The availability of genome sequences for hundreds of bacteria has helped reduce the disparity between the quality of physiological analyses available for these organisms and those possible for more traditional model organisms. Although genome-enabled approaches such as proteome and transcriptome analysis are now feasible for a great many systems, many of these non-traditional model organisms still have only limited genetic tools available for genetic manipulations. One particularly important tool is an allelic exchange vector to perform reverse genetic analysis. This generally involves homologous recombination of a selectable marker, such as an antibiotic resistance cassette, into the desired site on the chromosome. Issues arise, however, when the planned number of genetic manipulations outnumbers the number of markers available, or if the effect of mutations such as nucleotide substitutions, is desired.

Generating more genetic multiple manipulations than the number of markers requires a more sophisticated approach than just homologous recombination. One strategy used successfully in Escherichia coli is to use a combination of in vivo restriction of a donor plasmid with a transiently-expressed, highly active recombination system based on Lambda Red recombinase 1. The goal here is to perform recombination in the complete absence of selection at a frequency sufficiently high (>1%) to be screened for directly. Although this is a rather powerful system, it is unclear whether Lambda Red can be utilized outside of some taxa within the γ-proteobacteria. Attempts to introduce antibiotic markers into chromosomal loci the α-proteobacterium used in this work, Methylobacterium extorquens AM1, failed to produce any desired recombinants (Marx, unpublished). A second option for generating multiple mutations is to excise the markers utilized with an in vivo, site-specific recombinase system. Options that have been employed to this end include the recombinase/recognition site pairs cre/lox 2 and Flp/FRT 3. Although broad-host-range systems like this have been developed (for example 4), there remain some disadvantages. Notably, although the selectable marker itself is removed, recombination between the two, collinear recombinase recognition sites leaves behind a single copy of this site. Depending on the system in use, the minimal scar left behind is the recombination recognition sites of 34 (loxP) and 48 (FRT) bp in length. Additionally, multiple cloning sites between the two regions of homology cloned into the donor vector will also be left behind in the scar after marker removal. The introduction of scars is a particular problem when trying to cleanly assess the physiological effect of mutations that are more 'subtle' than knockouts or insertions, for example, a point mutation that results in an amino acid substitution in the encoded protein.

A third option for multiple genetic manipulations, which also avoids leaving behind undesired scars, is to use an "in-out" system (Figure 15). The basic idea behind these techniques is to first employ positive selection to select for single crossover integration of the entire donor vector due to recombination between a cloned region spanning the desired mutation in the vector and the corresponding chromosomal site. In the second step, negative selection is used to select for isolates that have recombined out the vector sequence. If the second recombination event excising the vector occurs on the same side of the introduced mutation as the first recombination event that introduced it onto the chromosome, the original chromosomal locus will be restored unchanged. If the second recombination event occurs on the opposite side of the introduced mutation, however, this results in excision of the original allele and the new mutation remains. As such, negative selection results in colonies with both resulting final states, as well as some percentage of false-positives that are resistant but have not excised the vector. As long as the false positives do not dominate, and the recombination rates to each side of the introduced mutation are reasonably balanced, screening of a modest collection of resulting recombinants will generate the desired unmarked mutation.

An "in-out" allelic exchange vector for generating unmarked mutations therefore must be able to be introduced into the recipient organism, be incapable of vegetative replication, and bear appropriate markers for positive and negative selection. Positive selection is generally accomplished using any number of antibiotic resistance genes, whereas comparably fewer options for negative selection generally exist. The most commonly used techniques are to use streptomycin (Sm) sensitivity, which comes as a pleiotropic effect of expressing the tetracycline (Tc) efflux pump 6, or sucrose-sensitivity that results from expression of levansucrase, encoded by sacB 7. Levansucrase activity is lethal in the presence of sucrose for most gram-negative bacteria. This paper presents a facile, broad-host-range "in-out" system based on sacB that has been specifically designed to allow facile unmarked allelic exchange in a wide variety of bacterial taxa. In order to test this system, allelic exchange has been performed at three different loci in M. extorquens AM1 89.

In order to generate a facile system for marker-free allelic exchange across a wide variety of bacterial species, the loxP-flanked kanamycin (Km) resistance cassette of the broad-host-range marker-recycling vector, pCM184 4 was first excised and replaced with a synthetic linker that introduced three new restriction sites to the extensive multiple-cloning sites. Subsequently, a fragment from pDS132 10 bearing sacB and cat (encoding levansucrase and chloramphenicol (Cm) acetyltransferase, respectively) was introduced, generating pCM433 (Figure 2). It may be noted that initial attempts were made to take advantage of the potential negative selection (Sm sensitivity) afforded by expression of the Tc efflux pump present on pCM184. Sm sensitivity was found to be enhanced in tet bearing cells, but the sensitivity was too modest to be utilized effectively for negative selection (Marx, unpublished results).

Three loci of interest in M. extorquens AM1 were chosen to test the utility of pCM433 for allelic exchange. These loci were hprA (encodes hydroxypyruvate reductase, a key enzyme of the serine cycle for assimilation of formaldehyde into biomass, 11), mptG (encodes β-ribofuranosylaminobenzene 5'-phosphate synthase, the first dedicated enzyme for the synthesis of tetrahydromethanopterin, the C₁-carrier molecule used for this organism's formaldehyde oxidation pathway 1213), and crtI (encodes phytoene desaturase, a necessary enzyme for carotenoid biosynthesis 14).

In all cases, constructs were made to convert the allele from wild-type (wt) to mutant, and the reciprocal reversion of mutant to wt. To accomplish this, both the ancestral, wt allele and the deletion (ΔhprA, ΔmptG) or insertion (crtI⁵⁰², generated by insertion of ISphoA/hah-Tc into crtI, followed by Cre-mediated excision of all but 132 bp of the IS 14) alleles were amplified by PCR, cloned into pCR2.1, sequenced, and then introduced into pCM433. Each of these donor plasmids were then introduced into the appropriate target strain via triparental conjugations and plated onto Tc plates (also containing Rif for counter-selection against E. coli). Tc^R transconjugants were obtained at a frequency of 10^-6 to 10^-7. In some cases, even these single-crossover recombinants that contained both the wild-type and mutant alleles exhibited a phenotype. For example, the pool of single-crossover intermediates from either pCM441 (wt crtI allele) inserted into the white CM502 strain or pCM440 (defective crtI⁵⁰² allele) inserted into the pink CM501 strain each contained Tc^R colonies of both colors. As such, one pink and one white isolate from the conjugation into each background were isolated (CM1263 (white) and CM1264 (pink) from CM502, and CM1265 (pink) and CM1266 (white) from CM501). A polar effect of pCM433 insertion into this site was clearly observed. Irrespective of whether the wt allele was being introduced into the mutant, or vise versa, strains with the wild-type allele upstream, proximal to the gene's promoter (as determined by PCR analysis for strains CM1264 and CM1265), were pink, carotenoid-containing colonies, whereas the other strains (CM1263 and CM1266) had the crtI⁵⁰² allele upstream of pCM433 and were white.

In order to select for recombinants that have excised the vector, suspensions of Tc^R isolates were diluted and plated onto plates containing various levels of sucrose (2.5, 5, and 10% w/v). At all sucrose levels sucrose-resistant colonies were obtained at a frequency of 10^-4 to 10^-5. These colonies were then screened for Tc sensitivity (indicating the expected loss of the pCM433-based construct), as well as the expected mutant phenotype (inability to grow on methanol for ΔhprA and ΔmptG, white colonies (versus pink) for crtI⁵⁰²). These were confirmed via PCR analysis using primers situated outside the region of the locus where recombination occurred. In the cases presented here, differences in the size of amplified products sufficed to distinguish the alleles used, but primers designed to distinguish single-nucleotide substitutions (or sequencing) have been used in subsequent studies (Chou and Marx, unpublished). Overall, a false positive rate of sucrose^R, Tc^R strains generated here in M. extorquens AM1 was 26% (105/402). It should be noted, however, that the range of frequencies varied from 0% to 78% for different construct/recipient pairs. This is likely related to the rate of recombination for the flanking regions of each locus as compared to the rate of generating sucrose-resistance from other mechanisms. For all three loci, wild-type alleles were replaced by mutant alleles, and vise versa. In subsequent work, dozens of allelic exchanges including the introduction of single-nucleotide substitutions have been successfully performed utilizing this system (Chou and Marx, unpublished).

M. extorquens AM1 strains were grown at 30°C on agar plates with "Hypho" minimal salts medium 15; E. coli were grown at 37°C on Luria-Bertani agar 16. Substrates and antibiotics were used at the following concentrations: methanol (125 mM), succinate (15 mM), sucrose (5% w/v unless otherwise stated), 50 μg/ml Ap (ampicillin), 20 μg/ml Cm, 50 μg/ml Km, 50 μg/ml Rif (rifamycin), 35 μg/ml Sm, and 10 μg/ml Tc.

Tri-parental conjugations were performed by mixing the E. coli strain with the donor plasmid, the M. extorquens AM1 recipient strain, and an E. coli strain with the helper plasmid pRK2073 17. This mixture was grown overnight on permissive Nutrient agar 16 plates at 30°C before introducing some of the mix (either by streaking with a loop or by washing with Hypho and re-plating) onto selective medium containing an appropriate C source, Rif for counter-selection against E. coli 9, and the selective antibiotic (Tc for pCM433-based donors; neither Ap nor Cm works effectively in M. extorquens AM1, Marx, unpublished). Sucrose selection was accomplished by suspending a loop of a given strain in 100 μl Hypho (approximately 10⁹ ml^-1) and plating 50 μl of a 10^-2 dilution of this suspension onto Hypho plates containing an appropriate C source (generally succinate) and 5% sucrose. Resulting strains were tested for Tc sensitivity, additional expected phenotypes (depending on the locus and allele being exchanged), and additionally, the chromosomal organization of all strains constructed was confirmed through PCR analysis. DNA concentrations were determined using a ND-1000 spectrophotometer (NanoDrop).

In order to generate the allelic exchange vector pCM433, the Km resistance cassette of pCM184 4 was excised with NdeI and SacII, and the remaining 5.4 kb vector backbone was ligated together with a synthetic linker designed to introduce three additional, unique cloning sites into the final vector (PstI, XhoI, and NotI). The linker was formed by boiling, and then slowly re-annealing at room temperature, a mixture of two oligos, CM-link1f (tatgctgcagctcgagcggccgc) and CM-link1r (ggccgctcgagctgcagca), which were designed to have complementary overhangs to NdeI and SacII. The resulting plasmid, pCM432, was then transformed into the dam dcm E. coli strain, C2925H (ara-14 leuB6 fhuA31 lacY1 tsx78 glnV44 galK2 galT22 mcrA dcm-6 hisG4 rfbD1 R(zgb210::Tn10) Tc^SendA1 rspL136 (Sm^R) dam13::Tn9 (Cm^R) xylA-5 mtl-1 thi-1 mcrB1 hsdR2, New England Biolabs), enabling digestion at an otherwise methylated, and therefore blocked, MscI site. The 2.7 kb XbaI-XmaI fragment of pDS132 10 containing sacB and cat was then purified, blunted with Klenow enzyme, and ligated with the MscI-digested pCM432 vector to generate pCM433 (see Figure 2). A construct with the sacB-cat fragment in the opposite orientation, pCM433r, was also obtained. All plasmids and strains used are referenced in Table 1.

A series of constructs and strains were generated in order to test the ability of pCM433 to enable unmarked allelic exchange at three distinct loci in the M. extorquens AM1 chromosome. Donor constructs for allelic exchange at the mptG locus were generated by first amplifying a region including mptG from CM501 (an isolate of wild-type, Rif^R M. extorquens AM1 9), or the corresponding region from the ΔmptG strain, CM508 (an isolate of CM253.1 18), each of which were ligated into pCR2.1 (Invitrogen) to generate pCM411 and pCM424, respectively. These PCR-amplified inserts (and all other alleles described below that were cloned into pCR2.1) were sequenced to confirm no PCR errors were introduced during amplification. The 2.1 kb ApaI-BamHI fragment of pCM411 containing the mptG region was then introduced into pCM433 that had been digested with ApaI and BglII, resulting in the donor vector pCM436. Similarly, the 1.3 kb SacI-XhoI fragment of pCM438 with the ΔhprA region was cloned into the same sites of pCM433 to generate the donor vector pCM439. This allowed the use of pCM436 (containing the wild-type mptG allele) to reverse the lesion found in CM508, while pCM437 (ΔmptG allele) was introduced into CM501 to do the opposite, generating the deletion in a single step.

Similarly, donor constructs for allelic exchange at the crtI locus were generated by first amplifying a region including crtI (encodes phytoene desaturase) from the pink CM501 strain, or the corresponding region from the white crtI::ISphoA/hah (i.e. crtI⁵⁰²) strain, CM502 (an isolate of AM1-W 14). These fragments were ligated into pCR2.1 (Invitrogen) to generate pCM417 and pCM426, respectively. The 1.6 kb BamHI-NsiI fragment of pCM411 containing the crtI region was then introduced into pCM433 that had been digested with BglII and NsiI, resulting in the donor vector pCM440. Similarly, the 1.7 kb BamHI-NotI fragment of pCM426 with the crtI⁵⁰² region was cloned between the BglII and NotI sites of pCM433 to generate the donor vector pCM441. This allowed the use of pCM440 (containing the wild-type crtI allele) to reverse the lesion found in CM502, while pCM441 (crtI⁵⁰² allele) was introduced into CM501 to do the opposite, generating the insertion allele.

Finally, for the third locus, hprA, an antibiotic-resistance free deletion strain was generated initially using a previously developed cre-lox system 4. In contrast to the system described here using pCM433, the process to generate the ΔhprA strain was substantially more involved (and resulted in leaving behind a loxP scar). First, the regions upstream and downstream of hprA, were amplified separately and cloned into pCR2.1 (Invitrogen) to generate pCM428 and pCM429, respectively. The 0.5 kb upstream region was then excised from pCM428 using BglII and NotI and ligated into the same sites of pCM184 to generate pCM430. Into this plasmid, the 0.6 kb ApaI-SacI fragment from pCM429 was cloned into the same sites to generate the donor plasmid pCM431. As previously described 4, this plasmid was introduced into both the wild-type (pink) M. extorquens AM1 strain, CM501, as well as the otherwise isogenic white strain with a crtI⁵⁰² allele, CM502, leading to the isolation of the hprA::kan strains CM1122 and CM1123, respectively. pCM157 (expressing Cre recombinase) was introduced into these two strains to catalyze the excision of the kan cassette, and was subsequently cured, ultimately resulting in the antibiotic-resistance free ΔhprA strains CM1203 and CM1204 used below.

Donor constructs for allelic exchange of the hprA locus were generated by first amplifying a region including hprA from CM501, or the corresponding region from the ΔhprA strain generated above, CM1203. Ligation of these fragments into pCR2.1 (Invitrogen) generated pCM434 and pCM438, respectively. The 2.2 kb ApaI-BamHI fragment of pCM434 containing the hprA region was introduced into pCM433 that had been digested with ApaI and BglII, resulting in the donor vector pCM434. Similarly, the 1.3 kb SpeI-NsiI fragment of pCM438 with the ΔhprA region was cloned between the XbaI and NsiI sites of pCM433 to generate the donor vector pCM439. This allowed the use of pCM434 (containing the wild-type hprA allele) to reverse the lesion found in CM1203, while pCM439 (ΔhprA allele) was introduced into CM501 to do the opposite, generating the deletion in a single step.

The sequence of pCM433 has been deposited [GenBank:EU118176] and the plasmid has been deposited with Addgene (http://www.addgene.org/Christopher_Marx, Plasmid 15670).