Purification of full-length and truncated periplasmic PBP2a
First, full-length PBP2a was expressed in E. coli with various N-terminal affinity tags (poly-histidine, calmodulin-binding peptide or Strep-tag). Isolated membranes were solubilized in Triton X-100 and the tagged proteins were purified on the corresponding affinity resins. None of the tagged PBP2a displayed GT activity (data not shown).
We hypothesized that the tags or the linker (formed mostly by a TEV protease cleavage site) could interfere with the activity. We therefore attempted to purify untagged full-length PBP2a by ion-exchange chromatography. We applied successively cation- and anion-exchange chromatography, and to our surprise, PBP2a bound to both types of resin (SP-Sepharose and Resource Q) at nearly the same pH, pH 8 for the cation-exchange and pH 7.5 for the anion-exchange chromatography. In both cases, PBP2a was eluted with ∼ 300 mm of NaCl. This unexpected behavior allowed purification to near homogeneity in two steps (Fig. 1A). The yield was ∼ 0.5 mg protein·L culture−1. The overall predicted isoelectric point (pI) of PBP2a is 6.0, with a net negative charge of -8 at pH 7.5–8. The retention of PBP2a on a strong cation-exchange medium is thus unexpected. However, a closer examination of the sequence reveals that the cytoplasmic region (residues 1–56) has a pI of 10.3 and a net positive charge of +7 at pH 8, whereas the extracellular region (residues 78–731) has a pI of 5.3 and a net negative charge of −16 at pH 7.5. This charge distribution is expected from the ‘positive inside’ rule that defines the transmembrane orientation of membrane proteins . We suggest that the two parts of the solubilized proteins are isolated from each other by the detergent micelle around the transmembrane segment, and that this particular arrangement may allow binding to the cation-exchanger by the cytoplasmic part, and to the anion-exchanger by the periplasmic domains. It would be of interest to test whether this behavior is general for bitopic membrane proteins that have a similar charge distribution.
Figure 1. Characterization of PBP2a and PBP2a*. (A) Coomassie Brilliant Blue- and bocillin-stained SDS/PAGE analysis of recombinant PBP2a and PBP2a*, following SP-Sepharose (SP) or Resource Q (RQ) chromatography. M is for solubilized E. coli membranes prior to chromatography. Boc is for bocillin. The Boc lanes were imaged under UV illumination from the same gels prior to Coomassie Brilliant Blue staining. The asterisk signals that 25% dimethylsulfoxide was present during the reaction with bocillin. (B) Partial tryptic proteolysis of PBP2a and PBP2a* analyzed by Coomassie Brilliant Blue-stained SDS/PAGE. (C) Cross-linking in vitro of PBP2a and PBP2a* with varying amounts of EGS, analyzed by Coomassie Brilliant Blue-stained SDS/PAGE.
Download figure to PowerPoint
The periplasmic region of S. pneumoniae PBP2a (PBP2a*) was prepared as described previously , with an additional anion-exchange chromatography following cleavage of the glutathione S-transferase-tag (Fig. 1A). This last step eliminated minor contaminants and allowed to exchange the detergent from Chaps to Triton X-100. The yield was ∼ 2 mg protein·L culture−1.
N-Terminal sequencing yielded the expected sequence for both PBP2a and PBP2a*, and electrospray MS confirmed the integrity of both proteins (PBP2a: 80 797 ± 2 Da measured vs 80 799 expected; PBP2a*: 73 810 ± 2 measured vs 73 812 expected). The TP domains of the full-length and truncated variants were functional for the reaction with the fluorescent penicillin bocillin (Fig. 1A).
Limited proteolysis of PBP2a and PBP2a* with trypsin produced a common major protein fragment with an apparent molecular mass of 44 kDa (Fig. 1B). N-Terminal sequencing showed that this fragment starts at residue 264 and thus comprises the TP domain. This main trypsin-resistant domain had been identified previously in PBP2a* . Mass spectrometry of this fragment had revealed that it consisted of residues 264–678 . The shorter minor tryptic fragment, which also appears with both PBP2a and PBP2a*, had been found in the previous study of PBP2a* to encompass residues 301–678 . The GT domain in full-length PBP2a is therefore susceptible to tryptic digestion, as in the truncated form. However, greater amounts of trypsin seem to be required to digest full-length PBP2a, indicating that the presence of the transmembrane and cytoplasmic regions may stabilize the GT domain to some extent, at least in the presence of detergent.
The oligomeric state of PBP2a was probed by chemical cross-linking followed by SDS/PAGE. Following incubation with the cross-linking agent [ethylene glycol bis(succinic acid) N-hydroxysuccinimide ester; EGS], additional bands were observed in the case of PBP2a, but not with PBP2a* (Fig. 1C). The apparent mass of the main species is consistent with the presence of a dimer (∼ 150 kDa). The presence of species with higher masses hints at the presence of higher orders oligomers or aggregates. The absence of cross-linking of PBP2a* by EGS (Fig. 1C) suggests that the self-association of PBP2a may depend on its transmembrane and/or cytoplasmic regions.
Size-exclusion chromatography in the presence of 0.02% Triton X-100 did not show the presence of aggregates with either PBP2a or PBP2a* (not shown). The presence of Triton X-100, which absorbs at 280 nm, precluded the monitoring by UV absorbance; instead, elution fractions were analyzed by SDS/PAGE. Both the full-length and truncated forms were eluted with comparable volumes of solution, corresponding to an apparent molecular mass of globular proteins comprised between 500 and 300 kDa. It is unclear why both proteins behave similarly in size-exclusion chromatography and differently in the cross-linking reaction. The presence of large micelles of detergent may mask size differences of the proteins.
The dimerization of PBP1b from E. coli has been observed by SPR with a dissociation constant KD of ∼ 100 nm . The dimerization was found to stimulate the GT activity of E. coli PBP1b and to be necessary for its TP activity . The oligomerization of class A PBPs has also been observed in vivo [31,32].
Optimization of GT reaction conditions
Reaction conditions are critical for PBP GT in vitro activity, for example, the nature of the divalent cation or the detergent and dimethylsulfoxide concentration [28,33,34]. Full-length PBP2a was incubated with labeled fluorescent lipid II in the presence of different concentrations of dimethylsulfoxide and Triton X-100 detergent. The detergent was either added to a concentration about fourfold its critical micellar concentration (CMC) in water, or only brought to the reaction mix with the purified protein, resulting in a concentration about fivefold below the CMC. The reaction products were analyzed by Tris/Tricine SDS/PAGE  (Fig. 2A,B). The lipid II migrates near the front and appears as dark thick bands at the bottom of the gel. The polymerized glycan chains appear as a smear above the lipid II band (see below for a more complete description of the assay). In this small 5-cm-long gel system, discrete chain lengths were not resolved. The reaction was optimal with 30–40% dimethylsulfoxide. The reaction started to be less efficient with 50% dimethylsulfoxide, as seen by the amount of unprocessed lipid II remaining after 40 min of reaction. The inhibitory effect of dimethylsulfoxide at high concentration may be due to denaturation of the enzyme.
Figure 2. Effect of dimethylsulfoxide and detergent on the GT activity of PBP2a. (A,B) Effect of the concentration of dimethylsulfoxide on the GT reaction in the presence of 0.002 and 0.04% Triton X-100, respectively. Reactions using dansylated lipid II at 30 °C were stopped after 40 min, analyzed by Tris–Tricine SDS/PAGE and imaged under UV light. The lane noted with Moe was with 0.5 mm moenomycin and no dimethylsulfoxide. (C) Comparison of the effect of different detergents on the reaction followed by the decrease of fluorescence, using dansylated lipid II as substrate. See Offant et al. and Schwartz et al. for a description of the assay [28,34]. All reactions contained 0.02% Triton X-100. Added detergents were: none (black solid line), 1% Chaps (black dotted line), 0.02% DDM (black short dashes), 0.02% Triton X-100 (black long dashes), 0.25% Cymal-5 (black dot-and-dashed line), 0.08% Cymal-4 (gray solid line) and 1%β-OG (gray dashes). The gray dotted line is without enzyme. Cymal-3 was omitted for clarity.
Download figure to PowerPoint
Interestingly, at a high concentration of Triton X-100, dimethylsulfoxide was required to observe an activity of PBP2a, whereas at a low detergent concentration, the reaction could occur to some extent in the absence of dimethylsulfoxide. Thus, dimethylsulfoxide appears to rescue the GT activity in the presence of detergent concentrations that are normally inhibitory, as observed previously with PBP1a from Thermotoga maritima and PBP1b from E. coli in decyl-poly(ethylene glycol) [28,34]. This effect of dimethylsulfoxide has been attributed to a greater solubility of the lipidic substrate and to faster exchange between detergent micelles . Note that the in vitro activity of the N-acetylglucosminnyl transferase MurG to convert lipid I into lipid II, the activity was highest with 35% dimethylsulfoxide in the presence of 0.1% deoxycholate .
To assess the effect of the nature of the detergent on the GT activity, we used a continuous fluorescence assay in a 96-well format [28,34]. In this assay, a modified lipid II is used, which has a fluorescent dansyl probe attached to the stem peptide. When glycan chains are polymerized, the probe is incorporated in the growing chains. However, a muramidase is present in the reaction mix that hydrolyzes the chains into disaccharide pentapeptides, which still harbor the fluorescent probe. Thus, the dansyl probe, which is initially bound to lipid II, is found in molecules devoid of lipid moiety as the reaction proceeds. This change of environment of the dansyl probe is accompanied by a decrease of fluorescence.
Detergents were tested at about twice their CMC in water. We observed no GT activity in the presence of n-octyl-β-d-glucopyranoside (β-OG), Cymal-3 and Cymal-4 (Fig. 2C). PBP2a was active in the presence of Chaps, n-dodecyl-β-d-maltopyranoside (DDM) or Triton X-100, and less so in Cymal-5. The highest activity was found with Chaps, Triton X-100 and DDM. In all cases, we used detergent concentrations above their CMC in water. The shortest detergents (β-OG, Cymal-3 and Cymal-4) were not suitable. Because short detergents have high CMCs, it is possible that the inhibitory effect results from the decreased probability that enzyme-containing micelles encounter substrate-containing micelles, as proposed previously . Considering these results, partially presented in Fig. 2, we chose for further work to use the intermediate dimethylsulfoxide concentration of 25% and 0.02 or 0.04% of Triton X-100 as detergent.
GT enzymatic activity
We aimed to determine the rate of transglycosylation catalyzed by PBP2a, PBP2a* and E. coli PBP1b, using the 96-well microplate assay. Because of substrate limitation, we were not able to measure enough rates to obtain accurate values of the enzymatic parameters. However, by measuring the initial rates of lipid II consumption at two concentrations of lipid II, we could estimate that the kcat for both PBP2a and PBP2a* is > 5 × 10−4 s−1. PBP2a* was somewhat more efficient. The specific activity measured with 10 μm dansylated lipid II was 0.38 nmol lipid II·min−1·mg enzyme−1 for PBP2a, and slightly faster at 0.5 nmol lipid II·min−1·mg enzyme−1 for PBP2a*. For comparison, we measured a kcat of (3.0 ± 0.5) 10−2 s−1 for E. coli PBP1b. The specific activity of E. coli PBP1b was 22.5 nmol lipid II·min−1·mg enzyme−1. These findings are comparable to published values .
Full-length PBP2a and PBP2a* showed comparable specific activities for the consumption of lipid II, indicating that the membrane anchor does not have an important effect on this aspect of the enzymatic activity. The catalytic constant was 104-fold higher than reported in a previous study of PBP2a* , but still 40–60-fold lower than that of E. coli PBP1b. The dimerization of E. coli PBP1b was found to promote GT activity . If this finding applies to PBP2a, one factor limiting the activity of PBP2a in vitro might be a concentration that is below its dissociation constant.
The Tris–Tricine SDS/PAGE assay developed by Barrett et al.  allows the visualization of the synthesized glycan chains length, using radio-labeled lipid II as precursor. In this study, we used a fluorescently labeled lipid II, with a dansyl group attached on the third position of the substrate pentapeptide (lysine amine). This kind of modification should not affect the kinetics parameters of the GT reaction [18,38]. However, because the lysine amine that bears the dansyl probe is normally the acceptor group of the transpeptidation, no TP activity can occur with this precursor. Gels were imaged on a UV transilluminator. Lipid II migrates near the front of the electrophoresis, chains of disaccharide units migrate more slowly as they are longer. For short chains, individual bands were resolved with this 20 cm gel system, up to ∼ 10 disaccharide units. Longer chains were not resolved and merged in a smear. Some fluorescent material, likely constituted of longer chains remained in the wells. Using this assay, we examined the distribution of products of full-length PBP2a and PBP2a*. We found that remaining free lipid II could not be detected after ∼ 2 h of reaction (Fig. 3). The lack of resolution of glycan chains of high molecular mass prevented precise determination of the chain length. However, by comparison with chains assembled by E. coli PBP1b, known to synthesize chains of 30–50 units [24,28,30], we estimated that PBP2a produced chains of 20–30 units. By contrast, PBP2a* produced much smaller chains of five to six disaccharides. These results indicate that full-length PBP2a from S. pneumoniae is more processive than its truncated form PBP2a*. Because both forms of the enzyme use up lipid II at similar rates, the different chain lengths that are produced suggest that PBP2a* starts assembling chains more easily, but also releases glycan chains more readily than the full-length proteins. This phenomenon could be due to the transmembrane and/or cytoplasmic region directly, or indirectly through the absence of dimerization of PBP2a*. Interestingly, the absence of the transmembrane segment of E. coli PBP1b was found to decrease fivefold the affinity for the drug moenomycin . Examination of the structure of E. coli PBP1b, the sole structure of a full-length class A PBP, led Sung et al. to propose that removal of the transmembrane segment would impact on the binding of lipid II and moenomycin . Alternatively, an indirect effect of the detergent micelle surrounding the hydrophobic transmembrane segment could change the dynamics of the enzyme–substrate–product interactions.
Figure 3. GT activity using dansylated lipid II, monitored by Tris–Tricine SDS/PAGE and imaged under UV light. (A) Comparison of the chain length distribution resulting from the activity of PBP2a and PBP2a*. As a positive control, a reaction catalyzed by E. coli PBP1b was included. As negative control, a PBP2a variant with the active site residue replaced E131Q was used. The last lane demonstrates the absence of effect of penicillin (Pen) on the GT activity. Reaction times are given below the lanes. (B) Kinetics of the polymerization of glycan chains by PBP2a, as in (A) with more time points. (C) The fluorescence in the lipid II band (•), the glycan chains smear (■) and the well (▲) were quantified and represented on the graph.
Download figure to PowerPoint
A former comparative study of several PBPs from different organisms showed that PBP2a from Enterococcus faecalis and PBP2 from Staphylococcus aureus produced chains of only 15 units. PBP1a from E. coli made longer chains than PBP1b from the same bacteria. The glycan chain length was not dependent on the enzyme-to-substrate ratio .
Some studies have reported a delay in the onset of the reaction catalyzed by E. coli PBP1b, which has been attributed either to a slow initial coupling step or to a slow rearrangement to the active conformation [34,39]. This lag phase has not been observed with PBP2a from S. pneumoniae (Fig. 3C). Note that in Fig. 3A,B, the images were overexposed to better visualize the glycan chains. This caused most of the lipid II bands to be saturated and to not reflect their actual amount. For the quantification shown in Fig. 3C, a shorter exposure was used, which revealed the concomitant disappearance of lipid II with the appearance of the glycan chains. The short exposure image is not shown as the glycan chains are then barely visible, although the amount of fluorescence integrated over the length of the gel remains constant.
Absence of TP activity in vitro
The TP assay was based on the SDS/PAGE assay, but using a mix of fluorescent (for detection) and nonfluorescent lipid II (for cross-linking). Pure labeled lipid II cannot be used to investigate the TP activity, because it cannot serve as an acceptor in the reaction. Indeed, the fluorescent probe is attached to the third residue-free amine that participates in the reaction. By using 95% of unlabeled and 5% fluorescent lipid II, we hoped to allow transpeptidation to occur, while incorporating enough labeled material to visualize the products after SDS/PAGE. The experimental setup was validated using E. coli PBP1b, which is known to display both GT and TP activities in vitro [30,40]. In the absence of β-lactam, using a mixture of lysine-dansylated (5%, 10 μm) and meso-diaminopimelate-containing lipid II (95%, 200 μm), E. coli PBP1b (0.5 μm) generated in 2 h a fluorescent product of high molecular mass that was retained on the top of the polyacrylamide gel (Fig. 4). We used meso-diaminopimelate-containing lipid II, the natural precursor of E. coli peptidoglycan, because PBP1b does not catalyze transpeptidation with lysine-containing precursors (W. Vollmer, unpublished). In the presence of penicillin, E. coli PBP1b produced glycan chains of the expected size, demonstrating that the large polymer obtained in the absence of penicillin resulted from transpeptidation. Under the same experimental conditions, using lysine-containing lipid II, PBP2a did not display transpeptidase activity, even with 10 times higher concentration of enzyme after 5 h.
Figure 4. Assembly of peptidoglycan in vitro by E. coli PBP1b monitored by Tris–Tricine SDS/PAGE. Reactions were carried out with a mixture of dansylated lipid II (5%) and unlabeled meso-diaminopimelate-containing lipid II, without and with moenomycin (Moe) or penicillin G (Pen). Cross-linked peptidoglycan does not enter de gel system and remains trapped at the level of the loading well.
Download figure to PowerPoint
Despite the absence of detectable transpeptidase activity of PBP2a and PBP2a*, both variants reacted with bocillin (Fig. 1A), and both catalyzed the hydrolysis of S2d, a thiol-ester analogue of the stem peptide. Hydrolysis of thiol-ester analogues has been used traditionally as a proxy for the TP activity in the absence of physiologically relevant substrate . Values given in Table 1 show that the activity of the pneumococcal PBPs is on par with that of E. coli PBP1b with respect to S2d hydrolysis. This observation raises the question of the absence of transpeptidase activity.
Table 1. S2d hydrolysis activity. The specific activity measured with 2 mm S2d.
|Protein||Specific activity (nmol·min−1·mg enzyme−1)|
|E. coli PBP1b||432|
One trivial explanation would have been that the DMSO required for optimal GT activity is inhibiting the TP activity. However, we showed that reaction with bocillin is not hampered by 25% DMSO (Fig. 1A, * lanes). Conversely, because we found that PBP2a exhibits some residual GT activity in the absence of DMSO at low a concentration of Triton X-100, we repeated the assay under these conditions (no DMSO, 0.002% Triton X-100) but did not observe TP activity (not shown).
To our knowledge, no transpeptidation catalyzed by purified PBPs from Gram-positive organisms has ever been reported. Early experiments with enzymes from Bacillus species showed that although GT activity was observed, no transpeptidation occurred with PBPs from Bacillus subtilis, and only minor amount with PBPs from Bacillus stearothermophilus . It is possible that the assembly of a thick peptidoglycan depends on a more controlled activity of the PBPs interacting with activating factors, and that isolated pneumococcal PBPs are inactive. Activators of peptidoglycan assembly are known in some Gram-negative species [43,44].
An alternative hypothesis is that the lipid II used in our study is not the proper substrate for PBP2a or for pneumococcal PBPs in general. The peptidoglycan of pneumococcus contains a significant fraction of ‘branched’ muro-peptides. In those, a Ser–Ala or Ala–Ala dipeptide is added onto the side chain of the third residue lysine. Cross-linking by transpeptidation then occurs through the free amine of the Ala. The presence of these branched peptides is necessary for the expression of β-lactam resistance . It is conceivable that the TP domain of PBP2a is specialized for branched peptides. Also, the lysine-containing lipid II used in our study contains a glutamate as second residue of the stem peptide . In most Gram-positive bacteria, the second residue is an iso-glutamine that results from the amidation of the glutamate by the recently identified essential amidotransferase MurT/GatD [46,47]. Depletion of the MutT/GatD amido transferase impairs growth and increases the susceptibility of Staphylococcus aureus to β-lactams. These observations suggested that one or more staphylococcal PBPs may require or prefer amidated lipid II [46,47]. This modification occurs at the stage of the lipid II, as shown in early studies with particulate extracts from Staphylococcus aureus . Finally, the correct substrate for PBP2a-catalyzed transpeptidation, at least as acceptor, might be the pre-existing peptidoglycan itself.
In conclusion, we aimed to purify and to characterize recombinant full-length and periplasmic PBP2a from S. pneumoniae, in order to reconstitute in vitro the peptidoglycan synthesis from this important Gram-positive pathogen. We found that the GT activity of PBP2a was very sensitive to the reaction conditions such as the concentration of DMSO and the detergent. Also, we observed that membrane anchor influences glycan chain length in vitro. The next challenge will be to demonstrate TP activities from pneumococcal PBPs, as it has never been obtained with recombinant enzymes from Gram-positive organisms.