|Home | About | Journals | Submit | Contact Us | Français|
The filamentous cyanobacterium Anabaena sp. strain PCC 7120 can form heterocysts for N2 fixation. Initiation of heterocyst differentiation depends on mutual regulation of ntcA and hetR. Control of hetR expression by NtcA is partially mediated by nrrA, but other factors must be involved in this regulation. Anabaena has two closely related PP2C-type protein phosphatases, PrpJ1 (formerly PrpJ) and PrpJ2; PrpJ1 is involved in heterocyst maturation. In this study, we show that PrpJ2, like PrpJ1, has Mn2+-dependent phosphatase activity. We further demonstrate that whereas prpJ2 is dispensable for cell growth under different nitrogen regimens tested, a double mutant with both prpJ1 and prpJ2 disrupted did not initiate heterocyst differentiation. Ectopic expression of hetR in the double mutant could rescue the failure to initiate heterocyst development, but the heterocysts formed, like those of a prpJ1 single mutant, were not mature. The expression of prpJ2 was enhanced during heterocyst development, and the upregulation of the gene was directly under the control of NtcA. Upregulation of both ntcA and hetR was affected in the double mutant. We propose that PrpJ1 and PrpJ2 together are required for mutual regulation of ntcA and hetR and are thus involved in regulation of the initiation of heterocyst differentiation.
Many cyanobacteria can fix N2 when combined nitrogen sources become limiting in the growth medium. The nitrogenase enzymatic complex responsible for nitrogen fixation is very sensitive to oxygen, and oxygen is produced by photosynthesis by cyanobacteria. The strategy used by some filamentous diazotrophic cyanobacteria to resolve this oxygen paradox is to perform photosynthesis and nitrogen fixation in two distinct cell types, differentiated cells called heterocysts that provide a microoxic environment for nitrogenase and vegetative cells which perform oxygenic photosynthesis (22, 36, 39). One such organism is Anabaena sp. strain PCC 7120. In this strain, heterocysts account for 5 to 10% of the cells and appear in a semiregular pattern along each filament. Therefore, the process of heterocyst differentiation provides a prokaryotic model to study developmental pattern formation. Three factors account for the microoxic environment in heterocysts: the heterocyst envelope composed of an inner layer of glycolipid surrounded by an outer layer of polysaccharides that limits oxygen penetration, the lack of oxygen-producing photosystem II, and an increased rate of respiration to consume oxygen (36).
The initiation of heterocyst differentiation and the formation of the heterocyst pattern are governed by multiple signals and the concerted actions of several proteins as positive or negative regulators (for a recent review, see 39). The accumulation of 2-oxoglutarate following limitation of combined nitrogen is a trigger that initiates heterocyst development by stimulating the DNA-binding activity of NtcA, a protein involved in the regulation of carbon and nitrogen metabolism, as well as initiation of heterocyst differentiation (7, 12, 13, 18, 20, 32, 35). HetR, a protease with DNA-binding activity, plays a central role in the early steps of heterocyst differentiation (14, 40). Both ntcA and hetR are autoregulated, and the expression of hetR and the expression of ntcA are mutually dependent because upregulation of one of theses genes is dependent on the other gene (3, 4, 23). How HetR regulates the expression of ntcA remains unknown. No NtcA-binding site has been found in the upstream region of hetR, and the regulation of hetR by NtcA could be partially due to the action of the response regulator NrrA (8, 9, 24). However, NrrA cannot be the only link between ntcA and hetR, because when nrrA was inactivated, both heterocyst differentiation and hetR upregulation were only delayed (8). Indeed, ccbP, encoding a calcium-binding protein, is regulated by NtcA, and it has been proposed that the pool of calcium affects the activity of HetR (31).
The genome of Anabaena sp. strain PCC 7120 contains a large number of genes encoding two-component signaling systems, protein Ser/Thr and/or Tyr kinases, and phosphatases, including eight genes encoding PP2C-type Ser/Thr phosphatases (16, 26, 34, 38). Some of these genes are involved in heterocyst development, mostly in heterocyst maturation and functioning (8, 11, 17, 19, 21, 25, 30, 37). We have shown previously that PrpJ is a PP2C-type protein phosphatase located on the plasma membrane (15). A prpJ1 mutant (strain S20) failed to grow under diazotrophic conditions and formed heterocysts lacking the major heterocyst-specific glycolipid (HGL), in contrast to other mutants whose mutations affect either the synthesis or the deposition of both the major and minor HGLs (1, 2, 10, 28) or only the minor HGL (30). Therefore, PrpJ represents a new regulatory branch for heterocyst maturation, possibly involving regulation of only a subset of genes involved in glycolipid synthesis. These observations indicate that multiple input pathways participate in the maturation of heterocysts. When proheterocysts were formed, filaments of the prpJ1 mutant, fragmented extensively at the junctions between proheterocysts and vegetative cells, resulting in free nonmature heterocysts and filaments that were 11 cells long on average (15).
Open reading frame all2470 encodes one member of the PP2C family of protein phosphatases in Anabaena sp. strain PCC 7120 (35). The deduced amino acid sequence of All2470 is similar to that of PrpJ, and these two proteins have similar architectures, with an N-terminal domain having an unknown function, a central domain similar to the catalytic domains of PP2C-type protein phosphatases, and a C-terminal domain with a putative transmembrane motif (Fig. (Fig.1).1). The amino acid sequences of these two proteins share 40% identity overall, and their catalytic domains are 45% identical. Because these two protein phosphatases are very similar, here we use the designations PrpJ1 (formerly PrpJ) for All1731 and PrpJ2 for All2470. In the present study, we show that PrpJ1 and PrpJ2 are involved in the initiation of heterocyst differentiation by acting on the mutual regulation of ntcA and hetR.
Anabaena sp. strain PCC 7120 and its derivatives were grown in BG11 with 18 mM NaNO3 (buffered with 10 mM HEPES, pH 7.5) as the combined nitrogen source or in BG110 without a source of combined nitrogen, as described previously (15). Cell growth was followed by monitoring the optical density at 700 nm. Strains of Escherichia coli were grown on LB medium. The ntcA mutant and hetR mutant hetR216 have been described previously (4, 12).
To inactivate nrrA, two PCR primers shown in Table Table11 were used to amplify a DNA fragment, which was followed by insertion of the Ω fragment (27) conferring resistance to streptomycin and spectinomycin into the EcoRV site (located 264 bp after the ATG sequence corresponding to the initiation codon in the coding region). After conjugation and selection, double recombinants were obtained (data not shown). To construct the plasmid used for inactivation of prpJ2, forward and reverse PCR primers incorporating PstI and XhoI restriction sites, respectively (Table (Table1),1), were designed. After PCR amplification of the coding region of prpJ2, the PCR product was cut with PstI and XhoI and cloned into pBluescript SK(−) (Stratagene) disrupted at a unique EcoRV site (located 1,266 bp after the ATG sequence corresponding to the initiation codon) by insertion of a neomycin resistance cassette (33). The whole insert was then cloned into the SacI/XhoI sites of the integrative vector pRL271 (6), resulting in plasmid pM19. This plasmid was transferred by conjugation into Anabaena sp. strain PCC 7120 and into the prpJ1 null mutant (S20) to obtain the single mutant MprpJ2 (S19) and the double mutant MprpJ1prpJ2, respectively. Complete segregation of the mutants was confirmed by PCR and reverse transcription (RT)-PCR using primers S19neo-f and S19neo-r (Table (Table1),1), which target regions of prpJ2 outside the resistance cassette.
For complementation of mutants, a DNA fragment covering the coding region of prpJ2 and the 266-bp upstream region was obtained after PCR using genomic DNA and two primers listed in Table Table11 (S19 into DM-f and S19 into DM-r), and the PCR product was cloned into the vector pCR 2.1-TOPO (Invitrogen). The accuracy of the sequence was confirmed by DNA sequencing. The DNA fragment was then subcloned into the XbaI/SacI sites of replicative plasmid pRL1272 to obtain plasmid pC19. pC19 was mobilized into the double mutant MprpJ1prpJ2 by conjugation, and the antibiotics neomycin, spectinomycin, and erythromycin were used for selection. For overexpression of the hetR gene, the coding region of hetR was amplified by PCR using primers hetR-f and hetR-r (Table (Table1).1). The amplified PCR product was introduced into the NdeI/EcoRI sites of the vector pBS-PetE downstream of the promoter of the petE gene (20). The hetR coding region together with the petE promoter was transferred to the NotI/HindIII sites of pRL1272, and the final construct was designated pPH1. The hetR gene with its own promoter region was also amplified by PCR and cloned into the SacI/KpnI sites of pRL1272 to obtain pPH2.
To produce the maltose-binding protein (MBP)-PrpJ2 fusion product, the prpJ2 gene was amplified by PCR with genomic DNA of Anabaena sp. strain PCC 7120 as the substrate using primers S19pMAL-f and S19pMAL-r (Table (Table1).1). The 1.96-kb PCR product was cloned into the BamHI/PstI sites of the pMAL-c2X vector to generate pMP. The accuracy of the cloned PCR product was confirmed by DNA sequencing. Protein overexpression was carried out as described previously for production of MBP-PrpJ1 (15). Crude cell extracts were centrifuged at 15,000 × g for 15 min. The soluble protein fractions were collected and mixed with amylose resin (New England Biolabs) for 1 h at 4°C. MBP-protein fusions were eluted with 10 mM maltose and dialyzed against phosphate-buffered saline. The phosphatase activity of PrpJ2 was assayed using several phosphorylated substrates as previously described (15).
RNA was prepared and real-time quantitative RT-PCR experiments were performed as described previously (30). All oligonucleotide primers used in these assays are listed in Table Table1.1. HetR antibody (used at a 1:2,500 dilution) was a kind gift from Liu Juan (Huazhong Agricultural University, People's Republic of China). Cells of the wild type or the double mutant were grown in BG11 to exponential phase and then transferred into BG110 medium without combined nitrogen and incubated for 1.5, 3, 6, 12, and 24 h.
For EMSA, the promoter region of prpJ2 was amplified by PCR using a pair of primers (prpJ2-f2 and prpJ2-r2) listed in Table Table1.1. The assays were then performed using previously described procedures (7). To mutate the NtcA-binding site in the upstream region of prpJ2, a PCR was carried out using primers prpJ2-f3 and prpJ2-r3 (Table (Table1)1) and then another PCR was carried out using primers prpJ2-f2 and prpJ2-r2.
Since PrpJ2 is highly similar to PrpJ1 (Fig. (Fig.1),1), we first tested the enzymatic activity of PrpJ2 and compared this activity with that of PrpJ1. PrpJ2 was purified as a fusion product with MBP. MBP-PrpJ2 was active when p-nitrophenylphosphate (pNPP) was used as the substrate (Fig. (Fig.2A),2A), and the activity was dependent on the divalent cation Mn2+. The activity of PrpJ2 reached a plateau when the concentration of Mn2+ was 10 mM (data not shown). PrpJ2 was also able to dephosphorylate synthetic phosphopeptides in which either phospho-Tyr or phospho-Thr was embedded in the peptide (Fig. (Fig.2B).2B). The specific activity of PrpJ2 was significantly higher with peptides containing phospho-Tyr than with peptides containing phospho-Thr. PrpJ2 could also dephosphorylate phospho-casein (Table (Table2).2). The phosphatase activity of PrpJ2 was resistant to both vanadate (an inhibitor of Tyr phosphatase) and nodularin (an inhibitor of protein phosphatases 1, 2A, and 2B). Thus, the overall characteristics of PrpJ2 were similar to those of PrpJ1 (14) (Table (Table2).2). From these results, we conclude that both PrpJ1 and PrpJ2 are PP2C-type protein phosphatases. However, PrpJ1 is more efficient than PrpJ2 in catalyzing dephosphorylation; when pNPP was used as a substrate, PrpJ1 displayed a lower Km value than PrpJ2 (0.77 versus 2.6 mM) and a much higher kcat value (0.95 versus 0.03 s−1).
The prpJ2 gene was cloned and disrupted by inserting a cassette conferring resistance to neomycin at an EcoRV site in a region corresponding to the catalytic domain of PrpJ2 (Fig. (Fig.1B).1B). The corresponding DNA fragment carried on the integrative plasmid pRL271 was transferred by conjugation into both the wild-type strain and the S20 mutant in which prpJ1 was inactivated by insertion of a DNA cassette conferring resistance to spectinomycin (15). Both the prpJ2 single mutant (MprpJ2) and the prpJ1 prpJ2 double mutant (MprpJ1prpJ2) were fully segregated, as demonstrated using PCR with oligonucleotide primers targeting either side of the antibiotic resistance cassette (data not shown). The MprpJ2 mutant grew like the wild type when different nitrogen sources were used (ammonium, nitrate, or N2). The heterocyst frequency and pattern were also not affected in this mutant (Fig. (Fig.3).3). Thus, the prpJ2 gene was dispensable for cell growth under the conditions tested.
The MprpJ1prpJ2 double mutant showed growth behaviors similar to those of the S20 mutant under similar culture conditions. Both mutants could use a combined nitrogen source such as nitrate or ammonium for growth but did not grow in the absence of a combined nitrogen source (data not shown). However, there were several important differences between the phenotypes of these two mutants. First, when the organisms were cultured under nitrogen fixation conditions, no heterocysts were observed for the MprpJ1prpJ2 double mutant even after staining with alcian blue, which binds specifically to the polysaccharide layer of heterocysts (Fig. (Fig.3),3), while premature heterocysts could be distinguished in the S20 mutant culture and were clearly stained with alcian blue (15). Second, filaments of the double mutant were not fragmented at 24 h after removal of combined nitrogen from the growth medium and lacked any sign of heterocyst differentiation, as described below; in contrast, filaments of the S20 mutant were extensively fragmented at the junctions between proheterocysts and vegetative cells at the same time and lacked a major heterocyst-specific glycolipid (15). Filaments of Anabaena sp. strain PCC 7120 break more easily at the junctions between heterocysts and vegetative cells, and the breakage occurs more frequently in mutants with partially developed heterocysts; in contrast, some mutants, such as ntcA or hetR mutants, which show no sign of differentiation, have long filaments (4, 12, 35). Taken together, our results suggest that heterocyst differentiation was affected at a earlier stage in the double mutant than that in the S20 single mutant (see below).
prpJ2 present on replicative plasmid pRL1272 (construct pC19) was transferred into the MprpJ1prpJ2 double mutant by conjugation. The strain obtained was still unable to grow in the absence of a combined nitrogen source. However, partially developed heterocysts, which were hardly distinguishable with a light microscope, could be efficiently stained using alcian blue (data not shown). Therefore, this phenotype resembled that of the S20 mutant (15), as expected, indicating that prpJ2 could complement a prpJ2 mutation. However, when the same plasmid was transferred into mutant S20, the phenotype was unchanged, suggesting that the function of prpJ1 could not be restored by the presence of extra copies of prpJ2.
HetR is a central regulator for initiation of heterocyst differentiation (4, 5). Since in the double mutant there was no detectable sign of heterocyst differentiation, we examined whether the ectopic expression of hetR could rescue the failure of double mutant MprpJ1prpJ2 to initiate heterocyst differentiation (Fig. (Fig.4).4). Plasmid pPH1 bearing hetR under control of the copper-inducible petE promoter was transferred into both the wild type and double mutant MprpJ1prpJ2. The wild-type strain bearing pPH1 frequently formed double heterocysts, although the increase in the number of heterocysts was not as dramatic as that reported previously (5), possibly due to the difference in the genetic backgrounds of the strains used in different laboratories. Under nitrogen fixation conditions, the growth rate was slightly lower than that of the wild type without extra copies of the hetR gene (data not shown). The double mutant bearing pPH1 could also form heterocysts, as shown in Fig. Fig.4;4; however, the heterocysts were not mature and lacked the polar granules that consist of the nitrogen reserve compound cyanophycin. Therefore, the ectopic expression of hetR restored the capacity of the double mutant to initiate the heterocyst differentiation process but did not rescue its failure to grow under diazotrophic conditions (Fig. (Fig.44 and data not shown). These results were consistent with the role of prpJ1 in heterocyst maturation, as we showed previously that when prpJ1 was inactivated, initiation of heterocyst differentiation still occurred but the differentiation stopped at a proheterocyst formation step (15). Similar results were obtained using pPH2, a plasmid with hetR under control of its own promoter region (data not shown). Therefore, extra copies of hetR restored the initiation of heterocyst differentiation in the double mutant.
The experimental results described above strongly indicate that hetR expression may be affected in the double mutant. Since hetR and ntcA are dependent on each other for expression and nrrA is partially responsible for the regulation of expression, we studied the expression of hetR, ntcA, and nrrA in the wild type and the double mutant (Fig. (Fig.5).5). Total RNAs were extracted from the wild type and the double mutant at 0, 1.5, 3, 6, 12, and 24 h after the removal of combined nitrogen, and real-time quantitative RT-PCR was performed to examine gene expression. The expression of hetR was enhanced 1.5 h after induction of heterocyst differentiation and reached a peak 6 h after induction, which was followed by a decline over time. The upregulation of hetR was strongly affected in the double mutant; there was a gradual increase but not the burst of expression seen in the wild type. The slow and gradual increase in the hetR expression level may be attributed to the autoregulation of hetR or the action of NrrA (Fig. (Fig.5A).5A). Similar results were obtained using antibodies against HetR, although the HetR protein levels were high even 24 h after induction in the wild type, while in the double mutant the level of the HetR protein increased slightly in the first few hours and then dropped to the basal level (data not shown). Similarly, the upregulation of ntcA observed in the wild type did not occur in the double mutant; we observed only a slight increase in the expression level of ntcA that might be attributed to autoregulation of ntcA (Fig. (Fig.5B).5B). In contrast, upregulation of nrrA was observed in both the wild type and the mutant (Fig. (Fig.5C).5C). Thus, the regulation of ntcA and hetR was strongly affected in the double mutant, while nrrA was expressed normally in the double mutant.
The strong negative effect on the expression of both ntcA and hetR observed in the double mutant could be explained in two ways. prpJ1 and prpJ2 could act either between, or at a step prior to, ntcA and hetR in the regulation of heterocyst initiation. We have reported previously that prpJ1 was expressed constitutively during the course of heterocyst differentiation (15). Thus, we examined whether the expression of prpJ2 could be regulated by removing combined nitrogen. We found that that there is a DNA motif (GTAGCCACAGGTAC) in the upstream region of prpJ2 that corresponds to the consensus binding site of NtcA, GTAN8TAC (13) (Fig. (Fig.6A).6A). Thus, we tested whether NtcA could bind to this DNA region by performing EMSA. As a control, we used the glnA promoter that was known to be regulated by NtcA (12, 35). This experiment showed that NtcA could bind to the upstream region of prpJ2, like it binds to the glnA promoter region (Fig. (Fig.6B).6B). To make sure that this binding activity was due to the presence of the NtcA-binding site in the upstream region of prpJ2, we replaced, in the same DNA fragment, the consensus binding site with CCGGCCACAGGGCA (the mutations are underlined). This DNA fragment did not interact with NtcA (Fig. (Fig.6B),6B), thus demonstrating that NtcA interacted specifically with this site.
To further confirm the regulation of expression of prpJ2 by NtcA, we compared the expression profiles of prpJ2 for the wild type and a ntcA mutant (Fig. (Fig.7).7). RNAs were extracted at 0, 1.5, 3, 6, 12, and 24 h after the combined nitrogen was removed from both strains. The results indicated that the mRNA levels for prpJ2 increased rapidly following the induction of heterocyst differentiation, reaching a peak 3 h after the combined nitrogen was removed, and then declined gradually. This result shows that the expression of prpJ2 is enhanced in the early phase of heterocyst differentiation. Induction of prpJ2 expression was not observed in the ntcA mutant under these conditions. These results, together with those obtained with the DNA-binding assay, indicate that NtcA directly regulates the expression of prpJ2. For comparison, the expression of prpJ2 in a hetR mutant strain (hetR216) and that in a nrrA mutant were also examined (Fig. (Fig.7).7). prpJ2 was not induced in the hetR mutant, consistent with the mutual dependence of ntcA and hetR expression (23). However, the prpJ2 expression profiles were similar for the wild type and the nrrA mutant. Thus, the expression of nrrA is not affected in double mutant MprpJ1prpJ2, nor is the expression of prpJ2 affected by inactivation of nrrA.
PrpJ1 and PrpJ2 are similar both at the level of the amino acid sequence and at the level of domain architecture. The only major difference between these two proteins is that PrpJ1 has a C-terminal extension that is not present in PrpJ2 (Fig. (Fig.1).1). We previously truncated this C-terminal region of PrpJ1 and found that this deletion had little effect on the phosphatase activity of PrpJ1 and on the function of PrpJ1 in vivo (15). In the present study, we compared the functions of PrpJ1 and PrpJ2. Both PrpJ1 and PrpJ2 could hydrolyze different phosphorylated substrates and displayed enzymatic activities characteristic of PP2C-type protein phosphatases, although the kcat values of the two enzymes differed by a factor of 31. When prpJ2 alone was inactivated, little effect on the growth was found under different nitrogen regimens, indicating that prpJ2 was not essential for cell growth under these conditions. For comparison, the S20 mutant failed to develop mature heterocysts and to grow under diazotrophic conditions. When both prpJ1 and prpJ2 were inactivated, the resulting double mutant, MprpJ1prpJ2, did not initiate differentiation of heterocysts. Several experimental results support this conclusion. First, no morphological sign of heterocyst differentiation was observed, even after staining with alcian blue specific for the polysaccharide layer of heterocysts. Second, expression of ntcA and hetR, both of which are critical for the initiation of heterocyst differentiation, was strongly affected in the double mutant. Consistent with this observation, ectopic expression of hetR in the double mutant could lead to the appearance of partially developed heterocysts which, like those found in the MprpJ1 mutant, could be visualized better by staining with alcian blue. Thus, although the ectopic expression of hetR did not lead to the formation of functional heterocysts that supported cell growth under diazotrophic conditions because of the lack of prpJ1, the early initiation steps were successfully restored. Since prpJ1 was necessary for heterocyst maturation, in particular the synthesis of one major heterocyst glycolipid (15), it is thought that the expression of hetR alone in the double mutant did not allow the formation of fully functional heterocysts. Together, these results established that prpJ1 and prpJ2 are new regulators that are involved in the initiation of heterocyst differentiation. If prpJ1 and prpJ2 acted downstream of hetR in the regulatory cascade during heterocyst differentiation, the ectopic expression hetR in the double mutant would have had no effect on the initiation of heterocyst differentiation.
Based on our results and data from other groups, a model involving PrpJ1 and PrpJ2 in the regulation of heterocyst development can be proposed. Our results indicate that NtcA directly regulates the expression of prpJ2. PrpJ1 and PrpJ2 are required for the upregulation of hetR, which in turn is required for ntcA expression (23). Thus, in the double mutant, the expression of both hetR and ntcA is affected. Therefore, prpJ1 and prpJ2 represent a newly discovered missing link between ntcA and hetR regulation. PrpJ1 alone is also involved in heterocyst maturation by regulating the synthesis of the major heterocyst glycolipid (15). How PrpJ1 and PrpJ2 regulate hetR expression is unknown. The simplest hypothesis is that HetR is phosphorylated and is a target for dephosphorylation by PrpJ1 and PrpJ2. We could not confirm this hypothesis by testing the phosphorylation of HetR. We used two strategies to study HetR phosphorylation. The first strategy was immunoprecipitation of HetR after radioactive labeling of the culture in the presence of radioactive phosphate. The protein band recognized by the antibodies was not labeled by radioactivity (data not shown). The second strategy was detection of HetR after immunoprecipitation by antibodies against phospho-Ser, phospho-Thr, or phospho-Tyr, because PrpJ1 and PrpJ2 are both PP2C-type protein phosphatases. We found no cross-reactivity of HetR with these antibodies. As no DNA-binding domain is present in PrpJ1 and PrpJ2, these two protein phosphatases should regulate the expression of hetR indirectly through their action on a phosphoprotein that remains to be identified.
Whereas prpJ1 and prpJ2 may cross talk in regulating hetR expression or HetR activity, these two genes must have distinct functions in late stages of heterocyst development. Indeed, prpJ2 on a replicative plasmid did not complement the phenotype of mutant S20; similarly, the same replicative plasmid bearing prpJ2 in double mutant MprpJ1prpJ2 allowed initiation of heterocyst differentiation but not maturation of heterocysts, which still required prpJ1. Therefore, these two genes are similar but not interchangeable. We have two possible explanations for the difference in the functions of prpJ1 and prpJ2. First, these two genes display different regulatory patterns; prpJ1 is constitutively expressed, and prpJ2 is induced during heterocyst development. Second, the enzymatic activities of PrpJ1 and PrpJ2 are dramatically different in vitro; PrpJ1 is much more efficient than PrpJ2 in dephosphorylation. Identification of the substrates of these two protein phosphatases would help us further understand their mechanisms of action.
This work was supported by a grant from the Agence National de la Recherche (Physique et Chimie du Vivant), by the CNRS, by the Aix-Marseille Université, and by the Natural Science Foundation of China (grant 30670046).
Published ahead of print on 24 July 2009.