![[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}
Department of Microbiology and Immunology, University of North Dakota School of Medicine, Grand Forks, North Dakota 58202-9037
Received 4 September 2002/ Accepted 6 January 2003
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To identify the morphologically important segments of domain I, we constructed hybrid proteins in which portions of domain I from PBPs 5 and 6 were interchanged with one another. A stretch of 20 amino acids around the KTG motif emerged as a major contributor to cellular morphology. PBPs 5 and 6 differ in 8 of these 20 residues, of which 5 were different compared with DacD. Site-directed mutagenesis of mosaic proteins established that amino acids Asp218 and Lys219 were especially important for creating and/or maintaining normal cell shape in E. coli.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
(deoR recA endA hsdR supE thi gyrA relA) were used as hosts for constructing recombinant plasmids. Strains used in the morphological experiments were derived from CS109 (W1485 rpoS rph) (C. Schnaitman) via CS701-1 (CS109 
[mrcA-yrfE-yrfF] dacB dacA dacC pbpG ampC ampH) and CS703-1 (CS109 mrcA dacB dacA dacC pbpG ampC ampH) (10, 20). PBP genes were expressed under the control of the arabinose promoter of pBAD18-CAM, provided by J. Beckwith (14). Strains were grown on Luria-Bertani (LB) broth or agar plates, with chloramphenicol (20 µg/ml) added as required to maintain selection of pBAD plasmids. Overnight broth cultures of E. coli strains were diluted 1:250 into fresh LB medium and incubated at 37°C until they reached an A600 of 0.2 (ca. five to six doublings) before complementation was assessed by microscopy. Cells were collected, prepared for microscopy, and photographed at 1,000x magnification as described previously (23). When necessary, glucose (0.2 to 0.4%, wt/vol) was added to the medium to inhibit gene expression from the arabinose promoter. To induce protein expression in complementation experiments, E. coli strains were grown in the absence of glucose or in the presence of arabinose (0.0005% or 0.001%) (22). For expression of DD-carboxypeptidases from other species in E. coli, the arabinose concentration for induction was increased (up to 0.1%). Unless otherwise noted, all chemicals were purchased from Sigma Chemical Co. (St. Louis, Mo.).
Molecular techniques. Plasmids were isolated from E. coli by with QIAprep Spin miniprep and midiprep kits (Qiagen Corp., Valencia, Calif.) according to the manufacturer's instructions. Competent cells were prepared and transformed by electroporation with the Gene Pulser apparatus from Bio-Rad (Hercules, Calif.) according to the manufacturer's instructions. CS109 chromosomal DNA for PCR amplifications was prepared by boiling 200 µl of overnight culture with 800 µl of distilled water for 10 min, followed by centrifugation at 14,000 x g for 1 min and collection of the supernatant. DNA agarose gel electrophoresis was performed as described before (24). DNA purification from agarose gels was performed with QIAquick gel extraction kits (Qiagen Corp.) as described by the manufacturer. Restriction digests and ligations were performed with enzymes purchased from New England Biolabs (Beverly, Mass.).
PCR. PCR was performed in a model 2400 Gene Amp thermal cycler (Perkin Elmer, Boston, Mass.). Oligonucleotide primers for PCR were from MWG Biotech Inc. (High Point, N.C.). Stock solutions of individual deoxynucleoside triphosphates for PCR were from Promega (Madison, Wis.). Deep Vent DNA polymerase was from New England Biolabs (Beverly, Mass.).
Construction of PBP gene fusions. With a PCR-based strategy, portions of the 5' terminus of the dacA gene, encoding domain I of PBP 5, were fused to the corresponding 3' terminus of dacC, encoding PBP 6. The PCR primers are described in Table 1, and their order of use to amplify and assemble each hybrid gene is listed in Table 2. Fragments of PBPs 5 and 6 were fused to one another at three positions where their amino acid sequences were identical or differed by only one of seven residues. Therefore, forward and reverse primers corresponding to amino acid positions 41 to 47, 197 to 203, and 220 to 228 of the mature form of PBP 5 were synthesized to match the nucleotide sequences of each separate gene (Table 1).
|
|
Construction of mosaic PBPs. Two mosaic PBPs were constructed in which amino acids around the active-site KTG motif were moved from one PBP to another. The nucleotides encoding PBP 5 amino acids 200 to 219 were spliced into PBP 6, creating plasmid pAG656-11, in the following manner. Gene fragments were amplified from plasmid pAGG-1 (primers C and J, Table 1) and from pAGS-29 (primers D and L, Table 1). The nucleotide sequence of the first PCR product encoded the first 196 amino acids of PBP 6 fused to residues 197 to 228 of PBP 5, and the nucleotide sequence of the second PCR product encoded residues 197 to 228 of PBP 5 fused to residues 229 to 378 of PBP 6. These two DNA fragments were isolated from agarose gels and used to prime one another in a second PCR, as described previously (21), by hybridization between nucleotides representing amino acid residues 197 to 228. Because the yield in this reaction was poor, the composite full-length PCR product was amplified further with primers C and D (Table 1). This DNA fragment was digested with restriction enzymes NheI and HindIII and cloned between the NheI and HindIII sites of pBAD18-CAM, as described above, creating plasmid pAG656-11. Thus, the final PCR product encoded a mosaic protein composed of PBP 6 residues 1 to 196, PBP 5 residues 197 to 228, and PBP 6 residues 229 to 378. Because amino acid residues 197 to 199 and 220 to 228 are identical in the two proteins, the final mosaic protein is the equivalent of inserting residues 200 to 219 from PBP 5 into full-length PBP 6 (see Table 3).
|
Site-directed mutagenesis. Site-directed mutagenesis of the amino acids in and around the conserved KTG motif was performed with the Quick Change mutagenesis kit from Stratagene (La Jolla, Calif.) as described previously (21). Mutagenesis on supercoiled double-stranded plasmid DNA was carried out exactly according to the manufacturer's instructions, with oligonucleotide primer pairs obtained from MWG Biotech, Inc. The primers ranged from 33 to 47 bases, depending on the number of individual nucleotides to be altered (one to three), and for each primer pair one or two codons at the center were altered to give the amino acid substitutions described in Table 3. The number of PCR cycles was from 12 to 18, increasing according to the number of bases altered. Mutated plasmids were transformed by heat shock into Epicurian XL1-Blue supercompetent cells (Stratagene, La Jolla, Calif.) and plated on LB-chloramphenicol plates. Mutagenesis was confirmed by DNA sequencing (MWG Biotech, Inc.).
PBP labeling, photography, and sequence analysis. Expression of hybrid PBP proteins from recombinant plasmids was confirmed by labeling equal numbers of cells with [125I]penicillin X, separating total cellular proteins by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and visualizing by autoradiography as described previously (16). Photography was performed and interpreted as described previously (22, 23). Homologous protein sequences were identified and compared by the BlastP 2.1.3 program (1) as supplied on the National Institutes of Health Entrez web site (URL http://www3.ncbi.nlm.nih.gov/Entrez/) and by the ClustalW program (version 1.81) (25) as supplied on the European Bioinformatics Institute web site (URL http://www2.ebi.ac.uk/clustalw/).
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Fusion proteins containing various portions of PBPs 5 and 6 reversed the shape defects of E. coli CS703-1 when residues 200 to 228 of the hybrids were derived from PBP 5 (Table 2 and Fig. 1 and 2). Residues 220 to 228 were identical in PBPs 5 and 6 (AGYNLVASA). Therefore, the ability to return CS703-1 to its normal shape was associated with the presence of a contiguous stretch of 20 amino acids consisting of residues 200 to 219 from PBP 5.
|
|
|
Five mutants derived from PBP 6/5/6 retained the ability to complement the aberrant shape of CS703-1: G200R (pAG656-X2A), N205S (pAG656-M1A), S206N (pAG656-M2A), I212M (pAG656-M3A), and H216T (pAG656-H) (Table 3). Converting the same positions in PBP 5/6/5 to PBP 5 residues did not restore normal morphology to CS703-1: R200G (pAG565-X1A and pAG565-X6A), S205N (pAG565-M6A), N206S plus M212I (pAG565-M7A), and T216H (pAG656-T) (Table 3). Therefore, the amino acids at positions 200, 205, 206, 212, and 216 did not distinguish the activities of PBPs 5 and 6.
On the other hand, mutation of either of two residues in PBP 6/5/6 destroyed its shape-complementing ability: D218A (pAG656-M4A), K219G (pAG656-M11A), and DK218/219AG (pAG656-M5A) (Table 3). Similarly, converting either of these two positions in PBP 5/6/5 to PBP 5 residues successfully reactivated the shape complementation ability of this hybrid: A218D (pAG565-M9A) and G219K (pAG565-M10A) (Table 3). Thus, at least in these mosaic proteins, the charged amino acids Asp218 and Lys219 were important differential determinants of the ability to create or maintain normal bacterial morphology.
Features other than Asp218 and Lys219 contribute to morphological activity of PBP 5. It appeared that the ability of PBP 5 to affect cell shape might be mediated by one or two individual residues guarding the entrance to the active site (see Discussion). However, the presence of either of these normally occurring residues (either Asp218 or Lys219) in PBP 5/6/5 created a functional enzyme (Table 3), indicating that one or the other amino acid was sufficient to confer morphological activity on this particular mosaic protein. On the other hand, mutating either or both of these residues in PBP 6/5/6 (e.g., D218A, K219G, or DK218/219AG) (Table 3) inhibited the morphological activity of this hybrid. Thus, while these two residues were obviously vital to the function of the mosaic proteins, their relative importance was still unclear. To understand this further, we mutated residues 218 and 219 simultaneously in wild-type PBP 5, changing them to amino acids normally present in PBP 6 (DK218/219AG, expressed from plasmid pPJ5-DK/AG) as well as in the presence of a third mutation (G200R, expressed from plasmid pPJ5-X4A) (Table 3, pPJ5 derivatives). Unexpectedly, these mutant proteins retained the ability to complement the shape defects of CS703-1 (Table 3). Therefore, although these two amino acids evidently play significant morphological roles in the mosaic proteins, the effects of these residues are moderated in the context of wild-type PBP 5.
Correlation of DD-carboxypeptidase activity with morphological effects of PBP 5. The lysine at position 213 of PBP 5 is the first residue in the canonical KTG motif of the PBPs (13). A K213E mutation in PBP 5 eliminates both penicillin-binding and DD-carboxypeptidase activity (28), whereas a K213R mutation eliminates the DD-carboxypeptidase activity but leaves intact the penicillin-binding and penicillin-hydrolyzing ability of PBP 5 (18). To see which activity contributed more strongly to the morphological abilities of PBP 5, we constructed mutant proteins containing these alterations and observed their effects on cell shape. Consistent with previous observations (18, 28), the K213E mutant did not bind 125I-labeled penicillin X, but the K213R mutant could still do so (data not shown). Neither mutant complemented the aberrant cell shapes exhibited by E. coli CS703-1 (Table 3, pPJ5 derivatives), indicating that maintenance of cell shape was correlated with DD-carboxypeptidase activity, not with penicillin binding.
As a further test of the supposition that the morphological ability of PBP 5 was related to its DD-carboxypeptidase action, we created a T217A mutant (Table 3, pPJ5 derivatives), which bound penicillin X very well (data not shown). This mutation decreases the DD-carboxypeptidase activity of PBP 5 to 0.5% of normal while leaving intact 45% of the penicillin binding activity (28). Nonetheless, the mutant protein corrected the morphological oddities in E. coli CS703-1 (Table 3, pPJ5 derivatives). Thus, if DD-carboxypeptidase activity is crucial to the function of PBP 5, then either a very small amount of activity suffices when the protein is supplied from a multicopy plasmid or else this particular mutation alters the activity of PBP 5 toward in vitro substrates but not toward the relevant in vivo substrate.
Shape complementation by DD-carboxypeptidases from other bacteria. Individual gram-negative bacteria express multiple low-molecular-weight PBPs related to PBPs 5 and 6 of E. coli. Until now, sequence comparison has been the only way to predict if one or more of these proteins might perform homologous functions within different species. The morphological assay afforded us the opportunity to test candidate proteins for functional equivalence.
A Blast search was performed to identify proteins most closely related to PBP 5 from E. coli (not shown). The most closely related DD-carboxypeptidase genes from five gram-negative bacteria were amplified by PCR, cloned under control of the arabinose promoter, sequenced, and expressed in E. coli CS703-1 (Table 4 and data not shown). DacA proteins from Salmonella enterica serovar Typhimurium, Vibrio cholerae (Cpase-1), Pasteurella multocida, Haemophilus influenzae, and Yersinia pestis reversed the morphological defects of CS703-1 (Table 4). The DacD protein of S. enterica serovar Typhimurium and the Cpase-2 protein of V. cholerae failed to complement the defects (Table 4). Thus, the shape phenotype of this multiply mutated E. coli provided the first explicit method for measuring homologous function among the DD-carboxypeptidases.
|
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In support of this idea, we discovered that splicing a 20-amino-acid segment from PBP 5 into PBP 6 transformed the latter protein into a morphological substitute for PBP 5. The simplest interpretation is that the enzymatic specificity of the mosaic PBP was altered to be more like that of wild-type PBP 5. This explanation is consistent with observations made by Chang et al., who performed an analogous experiment with RTEM ß-lactamase (7). When they replaced 28 amino acids around the active-site SxxN motif with homologous sequences from PBP 5, the enzyme lost its ability to hydrolyze penicillin and was transformed into a low-level DD-carboxypeptidase (7). Similarly, as shown here, PBP 6 acquired the ability to complement morphological abnormalities when 20 residues around its KTG motif were replaced with homologous residues from PBP 5.
Further evidence for an enzymatic explanation comes from site-directed mutagenesis of the PBP 5/6/5 and 6/5/6 mosaic proteins. In each case, the functional state of the protein could be toggled between the active and inactive forms by changing either of two amino acids at position 218 or 219, downstream of the conserved KTG motif (residues 213 to 215). There are precedents for considering these positions important determinants of enzymatic specificity. Ubukata et al. found that 9 of 25 ß-lactamase-negative, ampicillin-resistant clinical isolates of Haemophilus influenzae accumulated PBP 3 mutations that mapped to just this position, changing from KTGTARK to KTGTAHK (group I mutants) (26). Strains harboring these mutant PBP 3 molecules exhibited very different susceptibilities to a variety of ß-lactam antibiotics (26), indicating that the mutations changed the substrate specificity of this PBP. The locations of two charged amino acids at the third and fourth residues downstream of the KTG motif are equivalent to positions 218 and 219 in PBP 5 from E. coli, and the effect of the R-to-H mutation in the third downstream residue parallels the mutation in position 218 that we report for the mosaic proteins created in this work.
In addition, the other 16 isolates of Ubukata et al. (group II mutants) contain an N526K mutation which, compared with the homologous crystal structure of PBP 5, is equivalent to position 227 in PBP 5. This residue is one of nine very highly conserved amino acids that follow the KTG motif (compare positions 220 to 228 for the different species in Table 4). Residues 210 to 228 double back to form a hairpin structure in which amino acids 210 and 216 and 220 and 228 are in contact with one another as parallel ß-sheets, with residues 218 and 219 forming the "turn" of the hairpin (9). Thus, both mutations described by Ubekata et al. probably affect the active site in the vicinity of the KTG motif. These results were confirmed and extended by Dabernat et al., who found the same two mutant groups and alterations in ß-lactam susceptibility among 108 ß-lactamase-negative, ampicillin-resistant clinical isolates of H. influenzae (8). The parallels between these observations for PBP 3 in H. influenzae and the mosaic proteins described here strengthen the conclusion that the physiological difference between PBPs 5 and 6 lies in a specific enzymatic capacity and thus narrows the possible explanations for how E. coli maintains a uniform cell shape.
Given that the mutagenesis data support an enzymatic mechanism, what enzymatic property might explain the divergence in physiological function between PBP 5 and other noncomplementing DD-carboxypeptidases? To begin with, the ability of a PBP to bind ß-lactams does not, by itself, create a functional protein. For example, the K213R mutant, which binds penicillin but has no DD-carboxypeptidase activity (18), did not complement the misshapen phenotype. Thus, the DD-carboxypeptidase reaction is necessary, although we cannot say it is sufficient in the absence of ß-lactam binding. On the other hand, a T217A mutant, which retains most of its penicillin binding ability but only 
0.5% of its DD-carboxypeptidase activity (28), remained capable of complementing the odd morphology of CS703-1. So, if DD-carboxypeptidase activity per se is the vital property, then only a small amount is sufficient. This would be consistent with the observation that in the absence of induction, a low-level background expression of cloned PBP 5 complements the morphological phenotype (21, 22). Also, expressing the T217A mutant from a multicopy plasmid may have provided enough activity to restore proper morphology. A final possibility is that the chemical state or environment of the in vivo substrate of PBP 5 may be different than the artificial in vitro substrates used to assay bulk DD-carboxypeptidase activity.
The fact that inserting Asp218 or Lys219 residues changed a nonfunctional PBP 5/6/5 mosaic into a protein that complemented the morphological defects of E. coli CS703-1 suggests that there is a subtle biochemical explanation behind the differences in physiological function. The 20-amino-acid segment surrounding the conserved KTG extends almost linearly across one side of the active site of PBP 5, and the Asp218 and Lys219 residues within this sequence are located in a small loop a short distance away from the active site proper (9). These amino acids make no direct contact with the D-Ala-D-Ala terminus of the peptide side chain, but residues at these positions might modify the structure of the active site or interact with other portions of the peptide side chain or glycan polymer. In the simplest case, substitutions at these positions would change the kinetics of PBPs 5 and 6 toward a common substrate; for example, by moderating substrate access or affinity.
The two PBPs do have different in vitro activities: one group reports that PBP 5 is three to four times more active toward artificial substrates than is PBP 6 (2), though another lab reports that PBP 6 exhibits no activity at all towards the same compounds (27). In vivo evidence of such a difference also exists: mutants lacking PBP 5 accumulate muramyl pentapeptides, while PBP 6 mutants do not (B. Glauner, Ph.D. thesis, quoted and referenced in reference 27). In any case, the physiological difference between PBPs 5 and 6 may reflect nothing more than the enzymatic superiority of PBP 5. Alternatively, the presence of Asp218 and Lys219 might change the substrate specificity entirely. Unfortunately, in vitro assays cannot be trusted to distinguish among these possibilities because no one knows the relevant characteristics of the true in vivo substrates, which may be modified by the three-dimensional structure of peptidoglycan or by the degree of extension of peptide side chains (17). Perhaps these mosaic proteins may be used to discover substrates that mimic more closely what occurs in vivo.
The above considerations beg the following question: if the Asp218 and Lys219 residues are so important, why does mutating these amino acids eliminate the morphological activity of PBP 6/5/6 but not affect the complementation ability of wild-type PBP 5? First of all, it is possible that the mutations in wild-type PBP 5 produced a subtle or quantitative morphological change that was not detected by the gross microscopic assay that we employed. However, we easily detected that mutation of the mosaic proteins dramatically altered their ability to carry out a normal level of cell shape maintenance, indicating clearly that these enzymes were not exactly equivalent to wild-type PBP 5. Instead, the most straightforward explanation for the differences between PBP 5 and the mosaic proteins is that the wild-type PBP 5 active site retains partial activity or substrate specificity because of the nature or strength of interactions elsewhere in the molecule. This is easily understandable if only 0.5% activity is required to correct morphological deficiencies, as exhibited by the T217A mutant of PBP 5.
The active site of the PBPs and ß-lactamases contain variations of three major motifs — SxxK, SxN, and KTG in this case — which are brought into close proximity by protein folding. In the hybrids that we created previously (21) and in those reported here, both the SxxK and SxN motifs are derived either from PBP 5 or from PBP 6, and only the KTG motif comes from a different source. So, one possibility is that in PBP 6/5/6, the combination of SxxK and SxN from PBP 6 might make the protein sensitive to alterations in the PBP 5-derived KTG region, whereas mutations near the KTG segment in wild-type PBP 5 are offset by interactions in other parts of the molecule. This interpretation is consistent with the spectrum of ß-lactamase-negative, ampicillin-resistant PBP 3 mutants of H. influenzae: in many cases, mutation of residues in the KTG region were accompanied by additional unlinked mutations, which further affected the ß-lactam sensitivity of these mutant PBPs (8, 26).
The fact that DD-carboxypeptidase PBPs from other species complement the shape deficiencies of these E. coli mutants enhances the idea that subtle differences in enzymology are responsible. First of all, the degree of homology, as measured along the full length of these enzymes, does not predict which ones will function like PBP 5. For example, two noncomplementing PBPs (S. enterica serovar Typhimurium DacD and V. cholerae Cpase-2) are more closely related to E. coli PBP 5 than is DacA from Y. pestis, which did complement the phenotype (Table 4 and data not shown). The opposite is true for E. coli PBP 6. This protein is more closely related to E. coli PBP 5 than every protein except DacA from S. enterica serovar Typhimurium (data not shown), yet E. coli PBP 6 does not complement the phenotype.
Instead of overall similarity, the sequences equivalent to residues 204 to 219 of mature PBP 5 may be a more predictive measure of function. In this region, complementing DD-carboxypeptidases are generally more similar to PBP 5 than are PBPs that do not complement (Table 4). A second way in which the heterologous enzymes support the idea of a subtle enzymatic mechanism is that the Asp218 and Lys219 residues are present only in DacA from Pasteurella multocida. Nevertheless, the DD-carboxypeptidases from four organisms complemented the shape phenotype even though these enzymes have other amino acids at these positions (Table 4). This indicates, again, that although these two amino acids may play important roles in the PBP 5/6 mosaics, other factors contribute to the function of wild-type enzymes.
An interesting side observation is that the region defined by residues 204 to 219 in PBP 5 is shared by the DD-carboxypeptidases and ß-lactamases, and in both families this region is bracketed on either side by sequences having almost complete identity to that of PBP 5 (Table 4 and data not shown). In particular, the neighboring sequences are nearly identical in the DD-carboxypeptidases of E. coli as well as other bacteria (Table 4). Thus, though the KTG region may vary from one enzyme to another and therefore affect enzyme activity, the adjacent structures that hold it in position may be more highly conserved.
Finally, since it seems that the DD-carboxypeptidase activity of PBP 5 moderates bacterial morphology, the following difficult question arises: How does removing the terminal D-alanine from peptide side chains create a uniform cell shape? This minor change evidently regulates the gross structure of peptidoglycan, but the mechanism by which this occurs is unclear. One possibility is that the pentapeptide (which retains the terminal D-alanine) and the tetrapeptide or tripeptide side chains (which have lost one or two D-alanines, respectively) are preferred substrates of different synthetic PBPs. Such an idea was proposed to explain how enzymes might distinguish the synthesis of cylindrical cell wall versus dividing septum (4, 5). In this view, septation mediated by PBP 3 prefers substrates containing tripeptide side chains, and cell elongation mediated by PBP 2 prefers substrates with pentapeptides (4, 5). Thus, deletion of PBP 5, by increasing the amount of pentapeptides, should preferentially increase the relative activity of PBP 2. However, the morphological defects of PBP 5 mutants apparently emerge as a consequence of inappropriately placed patches of "septal" (inert) peptidoglycan (11), which should represent PBP 3 products. On the surface, this suggests that in these mutants PBP 3 activity is enhanced rather than decreased by elevated levels of pentapeptides. If so, this would contradict the previously proposed model and imply that PBP 3 might prefer pentapeptide substrates. On the other hand, if PBP 3 does indeed prefer tripeptides, then the enzyme might react to a relative abundance of pentapeptides with ill-timed or awkward cross-linking of this unfavorable substrate, resulting in the observed morphological oddities.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Appl. Environ. Microbiol. | Infect. Immun. | Eukaryot. Cell |
|---|---|---|
| Mol. Cell. Biol. | J. Virol. | Microbiol. Mol. Biol. Rev. |
| ALL ASM JOURNALS |
