Evolutionary expansion of metabolic networks entails the emergence of regulatory factors that become sensitive to new chemical species. A dedicated genetic system was developed for the soil bacterium Pseudomonas putida aimed at deciphering the steps involved in the gain of responsiveness of the toluene-activated prokaryotic regulator XylR to the xenobiotic chemical 2,4 dinitrotoluene (DNT). A mutant library of the A domain of XylR was screened in vivo for those variants activated by DNT through coupling the cognate promoter Pu to the P. putida yeast URA3 homologue, pyrF. All DNT-responsive clones maintained their sensitivity to ordinary effectors of XylR and broadened the range of inducers to unrelated aromatics. Yet, none of the altered amino acids lay in the recognizable effector binding pocket of the polypeptide. Instead, mutations appeared in protein surfaces believed to engage in the conformational shifts that follow effector binding and modulate signal transmission between XylR domains. It thus seems that transcriptional factors are likely to regress into functionally multipotent forms (i.e. stem protein types) as a first step towards the divergence of a new specificity.
Bacteria that colonize sites polluted by industrial waste often develop the ability to metabolize synthetic chemicals that have been in the biosphere for only a few years (Diaz, 2004). Such sites thus offer splendid experimental systems for examining the mechanisms behind the evolving complexity of metabolic networks (Cases and de Lorenzo, 2005). There are two aspects to the evolutionary challenge of attaining new metabolic capabilities in sites polluted with xenobiotic carbon sources. First, catabolic enzymes must act on new substrates. But enzymes are not originated with a given specificity for one chemical. The activity and affinity of an enzyme for a new substrate are believed to mature after a period of catalytic promiscuity in which the old and the new specificities are kept at tolerable fitness costs (Jensen, 1976; Aharoni et al., 2005). The second aspect is the regulation of expression by the new substrate(s) of the relevant gene(s). In environments with long-term exposure to the same C sources, very specific enzymes often become expressed under the control of promoters stringently regulated by cognate substrates. In this way, metabolic spills, futile cycles and dead-end (if not toxic) products brought about by side-reactions (Blasco et al., 1995; Skiba et al., 2002) are avoided. In contrast, enzymes of recently evolved pathways (i.e. those for metabolism of compounds which have been in the biosphere only since the era of synthetic chemistry; Pazos et al., 2003) are generally active on a large number of compounds (Wackett, 2004; Janssen et al., 2005). Similarly, new catabolic operons found in xenobiotic-degrading bacteria are typically inducible by chemical species which are not pathway substrates (de Lorenzo and Perez-Martin, 1996). The question therefore arises as to how transcriptional factors evolve an ability to respond to new signals, for instance, chemical structures that they have not been exposed to before.
In this work, we have set out to examine the ability of the toluene-activated XylR regulator to acquire a new responsiveness to the synthetic chemical 2,4 dinitrotoluene (DNT) by means of an experimental evolution/selection setup. The XylR protein controls the activity of the Pu promoter of the TOL plasmid pWW0 of the soil bacterium Pseudomonas putida mt-2 for biodegradation of toluene, m-xylene and p-xylene (Ramos et al., 1997). XylR belongs to the class of transcriptional regulators that act at a distance on prokaryotic promoters dependent on the alternative sigma factor σ54 and whose members are generically known as the NtrC family of enhancer binding proteins (Perez-Martin and de Lorenzo, 1996a). Like their eukaryotic counterparts, this type of regulator has a conserved modular structure (Fig. S1A) that includes an amino-terminal region (A domain) that, in the case of XylR, interacts directly with an inducer molecule (e.g. toluene, m-xylene), an event that leads to the conversion of the regulator into a form able to promote transcription from the Pu promoter (Fernandez et al., 1995; Perez-Martin and de Lorenzo, 1996c). Such a conversion requires the release of the intramolecular repression (or anti-activation) caused by the A domain itself on the rest of the protein (Garmendia et al., 2001; Devos et al., 2002; see Fig. S1B). As explained below, the experimental setup for evolution of DNT-activated variants was such that XylR is entered in the system as the only precursor of any new protein responsive to the nitroaromatic chemical species. Our results suggest that, instead of switching effector specificities, the protein meets the selection challenge by relaxing both the activity and the specificity of the transcriptional factor, to originate multipotent, stem protein forms, in which the original and the new function coexist. The data thus incorporate transcriptional factors to the classical hypothesis by Jensen (1976) that contemporary enzymes originated from ancestral forms with broad substrate specificities. Furthermore, our results provide a rationale for artificially evolving regulators responsive to synthetic chemical species.
A genetic trap for XylR effector-dependent growth of P. putida
In our previous attempts to introduce novel specificities in XylR by DNA shuffling, two experimental bottlenecks were experienced (Skarfstad et al., 2000; Garmendia et al., 2001). First, while the procedure was successful for isolating XylR variants responding to non-natural inducers, the sequence hybrids resulting from the shuffling had quite low diversity, yielding little information concerning the protein segments involved in effector specificity. Second, the variants isolated had undergone extensive rearrangements between the shuffled sequences, making it difficult to relate the chemical structure of the tested effectors to discrete changes in the protein structure. To circumvent these limitations, we developed a novel selection setup in which we combined single-base mutagenesis of the inducer binding domain of XylR (Fig. S1A) with a selection/counterselection system analogous to the yeast URA3 marker (Galvão and de Lorenzo, 2005). The rationale for this procedure for identifying XylR specimens that respond to synthetic inducers is sketched in Fig. 1. The key component of the system is P. putida TEC3 (see Experimental procedures and Fig. 1A). This strain is ΔpyrF and thus would be a uracil auxotroph (due to the lack of the gene encoding orotidine-5′-phosphate decarboxylase, the counterpart of URA3) except for the fact that the strain also contains on its chromosome a transcriptional fusion between the XylR-activated promoter Pu and the bicistronic segment pyrF–lacZ. The strain also carries on its chromosome a Pu→km-resistance marker. As conditional expression of Pu→pyrF both reverts the uracil auxotrophy of the ΔpyrF deletion and causes sensitivity to 5-fluoroorotic acid (FOA, a uracil analogue), this genetic setup provides a powerful selection system allowing both selection and counterselection of pyrF expression. Under this scheme, activation of the Pu→pyrF fusion of P. putida TEC3 by XylR allows selection through growth in medium lacking uracil or with kanamycin (Km; Fig. 1B). To validate such a selection system, we verified that the Pu→km/Pu→pyrF–lacZ fusions born by P. putida TEC3 allowed selection of pyrF-associated phenotypes. As shown in Fig. 1B, introduction of pCON916 (xylR+) in P. putida TEC3 led to an altogether toluene-dependent growth in a minimal (uracilless) medium with Km. By the same token, induction of the pyrF gene by toluene in strain P. putida TEC3 (pCON916, xylR+) inhibited cell growth in medium with uracil and FOA.
Mutagenesis and selection of XylR variants responsive to 2,4DNT
With the genetic selection system outlined above, mutagenic polymerase chain reaction (PCR) of the XylR A domain (positions 1–633; amino acids 1–211) and the B connector (positions 633–699; amino acids 212–233) was carried out to introduce an average of 2 ± 1 mutations per kilobase of DNA. The resulting PCR products were cloned in broad-host-range plasmid pURXAv so as to reconstitute full-length XylR under control of Pr (its native promoter; see Experimental procedures), and the ligation mixture was rescued in Escherichia coli. The library thereby generated contained a limited diversity of approximately 104 clones, which were transferred in masse to the selection strain, P. putida TEC3, by means of a triparental mating with helper strain E. coli HB101 (RK600). The conjugation was carried out in such a way that the number of P. putida exconjugants exceeded the E. coli coverage by at least 4-fold (50 000 P. putida colony-forming units per mating). This ensured that every plasmid born by E. coli cells was captured in the receptor strain. The resulting pool of P. putida TEC3 cells carrying the XylR AB library was then plated on M9 minimal media containing 2 mM of the prospective synthetic inducer 2,4DNT, uracil and Km. Colonies growing in these conditions were replica plated on media without uracil or containing uracil and Km; those clones that did not grow in either condition and were thus not constitutive were kept for further analysis. The Km-resistant clones were subsequently checked for uracil prototrophy by plating on media containing DNT but not uracil. As shown in Fig. 3, four clones grew in these conditions but not in the absence of DNT: V3, V9, V14 and V17. Alternatively, the library of XylR mutants was transferred to a P. putida carrying a mini-Tn5 chromosomal insertion with a Pu→gfp fusion, and spread on Luria–Bertani (LB) plates containing DNT. Fluorescent colonies were replica plated in the presence or absence of DNT, and those displaying fluorescence only in the presence of DNT were transferred to P. putida TEC3 to test their ability to activate the Pu→pyrF fusion as before. Upon transfer, one plasmid clone (V101) conferred prototrophy to the strain (Fig. 2) and was kept along with the others (Fig. 3). To confirm the phenotypes endowed by the corresponding protein variants, their ability to induce the Pu→kan fusion in the presence of DNT was re-evaluated. While V3, V9 and V17 reproduced the DNT-inducible Km-resistance phenotype screened for, V14 and V101 (from the Pu→gfp screen) had basal levels of activity high enough to confer Km resistance in the absence of DNT. Such levels could not be suppressed for these two variants by the negative selection system based on FOA (not shown). In any case, conditions in which simultaneous induction of the Pu→pyrF and Pu→kan fusions were tested in the absence of uracil, the five variants displayed the DNT-inducible phenotype associated with uracil prototrophy (Fig. 2). In every case, the control carrying wild-type XylR did not confer growth in the presence of DNT. The genetic and phenotypic characterization of the mutants is presented in the following paragraphs.
Mutations born by DNT-responding variants of XylR
Sequencing of the xylR variants conferring DNT-inducible uracil prototrophy revealed the mutants to have diverse alterations across the sequence of the A domain along with recurrent changes in the B linker at Leu222, either to proline or to arginine (Fig. 3). Changes in this amino acid, predicted to be in a coiled coil domain of the protein (Fig. 3), are likely to restrict severely the flexibility of the B domain (Garmendia and de Lorenzo, 2000). However, site-directed mutagenesis of residue L222-only to P or to R was insufficient to cause DNT responsiveness under the test conditions (data not shown). On the contrary, the combination of changes in L222 with other mutations exposed several remarkable features. The mutation in four positions (Glu12, Phe48, Leu83 and Tyr159) was to an amino acid observed in XylR homologues with different natural effector specificities: Glu12 is Asp in PoxR; Leu83 is Val in TbuT, DbtR and PhlR; Phe48 is Ile in HbpR and PcuR; Tyr159 is Phe in BphR (Fig. S2). Matching the location of the changes with an alignment of 21 A domain homologues highlights the relevance of Tyr159 and Phe48. Tyr159 is conserved in 20 proteins of this type, likely meaning that this residue has an important structural role. Phe48 is also conserved in the closest XylR relatives, but is nearly always a leucine in more distant proteins of the family (Fig. S2). Thus, the mutation Phe48Ile found in mutant V17 may expose one key determinant of effector specificity in XylR-like regulators. The other changes found in the DNT-responsive mutants involved (with one exception Ile208) alterations of amino acids at conserved positions or adjacent to conserved sites. Glu12 is next to semiconserved Asp13; Phe48 is semiconserved and is adjacent to conserved Arg49; Leu83 is semiconserved and is next to highly conserved Arg84; Pro97 is semiconserved; Leu99 is semiconserved and is next to conserved Tyr100; Asp119 is next to conserved Gly120; and Tyr159 is conserved in a highly conserved region (Fig. S2). In general, amino acids introduced by the mutagenesis kept the chemical nature of the same position. This was not unexpected, as single-base changes in the genetic code more often than not cause exchanges between chemically similar residues. Except for Pro97Ala, Tyr159Phe and Leu222Arg/Pro, every mutation is from a non-polar, bulky amino acid to another similar residue (Phe48Ile; Leu83Val; Leu99Phe; Ile208Phe), or from a charged side-chain to its corresponding charged (Glu12Asp) or uncharged side-chain (Asp119Asn).
To quantify DNT activation of the variants conferring Km resistance (V3, V9, V14 and V17) or inducing gfp expression (V101), the corresponding plasmids (pURXV3, V9, V14, V17 and V101) were transferred to P. putida SF05. This strain bears a chromosomal mini-Tn5 Pu→lacZ transcriptional fusion (Fernandez et al., 1995), allowing quantification of XylR activity by recording accumulation of β-galactosidase. Variant activation was assayed in 96 deep well plates through a range of DNT concentrations (Fig. 4A). As a control, P. putida SF05 carrying pCON916 (xylR+) was included, showing the wild-type regulator to be unresponsive to DNT. Under the conditions of the assay, V3, V17 and V101 induced Pu expression to approximately the same level, near 3-fold over when not induced. Interestingly, these variants were activated at the lowest DNT concentration used, ∼125 μM, yielding response levels near 2-fold over the non-induced controls. However, while they were activated by low DNT concentrations, the increase in response to higher effector concentrations was not proportional. These results argue that the mutations found in the corresponding A domains created a new structural capacity for a regulatory commitment of these variants upon DNT binding, leading to transcription. However, the dispersion (rather than the clustering) of the changes through much of the A domain sequence raised the question on whether such a new capacity was related to modifications in the effector binding pocket of the protein.
Properties acquired by DNT-responsive XylR mutants
Responsiveness of XylR observed in the DNT mutants may originate either from changes in the constellation of the amino acids that form the aromatic binding site of the A domain, or from the habilitation of the existing pocket to a new function brought about by changes in a different region of the protein. To distinguish these possibilities, we examined the response of each of the mutants to the natural XylR effector 3-methylbenzylalcohol (3MBA), using as a reference the isogenic strain P. putida SF05 bearing a plasmid encoding the wild-type regulator (pCON916). The results of such an experiment are shown in Fig. 4B and C, in which two features are noteworthy. First, to different extents, all DNT-responsive mutants kept an ability to respond to 3MBA, the V101 variant being the one that reached a more noticeable level (Fig. 4C). Moreover, V101 triggered Pu induction by 3MBA at lower concentrations of the effector than the wild-type regulator (Fig. 4B). This seems to be a unique feature of this protein, as all the other variants were far less responsive to 3MBA (Fig. 4B; see Discussion below). The second peculiarity of these induction assays was the remarkable decrease of the full level of promoter output caused by 3MBA in the mutants as compared with the wild-type (Fig. 4C). A closer inspection of the data of Fig. 4C in comparison with those of Fig. 4A indicated that the induction of the mutants by 3MBA was higher than that triggered by DNT, but quite distant from the approximately 25-fold induction of the wild-type protein exposed to 3MBA (Fig. 4C). In summary, the screening procedure described above yielded XylR variants that did acquire an ability to respond to DNT, while they kept a lower-level, but still significant, response to an inducer of the wild-type protein, such as 3MBA.
That these protein variants responded both to 3MBA and to DNT argued against a genuine exchange of specificity in the mutants in respect to the wild-type protein. Instead, the data raised the possibility that the XylR A domain contained an effector pocket that binds effectively DNT but failed to translate such a binding into any further action. To address this question, co-induction assays with wild-type XylR were performed using 3MBA and DNT simultaneously (Fig. 5). The logic of such an experiment is that a non-binder would not affect XylR induction by a bona fide effector (3MBA), while a non-productive binder of the same target site in the protein would inhibit XylR activity. Cultures of P. putida SF05 strains bearing the wild-type protein were exposed to a concentration of 3MBA optimal for Pu induction (0.5–1.0 mM), and increasing levels of DNT, β-galactosidase activity being measured after 3 h. As observed in Fig. 5, Pu activity decreased as the concentration of DNT went up: at the highest DNT concentration used, 2 mM, Pu output was 60% of that in the absence of the compound. It should be noted that 2 mM DNT did not inhibit the effector-independent activity of the constitutive XylR variant XylRΔA, which was altogether deleted from the A domain (not shown). This indicated that the inhibition caused by DNT on the induction of wild-type XylR by 3MBA was due to interference with effector binding, and not to a generic inhibitory effect on the transcription machinery. This result suggested that DNT binds XylR but is unable to trigger activation of the regulator, by competing with 3MBA for the same site and thus reducing Pu output (Garmendia et al., 2001). We thus argue that DNT can bind the wild-type regulator efficiently without activating it, and mutation leads to creating the structural arrangements needed for these compounds to cause transcriptional competency. Although the experiment of Fig. 5 does not determine whether inhibition of XylR by DNT is truly competitive, it hints at XylR having an existing, but unproductive binding site for DNT present in the wild-type XylR, which becomes habilitated in the mutants as a functional effector pocket. Further, the experiment of Fig. 5 suggests that the changes borne by mutant proteins produce the relief of constraints of the structural transmission pathway that renders effector binding into protein activation.
Dinitrotoluene-responsive mutants upgrade the capacity of the XylR effector pocket to engage aromatic effectors
If the effector binding site of the DNT-responsive mutants is the same as that of the wild-type protein for 3MBA, the next question was whether such a new capacity is limited to the aromatic compound used for the genetic screening. Alternatively, it is conceivable that using the same binding site, other unrelated aromatics may now become effectors. To probe the structural tolerance of the aromatic binding pocket to diverse aromatic molecules, we carried out similar Pu induction experiments with each of the mutants, exposed separately to the 3-mono-substituted nitrotoluenes. These bear the very electronegative NO2 group in position ortho, meta or para. As the mutants were raised to respond to 2, 4 DNT, one would expect to keep an induction by 2-nitrotoluene (2NT) and 4-nitrotoluene (4NT), but not by 3-nitrotoluene (3NT). The results of these experiments are shown in Fig. 6A–C. The pattern of variant response to 2NT showed that maintaining the wild-type response to this effector correlates with activation by DNT (Fig. 4A), in that the variants with greatest induction by DNT, V3 and V17, also display 2NT response levels similar to those of XylR, while V9 and V14 yield lower Pu output by these effectors. Despite relatively small differences, t-tests carried out on the data from independent experiments with 2NT indicated that variations of the response between V3 and V14, as well as between V17 and V9 or V14, shown in Fig. 4A, are significant. The explanation for the correlation between response to DNT and 2NT may lie in the fact that, by virtue of sharing a nitro-substitution at the same ring position, these effectors use a similar transduction path in signal activation. In other words, the ability of XylR to be activated by mono-substituted rings is taken advantage of, so as to poise induction by aromatic rings dinitro-substituted at positions 2 and 4. In addition, the results of Fig. 6B revealed that 3NT was as active as an effector as the di-substituted predecessor. This piece of data indicated that the mutations born by the XylR variants had upgraded the effector pocket to bind productively not only DNT and its mono-substituted precursors, but also the unrelated isomer 3NT.
To examine the extent to which the DNT XylR mutants could productively engage other aromatics, we tried a small collection of inducers, including 1,2,4-trichlorobenzene (TCB), 4-chlorophenol (ClPh) and 2,4-dichlorophenol (DClPh). As shown in Fig. 6D–F, while the behaviour of the singular mutants was not identical, the general trend was to respond to the new effectors as well. A conservative estimate of all experiments shown in Fig. 6D–F is that DNT response is, in every case, accompanied by activation by two effectors that do not activate XylR, 3NT and ClPh, as well as to reduced activation by 3MBA and 4NT (albeit V101 is an exception, showing greater activation than XylR by this effector; compare Figs 4B and 6C). In fact, the V101 variant yield the greatest levels of Pu activation with every non-natural effector. This mutant may be said to be happy trigger, or highly sensitive to induction. Its mutation of a highly conserved amino acid, Tyr159Phe, constitutes the removal of a hydroxyl group that may be implicated in stabilizing a contact whose transition is important in signal transmission activation. In the absence of the hydroxyl group, enough stability might be maintained to keep the protein inducible, but a variety of effectors set off the structural transitions leading to transcriptional competency (see below). In any case, a degree of structural similarity of the new inducers with the natural effectors should be maintained for activation of the mutant XylR variants, as several non-aromatic, but hydrophobic molecules were incapable of triggering Pu transcription (not shown).
Structural context of mutations leading to DNT responsiveness
To further understand how the mutations observed in the XylR variants lead to activation by DNT, the structural context of the mutated amino acids was visualized in a state-of-the-art XylR A domain model (Devos et al., 2002). This model encompasses amino acids 1–207 and thus includes all except I208F and L222R/P positions (Fig. 7). Visualization of Glu12, Phe48, Leu83, Pro97, Leu99, Asp119 and Tyr159 shows the mutations to fall into three classes based on their location in the model. First, Glu12, Pro97 and Asp119 colocalize in one face of the domain, suggesting that they are on a region of interaction with another protein or domain whose interface plays an important role in activation by effectors (Fig. 7A). Second, Phe48 of V17 is on the opposite face of the domain, immediately adjacent to the C-terminal region, and in apparent contact with Asp193 and Trp195 in the next-to-last β sheet of the domain (Fig. 7B). We envisage that Phe48 is involved in XylR activation as a signalling post next to the domain C-terminus, working to modulate movement between the A, B and C domains. Mutation to Ile at this position may thus change effector specificity by allowing an initial event triggered by DNT to pass through to the B linker, which does not happen with the wild-type Phe. Third are Leu83, Pro99 and Tyr159, which form a group of residues whose side-chains are buried in the domain (Fig. 7A and B). It is likely that I208 also falls in this group, as the last amino acid in the model (Pro207) is buried. It is only possible to see their side-chains when visualizing the secondary structure elements of the domain, with the exception of a small part of Leu83 that is on the domain surface. The mutations found in variants isolated in search for variants with effector specificity for other aromatics also fall in these classes (T.C. Galvão and V. de Lorenzo, in preparation), further strengthening the functional meaning of the mutations. The proposed effector binding site is made up of five loops in distant regions of the primary sequence (Devos et al., 2002), and does not include any of the mutated amino acids in the variants (Fig. 7C), although Asp119 is adjacent to one of its amino acids. Indeed, the position of most mutations in the XylR (and the similar protein DmpR) A and B domains altering specificity or activation does not coincide with the recognized effector binding pocket (Devos et al., 2002 and references therein).
We have exploited a novel selection system for the isolation of XylR variants activated by 2,4DNT, aimed at understanding the process by which transcriptional regulators acquire the ability to respond to new effector molecules. To this end, we generated a limited library of XylR variants with mutated AB domains, and screened through it for those activated by DNT through coupling Pu activation to the P. putida yeast URA3 homologue, pyrF. Although every DNT-responsive XylR variant resulting from this screening carried a mutation at Leu222, changing this position alone was not sufficient for activating Pu transcription by the new effector under the assay conditions (not shown). This suggested that the structural constellation created by the mutations was essential in orchestrating the regulator's ability to turn on promoter activity. One first feature of our results is that the mutated amino acids that originate a response to DNT are spread throughout the A domain protein sequence (Fig. 3) and are thus unlikely to cluster in the effector binding pocket, a notion that is consistent with the available structural prediction (Fig. 7). A useful reference to comprehend this can be found in the work by Hillen's group on TetR (Scholz et al., 2003; 2004; Henssler et al., 2004). In this case, TetR some mutations leading to regulators of changed specificity could not have been predicted from the structural data, as they are found in the secondary shell around the tetracycline binding pocket. In our case, the assay of Fig. 5 suggests that the inducer pocket in the wild-type XylR is the same for both a natural effector (3MBA) and a new effector (DNT). Indeed, affinity for an effector is not always proportional to the resulting transcriptional output of the regulated promoter. As seen also with TetR, inducibility is not equivalent to effector affinity: some compounds exhibit greater inducibility of TetR mutants without changes in binding constants, while others produce similar induction levels in spite of varying affinities for the regulator. In some cases both phenomena are observed, of mutation increasing binding constant and influencing the conformational transition that happens in induction (Scholz et al., 2003; 2004; Henssler et al., 2004; 2005). In other words, mutation can cause either inducibility changes without altering affinity by poising the regulator into a state of activation, or a change in the affinity for a compound without that being coupled to enhanced response. Similarly, a DmpR hybrid derivative has enhanced binding affinity for phenol but is completely deficient in activation, suggesting that A domain conformation upon effector binding is crucial for productive interactions with the transcriptional apparatus (O'Neill et al., 1999). Furthermore, our experiments with XylR (this work; Garmendia et al., 2001) and others' with DmpR (Wikstrom et al., 2001) showing that non-effector molecules can bind the regulator argue that: (i) the wild-type protein may have similar affinities to both productive and non-productive aromatics, and (ii) effector specificity relies not only in the geometry of the binding pocket, but in the structural transmission events that precede or follow such binding. Like a lock which accommodates many keys with the same ease, although only one of them allows the next steps for dislodging the mechanism. In this respect, the structural framework of small molecule binding and function is reversed in XylR (and its homologues) with respect to enzymes. In one case (enzymes), amino acids of the catalytic site and its surroundings determine specificity and, with relative ease, can be engineered to modulate substrate and catalytic properties. In XylR (and possibly in many other regulators as well), the effector binding pocket itself seems secondary to regulator function, with the context of the full-length protein having a greater role in determining whether binding triggers the adequate molecular movements leading to transcriptional competency.
The second salient feature of the results presented in this work is the high frequency at which DNT-responsive variants of XylR were found with our screening procedure. Even in a limited mutant library of 104 clones, we found ≥ 5 genuine protein species able to pass the harsh genetic test. The work of Hellinga's group on computer-aided manipulation of effector binding sites in periplasmic sugar binding proteins of E. coli (Looger et al., 2003) calculates that the number of combinatorial possibilities to alter the native sugar binding pocket of such proteins to recognize specifically a different effector is in the range of 1076, approximately the number of atoms of the Universe. It is thus clear that evolution of specificities of transcriptional factors is virtually impossible to happen in a single step. One possible way to meet such an evolutionary problem could involve the high-frequency occurrence of protein forms in which both the activity and the specificity of the factor relax to form intermediates that respond well – albeit to a lower level – to a whole range of structurally related inducers. In this sense, it is revealing that the mutated XylR variants generally displayed a reduced response to the native effector of the wild-type XylR (3MBA) along with a definitive promiscuity for a range of aromatic compounds. It is thus likely that moving between peaks of effector specificity is bound to occur through valleys of non-specificity with a tolerable fitness cost. We argue that these events are not casual, but they reflect an intermediate step in the process by which a given transcriptional regulator evolves the ability to respond a new small molecule. While the data available so far may not be enough to generalize such an evolutionary mechanism, the rise of promiscuous regulators after mutagenesis has been recurrently observed in other transcriptional factors as well (Ramos et al., 1990; Cebolla et al., 1997; Skarfstad et al., 2000; Sarand et al., 2001). Such a process probably entails the easy surfacing of a latent promiscuity (Afriat et al., 2006) built in the protein, followed by a slower takeover of the specificity for one effector versus the others (Fig. 8). We entertain that such relaxed forms recreate ancestral, developing molecular species (stem protein types), which may reappear at high frequencies under suitable selective pressure prior to differentiate fresh functions (Fig. 8). While a suite of single-step mutations may easily cause regression to such a stem form, surmounting promiscuity towards a singular, new specificity surely requires multiple, stepwise and less frequent amino acid changes. Bershtein et al. (2006) have recently theorized that overcoming promiscuity in enzymes might be far less probable than acquiring it, as the evolving protein must attain specificity for one small molecule while keeping its overall thermodynamic stability. To be sure, our attempts to use the DNT XylR mutants described here as the starting material to suppress promiscuity for non-desired effectors have originated single amino acid change variants with only small incremental improvements in their specificity (data not shown).
A current paradigm in the creation of new biological functions is that of gene duplication followed by mutation giving rise to novel traits and the potentially associated evolutionary fitness (Kondrashov, 2005). Recently, key evidence has been reported towards a model of plasticity in evolvability, whereby new function evolution takes place before gene duplication, after which specialization can happen. The observation supporting the model is that new enzymatic functions can be easily obtained through mutation without affecting the original specialized role (Matsumura and Ellington, 2001; Aharoni et al., 2005). Thus, new and original functions can coexist without one being deleterious to the other, permitting further specialization after a gene duplication event. In the context of XylR-type regulators, these concepts argue that there is evolutionary fitness associated with the ability to respond to novel effectors through A domain mutation. Our observations with XylR (and also those made with DmpR) are also consistent with specialization taking place independently of gene duplication, so long as the mutations acquired confer functions endowed with fitness values greater than the original as a consequence of environmental changes. Some in situ biodegradation experiments support this notion. Pseudomonas sp. CF600 cannot degrade 4-methylphenol, because wild-type DmpR does not recognize this compound as an effector for activation of the catabolic dmp genes (Sarand et al., 2001). However, exposure of the wild-type strain to 4-methylphenol-contaminated soil led to rapid selection of a subpopulation exhibiting enhanced capacities to degrade this compound. All successful strains harboured a single mutation in the A domain of DmpR that mediated the phenotype and was able to induce transcription of dmp genes in response to a large variety of effectors. It is likely that prolonged monitoring of 4-methylphenol degradation in that soil would reveal the rise of a new dedicated regulator for that compound. We argue that the promiscuous DmpR variants isolated in that study are functionally equivalent to our DNT-responsive XylR mutants, and to those mutant enzymes observed by Matsumura and Ellington (2001) or Tawfik and colleagues (Aharoni et al., 2005; Bershtein et al., 2006) in the coexistence of new and original functions. Furthermore, that the polyvalent properties of XylR and DmpR variants in responding to different compounds can be selected for in a given environment is an indication that evolution of dedicated functions can take place through moving along a continuous adaptive ridge on a multidimensional fitness landscape, avoiding valleys of low fitness (Bershtein et al., 2006; Whibley et al., 2006; Yoshikuni et al., 2006).
Strains and culture conditions
The strains and plasmids used in this study are listed in Table S1. P. putida KT2442 (Herrero et al., 1990) P. putida SF05 (Fernandez et al., 1994), P. putida TEC1 (Galvão and de Lorenzo, 2005), P. putida TEC2, P. putida TEC3 and E. coli strains were grown in Luria–Bertani (LB) broth and maintained by standard procedures. P. putida strains were also grown in M9 minimal media with MgSO4 (2 mM) and citrate (2 g l−1) as carbon source and, in the case of ΔpyrF strains, amended with uracil (20 μg ml−1, Sigma Aldrich). Antibiotics were used at the following concentrations: ampicilin, 150 μg ml−1; piperacillin, 40 μg ml−1; Km, 50 μg ml−1; streptomycin, 50 μg ml−1 (E. coli) or 100 μg ml−1 (P. putida); potassium tellurite (Tel.); 80 μg ml−1. Plasmids were transferred from E. coli to P. putida by tripartite mating (de Lorenzo and Timmis, 1994), using helper strain E. coli HB101 carrying plasmid RK600. After 5–8 h incubation at 30°C on LB agar, the conjugation mixture was plated on minimal selective media.
Construction of plasmids and strains
A transcriptional fusion between Pu, pyrF and lacZ was generated. First, the P. putida KT2442 pyrF gene (702 bp) was amplified with primers containing BamHI sites and introducing a ribosomal binding site (forward: 5′-CGTAGGATCCAGGACAGATATGTCCGCCTGCCAGACG-3′; reverse: 5′-GCTTGGATCCTTAATTACCCACGGA-3′). The PCR product was cloned as a BamHI fragment into pMAD (Cases et al., 1996), yielding pMP200. This plasmid was digested with NotI and the fragment containing the Pu→pyrF–lacZ fusion was ligated to pUT/mini-Tn5 streptomycin (de Lorenzo et al., 1990), generating pUMP200. This plasmid was transferred to P. putida TEC1 (Galvão and de Lorenzo, 2005) by tripartite conjugation with the helper strain E. coli HB101 (RK600) as above. Exconjugants were selected on M9 minimal media with MgSO4, citrate and streptomycin. A clone found to be sensitive to piperacillin was called P. putida TEC2. P. putida TEC3 was generated by conjugating P. putida TEC2 with CC118λpir carrying pUT-PuK (pUT/mini-Tn5Tel bearing a Pu→Km fusion, courtesy of J. Perez-Martín) along with E. coli HB101 (RK600). Exconjugants were selected on M9 minimal media with MgSO4, citrate and tellurite. A clone found to be sensitive to piperacillin was called P. putida TEC3.
The plasmid used for mutagenesis, pURXAv, was constructed as follows. RK2-based broad-host-range ampicillin-resistance vector pJB655 (Blatny et al., 1997) was digested to completion with BamHI and SphI. The oriV-containing segment was then ligated to a 1.3 kb BamHI–SphI fragment originated from pPrXylRΔ210 (Perez-Martin and de Lorenzo, 1996b), which encodes a xylR deletion mutant (XylRΔA, starting at amino acid 210), an EcoRI site adjacent to the leading ATG, and a truncated Pr promoter. The resulting plasmid, named plasmid pRX210, was cut with MunI, the ends filled in with Klenow enzyme and religated to obtain pURX210, that shows good expression of XylRΔ210 variant in P. putida. In a subsequent step, a PCR fragment comprising the entire xylR gene was amplified by PCR of plasmid pCON916 (Garmendia et al., 2001) with the oligos NXylREco (5′-AAAACATATGGAATTCTCGCTTACAT-3′) and BamTAG (5′-CGGAGGATCCTATCGGCCCATTGCTTTCC-3′). The amplified DNA was digested with EcoRI and BamHI and cloned into the corresponding sites of pURX210 to obtain pRX, a plasmid that efficiently expresses full-length XylR. To introduce a restriction site convenient for the mutagenesis procedure, pRX was entered with an AvrII site just after the B domain of the xylR gene sequence using the QuikChange kit from Stratagene and the oligos AvrDir (5′-CAGTATTACGGCCTAGGCCATTCGC-3′) and AvrRev (5′-GCGAATGGCCTAGGCCGTAATACTG-3′). This procedure originated pRXAv. Finally, XylR-containing EcoRI–BamHI fragments were excised from pRX and pRXAv, and recloned in pURX210 to obtain respectively pURX and pURXAv.
Random mutagenic PCR
Mutagenic PCR conditions were set as follows: 1× Taq Mg-less Boehringer reaction buffer, 3.5 mM MgCl2, 100 mM MnCl2, 1 μM each oligonucleotides JunATG (5′-ATGGAATTCTCGCTTACATACAAACCCAAGATGC-3′) and AvrRev (see above), 200 μM each dATP, dGTP, dTTP, and 40 μM dCTP or 200 μM each dATP, dCTP, dGTP, and 40 μM dTTP, 20 ng template DNA (pRXAv), 2.5 units of Taq polymerase in a final volume of 100 μl. The reaction was then allowed to proceed for 40 cycles of 94°C (1 min), 50°C (1 min) and 72°C (1 min). The resulting PCR product (720 bp) was digested with EcoRI and AvrII, and ligated to pURXAv (predigestion of which with the same enzymes removed wild-type AB domains). Finally, the ligation mixture was electroporated into E. coli DH5α.
Screening of the XylR A domain library
The pool of XylR sequences with mutations in its A and B domains captured in E. coli DH5α was transferred in masse to P. putida TEC3 through tripartite conjugation with E. coli HB101 (RK600). Exconjugants were selected on solid M9 minimal media with MgSO4, citrate, uracil, Km, piperacillin, and 2 mM 2,4DNT [from a 1 M stock dissolved in dimethylsulphoxide (DMSO)]. After 48 h at 30°C, colonies were replica plated on M9 minimal media with piperacillin and with or without uracill to identify constitutive clones. Those growing only in the presence of uracil were further analysed by DNA sequencing and transferred to the Pu→lacZ strain P. putida SF05 (Fernandez et al., 1994). Alternatively, the XylR variant library was conjugated with P. putida Pu→GFP (courtesy of J. Garmendia) and plated onto LB with Km, piperacillin and 2 mM DNT. In this case, fluorescent colonies were replica plated onto LB with Km, piperacillin in the presence or absence of 2 mM DNT. Clones fluorescing only in the presence of DNT were transferred to P. putida TEC3 and SF05 for further characterization.
Microtiter plate β-galactosidase assays
Overnight cultures of P. putida SF05 carrying plasmids were diluted in LB piperacilin to OD600 of approximately 0.05, and 190 μl of aliquots was loaded in 96 deep well plates. After ∼5 h shaking at 30°C cultures reached OD600 of approximately 1–1.2, at which point 10 μl of serial dilutions of the effector indicated dissolved in DMSO was added to a concentration ranging 0–2 mM. After 3 h of induction, the cultures were diluted 1:3 in LB, and the OD at 600 nm read. A total of 20 μl of the diluted samples was transferred to a second 96 well plate containing 80 μl of buffer Z (Miller, 1972), 0.025% SDS, and 0.014% Triton X-100. The cells were frozen at −80°C, and upon thawing, β-galactosidase assay was carried out by adding 20 μl of ONPG (0.4% w/vol). The reactions were stopped by adding 80 μl of 1 M Na2CO3. Absorbance at 420 and 550 nm was measured, and arbitrary units were calculated as described with Miller's formula (1972) to yield arbitrary units.
Batch culture β-galactosidase assays
Overnight cultures of P. putida SF05 carrying pCON916 were diluted in 10 ml LB piperacilin to OD600 of approximately 0.05 and shaken at 30°C. When cultures reached OD600 of approximately 1–1.2, 3MBA from a 1 M stock was added to a fixed concentration of 0.5 mM along with varying concentrations of DNT (0–2 mM) from a 0.5 M stock (also in DMSO). After 3 h of incubation, β-galactosidase activity assays were carried out with cells permeabilized with chloroform and SDS as described (Miller, 1972). The values reported in this work represent the average of at least three independent experiments, with deviations of less than 25%.
We thank Dan Tawfik and Jan Roelof van der Meer for critical reading of the manuscript. We are grateful also to Ana Isabel Martínez and Esther Rey for technical support, and David Cánovas and Carlos O. Sanchéz for advice on statistical analysis. This work was supported in part by EU grants of the 6th Framework Program. Mario Mencía was the recipient of a Ramon y Cajal Contract.