Origin and Evolution of the AmpC β-Lactamases of Citrobacter freundii

Bacterial strains and plasmids.

E. coli K-12 strain JM109 (33) was used as the host for recombinant plasmids. Plasmid pACSE2 was used as the vector for cloning and expression of the different ampC alleles. pACSE2 (5,201 bp) was created in two steps. Step 1 involved replacing the MscI fragment (bp 2938 through 3980 of pACYC184) with bp 6 through 622 of pSE380 (Invitrogen) to yield plasmid pACSE. Step 2 involved amplifying bp 3269 through 167 of pSE380 by using primers with SpeI and XbaI sites and ligating the amplicon, which contains lacI^q, into the SpeI site of pACSE. pACSE2 carries the p15A origin of replication (5 to 10 copies/cell); a gene for tetracycline resistance, lacI; and the pTrc promoter-multiple cloning region of pSE380. Expression from the pTrc promoter is regulated by the lacI^q gene product. Part of the chloramphenicol resistance gene of pACYC184 is deleted in pACSE2.

C. freundii strains 6879 and 8454 were obtained from the American Type Culture Collection. Both were collected in the 1920s or early 1930s, prior to the clinical introduction of penicillin (28, 32). Cultures of both strains were spread onto Mueller-Hinton agar (Difco). BBL antibiotic disks containing 30 μg of aztreonam, 30 μg of cefepime, 10 μg of meropenem, 10 μg of imipenem, 100 μg of piperacillin, 30 μg of ceftazidime, 30 μg of cefotaxime, or 30 μg of cefuroxime were applied to the plates. Both strains were sensitive to all of these antibiotics because, following overnight incubation at 37°C, inhibition zone diameters were at least 33% greater than the NCCLS (22) threshold for sensitivity.

G. Jacoby supplied E. coli J53/pMG263 (CMY-2). E. Tzelepi and L. S. Tzouvelekis supplied E. coli XL1Blue/pPHPI-5/XL (LAT-1), E. coli XL1Blue/pMEL/XL (LAT-2), E. coli XL1Blue/pMEL-1/XL (LAT-3), and E. coli XL1Blue/pMEL-2/XL (LAT-4). A. Philippon supplied Proteus mirabilis 34955 (CMY-4) and P. mirabilis LAR (CMY-3). A. Fosberry and D. Payne supplied E. coli TC446.891 (BIL-1). D. Sirot provided P. mirabilis CF09 (CMY-3b).

Plasmids pMXCMY2, pMXCMY3, pMXCMY3b, pMXCMY4, pMXLAT3, pMXLAT4, pMX6879, and pMX8454 were created by amplifying the corresponding ampC alleles and cloning them into plasmid pACSE2.

Recombinant DNA methodology.

The ampC alleles were amplified with a Failsafe (Epicentre) polymerase system (buffer A) by using primers P1 (TTGCTTTTAATTACGGAACTGATTT) and P2 (GGAGCTCTAAGTGTAGATGACAGCAGGRAAA), which has a SacI site (boldface) at the 5′ end of the primer. Amplification reactions were carried out by using the following program: 25 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 30 s, and extension at 72°C for 90 s, followed by 1 cycle at 72°C for 5 min.

The resulting amplicons were purified with Qiagen PCR purification columns according to the manufacturer's instructions and digested with restriction endonucleases BspHI (New England Biolabs) and SacI (New England Biolabs). A BspHI site is located at the ATG start site of the C. freundii ampC gene. Plasmid pACSE2 was digested with restriction endonucleases NcoI and SacI and dephosphorylated with calf intestinal alkaline phosphatase from New England Biolabs.

Digested amplicons and the vector were purified with Qiagen PCR purification columns according to the manufacturer's instructions, combined, ligated with T4 DNA ligase (Gibco BRL), and transformed into strain JM109 by electroporation.

Sequencing.

For each allele that was cloned, the sequence was verified by sequencing an amplicon of the region of the plasmid containing the ampC insert. Amplification reactions were carried out by using Taq PCR Master Mix (Qiagen) under the conditions described above. The amplicons were purified with Qiagen PCR purification columns and sequenced on an ABI Prism 377 DNA sequencer with an ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction kit (PE Applied Biosystems) according to the manufacturer's protocol.

For the pre-antibiotic era alleles, the sequence was determined for a single strand of both the original amplicon and the cloned copy of the gene. For both alleles, the sequences agreed.

We verified the published sequences of the cloned alleles by sequencing a single strand. When the cloned sequence differed from the published sequence, both strands of the original amplicon generated by primers P1 and P2 were sequenced to determine whether the differences were the result of amplification errors or of errors in the published sequence.

Expression of ampC.

Cloned ampC alleles from pACSE2-based plasmids are expressed from the TAC promoter, which is tightly regulated by the lacI gene product. Expression of the ampC alleles was induced by isopropyl-β-d-thiogalactopyranoside (IPTG). To determine the concentration of IPTG required to induce the expression of ampC at physiological levels, we determined the MIC of piperacillin for strain E. coli J53 carrying plasmid pMG263, a naturally occurring plasmid bearing the CMY-2 allele of ampC. We then determined the concentration of IPTG required to give an equivalent MIC of piperacillin for JM109 carrying plasmid pMXCMY2. IPTG at 62.5 μM induced the expression of CMY-2 to the appropriate level.

Antibiotics and determination of MICs.

The MICs of the following antibiotics were determined: ampicillin (Sigma), piperacillin (Sigma), piperacillin-tazobactam (8:1) (Lederle), temocillin (SmithKline Beecham), ticarcillin (SmithKline Beecham), ticarcillin-clavulanate (30:1) (SmithKline Beecham), cephalothin (Sigma), cefoxitin (Merck & Co.), cefuroxime (Sigma), ceftriaxone (Roche), ceftazidime (Glaxo Wellcome), cefotaxime (Sigma), cefepime (Bristol-Myers Squibb), imipenem (Merck & Co.), meropenem (Zeneca), and aztreonam (Bristol-Myers Squibb). Stock solutions of these drugs were prepared in 0.1 M NaPO₄ buffer (pH 7.0) and stored at −80°C in single-use aliquots.

MICs were determined with cultures at a concentration of 10⁵ cells/ml in Mueller-Hinton broth (Difco) containing 62.5 μM IPTG by application of twofold serial dilutions of each antibiotic to each tested strain. Cultures were incubated at 37°C on a rotating shaker for 24 h. The MIC was the lowest concentration of antibiotic that completely blocked the growth of the tested strain.

Phylogenetic methods.

The sequences of 59 genes that are closely related to the plasmid-borne CMY-2 ampC gene were identified by a BLAST search (1, 2) of the nonredundant protein database at the National Center for Biotechnology Information. Those protein sequences were aligned with the sequences of CMY-2, the sequence of CMY-3b (8), and the deduced amino acid sequences of AmpC from pre-antibiotic era C. freundii strains Cfre8454 and Cfre6879 by using the multiple alignment program ClustalX (14, 30, 31). The Gonet 250 similarity matrix was used with a gap opening penalty of 35 and a gap extension penalty of 0.75 for the pairwise alignment stage and a gap opening penalty of 15 and a gap extension penalty of 0.3 for the multiple alignment stage. When sequencing revealed errors in the published sequences, the corrected sequences were used in the alignment.

The corresponding DNA coding sequences (accession numbers are given in Table 1) were aligned by introducing triplet gaps between codons corresponding to gaps in the aligned protein sequences with the program CodonAlign (version 1.0). CodonAlign for Macintosh and for PC (Windows) computers and a source code that can be compiled for other platforms are available at no charge at http://www.rochester.edu/College/BIO/labs/HallLab/index.html.

Phylogenies were constructed by the Bayesian method (19, 20, 27) as implemented with the program MrBayes (version 1.04). MrBayes is available from http://brahms.biology.rochester.edu/software.html. The Bayesian method seeks the most likely trees given the data (the alignment) and the evolutionary model. The posterior probabilities of the phylogenies, branch lengths, and substitution parameters cannot be calculated directly; however, they can be approximated by the Markov chain Monte Carlo (MCMC) process by sampling trees from the posterior probability distribution. A variant of the MCMC process called the metropolis-coupled Markov chain Monte Carlo (MCMCMC) process runs several chains, some of which are heated. A heated Markov chain has the posterior probability of a tree raised to some power i. Heated Markov chains can more easily cross deep likelihood valleys. The effect of heating is to fill in valleys and lower peaks; hence, a heated Markov chain can better explore the parameter space. Using the MCMCMC algorithm, a swap of the states between two chains is attempted at each step. If the swap is accepted, then the states for the two chains are exchanged. If a swap occurs between a heated chain and the cold chain, the cold chain might cross a large valley that it would normally cross with only a very small probability. A more detailed discussion of the method and its implementation is provided in reference 13.

Trees were saved every 100 generations. At the end of the run, a consensus tree with branch lengths was calculated from the saved trees. As the chains ran, the ln likelihoods of the trees converged on a stable value. Only trees saved well after the ln likelihoods had converged were used to calculate the consensus tree.

The evolutionary model was the general time reversible model (29), and among-site variation in the evolutionary rate was estimated separately for first, second, and third positions of sites within codons. Four chains, with a “temperature” of 0.2 for the heated chains, were run for 1,010,000 generations, with sampling of trees every 100 generations. The ln likelihoods of the trees had converged on a constant value by generation 10,000, i.e., after saving of 100 trees. The consensus tree was calculated from the final 10,000 trees visited, well after convergence had occurred.

One of the advantages of Bayesian inference of phylogeny is that the results are easy to interpret. For example, the sum of the posterior probabilities of all trees is 1. Moreover, the posterior probability of any single clade is simply the sum of the posterior probabilities of all trees that contain that clade. The consensus tree calculated by MrBayes does not include the posterior probabilities of the clades; thus, the entire set of trees was imported into PAUP* (version 4.0b4a; Sinauer Associates, Sunderland, Mass.), and the same trees used by MrBayes to calculate a consensus were used to calculate a 50% majority rule consensus in PAUP*. The resulting tree shows the posterior probabilities of the clades. The consensus tree calculated by MrBayes was imported into PAUP* for the purposes of displaying and printing the tree.

Sequence corrections.

Analysis of the sequences for LAT-1, LAT-2, LAT-3, LAT-4, and BIL-1 revealed several errors in the published sequences for their alleles (Table 2).

Correction of those sequences showed that the sequences for BIL-1, LAT-2, and CMY-2 are identical (BIL-1-LAT-2-CMY-2). LAT-1 is identical to LAT-4, and LAT-4 and LAT-1 differ from CMY-2 by one nonsynonymous substitution. LAT-3 is the same as CMY-6. The differences are summarized in Table 3.

Phylogenetic analysis.

Figure 1 shows a Bayesian consensus phylogeny of the ampC genes. The phylogeny was rooted by using the alleles for Oant82116 and Oant2171 as the outgroup. These alleles come from the chromosome of Ochrobactrum anthropi, a species in the alpha group of the Proteobacteria, whereas all of the other ampC genes come from the chromosomes of species in the gamma group. This phylogeny is largely consistent with the taxonomic groupings of the gamma group. All of the Enterobacteriaceae are contained within a single clade, the members of which are descended from a single common ancestor. The Aeromonadaceae and Pseudomonadaceae are contained within a different single clade. For clades that had a probability of less than 90%, the most probable clade was used, and the probability of that clade is shown at its root.

Figure 2 shows a Bayesian consensus phylogeny of the C. freundii ampC alleles. CfreI113 was used for the outgroup because it is the outgroup for this clade in the multispecies ampC phylogeny shown in Fig. 1. The Bayesian estimates for the number of nucleotide changes that have occurred along each of the branches are shown, as are the probabilities of clades that are less than 100%.

Phenotypic analysis.

The MICs of ampicillin, ticarcillin, ticarcillin-clavulanate, piperacillin, pipercillin-tazobactam, temocillin, cephalothin, cefoxitin, cefuroxime, ceftriaxone, cefotaxime, ceftazidime, cefepime, imipenem, meropenem, and aztreonam were determined for each strain carrying the cloned C. freundii ampC alleles (Table 4). Each of the alleles was cloned into the same vector and transformed into the same host background so that the resistance levels conferred by the alleles could be directly compared. Because the CMY-2-BIL-1-LAT-2 allele is the ancestor of the other plasmid-borne alleles, any evolution of increased activity toward an antibiotic would result in an MIC higher than that for CMY-2-BIL-1-LAT-2; therefore, the MIC for the CMY-2-BIL-1-LAT-2 allele should be used as a reference for comparing the MICs for the other alleles.

For the antibiotics tested, there was very little difference between the MICs for JM109/pMXCMY2 and those for the other strains containing the cloned alleles. There was a 4- to 16-fold increase in the cefuroxime and aztreonam MICs relative to those for JM109/pMXCMY2.