![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Departmento de Biotecnología Microbiana, Centro Nacional de Biotecnología, CSIC,1 Departamento de Biología Molecular, Universidad Autónoma de Madrid, Madrid, Spain,3 Max-Planck-Institut für Molekulare Genetik, Berlin, Germany2
Received 5 February 2004/ Accepted 21 May 2004
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
(ruvA, 
ruvB, recU, and recD) and 
(recG) epistatic groups, in an otherwise rec+ background, render cells impaired in chromosomal segregation. A less-pronounced segregation defect in recA and sms (radA) cells was observed. The repair deficiency of addAB, recO, recR, recH, recS, and subA cells did not correlate with a chromosomal segregation defect. The sensitivity of epistatic group mutants to DNA-damaging agents correlates with ongoing DNA replication at the time of exposure to the agents. The sms (radA) and subA mutations partially suppress the DNA repair defect in ruvA and recD cells and the segregation defect in ruvA and recG cells. The sms (radA) and subA mutations partially suppress the DNA repair defect of recU cells but do not suppress the segregation defect in these cells. The recA mutation suppresses the segregation defect but does not suppress the DNA repair defect in recU cells. These results result suggest that (i) the RuvAB and RecG branch migrating DNA helicases, the RecU Holliday junction (HJ) resolvase, and RecD bias HJ resolution towards noncrossovers and that (ii) Sms (RadA) and SubA proteins might play a role in the stabilization and or processing of HJ intermediates. |
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In Bacillus subtilis, the recombination genes other than recA have been classified into six different epistatic groups (
, ß, , 
, 
, and ). Mutations in genes classified within the (recF, recL, recO, and recR [known collectively as recFLOR] and recN), (recU, recD, and ruvA [formerly termed recB] and ruvB), and (recG) epistatic groups markedly affect the viability of cells exposed to DNA-damaging agents, whereas mutations in genes classified within the ß (addA and addB [collectively known as addAB]), (recH and recP), and (recS) epistatic groups slightly reduce the viability of cells exposed to DNA-damaging agents (reference 16 and this study). The recA, recF, recO, recR, recN, ruvA, ruvB, and recG genes have their counterparts in E. coli in genes with identical names, whereas the addAB, recU, and sms genes have their counterpart in the recBCDEco, ruvCEco, and radAEco genes, respectively (3, 16). The B. subtilis recL, recD, recH, recP, recS, and subA genes have no obvious counterpart in genes in E. coli. The products classified within the , ß, , and groups have their functional counterparts in the RecN-FOREco, RecBCDEco, RuvABCEco, and RecGEco products, respectively (3, 8, 10, 16, 25). The role of the functions classified within the and epistatic groups in DNA repair and HR remains unknown (16). Unless otherwise stated, the indicated genes and products are of B. subtilis origin.
In E. coli cells, 18 to 50% of cells require replication fork reloading during a single round of chromosomal replication in the absence of any exogenous DNA-damaging agent (13, 34). Using an indirect measurement (measurement of repair centers as a measurement of blocked replication forks), we assumed that replication fork reloading might occur with a similar frequency in B. subtilis cells (25). The rate of formation of RecN-RecOF repair centers in the absence of any exogenous DNA-damaging agent was found to be about 35 and 5% in exponentially growing recA and recU cells, respectively (25).
A defect in the HJ resolvase RecU (3) (also termed penicillin-binding protein [PBP]-related factor A [designated PrfA]) or in the DNA organizer SMC complex (formed by the Smc, ScpA, and ScpB proteins) in an otherwise wild-type (wt) background, leads to the accumulation of anucleate cells (
3 and 10%, respectively) (7, 20, 35, 42, 45, 56). The recU smc double mutant does not seem to be viable. Genetic analysis of a synthetic conditional recU mutant combined with the smc mutant at a permissive temperature indicated the accumulation of 24% anucleate cells (45). These data suggest a role for the SMC complex and RecU in chromosomal segregation. Finally, it has been shown that the recU segregation phenotype is greatly exacerbated by the additional loss of PBP1 but not by the loss of other PBPs (e.g., PBP2c or PBP4), suggesting a possible role for recU in septum formation or as a chaperone in DNA-cell wall interaction (24, 45). Furthermore, genetic evidence suggests that the sms (also termed radA) and subA mutations partially suppress the DNA repair and recombination defect of epistatic group mutants (8).
In this paper, we analyze the effect on segregation of the different repair-deficient B. subtilis epistatic groups, as well as the putative suppression of the segregation phenotypes by the sms (radA) and subA mutations. Our results indicate that the functions of genes classified within the and epistatic groups, which are involved in the processing of an HJ, are required for proper chromosomal segregation in wt cells under normal growth conditions. It is likely that the replication and subsequent segregation of chromosomes bearing unrepaired DNA lesions can seriously compromise genome stability. This is consistent with the hypothesis that B. subtilis RuvAB-RecU-RecD and RecG proteins in an otherwise wt background under normal growth conditions (this work) and E. coli RuvABC proteins in UV-irradiated, rep or recBC sbcBC backgrounds (22, 36, 41) prevent dimer formation in vivo. Finally, the suppression of the segregation defect of HJ processing functions by sms (radA) and sub mutations point to the role for both proteins in the stabilization or processing of branched DNA molecules.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
recG and ruvAB strains. Their isogenic rec-deficient derivatives, as well as the recA recU and recU recO double mutants, were generated by a double-crossover event as previously described (1). Expression of the ß gene mediated deletion of the cat gene. The attSKIN and attPBSX (62) regions were moved into wt and recU backgrounds by chromosomal transformation as previously described (2).
|
Fluorescence and electron microscopy of B. subtilis cells. Exponentially growing cells were obtained by inoculation of overnight cultures in fresh LB medium and growing them to an optical density at 560 nm of 0.4 at 37°C. The mid-log-phase cells were then fixed with 2% formaldehyde, 4',6'-diamino-2-phenylindole (DAPI) (1 µg/ml) was added for nucleoid visualization, and cells were analyzed by fluorescence microscopy as previously described (7). For electron microscopy sectioning, cells were fixed with glutaraldehyde, treated with osmium tetroxide, and embedded in Spurr's low-viscosity medium (57).
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
[recO and recR], ß [addA5 and addB72], [recH342], [recU, ruvAB, and recD41], [recS], and [recG]) as well as the recA strain (Table 2) were collected during exponential phase, and as a measure of a segregation defect, the frequency of anucleate cells was quantified after the cells were stained with DAPI. Anucleate cells in the addA5 addB72, recH342, and recS strains were rare (Table 2). However, diffuse and "linked" nucleoids that occupied almost the whole cell were visible in 6% of addA5 addB72 cells, 10% of recO cells, and 26% of recH342 cells (Fig. 1). Very little is known about the biochemical role of RecH on DNA repair and recombination.
|
|
ruvA, recU, and recG cells and the uncharacterized activity of recD41 cells, all impaired in the processing of HJs, showed a clear defect in chromosomal segregation (Table 2). The most severe segregation defect was found in recG cells: >10% of recG cells had abnormally condensed nucleoids, and 7% of the cells were anucleated (see below) under normal growth conditions.
In all experiments, the lysogenic prophage SKIN encoding a RusA-like HJ resolvase protein (52) was present in the genetic background used. To learn whether the RusA-like protein could play any role in chromosomal segregation, SKIN-free wt and recU strains were constructed. Similar segregation patterns were observed with SKIN-free and SKIN-containing cells (Table 2). The percentage of anucleate cells in the recU mutant that lacks bacterially encoded HJ resolvase was unaffected by the absence of the SKIN prophage (Table 2). Therefore, it is likely that the SKIN-encoded RusA-like protein either is not expressed or has no effect on chromosomal segregation under normal growth conditions.
Unlike a recAEco mutant that shows 10% anucleate cells (64), a recA mutant shows a moderate segregation defect (1% of cells) (Table 2) (28). The presence of the recA null allele in the recU background suppressed the segregation phenotype (Table 2). This is consistent with the observation that in both E. coli and B. subtilis cells, chromosome dimer formation is not observed and the Xer-like site specific recombinase is not needed in the absence of the RecA protein (6, 27, 28).
The RecOEco, RecO, and RecU proteins can catalyze D-loop formation (3, 32). A recO recU double mutant strain was constructed to assess whether the absence of DNA strand invasion could suppress the chromosomal segregation phenotype. The recO recU double mutant strain showed a segregation defect similar to that of the recU single mutant (4% of anucleate cells) (Table 2). These results suggest a strand-invading accessory role for both RecO and RecU proteins and confirm that RecA is primarily responsible for the formation of HJ in vivo.
The DNA damage sensitivity of recU, ruvAB, and recD cells correlates with DNA replication.
Previously, it has been shown that UV-generated DNA damages are removed by the nucleotide excision repair (NER) machinery in E. coli wt cells (12). The NER proteins are involved in the repair of UV-generated DNA damage independently of the replication state of the cell. UV-irradiated cells resume DNA synthesis after a transient inhibition by a process called replication restart that has been shown to involve recFEco, recOEco, and recREco gene products (11). These results suggest a close interplay between recombination repair and DNA replication and suggest that the failure of recFOREco and perhaps recFLOR cells arises from a defect in rescuing a stalled replication fork (12, 13, 26). To learn whether the high sensitivity of recU, ruvA, and recD cells to DNA-damaging agents also correlates with ongoing DNA replication, different assays were undertaken. First, wt, recR13, uvrA42 (uvrA42 is the counterpart to uvrBEco mutants), and recU40 cells were grown in LB medium until mid-exponential or stationary phase and exposed to 100 µM 4NQO for various times, and then the numbers of viable cells were measured. Independently of the growth phase, wt cells were resistant to the killing action of 100 µM 4NQO, whereas uvrA42 cells, deficient in NER, were sensitive (1). As previously reported, exponentially growing recU40 and recR13 cells were sensitive to DNA-damaging agents (1), but stationary-phase recU40 and recR13 cells were 100-fold more resistant to 4NQO (Fig. 2) than were the exponentially growing cells (1, 18). Stationary-phase ruvA2 and recD41 cells were also 100-fold more resistant to 100 µM 4NQO than were exponentially growing cells (8) (data not shown). Furthermore, stationary-phase recU40, ruvA2, recD41, and recR13 cells were also 80- to 100-fold more resistant to other DNA-damaging agents, such as 10 mM MMS, than were exponentially growing cells (data not shown).
|
100-fold more resistant to the DNA-damaging agent than were cells untreated with CM (Fig. 2). Similar results were observed when the ruvA2, ruvAB, recD41, or recO cells were pretreated with CM for 120 min before exposure to 4NQO (data not shown). Pretreated uvrA42 cells were as sensitive as untreated cells (Fig. 2B), whereas wt cells were resistant to 100 µM 4NQO exposure. Therefore, it is likely that (i) the failure of recU40, ruvA2, and recD41 cells arises from a defect in rescuing a stalled replication fork until the lesion can be removed by NER, (ii) the defect of both recREco (11) and recR13 cells correlates with ongoing DNA replication at the time of exposure to the agent, and (iii) DNA-damaged uvrA42 cells do not recover normally independently of the growth phase. Previously, a direct correlation between increased damage sensitivity of recFOREco cells and DNA replication has been established (11). Therefore, it is likely that the rescue of arrested replication forks in exponentially growing cells occurs via HR in both E. coli cells (11, 12) and B. subtilis cells (Fig. 2).
Nucleoid and cell morphology phenotypes of recU, recD, and ruvA cells during exponential-phase growth.
To investigate whether the chromosome segregation defect may be due to a defect in replication fork progression, the wt strain and its isogenic derivatives (recU, recU40, ruvA2, and recD41 cells) were grown to mid-exponential phase in rich medium and either stained with DAPI, fixed, and visualized by fluorescence microscopy or fixed, processed, and visualized by electron microscopy.
Previously, it was shown that 3 to 5% of recU cells have a chromosomal segregation phenotype (45). A similar chromosomal segregation defect was observed with the ruvA, ruvAB, and recD mutants (Fig. 3; Table 2). This observation is consistent with the classification of ruvA, ruvB, and recD in the same epistatic group as recU (3). An absence of DAPI-stained material was observed for 3 to 5% of the ruvA2 and recD41 cells, whereas <0.05% of wt cells were anucleate under identical growth conditions (Fig. 3A). In addition to a higher abundance of cells showing no nucleoids, a high proportion of recU, recU40, ruvA2, and recD41 cells had defects in nucleoid structure. The one or two normally compact, condensed, and regular nucleoid bodies seen in fixed wt cells often appeared as highly condensed nucleoids asymmetrically located in recU40, ruvA2, and recD41 cells (Fig. 3A). recU, ruvA2, and recD41 cells had many more nucleoids of much higher DNA content and with large cytoplasmic spaces free of nucleoid bodies than did wt cells. Similar results were obtained when the ruvAB strain was analyzed.
|
The recU gene maps upstream and forms an operon with ponA, which encodes PBP1. As shown in Fig. 3, the absence of the genetically unlinked recU, ruvAB, and recD genes has the same profound effect on both chromosomal structure and segregation. It is likely, therefore, that the segregation defects observed with recU, recD, and ruvAB cells are unlinked to PBP1 and therefore not due to the PBP1 defect in septation and its localization at sites of cell division (45).
The ruvA2 and recD41 segregation defect is partially suppressed in sms (radA) cells.
Previously it was shown that Sms (RadA), the counterpart of RadAEco, and SubA proteins play an active role in recombinational repair, most likely through the stabilization and/or processing of branched DNA molecules or blocked replication forks (5, 8). Mutations in both proteins partially suppress the recombination defect of mutations in proteins expressed by genes of the epistatic group (8). To learn whether the chromosomal segregation defect of recU, recD41, and ruvA2 cells may be also suppressed by mutations in the sms (radA) and subA genes, we constructed double mutant strains and investigated their segregation phenotypes. The sms (radA), subA, recU sms (radA), recD41 sms (radA), ruvA2 sms (radA), recU subA, recD41 subA, and ruvA2 subA cells were grown to mid-exponential phase in rich medium and stained with DAPI and either fixed and visualized by fluorescence microscopy (Fig. 4) or fixed, processed, and visualized by electron microscopy (data not shown). The sms (radA) strain contains a low number of anucleate cells (0.5% of total cells). The chromosomal segregation defect observed with ruvA2 and recD41 cells was partially suppressed if the sms (radA) mutation was present in the background (Fig. 4). In contrast, the recU segregation defect was not suppressed by the presence of the sms (radA) mutation.
|
subA strain were rare. The presence of the subA mutation suppressed the segregation phenotype of ruvA cells, but subA did not suppress the segregation defect of recU and recD cells (Fig. 4). This finding is consistent with the observation that subA partially overcomes the repair defect of ruvA2 cells but fails to suppress the recombinational defect of recD41 cells (8).
The subA mutation partially suppressed DNA repair and segregation phenotypes of recG cells.
The RuvABEco and RecGEco helicases, in concert with the HJ endonuclease RuvCEco, are involved in the formation and processing of branched recombination intermediate structures (38, 40). Above, we showed that ruvA, ruvAB, recU, and recD cells have a segregation defect that can be, in some cases, partially suppressed either in sms (radA), subA, or both genetic backgrounds. To determine whether recG cells show any segregation and DNA repair phenotype and if the sms (radA) or the subA null mutation has any influence in the segregation pattern of recG cells, recG single and double mutant strains (recG sms [radA] and recG subA mutants) were constructed and analyzed.
The recG strain failed to form colonies in the presence of 20 µg of MMS/ml. subA and sms (radA) strains formed colonies in the presence of 250 µg of MMS/ml, and the wt strain formed colonies in the presence of 300 µg of MMS/ml (8). The recG sms (radA) strains failed to form colonies in the presence of 20 µg of MMS/ml, whereas the recG subA double mutant strain was able to form colonies in the presence of 250 µg of MMS/ml. Therefore, it is likely that the subA mutation partially suppresses the recombinational defect of recG cells.
The absence of DAPI-stained material was observed for 7% of exponentially growing recG cells, and >30% of these cells had abnormally condensed nucleoids (Fig. 5). Similar results were observed with recG sms (radA) cells (Fig. 5). The chromosomal segregation defect observed with recG cells was partially suppressed if the subA mutation was present in the background. The presence of the subA null allele in the recG background reduced the number of anucleated cells to 1% (Fig. 5). This finding is consistent with the observed partial subA suppression of the DNA repair defect of recG cells.
|
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
epistatic group), addAB (ß), recH (), and recS () mutant cells did not show any chromosomal segregation phenotypes. We favor the hypothesis that recombination events catalyzed by RecFLOR, AddAB, RecH-RecP, and RecS might occur predominantly in the absence of crossing over. This hypothesis is consistent with the observation that RecFOREco-dependent recombination events occur in the absence of crossing over (14) and with the viability of E. coli rep mutants in the absence of XerC (41). Alternatively, all presynaptic proteins can be considered RecA accessory proteins, and mutations in only one of the genes will never lead to a strong segregation phenotype.
Genes classified within the epistatic group are required for replication fork repair and chromosomal segregation.
Genetic and biochemical evidence suggests that the genes classified within the epistatic group (ruvAB, recU, and recD) are involved in DNA repair and HR (3, 8, 18). We show here that the recU, ruvA, ruvB, and recD gene products are involved in recombinational repair of replicating cells and in proper chromosomal segregation. Furthermore, the results presented suggest a postsynaptic role for the unknown activity associated with the recD41 mutation.
The recombinational repair of stalled or collapsed replication forks leads to the production and resolution of an HJ. In both E. coli and B. subtilis, the HJ resolvases RuvC and RecU, respectively, bind and resolve the HJ (3, 61). Depending on the particular binding orientation, RuvCEco or RecU can resolve the symmetric HJs to crossover or noncrossover status. The defect of recU, ruvAB, and recD41 mutations in chromosomal segregation might be a consequence of their inability to bias HJ resolution toward noncrossovers. In that case, the crossover product will produce a dimeric chromosome. Alternatively, the dimer is formed because the HJ remains unresolved in both ruvABCEco cells (41) and ruvAB recU recD cells (this work). In both cases, dimers need to be resolved before cell division can occur. In E. coli and B. subtilis cells, specific site-specific recombinase systems, the XerCD/FtsK and CodVRipX/SpoIIIE complexes, respectively, act at dif to ensure the resolution of dimeric chromosomes (6, 27, 48, 49). This is consistent with the observations that for both E. coli and B. subtilis, the segregation defect of xerCEco dif and ripX mutants is suppressed by inactivation of the RecA protein (6, 27, 28) and that the absence of the RecA protein also suppresses the segregation defect of recU cells. We propose that the HJs made in the absence of the RecA protein are resolved to noncrossovers. This proposal is consistent with the observation that chromosome dimer formation (crossovers) is prevented in recA recU mutants, in repEco ruvABCEco difEco recAEco or priAEco recAEco mutants, or in UV-irradiated ruvCEco recAEco cells (22, 36, 41).
Previously, it has been shown that the sms (radA) and subA mutations partially suppress the DNA repair defect of genes classified within the epistatic group (8). As shown in Fig. 4, the sms (radA) mutation suppressed the segregation phenotype of ruvA2 and recD41 cells but failed to suppress the segregation defect of recU cells. We propose that, in the absence of the Sms (RadA) and RuvAB or RecD proteins, the branch migration RecG protein bound to an HJ intermediate will dictate the RecU resolution of the HJs in a way that should allow replication restart and noncrossover formation. This proposal is consistent with the observation that in E. coli, the sms (radA) and ruv mutations are synergistic with the recG mutation (5, 30). Alternatively, as previously proposed by McGlynn and Lloyd (38) for E. coli cells, the RecG protein in the sms (radA) ruvAB background would reestablish the fork ready for PriA-dependent reloading of the replisome. The Sms (RadA) protein shares a significant degree of identity with the RecA protein at its central region and with the Lon protease at its C-terminal region and plays a role in recombinational repair (5, 8). At present the biochemical activity(ies) associated with the Sms (RadA) protein remains to be elucidated.
The recG gene product is required for chromosomal segregation.
The RecGEco protein plays an essential role in the processing of recombination intermediates in E. coli cells (38, 40). Unlike the recGEco mutation that confers moderate sensitivity to DNA-damaging agents (31), the recG mutation markedly affects the viability of cells exposed to 20 µg of MMS/ml (M. C. Cozar and H. Sanchez, personal communication). Furthermore, recG cells show a chromosomal segregation phenotype (Fig. 4), suggesting that the recG mutant failed to repair stalled or collapsed replication forks. Furthermore, as observed with E. coli cells, if positive supercoiling is allowed to accumulate ahead of the replication fork, the forks may be converted to HJs, which have to be converted back to forks if replication is to be completed (43, 47). Hence, in both ruvAB recU (recD) and recG cells, replication should be stalled and anucleate cells should accumulate.
It has been suggested that the Sms (RadA) and SubA proteins are involved in the formation, stabilization, or processing of branched DNA molecules or blocked replication forks (5, 8). Here, we show that the subA mutation also partially suppresses the DNA repair and segregation phenotypes of recG cells, but the sms (radA) mutation suppresses neither the DNA repair nor the segregation defect of recG cells. Interestingly, the subA mutation suppresses the DNA repair and segregation phenotypes of both previously described branch-migrating DNA helicases (RuvAB and RecG).
What is the role of the SubA protein? The subA and mfd genes form an operon (4, 8), and a subA counterpart in E coli is apparently absent. SubA shares a low degree of identity with the UvrA protein, and Mfd shares a significant degree of identity with the RecG and PriA proteins (4, 8, 33). Both Mfd and MfdEco proteins recognize a stalled RNA polymerase (RNAP) at UV-induced lesions in the template DNA, dissociate RNAP from the DNA, and recruit UvrA to the site of damage, thereby facilitating excision repair of the transcribed strand (4, 44, 51). RNAP molecules stalled at lesions in the DNA are major obstacles to replication fork progression, and RuvABCEco is required to promote the rescue of the stalled replication forks (39, 50, 59). With E. coli, it has been shown that elevation of ppGpp levels or certain RNAP mutations improves the survival of UV-irradiated RuvABC mutants, probably by minimizing stalling of RNAP at lesions (39).
PriAEco loads the replisome at recombination intermediates to rescue arrested forks (29, 37). Although a mutation in the helicase motif of PriAEco reduces the ability of ruv mutants to survive DNA damage, it suppresses the DNA repair defect in recG cells (23). Since (i) the subA mutation suppresses the phenotype of mutations (ruvA2, ruvAB, and recG) in genes encoding the major branch migrating helicases and (ii) ruvAB and recG suppressors in E. coli are helicase-defective proteins, we hypothesize that Mfd alone or in concerted action with another factor(s) may recognize branched structures and translocate in such structures in the presence of SubA. This hypothesis is consistent with our previous failure to detect Mfd-specific binding to HJs and promotion of branch migration (4) and with the fact that the DNA translocation motifs of RecGEco and MfdEco are conserved (33). However, E. coli mfd recG and mfd ruvAB cells were two- to threefold-more UV sensitive than the recG or ruvAB cells (53). Furthermore, we predict that the low degree of identity of SubA with UvrA might correspond to the domain of interaction with Mfd. At present, the MfdEco-interacting domain in UvrAEco remains unknown.
A direct effect due to the absence of Mfd in the subA strain can be ruled out because (i) the downstream mfd gene is under the control of an inducible promoter in the subA cells (8) and (ii) the mfd mutation increased the sensitivity to DNA-damaging agents of recU cells (4), whereas the subA mutation partially suppressed its defect (8). Alternatively, the subA gene might code for an Mfd repressor. However, a suppression of the recU segregation defect was observed with subA cells, even in the absence of induction, which will render low levels of Mfd. At present, the biochemical activity(ies) associated with the SubA protein remains to be elucidated.
|
ACKNOWLEDGMENTS |
|---|
We thank H. Sanchez for communicating unpublished results.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||
