., 2015).

We sequenced the genome of S. eriocheiris, a single, circular DNA of 1,364,757 bp (Figure 1). S. eriocheiris showed different genomic characters from the other 13 Spiroplasma species, harboring the second largest genome in the Spiroplasma genus, only 100 kb smaller than corn parasite S. kunkelii, and at least 100 kb larger than other Spiroplasma species (Table S1). The genome similarity analysis by GGDC (Meier-Kolthoff et al., 2013) of the Spiroplasma genomes also indicates the high divergence between S. eriocheiris and other Spiroplasma species, with 70% similarity to only S. mirum (Table S2).

The cytoskeletal proteins, Fibril (SPE_0666) (Williamson et al., 1991), five MreB proteins (Ku et al., 2014) (SPE_0470, SPE_1224, SPE_1228, SPE_1230, and SPE_1231) and FtsA (SPE_0753)—which are related to eukaryotic actin—and FtsZ (SPE_0752)—related to eukaryotic tubulin—were found.

To date, all bacterial species that undergo chemotaxis are known to have genes classified as TCSs. To identify proteins related to the conventional TCS in the genome, a homology search against all the “Prokaryotic 2-Component System” (P2CS) database (Barakat et al., 2011) was firstly performed. However, no TCS proteins were found in the S. eriocheiris genome. We then analyzed all of the domains encoded by the genome, and failed to find any known domains of histidine kinases (HKs), response regulators (RRs), and other protein components that are involved in well-studied chemotaxis systems, like Methyl-accepting chemotaxis protein (MCP) signaling domain.

Next, we examined the chemotactic ability of S. eriocheiris cells by using a capillary assay. The cells cultured in either aerobic or anaerobic conditions were suspended in a 100 μl PSF buffer containing 0.6% MC in a test tube. Pipet tips filled with various attractants, which were reported as an attractant for other microorganisms (Daniels et al., 1980; Daniels and Longland, 1984), were placed into the buffer. After incubation for 30 min at 30°C, the contents of tips were recovered and titrated for ATP concentration, which is known to be linearly related to the cell number (Dexter et al., 2003). ATP concentration is useful as a maker of cell number for Mollicute species which are not suitable for counting CFU (colony forming unit) because of their sticky cell property and slow growth rates. The cell numbers estimated from the ATP concentrations were obviously higher in the capillary containing the putative attractants, showing that more spiroplasmas swam into the pipet tips containing these substrates than in the control (Figure 2). It is unlikely that they propagated in the capillary because they need 150 min for doubling even in the growth medium. Interestingly, the chemotactic activity was more obvious when the spiroplasmas grew under anaerobic conditions vs. aerobic ones. These results, together with the genome analysis, show that S. eriocheiris performs chemotaxis dependent on aerobic conditions without using conventional TCS.

To trace the cell behaviors responding to the attractants, an agar-drop assay was used, in which the density and movements of cells around an agar drop containing the attractant were traced by optical microscopy (Islam et al., 2014). The solutions containing 1% agar were placed at the center of a chamber that was assembled using a glass slide, coverslip, and double-sided tapes (Figure 3A). Then, the remaining space of the chamber was filled with the cells suspended in PSF containing 0.6% MC. After 40 min at RT, the cell images were video recorded, and the numbers and the behaviors of cells were analyzed at various positions from the agar drop. We examined the same set of chemicals with those examined in the capillary assay and found that the cells distributed differently depending on the chemicals, although they distributed evenly throughout when the drop contained only PSF (Figure 3B). In the presence of attractants, the cell density at the position immediately adjacent to the agar drop was lower than that without attractant and increased with distance from the agar drop, reaching a peak at a point that may represent an optimal concentration in the diffusion gradient of the attractant. The peak positions were around 1.5, 1.5, 2.2, and 2.5 mm, respectively, for 150 mM serine, 150 mM proline, 100 mM methionine, and 200 mM glucose, suggesting that the cells came to the optimal concentration for each attractant (Figure 3B). These results are not inconsistent with the above results that the cells moved into the capillary filled by agar (Figure 2), because higher concentrations of chemicals were applied to the agar drop assay. Next, we examined the time dependency of the cell distribution (Figure 3C). The cell density at different positions from the agar drop containing 100 mM methionine was examined after various times. The cells distributed evenly at time 0, and then formed a density peak with time, and the peak moved. This observation shows that the cells moved synchronously with the diffusion of methionine, by tracing the most optimal concentration as it diffused outward from the agar drop. Then, we examined the cell density at various positions, with 10 and 100 mM methionine, the most effective attractant at 40 min (Figure 3D). The cells distributed with a density peak at 0.3 and 1.1 mm in the presence of 10 and 100 mM methionine, respectively, showing that the cells moved to the zones of the optimal concentration of methionine. A similar result was obtained also for glucose (Figure 3E). Next, we traced the swimming of individual cells at three positions, 0.5, 1.0, and 3.0 mm, from the agar drop containing 100 mM methionine (Figure 3F). The cells showed longer trajectories in the low-density area than in the high-density area. These results suggest that S. eriocheiris cells can sense attractive chemicals and swim to the most favorable concentration by changing their swimming behaviors.

Most motile bacteria have cell polarity and exhibit chemotactic activity by switching moving directions (Adler, 1966; Porter et al., 2011; Typas and Sourjik, 2015). Generally, flagellated bacteria sense the change in concentrations of attractants and repellents with time, and control the reversal frequency between counterclockwise (CCW) and clockwise (CW) in flagella rotation. Interestingly, S. eriocheiris cells have polarity defined as a tapered end, also called tip, which is observed in a few species (Garnier et al., 1981; Ammar et al., 2004; Wang et al., 2004a). Cells mostly swam in the direction of the tip (Figure 4A, Movies 1, 2). For example, 30 in 35 swimming cells moved in the direction of the tip in 1 s. In addition, the cells changed swimming directions frequently (Figure 4B and Movie 3). The difference in the trajectory lengths among different positions in the agar drop assay likely resulted from the reversal frequency of the swimming direction, because the swimming speeds measured every 1 s did not change significantly by the cell position in the chamber (Figure 3D). To examine this assumption, we counted the reversal events of cells in different conditions. The frequency of reversals clearly depended on the methionine concentration as 13, 20, and 29 reversal events were observed for 0, 3, and 30 mM of methionine, respectively. The cells did not show any adaptive response in 4 min, namely the reversal frequency did not change significantly (Figure 4C). The frequency depended on the aerobic condition for cultivation (Figures 4D,E, Movies 1, 4), consistent with the results of the capillary assay shown in Figure 2.

Next, we tried a tethered cell assay, in which the behavior of a single cell fixed on a solid surface was analyzed. We inserted the cells suspended in PSF into a tunnel slide, and found a cell occasionally attached to the glass surface (Figure 4F). We replaced the PSF in the tunnel slide with fresh PSF containing 20 μM methionine, started to record the cell images as soon as possible, and analyzed the reversal behaviors. When the cells were exposed to methionine, they switched their behaviors from one to both directions, demonstrating reversal behavior (Figure 4F and Movie 5). The kymograph of a cell image shows two reversals at 1.0 and 2.3 s after the cell was exposed to methionine at time 0 (Figure 4G).

The EM images of negatively-stained intact cells revealed helical cell shapes (Figure 5A) with dimensions 2.0 to 10.0 μm long, 77.6 ± 8.8 nm wide, 806.7 ± 140.4 nm helix pitch, and 228.2 ± 55.3 nm helix diameter (n = 200) (Figures 5B–D). A pole of an intact cell was featured with a tapered tip structure (Figures 5A,B), as observed by optical microscopy (Figure 4A). The length of an intact tip structure distributed around 379.0 ± 115.4 nm (n = 110; Figure 5E).

To clarify the architecture of the internal structure of S. eriocheiris, we observed the cell after treatments with various concentrations of Triton X-100 (Figures 5F–J). After treatment with 0.05% Triton X-100, some filamentous structures were observed on the surface and in the cell (Figure 5G). Next, we treated the cells with 0.1% (Figures 5H,J) and 0.3% (Figure 5I) Triton X-100. After treatment with 0.3% Triton X-100, the major parts of cell membrane were removed, and the intracellular structures were exposed (Figures 5I,K). We observed that the internal structure of a spiroplasma cell is composed of a dumbbell-shaped structure and a flat ribbon-like structure. The dumbbell-shaped structure at the tip was connected by a flat ribbon that traces a short line in the helical cell from the tip to the other pole (Figure 5K). Interestingly, one spiral filament was found around the dumbbell structure (Figure 6A). The length of the dumbbell was approximately 237.1 ± 101.7 nm (n = 110; Figures 6B–D). The ribbons were composed of five to eight filaments (Figures 6E,F), consistent with a previous study of S. melliferum (Kürner et al., 2005; Cohen-Krausz et al., 2011). Next, to obtain clearer images of the filaments, we treated the cells with 0.8% Triton X-100, recovered the insoluble fraction by centrifugation at 19,000 × g for 40 min, and observed the fraction by EM (Figures 6G–I). The filamentous structures were clearly visible as connected ring units. Image averaging showed that the filaments can be classified into three structures, namely square, oval and ring, averaged for 753, 1045, and 857 particles, respectively (Figure 6J). The dimensions were 8.49 long, 9.25 nm wide with 9.69 nm intervals for square; 9.21 long, 9.58 nm wide with 12.02 nm intervals for oval; and 9.09 long, 9.77 nm wide with 12.23 nm intervals for ring. The oval and the ring are consistent with previous studies of S. melliferum (Cohen-Krausz et al., 2011) but the square was only identified in this study. More experimental information is needed to clarify the relationship among these structures.

To identify the protein components of the internal structures, the cells were treated with various concentrations of Triton X-100. The insoluble fractions were collected and analyzed by SDS-PAGE (Figure 7). Fewer proteins remained in the insoluble fractions after the treatments with higher concentrations of Triton X-100. The 14 protein bands in the insoluble fraction after 0.8% Triton X-100 treatment were excised, and subjected to protein identification by MALDI-TOF mass spectrometry. Sixteen proteins were identified (Table 1), including Fibril protein (SPE-0666) (Townsend et al., 1980; Williamson et al., 1991) and four MreB proteins (SPE-0470, SPE-1224, SPE-1230, and SPE-1231) (Ku et al., 2014). The band positioning between m and n was not identified because of weak mass spectrometry signals, which may be caused by the mixture of source proteins. Nine protein bands through i to ix from the whole cell lysate at the corresponding positions with those in the insoluble fractions were also analyzed by mass spectrometry (Table S3). DnaK (SPE-0178) detected as band d might be a contamination because it was abundant in the whole cell fraction. The other 15 proteins should be involved in the filaments observed by EM as the components. The identified proteins were classified into five categories, i.e., (I) probable subunits of filaments including Fibril and MreB proteins, colored red in Figure 7 and Table 1; (II) proteins containing transmembrane segments, colored green; (III) lipoproteins, colored blue; (IV) proteins predicted for ATPase activity, colored purple; and (V) other, colored black.

The polarized cell morphology and the tip structure have been previously reported for spiroplasmas (Garnier et al., 1981; Ammar et al., 2004). In addition, the relationship of the cell polarity to the cell division cycle has been reported (Garnier et al., 1981), but the relationship between the tip and swimming direction has not been examined. In the present study, we clearly showed that S. eriocheiris cells swim in the direction of the tip, at least for the most part (Figure 4A and Movie 2). Considering that spiroplasma swims by the switch of helicity traveling from front to back (Shaevitz et al., 2005; Wada and Netz, 2007, 2009), the tip may play a role in changing the helicity.

Moreover, we showed that the reversal of the moving direction is linked to chemotactic activity (Figure 4). The reversal motion may be common amongst the species since it was previously observed that the chemotaxis of S. melliferum is linked to the “twitch frequency” (Daniels et al., 1980; Daniels and Longland, 1984). The chemotaxis to phosphatidylcholine found here in S. eriocheiris may be evidence of neurotropic characteristics (Figure 2) (Wang et al., 2011). Interestingly, the chemotaxis of S. eriocheiris was more obvious when it was cultivated under anaerobic conditions vs. aerobic ones (Figures 2, 4). This may be the cause of the phenomenon that S. eriocheiris is more epidemic in summer when the oxygen concentration in fresh water of their habitat is lower than other seasons (Benson and Krause, 1984; Wang et al., 2004b). The anaerobic conditions may induce the expression of genes essential for the chemotaxis or switch the unknown chemotaxis mechanism in S. eriocheiris.

Even though genes for TCS were not found in the genome of spiroplasmas, S. eriocheiris behaves in a similar way with respect to chemotaxis as other motile bacteria; the bacteria manage to reach and/or avoid chemicals by changing the reversal frequency of propulsion (Porter et al., 2011; Typas and Sourjik, 2015). These facts can be explained by two possibilities, (i) spiroplasmas also have a TCS distantly related to other motile bacteria, or (ii) they perform chemotaxis by a mechanism different from other motile bacteria.

In the present study, we identified an internal structure in the tip, designated as dumbbell, by Triton X-100 extraction (Figures 5, 6A). Although this structure is reminiscent of the core structure of the M. pneumoniae tip (Nakane et al., 2015; Kawamoto et al., 2016), none of the protein encoding genes in the S. eriocheiris genome showed sequence similarities to the 15 components of the M. pneumoniae tip structure (Figure 7 and Table 1; Nakane et al., 2015; Miyata and Hamaguchi, 2016). Some groups of mycoplasmas are known to bind to sialylated oligosaccharides on solid surfaces through the tip structure (Kasai et al., 2013, 2015). The spiroplasma tip may have a similar role to those of mycoplasmas, because spiroplasmas appear to bind to host cells through the tip in sectioning images (Ammar et al., 2004).

Our results confirmed that Fibril, an abundant internal protein forms the ribbon, consistent with previous studies of S. melliferum (Kürner et al., 2005; Trachtenberg et al., 2008). We also showed that at least four of five MreB proteins encoded by the genome were found in the insoluble fraction (Figure 7 and Table 1), consistent with a previous study which identified an MreB protein in an insoluble fraction of S. melliferum (Trachtenberg et al., 2008). Fibril and the four MreB proteins may be involved in the ribbon structure. Interestingly, five classes of MreB proteins were found in all Spiroplasma genomes sequenced, although most bacteria have only one MreB homolog and Mollicute species other than spiroplasmas do not have MreB genes (Table S4), suggesting that the combination between Fibril and MreB proteins is essential for the unique swimming of spiroplasmas (Ku et al., 2014).

The other 11 proteins that remained after 0.8% Triton X-100 treatment may also be components of the internal structure. However, we need additional experiments to conclude this, because the proteins containing transmembrane segments detected as protein bands a, c, and g may remain in the fraction because of their insolubility. The proteins predicted for ATPase activity may be candidates for the motor of spiroplasma swimming, although the direct energy source has not been specified so far.

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.