Pseudomonas aeruginosa β-lactamase induction requires two permeases, AmpG and AmpP

Background In Enterobacteriaceae, β-lactam antibiotic resistance involves murein recycling intermediates. Murein recycling is a complex process with discrete steps taking place in the periplasm and the cytoplasm. The AmpG permease is critical to this process as it transports N-acetylglucosamine anhydrous N-acetylmuramyl peptides across the inner membrane. In Pseudomonadaceae, this intrinsic mechanism remains to be elucidated. Since the mechanism involves two cellular compartments, the characterization of transporters is crucial to establish the link. Results Pseudomonas aeruginosa PAO1 has two ampG paralogs, PA4218 (ampP) and PA4393 (ampG). Topology analysis using β-galactosidase and alkaline phosphatase fusions indicates ampP and ampG encode proteins which possess 10 and 14 transmembrane helices, respectively, that could potentially transport substrates. Both ampP and ampG are required for maximum expression of β-lactamase, but complementation and kinetic experiments suggest they act independently to play different roles. Mutation of ampG affects resistance to a subset of β-lactam antibiotics. Low-levels of β-lactamase induction occur independently of either ampP or ampG. Both ampG and ampP are the second members of two independent two-gene operons. Analysis of the ampG and ampP operon expression using β-galactosidase transcriptional fusions showed that in PAO1, ampG operon expression is β-lactam and ampR-independent, while ampP operon expression is β-lactam and ampR-dependent. β-lactam-dependent expression of the ampP operon and independent expression of the ampG operon is also dependent upon ampP. Conclusions In P. aeruginosa, β-lactamase induction occurs in at least three ways, induction at low β-lactam concentrations by an as yet uncharacterized pathway, at intermediate concentrations by an ampP and ampG dependent pathway, and at high concentrations where although both ampP and ampG play a role, ampG may be of greater importance. Both ampP and ampG are required for maximum induction. Similar to ampC, ampP expression is inducible in an ampR-dependent manner. Importantly, ampP expression is autoregulated and ampP also regulates expression of ampG. Both AmpG and AmpP have topologies consistent with functions in transport. Together, these data suggest that the mechanism of β-lactam resistance of P. aeruginosa is distinct from well characterized systems in Enterobacteriaceae and involves a highly complicated interaction between these putative permeases and known Amp proteins.


Background
Pseudomonas aeruginosa is a Gram negative opportunistic pathogen. As a frequent colonizer of catheters and the most frequent fatal causative agent of ventilatorassisted pneumonia, it is one of the most common agents in health-care associated infection [1]. Lung deterioration due to chronic infection by P. aeruginosa affects patients with chronic obstructive pulmonary disorder and is a leading cause of morbidity and mortality in cystic fibrosis patients [2]. P. aeruginosa infection treatment is often difficult because of the organism's intrinsic and acquired antibiotic resistance. This is due to the presence of multidrug efflux pumps [3], low outer membrane permeability [4], hypermutability [5], biofilm formation [6], and β-lactamase expression [7,8].
Comprehensive and elegant studies using Enterobacteriaceae established the paradigm of the β-lactamase induction mechanism. Orthologs of ampR, ampD, and ampG are found in numerous Gram-negative species [18]. Whether similar mechanisms are employed in all these organisms has not been established. It is possible that the induction mechanism could differ. The β-lactamase induction mechanism of P. aeruginosa has not been well-defined; however, it is known that P. aeruginosa AmpR regulates expression of ampC as in other organisms [8][9][10]. Similar to other systems, ampR is located upstream of the ampC gene [10]. Additionally, P. aeruginosa AmpR controls transcription of the oxacillinase, poxB, and several genes involved in virulence [8][9][10]. Loss of AmpR in P. aeruginosa causes a significant elevation in β-lactamase activity and other virulence factors [10]. P. aeruginosa also differs from other previously studied systems in that its genome has two ampG orthologs, PA4218 and PA4393 [19]. The current study reveals that these two genes, PA4218 and PA4393, are required for β-lactamase induction, hence they have been named ampP and ampG, respectively. Consistent with their putative roles as permeases, fusion analysis suggests that AmpG and AmpP have 14 and 10 transmembrane helices, respectively. Expression of ampP is dependent upon AmpR and is autoregulated. Together, these data suggest the distinctiveness of P. aeruginosa β-lactamase induction, as it is the first system that potentially involves two permease paralogs, and contribute to the general understanding of the induction mechanism.

Results
Genome Sequence Analysis of the PA4218 and PA4393 Operons E. coli AmpG has been shown to be a permease that transports GlcNAc-anhMurNAc peptides from the periplasm to the cytoplasm [13,17]; however, the AmpG function in P. aeruginosa has not been described. BLAST analysis of the E. coli AmpG sequence against the sixframe translation of the PAO1 genome identified two open reading frames, PA4218 and PA4393, with significant homology [20,21]. Global alignment using the Needleman-Wusch algorithm [22] demonstrated that PA4218 is 21.8% identical and 34.8% similar, while PA4393 is 23.2% identical and 34.3% similar to AmpG (Figure 1). The Pseudomonas Genome Database identifies PA4393 as encoding a putative permease with an alternate name of ampG, while PA4218 is identified as encoding a probable transporter [23]. Thus, PA4393 will be referred to as ampG and PA4218 as ampP (P for permease).
Analysis of the sequences around ampG and ampP revealed that they were in close proximity to two respective upstream ORFs. Based upon sequence analysis, it is likely that ampG and ampP constitute two twogene operons with their respective upstream ORFs (Figures 2A and 2B). PA4219 (ampO) overlaps the first seven base pairs of ampP ( Figure 2A). AmpO is a putative seven-transmembrane protein with a strong lipoprotein signal peptide that has a potential cleavage site between amino acids 18 and 19 [23]. The ampG gene is located 43 bp downstream from PA4392 (ampF), which encodes a putative protein with a DNA-protein cysteine methyltransferase domain ( Figure 2B). The function of this domain remains unknown. No lipoprotein signal was detected in AmpF.
To determine if ampG and ampP constitute two-gene operons with their upstream ORFs, RNA isolated from PAO1 was analyzed by reverse transcription polymerase chain reaction (PCR) using primers flanking the intergenic (ampF-ampG) ( Figure 3A) and the overlapping (ampO-ampP) region ( Figure 3B). The expected amplicon sizes are 136 and 158 bp for the ampF-G junction and ampO-P junction, respectively [23]. As expected, amplification was observed with genomic DNA ( Figures 3A and 3B, Lane 3). In the RNA analyses, PCR products were observed in reverse transcription PCR when the template was prepared in the presence of reverse transcriptase ( Figures 3A and 3B, Lane 1), but not in the control reaction when reverse transcriptase was omitted ( Figures 3A and 3B, Lane 2). This confirms that ampO and ampP constitute a two-gene operon and ampF and ampG constitute another. In addition, reverse transcriptase real time PCR data is in agreement with ampO and ampP belonging to the same operon and ampF and ampG comprising another operon (data not shown).

Topology analysis of AmpG and AmpP
The ampG and ampP genes encode predicted proteins with 594 and 414 amino acids, isoelectric points of 9.3 and 9.4, and calculated molecular weights of 64.6 kDa and 43.2 kDa, respectively. Hydrophobicity plots predict that AmpG has 16 or 14 predicted transmembrane (TM) helices, depending upon the algorithm used and AmpP has 10 [23]. To determine the membrane topology of AmpG and AmpP, phoA or lacZ was cloned downstream of the ampG and ampP genes. The 3'-end of the ampG and ampP genes were progressively deleted using exonuclease III. At various time-points, the truncated genes were ligated and assayed for PhoA and LacZ activities in E. coli. Clones were also sequenced to determine the reporter and amp gene junctions. Figure 1 Alignment of E. coli AmpG, PA4218 and PA4393. The primary sequence of E. coli AmpG, PA4218 (AmpP) and PA4393 (AmpG) were used as an input to M-Coffee, which combines multiple sequence alignments using the T-Coffee platform [45,46]. Identical and similar amino acids were shaded black and gray, respectively, using BOXSHADE.
b-lactamase activity in strains containing mutations in ampG and ampP The failure to induce C. freundii ampC in the absence of E. coli ampG suggested that AmpG is essential for  [41]. The Gm cassette (black inverted triangle) was inserted into the HincII and AscI sites of pKKF03 and pKKF04, respectively.  the induction of chromosomal β-lactamases [24,25]. To ascertain the role of the permeases in P. aeruginosa, isogenic ampG and ampP insertional inactivation mutants were constructed in the prototypic P. aeruginosa strain PAO1, referred to as PAOampG and PAOampP, respectively. The β-lactamase activity in the two isogenic mutants, PAOampG and PAOampP, was compared to PAO1. In the absence of β-lactam antibiotics, all strains showed a basal level of β-lactamase activity (Table 1). Upon challenge with 500 μg/ml of benzyl-penicillin, this level was elevated 10-fold (p < 0.05) in PAO1 (Table 1). However, the β-lactamase activities of PAOampP and PAOampG remained low in the presence of β-lactam antibiotic, indicating a loss of β-lactamase induction ( Table 1). The loss of inducibility in PAOampG could be partially restored by expressing ampG in trans, whereas the β-lactamase inducibility of PAOampP was completely recovered when ampP was supplied in trans (Table 1). Both PAOampP and PAOampG mutants had the other copy of the permease gene intact. These observations suggest that ampG and ampP are individually important members of the β-lactamase induction system. To confirm that ampG and ampP play independent roles, cross-complementation of PAOampP with pAmpG, and PAOampG with pAmpP was performed. Similar to the mutants, the cross-complemented strains did not show inducible β-lactamase activity (Table 1).
To further understand the role of ampG and ampP in β-lactamase induction, β-lactamase activity was assayed at different concentrations of benzyl-penicillin in PAO1, PAOampG and PAOampP ( Figure 5). Upon encounter with the inducer (25 μg/ml), there was approximately 38% induction ( Figure 5). For strain PAO1, this increase in β-lactamase activity continued in a dose-dependent manner until the maximum level of β-lactamase activity was reached when 100 μg/ml of benzyl-penicillin was added ( Figure 5). A higher concentration of inducer did not result in a concomitant increase in the expression of the β-lactamase ( Figure 5 and data not shown).
In PAOampG, the initial increase of β-lactamase activity was observed at 25 μg/ml, suggesting that this burst of β-lactamase production is ampG-independent ( Figure 5). However, unlike PAO1, the induction level failed to increase after 25 μg/ml of benzyl-penicillin and even significantly decreased with addition of increased concentrations of benzyl-penicillin ( Figure 5).
Mutation of ampP also prevented maximum induction of β-lactamase, but the defect was not quite as severe as in PAOampG. In PAOampP, the pattern of β-lactamase induction was very similar to PAO1 at concentrations of benzyl-penicillin up to 50 μg/ml ( Figure 5). However, unlike PAO1, addition of benzyl-penicillin at concentrations greater than 50 μg/ml failed to further induce production of β-lactamases ( Figure 5). Thus, low induction is independent of ampG or ampP. The observation that PAOampP exhibited higher levels of β-lactamase expression at higher concentrations of benzyl-penicillin may suggest that ampG plays a greater role at higher concentrations of β-lactam.
Most of the β-lactamase activity of P. aeruginosa can be attributed to AmpC, however, P. aeruginosa does contain another chromosomally encoded β-lactamase, PoxB [9,26]. To further analyze if the loss of β-lactamase induction in the PAOampG and PAOampP strains was due to loss of AmpC function, the ampC promoter (P ampC ) activity was measured in PAO1, PAOampG, and PAOampP. As expected, upon treatment with benzylpenicillin, P ampC -lacZ activity increased approximately 15-fold ( Figure 6). Benzyl-penicillin dependent induction of P ampC -lacZ was lost in PAOampG or PAOampP ( Figure 6).
To further characterize the role of ampG and ampP, the sensitivity of PAO1, PAOampG, and PAOampP to several β-lactams was determined (Table 2 and data not shown). Inactivation of ampG led to a significant decrease in resistance to amoxicillin (> 16-fold) and imipenem (> seven-fold). No difference was observed with ampicillin/sulbactam, cefaclor, cefepime, oxacillin, piperacillin, piperacillin/tazobactam, or ticaricillin/clavulonic acid (data not shown). Inactivation of ampP in PAO1 did not alter its resistance profile with these β-lactams (Table 2 and data not shown).

AmpR regulation of P ampFG and P ampOP
In inducible amp systems, the expression of ampC is tightly regulated by the transcription factor, AmpR [27]. In order to investigate the role, if any, of AmpR in the regulation of P. aeruginosa ampG and ampP, P ampFG -lacZ and P ampOP -lacZ promoter fusions were generated and integrated into the chromosome of PAO1 and PAOampR via attB-attP site-specific recombination. These constructs are likely to mimic the chromosomal regulation of the ampFG and ampOP operons. In the absence of inducer in PAO1 and PAOampR, there was a detectable basal level of promoter activity (Figure 7). The expression of the P ampOP -lacZ promoter fusion was significantly increased in the presence of inducer in the wild-type PAO1, and this induction was lost completely in PAOampR (Figure 7). However, the activity of the P ampFG -lacZ promoter fusion was comparable to the basal level in the absence and presence of inducer in PAO1 and PAOampR.

Autoregulation of the ampG and ampP genes
To determine if ampG or ampP affected their own or each other's expression, P ampFG -lacZ and P ampOP -lacZ promoter fusions were introduced into the chromosomes of PAOampP and PAOampG. Interestingly, the activity of the P ampOP -lacZ promoter fusion was significantly de-repressed in PAOampP in the absence and presence of inducer ( Figure 7). The activity of the P ampFG -lacZ was unchanged in PAOampG in either the absence or presence of benzyl-penicillin. Interestingly, P ampFG -lacZ activity was significantly increased in the presence, but not absence of β-lactam in the PAOampP mutant, suggesting β-lactam-dependent repression of P ampG by ampP. No change of the promoter activity of the ampOP operon was observed in the PAOampG mutant.

Discussion
Members of the Pseudomonadaceae family are intrinsically resistant to β-lactam antibiotics. Earlier reports Figure 5 Relative b-lactamase activity in PAOampP and PAOampG mutants. Assays were performed on the parental PAO1, and the mutants, PAOampP and PAOampG in the presence of benzyl-penicillin at a concentration gradient of 0 to 125 μg/ml. Cultures at OD 600 of 0.6-0.8 were induced for three hours before harvesting. Assays were performed on sonicated lysate using nitrocefin as a chromogenic substrate. The β-lactamase activity of PAO1 at 100 μg/ml of benzyl-penicillin was taken as 100%. Each value is the mean of at least three independent experiments. The asterisk refers to p-values of < 0.05 with respect to PAO1, which were calculated using the two-tailed Student's t-test. Figure 6 Activity of the ampC promoter. Promoter activity of the ampC gene was analyzed using lacZ transcriptional fusions integrated at the att locus of PAO1, PAOampR, PAOampG and PAOampP (see Materials and Methods and text for details). Cells were grown to an OD 600 of 0.6 -0.8, at which time cultures were divided into two and one set treated with 100 μg/ml benzylpenicillin. After three hours, cells were harvested and βgalactosidase activity assayed as described [10]. Each value is the mean of at least three independent experiments.
successfully identified ampC, ampR, ampD, and ampE as genes involved in the β-lactamase induction mechanism. However, the question of how chromosomal β-lactamase is induced remains elusive. This study examines the role of two previously uncharacterized P. aeruginosa putative permeases.

P. aeruginosa harbors two distinct and independent AmpG orthologues
In Enterobacteriaceae, besides AmpR, AmpD and AmpE, AmpG has also been implicated in the ampCencoded β-lactamase induction, acting as a membrane permease that transports 1,6-anhMurNAc-tripeptide and 1,6-anhMurNAc-pentapeptide [17]. In P. aeruginosa, two paralogs, PA4393/ampG and PA4218/ampP, were found ( Figure 1) [28]. Both ampG and ampP appear to be one member of two independent two-gene operons (Figures 2 and 3). PFAM analysis of AmpP identifies a Major Facilitator Superfamily (MFS1) domain between amino acids 14 and 346, in agreement with a role in transport [23,29,30]. Upstream from ampP is PA4219/ ampO, a gene that has seven putative transmembrane domains [23,31]. Together, these genes form an operon  Figure 7 Activity of the ampG and ampP promoters. Promoter activity of the ampG and ampP genes was analyzed using lacZ transcriptional fusions integrated at the att locus of PAO1, PAOampR, PAOampG and PAOampP (see Materials and Methods and text for details). Cells were grown to an OD 600 of 0.6 -0.8, at which time cultures were divided into two and one set treated with 100 μg/ml benzyl-penicillin. After three hours, cells were harvested and β-galactosidase activity assayed as described [10]. All 16 conditions were assayed at the same time but are divided into two panels for visualization purposes. Each value is the mean of at least three independent experiments. The asterisk refers to pvalues < 0.05, which were calculated using the two tailed Student's t-test.
( Figure 3) that is conserved in P. aeruginosa PA14, LES, PACS2, and PA2192 [23,32]. In contrast, PFAM analysis of AmpG does not reveal any significant hits, however, there was an insignificant match to the MFS1 domain (E = .00018) [29,30]. The ampG gene is downstream from PA4392/ampF, which encodes a protein with a putative 6-O-methylguanine-DNA methyltransferase domain [23,33]. These two genes also form an operon ( Figure 3) that is conserved in P. aeruginosa PA14, LES, and PA7 [23]. The topology of the E. coli AmpG permease has been analyzed using β-lactamase fusion proteins [15]. It was shown that AmpG has ten transmembrane domains with the amino-and carboxyl-termini localized to the cytoplasm [15]. In accordance with roles as transporters, AmpG and AmpP have 14 or 16 (depending upon the algorithm used) and 10, respectively predicted TM domains. PhoA and LacZ fusion analysis corroborates the existence of 14 and 10 TM domains in AmpG and AmpP, respectively ( Figure 4). In AmpG, the predicted transmembrane helices between amino acids 440 and 460 and either 525 and 545 or 555 to 575 of PA4393 are likely false positives. AmpG fusions at amino acids 438, 468 and 495 indicate that these amino acids are cytoplasmic (Figure 4), suggesting that if the region between amino acids 440 and 460 is membrane associated, it may be an integral monotopic domain. Similarly, AmpG fusions at residues 495 and 594 are cytoplasmic, while that at 540 is periplasmic, suggesting that if the region between amino acids 525 and 545 is membrane associated, it may be an integral monotopic domain. The fusion data indicates that the carboxyl-termini of both AmpG and AmpP are cytoplasmic ( Figure  4). Bioinformatic analysis predicts that the amino termini of both proteins are also cytoplasmic. Thus, like E. coli AmpG, both the amino and carboxyl termini would be cytoplasmic [15] (Figure 4).
Consistent with a role in transport, AmpP has an MFS domain [23,30]. The Major Facilitator Superfamily domain is present in approximately one-fourth of all known prokaryotic transport proteins [34]. Interestingly, most MFS proteins have 12 TM domains, while AmpP, like E. coli AmpG, has only 10 [35]. The topology analysis suggests PAO1 AmpG has 14 TM domains. PAO1 AmpG also has an insignificant MFS1 domain. A few MFS proteins have also been shown to have 14 TM domains [29,35].
The ampG and ampP genes are essential for maximum b-lactamase induction Because of the similarity between AmpG from Enterobacteriaceae and PAO1 AmpG and AmpP, β-lactamase levels of single ampG and ampP mutant isogenic strains were determined. Although an increase in β-lactamase activity was observed, neither the ampG nor ampP mutant strain produced the same level of β-lactamase in the presence of benzyl-penicillin as PAO1 (Table 1, Figure 5). Moreover, inactivation of ampG or ampP abolishes induction of P ampC (Figure 6). This indicates that both ampG and ampP are essential for chromosomal β-lactamase induction. These genes did not cross-complement or exhibit gene dosage effects indicating that they play different roles in the induction pathway (Table 1). These results are consistent with recent data demonstrating that mutation of ampG affects induction of β-lactamase and failure of ampP to complement an ampG mutation [28]. Furthermore, the analysis using increasing benzyl-penicillin concentrations, shows that ampP plays an important role at lower inducer concentrations, whereas ampG is crucial at higher concentrations ( Figure 5). Mutation of ampG affects PAO1 β-lactam resistance ( Table 2) [28]. Recent studies by Zhang et al., in which deletion of ampG results in increased sensitivity to ampicillin [28], are consistent with results presented here ( Table 2). In addition, ampG inactivation increases imipenem sensitivity ( Table 2). Loss of ampP (also referred to as ampGh1) function did not affect β-lactam sensitivity in either study (Table 2) [28]. AmpP (PA4218) has previously been named FptX due to its homology to RhtX in Sinorhizobium meliloti 2011 [36]. PA4219 does not have a S. meliloti orthologue [36]. Mutation of ampP in a P. aeruginosa CDC5 derivative that produces pyochelin but not pyoverdine, resulted in loss of pyochelin utilization [36]. In agreement with a role in pyochelin utilization, ampP is located next to genes involved in pyochelin biosynthesis and transport [23,36]. Thus, the results presented in Table 1 and Figures 5 and 6 demonstrate that ampP is involved in β-lactamase induction in addition to its previously characterized role in pyochelin utilization [36].

Expression of ampP is induced by b-lactam addition in the presence of ampR
Despite the importance of ampG in β-lactamase induction, little is known about its regulation. E. coli ampG is also the second gene in a two gene operon. Upstream and divergently transcribed from the E. coli ampG operon, is the bolA transcriptional regulator [24]. Expression of bolA is dependent upon RpoS. Previous studies suggest the expression of the E. coli ampG gene is independent of bolA, rpoS or ampD [24]. Neither the P. aeruginosa ampG nor ampP gene is located near the bolA locus [23], thus P ampFG and P ampOP -lacZ transcriptional fusions were integrated into the chromosome of isogenic PAO1 strains to begin to understand ampG and ampP regulation.
In light of the requirement of ampG and ampP for maximum P. aeruginosa β-lactamase induction, it was of interest to determine if expression of either was affected by β-lactam addition (Table 1, Figure 5). In the absence of antibiotic, P ampFG and P ampOP were constitutively expressed. Expression of P ampOP significantly increased in the presence of inducer, while P ampFG did not (Figure 7).
The LysR type transcriptional regulator AmpR induces the expression of the AmpC β-lactamase in the presence of β-lactam antibiotics [27]. AmpR also affects the regulation of additional genes involved in P. aeruginosa antibiotic resistance and virulence [10]. Insertional inactivation of ampR, did not affect P ampFG -lacZ activity, however, the increase in P ampOP -lacZ activity previously observed upon β-lactam addition was lost in the absence of ampR (Figure 7). This indicates that ampP expression is regulated by AmpR. Future analyses will determine if this regulation is direct or indirect.
ampP affects regulation of both its own promoter and that of ampG Given that both ampG and ampP are required for maximum β-lactamase expression, both contain structural elements consistent with roles in transport, and the regulation of ampP expression by β-lactam and ampR, it was feasible that ampP could contribute to its own expression, perhaps by transporting potential effector molecules for AmpR. Indeed, ampP does appear to inhibit its own expression, as P ampOP activity increased ten-fold in PAOampP in the absence, and approximately seven-fold in the presence of β-lactam ( Figure 7). Insertional inactivation of ampP also resulted in increased expression of P ampFG in the presence of β-lactam (Figure 7).

Proposed model for regulation of b-lactamase induction
The results presented contribute to what is known concerning β-lactamase induction in P. aeruginosa. It is well established that induction of the expression of the AmpC β-lactamase is dependent upon AmpR. Although the exact mechanism has not been well characterized in P. aeruginosa, it is believed that the induction is triggered by conversion of AmpR from a repressor to an activator (Figure 8). Evidence from Enterobacteriaceae suggest that this is a result of inhibition of the cell wall remodeling process by β-lactam antibiotics and subsequent accumulation of intermediates which in turn serve as effector molecules for AmpR. Consistent with previous findings suggesting that AmpR acts as a positive regulator of amp genes [10], activation of ampP expression required the presence of AmpR and β-lactam antibiotic (Figure 7). Based upon glycopeptide accumulation studies in other organisms, these findings suggest that the accumulation of 1,6-anhMurNAc-tripeptide and 1,6-anhMur-NAc-pentapeptide in the presence of β-lactam antibiotics activates AmpR that in turn up-regulates the expression of ampP. However, P. aeruginosa appears to use two non-redundant permeases in β-lactamase induction, suggesting, one may be involved in the import of muramyl peptides and the other in an as yet unknown function. The second permease may be involved in export of muramyl peptides or import of different muramyl peptides. Further studies to determine the identity of these peptides and how they regulate AmpR will be a critical next step in deciphering βlactam resistance in P. aeruginosa.
The ampP gene is also auto-regulated via an unknown mechanism. If AmpP performs a similar function as E. coli AmpG, the absence of ampP would result in the accumulation of the periplasmic pool of GlcNAc-anhMurNAc peptides or the reduction in the cytoplasmic pool of 1,6-anhMurNAc-tripeptide and 1,6-anhMurNAcpentapeptide alerting the cell that the peptidoglycan recycling process is inhibited. This signalling could result in a positive feedback mechanism that up-regulates the expression of ampP. The accumulation of the periplasmic pool of 1,6-anhMurNAc-tripeptide and 1,6-anhMurNAcpentapeptide in PAOampP is also likely to up-regulate the expression of P. aeruginosa PAO1 ampG in the presence of β-lactam.
Currently, it is not known if PAO1 AmpG and AmpP function similarly to E. coli AmpG, however, like ampG, the PAO1 ampG and ampP are important for β-lactamase induction [14] ( Figure 5, Figure 6, Table 1). Moreover, ampG and ampP are not functionally redundant as both are required for maximum induction. Their expression is also differentially regulated. An ampP promoter-lacZ fusion exhibited increased activity in the presence of ampR and β-lactam or the absence of ampP. An ampG promoter-lacZ fusion was unaffected by the absence or presence of ampR or ampG. Increased β-galactosidase activity was observed from the ampG promoter fusion in the presence of β-lactam in an ampP mutant (Figure 7). It is not known if this is dependent upon ampR, related to an ampR-independent function of ampP in β-lactamase induction or the function of ampP in pyochelin utilization.

Conclusions
P. aeruginosa appears to have two ampG paralogs, ampG and ampP, which encode proteins with 14 and 10 transmembrane domains. Both are required for maximum induction of chromosomal β-lactamase and induction of the ampC promoter. Expression of ampP did not restore maximum β-lactamase induced activity in an ampG mutation nor did expression of ampG complement an ampP mutation, indicating that ampG and ampP have distinct functions in β-lactamase regulation. In addition to being autoregulated, ampP is regulated by AmpR and β-lactam. ampP is also involved in the regulation of ampG in the presence of β-lactam. In summary, the presence of two distinct permeases required for β-lactamase induction suggests that the P. aeruginosa β-lactamase resistance mechanism is more complex and distinct from the current paradigm.

Bacterial strains, plasmids and media
Bacterial strains, plasmids and primers employed in this study are shown in Table 3. E. coli and P. aeruginosa were routinely cultured in Luria-Bertani medium (10 g tryptone, 5 g yeast extract, 5 g NaCl, per liter). Pseudomonas Isolation Agar (PIA, Difco) was used in triparental mating experiments. Mueller-Hinton agar (Difco) was used in E-test experiments. Antibiotics, when used, were at the following concentrations (per liter) unless indicated otherwise: ampicillin (Ap) at 50 mg, tetracycline (Tc) at 20 mg, gentamycin (Gm) at 30 mg for E. coli and carbenicillin (Cb) at 300 mg, Gm at 300 mg and Tc at 60 mg for P. aeruginosa.

DNA manipulations
Standard procedures in molecular biology were performed as previously described [37].

Insertional inactivation of the ampG and ampP genes
A 2904-bp ampG fragment was PCR-amplified from PAO1 genomic DNA using KKF01ampGFor and KKF04ampGRev (Table 3). Similarly, KKF05ampPFor and KKF08ampPRev were used to PCR-amplify a Figure 8 Model for regulation of AmpC b-lactamase induction by AmpR, AmpP and AmpG in P. aeruginosa. In Enterobacteriaceae as well as P. aerugniosa, the induction of β-lactamase expression is due to the action of the LysR transcriptional regulator, AmpR. In vitro studies suggest that AmpR can act as either a repressor or an activator, depending upon the presence of different peptidoglycan remodelling intermediates. In this study, it is shown that unlike previously characterized systems, P. aeruginosa has two putative AmpG permease paralogs, AmpG and AmpP. Expression of AmpP is inducible by β-lactam in an ampR-dependent manner. The ampP gene also appears to repress its own expression independent of β-lactam through an unknown mechanism. Although not observed to be induced by β-lactam in a PAO1 background, expression of ampG also appears to be repressed by ampP in the presence of β-lactam (see text for details).  [38] pEX100T Ap R ; sacB oriT [39] pMF54 ColE1-SF broad-host replicon [48] pME6030 Tc R ; oriV pVS1 oriV p15A oriT Km R ; ColE1ori-Tra (RK2) + [40] pTrcphoA Ap R ; low-copy trc promoter expression vector carrying the lacI q and phoA [43] pTrclacZ Ap R , low-copy trc promoter expression vector carrying the lacI q and lacZ [43] pSJ10 Tc R ; CTX-lacZ fused with ampC promoter, P ampC [10] pKKF003 2779-bp ampP fragment. The ampP and ampG PCR products were cloned into pCRII-TOPO according to the manufacturer's instruction (Invitrogen, CA), generating pKKF04 and pKKF03, respectively. A Gm cassette carrying the aacCI gene was retrieved from pUCGm [38]. The cassette was inserted into the unique HincII and AscI restriction sites of ampP and ampG, respectively, creating pKKF145 and pKKF149 ( Figure 2). These insertions created a polar mutation in the 5'-ends of ampP and ampG ORFs in pKKF04 and pKKF03, respectively. Subsequently, the ampP::aacCI and ampG::aacCI from pKKF145 and pKKF149, respectively, were subcloned into the SmaI site of pEX100T [39], a mobilizable suicide plasmid. These plasmids were conjugated into P. aeruginosa PAO1, with a helper strain harboring pRK2013 [40]. The merodiploids, resulting from homologous recombination, were selected with PIA containing Gm. These Gm R colonies were then screened for Gm resistance and Cb sensitivity by replica plating. The insertions were confirmed by PCR and restriction analysis of the PCR product (data not shown). The PAO1 isogenic strains with defective ampP and ampG are henceforth referred to as PAOampP and PAOampG, respectively.

Construction of ampP and ampG complementing plasmids
Plasmids containing ampP and ampG, pKKF73 and pKKF69, respectively, were generated by inserting the EcoRI fragment with ampP and ampG from pKKF004 and pKKF003 into a broad-host range, low copy number vector, pME6030 [41]. These were later conjugated into PAOampP and PAOampG for complementation analysis.

Promoter-lacZ fusion constructions
The putative promoter regions of ampG and ampP were subcloned from pKKF003 and pKKF004 into pGEMEX-1, respectively, generating pKKF091 (P ampFG -lacZ) and pKKF087 (P ampOP -lacZ) ( Table 3). This suicide vector contained the integration-proficient attP site, which recombines into the chromosomal attB site to generate a single-copy reporter fusion [42]. The resulting clones were mobilized into PAO1 and PAOampR ( Table 3). The presence of the chromosomal insertions was confirmed by PCR and restriction analysis of the product.

Topological analysis of AmpP and AmpG
The topology of AmpP and AmpG were investigated using two markers, phoA and lacZ, that function in the periplasm and cytoplasm, respectively. The entire ampP gene was PCR amplified using primers KKF13ampP2For and KKF14ampP2Rev and cloned into pTrcphoA [43]. The entire ampG gene was PCR amplified using primers KKF09ampG2For and KKF10ampG2Rev and cloned into a pTrcphoA plasmid which had been modified by insertion of a broad host range stabilization fragment from pMF54 (Table 3). Both ampG and ampP genes were cloned into pTrclacZ [43]. The erase-a-base system (Promega, WI) was used to generate deletions of the genes from the 3'-ends. The resulting clones were then sequenced to determine the fusion junctions. The phoA and lacZ activities were determined as previously described [44].

Determination of minimal inhibitory concentrations (MICs)
MICs were determined using E-test strips (Biomerieux, Marcy l'Etoile, France) according to the manufacturer protocols.

Reverse transcription PCR
For the reverse transcription PCR, RNA was isolated from PAO1 using the RNAeasy mini kit (Qiagen, Valencia, CA) according to the manufacturer protocol. DNA was removed by two sequential 1 hour treatments at 37°C with RQ DNaseI (Promega Corporation, Madison, WI) followed by heat inactivation at 65°C for 10 minutes. Synthesis of cDNA was performed with Superscript III reverse transcriptase (RT) (Invitrogen, Carlsbad, CA) using a (NS) 5 random primer and 5 μg RNA according to the manufacturer protocol. A control reaction containing all components except for Superscript III RT was performed in parallel. After cDNA synthesis, RNA was removed by treatment with 0.2 N NaOH for 30 minutes at 65°C. The reactions were neutralized by addition of 0.2 N HCl and cDNA was purified using the QIAquick PCR purification kit (Qiagen, Valencia, CA) according to the manufacturer protocol. PCR reactions to amplify the ampF-ampG intergenic region were performed using primers PA4392_3junc-tionRTF and PA4392_3junctionRTR (Table 3) using GoTaq Flexi (Promega Corporation, Madison, WI). PCR reactions to amplify the ampO-ampP overlapping region were similarly performed with the exception that primers PA4218_9junctionRTF and PA4218_9junctionRTR (Table 3) were used. PCR products were analyzed by electrophoresis on a 10% polyacrylamide/1× TBE gel followed by staining with SybrSafe (Invitrogen, Carlsbad, CA).