Cooperative amino acid changes shift the response of the σ54-dependent regulator XylR from natural m-xylene towards xenobiotic 2,4-dinitrotoluene


E-mail; Tel. (+34) 91 585 45 36; Fax (+34) 91 585 45 06.


XylR is a σ54-dependent transcriptional factor of Pseudomonas putida that activates the Pu promoter of the TOL plasmid upon binding its natural effector, m-xylene. The search for mutants of the signal-sensing module of XylR that respond to the xenobiotic compound 2,4-dinitrotoluene recurrently yields protein variants with a broad effector range. These mutants had amino acid changes not only in the effector recognition moiety (A module), but also in the inter-domain B linker of the protein. A random mutagenesis and selection/counterselection setup was adopted to optimize the 2,4-DNT reaction of XylRv17, one of the best 2,4-DNT responders and thus recreate how this regulator can adjust its specificity to novel effectors by individual changes on the evolving protein. Site-specific mutagenesis was then used to decipher the contribution of individual mutations in XylRv17 and in one of the mutants evolved from it (XylR28) to the 2,4-DNT response. This approach allowed us to capture a new XylR version with novel mutations that fixed the protein in an intermediate stage of the progress from an effector-promiscuous, pluri-potent protein type to a more specific form where the natural response to m-xylene was decreased and the non-native acquired response to 2,4-DNT was increased.


Bacteria that colonize sites polluted by industrial waste are often capable of metabolizing recalcitrant chemicals and synthetic compounds and that have been in the biosphere for only a few years (Seo et al., 2009). This capability is orchestrated by the integration of environmental and physiological signals into regulatory systems that tightly control the expression of genes that are in charge of metabolizing such molecules (Shingler, 2003). The emergence of novel catabolic aptitudes is due not only to the adjustment of enzymes for recognition of new substrates, but also to the onset of new regulators that control the expression of the genes that determine the corresponding catabolic functions. 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 coexist with no or little apparent fitness cost (Aharoni et al., 2005; Bloom and Arnold, 2009; Tracewell and Arnold, 2009). In this way, evolving enzymes can often readapt to act on large number of compounds (Wackett, 2004; Janssen et al., 2005). More recently the related question as to how transcriptional factors (TFs) evolve to respond to new signals (e.g. small molecules) has been addressed, among others, by using in vitro evolution techniques to modify the specificity of the prokaryotic protein called XylR (Fig. 1A; Galvão et al., 2007). This TF is the main regulator of the so-called TOL pathway for degradation of m-xylene encoded in plasmid pWW0 of the soil bacterium Pseudomonas putida mt-2 (Ramos and Marques, 1997). XylR belongs to the NtrC family of TFs dependent on the alternative sigma factor σ54 (Morett and Segovia, 1993). This type of regulators have a modular structure that includes an amino-terminal region, signal-sensing (A domain) that interacts with the inducer molecule (Pérez-Martín and de Lorenzo, 1995), the B domain that is involved in fixing the protein in an appropriate form for transcriptional activation, the central C domain responsible for binding and hydrolysis of the ATP, and the C-terminal D domain that binds DNA (Fig. 1A; Ramos and Marques, 1997). After the recognition and binding of its natural effectors (e.g. m-xylene), XylR adopts an active form that is capable of triggering the transcription from the Pu promoter (Fernandez et al., 1995). Such a conversion requires the release of the intra-molecular repression (or anti-activation) caused by the A domain itself on the rest of the protein (Garmendia et al., 2001). The modular structure of this regulator (Fig. 1A) makes it possible to generate new effector-responsive variants of XylR by modifying only the A domain, a characteristic shared by other members of the NtrC family (Wise and Kuske, 2000; Beggah et al., 2008).

Figure 1.

Components of the in vitro evolution system for the XylR regulator.
A. Domain organization of the XylR protein. The boundaries between the functional modules and the localization of the relevant functions within the protein sequences are shown: A, signal reception and inducer binding; B, interdomain linker region; C, binding and hydrolysis of ATP and contacts with the sigma 54-RNAP and D, which includes a helix–turn–helix motif, for binding to the upstream activating sequences (UAS) of the target Pu promoter. The location of mutations F48I and L222R borne by variant XyRv17, the starting material of this work, is indicated.
B. 2,4-DNT responsive XylR mutants were generated by either site-directed mutagenesis or by mutagenic PCR of the DNA sequence of the A domain. The resulting segments (or the pool of them) were cloned in matching sites of reference plasmid pBBxylR.
C. XylR variants were tested for activity by passing the corresponding plasmids to pyrF-lacZ, lacZ or lux reporter strains as indicated in the text.
D. Beta-galactosidase activity of strains E. coli CC118 Pu-lacZ, P. putida SF05 and P. putida TEC2 (all of them carrying a chromosomal Pu → lacZ transcriptional fusion) transformed with the reference pBBxylR plasmid, induced with 1.0 mM 3MBA, m- xylene and toluene, as indicated.

Previous work in our laboratory has focused on endowing XylR with a new responsiveness to the synthetic chemical 2,4 dinitrotoluene (2,4-DNT) by means of different experimental evolution/selection setups (Garmendia et al., 2001; 2008; Galvao and de Lorenzo, 2006; Galvão et al., 2007). Using these strategies, 2,4-DNT responsive mutant versions of XylR were obtained. These mutants shared two important characteristics that emerged during the generation of variability and the subsequent selection: (i) they responded to 2,4-DNT (and other compounds) but still retained the capability to react to the natural effectors and (ii) they encompassed changes in the Leu residue placed in position 222 (L222) in the B domain. These attributes of the XylR mutants provide a good scenario to study how transcriptional regulators can evolve to give more specific responses and the significance of the effects of individual changes on global behaviour of the evolved TFs. In order to address these issues we have used in this work two different strategies of mutagenesis. The first one was based on site-directed amino acid changes to decipher the contribution to the 2,4-DNT response of the individual mutations in one of the best 2,4-DNT responders, XylRv17 (Galvão et al., 2007). The second strategy involved a new random mutagenesis/selection setup to improve the specificity of XylRv17 for 2,4-DNT. This approach allowed us to generate a XylR variant with novel mutations that contributed to pass from a protein type able to respond equally well to a large diversity of inducers (Galvão et al., 2007) to a more specific form where the natural response to m-xylene is decreased and the non-native acquired response to 2,4-DNT is increased. It thus seems that while TFs can easily relapse into inducer-promiscuous forms (perhaps reflecting an ancestral form of the protein; Khersonsky and Tawfik, 2010) the build-up of new effector specificities require the cooperative concourse of successive amino acid changes. These results generalize current models of enzyme evolution (Romero and Arnold, 2009; Tracewell and Arnold, 2009; Khersonsky and Tawfik, 2010; Soskine and Tawfik, 2010) to the realm of inducer-responsive transcriptional regulators.


Standarization of the input/output function of XylR mutants

Previous efforts have demonstrated that it is possible to endow bacterial transcriptional regulators with new specificities by using in vitro evolution approaches (Cebolla et al., 1997; Skarfstad et al., 2000; Garmendia et al., 2001; Galvao and de Lorenzo, 2006). In the case of the multi-domain TF XylR, it is possible to generate pools of mutants only by modifying the ligand-binding A module. Efforts to get 2,4-DNT responsive XylR mutants were carried out in our laboratory by using two different strategies. The first one implied large rearrangements in the A domain of XylR (Garmendia et al., 2001). The mutants obtained, although were effective in the detection of non-native compounds were inadequate to identify residues of XylR involved in the recognition of the effectors. The second strategy was based on the introduction of random point mutations in the A and, because of practical requirements, also in the B domain (Fig. 1B). All the 2,4-DNT responsiveness mutants of XylR obtained by this procedure, apart from bearing different point mutations in the A domain shared a L222 mutation in the B domain (Galvão et al., 2007). Although preliminary data suggested that changing this position alone was not sufficient for explaining the behaviour of these XylR mutants the experiments used at that time two different plasmid vectors (pURXav and pCON916) for expression of the different versions of this TF. For this reason the effect of the B domain change on the ability of these mutants to recognize and respond to 2,4-DNT remained ambiguous. To clarify this issue, we first standarized the vector system used for expression of each of the XylR variants tested in this work (Fig. 1C). This was made by adopting as a reference plasmid pBBxylR (de Las Heras et al., 2008; Fig. S1), which is based on the pBBRMCS1 vector series (Kovach et al., 1995). In order to check the robustness of the input/output function mediated by XylR (de Las Heras et al., 2010), the ability of pBBxylR to activate Pu in the presence of three natural effectors of XylR was tested in three different genetic backgrounds employed later for various procedures. One was Escherichia coli CC118 Pu-lacZ, in which a translational fusion xylU::lacZ is engineered in the genome behind Pu (xylU is the first gene of the operon driven by this promoter; de Lorenzo et al., 1991). A second host strain of pBBxylR was P. putida SF05, which carries in its chromosome a transcriptional reporter fusion Pu → trp::lacZ (Fernandez et al., 1994; Galvão et al., 2007). Finally, the same plasmid was placed in P. putida TEC2, a ΔpyrF strain that bears in its chromosome an engineered pyrF-lacZ operon under the control of Pu (Fernandez et al., 1994; Galvão et al., 2007). The results of Fig. 1D showed that, despite various levels of output, the relative pattern of responses to toluene, m-xylene and 3-methylbenzylalcohol (3MBA) was comparable in the three strains. This warranted pBBxylR as a suitable plasmid frame for expression of wild-type xylR and its variants in the experiments described below.

Deciphering the effects of the individual mutations F48I and L222R of XylRv17

Following standarization of the test system for XylR variants we investigated the individual effects of the amino acid changes contained in XylRv17, the best 2,4-DNT responder variant isolated by Galvão et al. (2007). The two mutations borne by XylRv17 (F48I and L222R; Fig. 1A) were segregated into two derivatives of pBBxylR carrying separately each one of the two mutations: pBBxylR[F48I] and pBBxylR[L222R]. These two plasmid, along with their precursors pBBxylR[wt] and pBBxylRv17[F48I/L222R] were introduced in the P. putida strain TaPuLUX which bears a Pu → luxCDABE transcriptional fusion in the chromosome that allows a non-disruptive monitoring (i.e. bioluminescence production) of transcriptional activity. To this end we exposed the four plasmid-containing P. putida TaPuLUX strains (Fig. 2A) to different concentrations of 2,4-DNT and measured the bioluminescence production 4 h after the addition of the inducer (Fig. 2B). The results indicated that the XylR mutant version with the sole L222R change had a similar response to 2,4-DNT that the double mutation F48I/L222R carried by XylRv17 (Galvão et al., 2007). On the other hand, the mutant carrying the F48I-only change did produce a detectable level of bioluminescence in the presence of 2,4-DNT that the wild type lacked entirely, but this change alone was not enough to mimic the effect of XylRv17 [F48I/L222R] (Fig. 2B). Since changes in either residue F48 and L222 separately brought about a degree of response to 2,4-DNT we reasoned that they could form part of either the effector recognition pocket or participate in the propagation of the alosteric change that enables XylR for transcription activation (or both). To explore this possibility, we site-directed mutagenized either position (see Experimental procedures) and screened each of the resulting pools for activity.

Figure 2.

Disengagement of the mutations carried by XylRv17 variant.
A. xylR region corresponding to B and A domain present in the plasmid pBBxylR was changed for the equivalent segment of the plasmid pURXV17 that bears the mutant XylRv17 (Galvão et al., 2007), rendering pBBxylRv17[F48I/L222R]. The mutations F48I and L222R were then introduced independently in pBBxylR by site-directed mutagenesis producing the plasmids pBBxylR[F48I] and pBBxylR[L222R] respectively. To measure the effect of the individual mutations, the four plasmids were introduced in the P. putida TaPuLUX strain that bears chromosomal Pu → luxCDABE transcriptional fusion.
B. Bioluminiscence production of P. putida TaPuLUX carrying pBBxylR (XylR wild type), pBBxylRv17[F48I/L222R] (XylRv17), pBBxylR[F48I] and pBBxylR[L222R] in cultures with different 2,4-DNT concentrations.

In the first case we carried out a saturation mutagenesis of position F48 of the XylR A domain using the plasmid pBBxylR as the template of the method (see Experimental procedures). Following transformation of the mutant pool in reporter strain P. putida TaPuLUX and plating on media with and without 2,4-DNT, we picked those clones that produced more bioluminescence than those expressing XylR [F48I] or XylR [wt] in response to the inducer. The XylR variant that showed a highest induction with 2,4-DNT (Fig. 3C) had the F48 position exchanged by a Thr (T) residue. To discriminate whether the new mutant F48T had switched inducer specificities or had instead become promiscuous to various effectors, the mutant was also examined for responses to the natural inducer of wild-type XylR, 3MBA, to suboptimal inducers, e.g. 1,2,4-trichlorobenzene (TCB) or 4-nitrotoluene (4-NT) and to non-inducers (e.g. 3-nitrotoluene, 3-NT). The results were compared with those of XylR [wt] as a negative control, and XylRv17 [F48I/L222R], which is known to be promiscuous (Galvão et al., 2007). The data of Fig. 3 show that the F48T resulted in a XylR protein that – to various degrees – responded to all inducers tested. The mutant F48T is therefore a promiscuous protein variant (Garmendia et al., 2001; Galvão et al., 2007) that has been created by mutation of one residue in the A domain of XylR rather than in the linker B sequence. Also, these results ruled out that F48 participates directly in shaping the effector pocket, which is consistent with structural predictions (Devos et al., 2002).

Figure 3.

Effector profile of XylR mutants bearing changes the A and B domain.
A. Pattern of bioluminescence production of P. putida TaPuLUX carrying plasmid pBBxylR.
B. Same, pBBxylRv17[F48I/L222R].
C. Same, pBBxylR[F48T]. Inducers: None (NI); 3-methyl-benzyl-alcohol (3MBA); 2-nitrotoluene (2NT); 3-nitrotoluene (3NT); 4-nitrotoluene (4NT); 2,4-dinitrotoluene (2,4-DNT); 1,2,4-trichlorobenzene (TCB). Note the different range of specific bioluminescence caused by wild-type XylR and the different XylR variants.

The second round of site-specific saturation mutagenesis was focused on residue L222, which, when changed to L222R, sufficed to endow XylR with a response to 2,4-DNT indistinguishable from that of XylRv17 [F48I/L222R]. Because L222 lies in a coiled-coil region that, if disrupted, results in a semi-constitutive and promiscuous protein variant (Garmendia and de Lorenzo, 2000), we wondered whether the phenotype of the L222R mutant was just reflecting a structural disruption in the B domain of XylR. To examine this, we followed an strategy similar to that used for residue F48, i.e. constructing a combinatorial library of the position L222 and seeking clones affording higher bioluminescence in the presence of 2,4-DNT than those with XylRv17 [F48I/L222R]. The ensuing screening produced two types of clones. First, those bearing recurrently the L222R change that was already present in the mutant of reference. Second, we found a number of high-luminiscence mutants that contained large sequence rearrangements through the B domain (Fig. S2). These mutants also showed an increased basal level and a higher response to nitroaromatic compounds and TCB as compared with XylRv17 [F48I/L222R] and the XylR variant F48T (Fig. S2; compare with Fig. 3). These unexpected XylR variants suggested additional scenarios that enable XylR to acquire a promiscuous state, but they were not useful to trace evolutionary trajectories and were not studied further. However, they strengthened the notion that selective pressure for a new specificity (i.e. response to 2,4-DNT) systematically results in a broadening of the effector profile to include the new inducer rather than the swap of specificities from one small molecule to the other. But, if that is so, how can TFs surmount the valley of promiscuity towards a new peak of specificity (Tracewell and Arnold, 2009)? The experiments below were done to shed light on this question.

Mutagenesis and selection of further optimized 2,4-DNT responsive XylR variants

The above mentioned mutant XylRv17 [F48I/L222R] was generated (Galvão et al., 2007) by applying to XylR a selective pressure in vivo, which rewarded responsiveness to 2,4-DNT while penalizing constitutive activity. But, as discussed above, XylRv17 also conserved the capability to react to natural XylR effectors and responded to many other non-natural inducers as well. Because XylRv17 was the best 2,4-DNT responder obtained at that time, we wondered whether further rounds of mutagenesis penalizing responses to natural effectors -while maintaining the reward for 2,4-DNT could result in protein variants with a superior specificity. To this end, we generated diversity in the A and B domains of XylRv17 followed by a pyrF-based selection/counterselection system (Galvao and de Lorenzo, 2005). For this, we first amplified the segment of the xylRv17 sequence of pBBxylRv17 corresponding to the target domains with a PCR reaction run by an error-prone DNA polymerase and the mutated fragments cloned in the corresponding sites of pBBxylR as described in the Experimental procedures section. Transformation of the plasmid pool into strains E. coli CC118 Pu-lacZ (see above) or E. coli DH5α (Experimental procedures) originated a library of mutants with a diversity estimated in the range of ∼ 106. Random sequencing indicated that the mutated fragments contained an average of 2 ± 1 mutations in the 713 bp segment encompassing the A and B domains of XylR. Finally, the library was transferred in masse by conjugation into the selection strain P. putida TEC2 (ΔpyrF, Pu → pyrF-lacZ; Galvão et al., 2007). This strain is a uracil auxotroph because of the lack of pyrF, the gene encoding orotidine-5-phosphate decarboxylase (the counterpart of the yeast URA3). On the other hand, activation of pyrF-lacZ both reverts the auxotrophy and causes sensitivity to 5-fluoroorotic acid (FOA, an uracil analogue). As explained in Galvão et al. (2007), this genetic setup enables selection and counterselection of pyrF expression in the same reporter strain (Fig. 4A). On this basis, the pool of P. putida TEC2 exconjugants carrying the library of mutants derived from XylRv17 was plated in M9 mineral medium containing 1 mM 2,4-DNT and no uracil. Under these selection conditions, only the clones carrying constitutive versions of XylR or variants that respond to 2,4-DNT can grow. Note that the concentration of the inducer used in the selection plates was lower than that formerly used for isolation of XylRv17 (Galvão et al., 2007), the expectation being the capture of mutants more sensitive to 2,4-DNT (P. putida TEC2 transformed with pBBxylRv17 gave only very small colonies in this selective medium; Fig. 4B). In contrast, plating of P. putida TEC2 bearing mutants derived from XylRv17 originated colonies of various sizes (Fig. 4B). Those which were manifestly larger than the control P. putida TEC2 (pBBxylRv17) were then subjected to two rounds of counterselection with FOA and m-xylene in order to suppress possible constitutive mutants and enrich in variants with a decreased response to the native effector of XylR. Four clones that fulfilled all these requirements were finally picked for further analyses as explained below.

Figure 4.

Second round of mutagenesis of the XylRv17 protein.
A. Mutagenic PCR. Following error-prone amplification of the DNA sequence of the xylRv17 A and B domains, the resulting PCR products were cloned in the corresponding sites of plasmid pBBxylR, thereby reconstituting full-length xylR. The ligation pool was then passed to P. putida TEC2 that bears a chromosomal Pu → pyrF-lacZ transcriptional fusion. This allowed selection of the clones responsive to 2,4-DNT by means of the conditionally expressed pyrF as explained in the text.
B. Morphology and growth of P. putida TEC2 (pBBxylRv17 [F48I/L222R]) colonies plated on M9 minimal media without uracil, but containing MgSO4, citrate, Km, 1 mM 2,4-DNT (top), or the library of mutants (pBBLib17-TEC2) achieved by subjecting xylRv17 to a new round of mutagenesis.
C. Bioluminescence production P. putida TaPuLux carrying pBBxylR derivatives expressing the XylR variants indicated when exposed to either 1 mM 2,4-DNT or m-xylene vapours. The lower part of the graph shows a Western blot of cells collected from each of the cultures and probed with an anti-XylR antibody.

Identification of XylR28 as a bona fide 2,4-DNT responder

The plasmids carried by the four clones isolated in the previous screening for enhanced 2,4-DNT response and lesser m-xylene sensitivity were recovered from the selection strain P. putida TEC2 (Galvao and de Lorenzo, 2005) and passed to the a Pu → lacZ reporter P. putida SF05 (see above; Fernandez et al., 1994). Out of four plasmids derived from pBBxylRv17, only the one named pBBxylR28 appeared to encode a XylR variant that consistently kept the expected characteristics in the fresh host (i.e. higher response to 2,4-DNT, lower response to m-xylene). In order to examine this new mutant in a genetic background with an output that could be followed more accurately, we resorted to lux rather than lacZ as the proxy of choice to continue with the analysis of this XylR mutant version as it has been shown to be more sensitive (de Las Heras et al., 2008). To this end, pBBxylR28 was transferred to the P. putida strain TaPuLUX that is engineered to bear in its chromosome a Pu → luxCDABE operon. The resulting strain carrying pBBxylR28 was exposed to 2,4-DNT or m-xylene as shown in Fig. 4C. Controls included P. putida TaPuLUX (pBBxylR, wild type) and P. putida strain TaPuLUX (pBBxylRv17, precursor). To rule out dosage effects, the relative intracellular concentrations of the regulator were followed with an anti-XylR antibody. The results of Fig. 4C were not only consistent with the phenotypes predicted from the screening of mutants in P. putida TEC2 (see above), but they also suggested that the reduced response to m-xylene produced by XylR28 was due to a genuine malfunction of its ability to respond to this ordinary XylR effector. This indicated that after a second round of mutagenesis it was possible to obtain new versions of the regulator with a reduced natural function, favouring at the same time, the evolved responsiveness to 2,4-DNT. To examine the sensitivity of XylR28 to 2,4-DNT, we exposed the three P. putida TaPuLUX strains with each of the plasmids to different concentrations of 2,4-DNT, and measured bioluminescence production (Fig. 5A). The results revealed that XylR28 both produced a net promoter output higher than the precursor XylRv17 and was more sensitive to lower amounts of 2,4-DNT. Interestingly, when the response of XylR28 to other nitroaromatic compounds and to TCB was tested (Fig. S3), no significant increase in light emission was observed as compared with XylRv17. This suggested that the augmentation of the response to 2,4-DNT was genuine step towards specificity of XylR for the new effector rather than just an enhancement of the promiscuous phenotype of XylRv17 (Fig. 5B)

Figure 5.

Characterization of mutant regulator XylR28 [F48I/I136T/S174R/L222R].
A. Bioluminescence dose–response patterns of P. putida TaPuLux strain carrying pBBxylR, pBBxylRv17[F48I/L222R] and pBBxylR28[F48I/I136T/S174R/L222R] to varying concentrations of 2,4-DNT.
B. Share of Pu activity (fold induction) caused by wild type XylR, XylRv17 [F48I/L222R] and XylR28 [F48I/I136T/S174R/L222R] (from the centre to the edge) in the presence of 1 mM 2,4-DNT and m-xylene. The same share was observed when plasmids encoding each of the proteins were placed in either lux reporter strain P. putida TaPuLUX or in lacZ reporter strain P. putida SF05. Note the growing part of the new inducer at each step of the evolution.

Dissection of amino acids changes that cause XylR28 to respond to 2,4-DNT

In order to study the basis of the new phenotypes associated to the xylR28 alelle we determined the amino acid substitutions carried by the protein. As shown in Fig. 6A, XylR28 conserved the F48I and L222R changes present in XylRv17, but included two additional modifications in the A module, namely I136T and S174R. Using as a reference the XylR A domain model of Devos et al. (2002) shown in Fig. 6B, these mutations were located in the predicted tridimensional structure of the protein. It is revealing that changes mapped within or close to the minimal effector-binding region defined by analysing deletion mutants and domain swapping of XylR and its phenol/methylphenol-responsive homologue protein DmpR (Devos et al., 2002). Interestingly, S174R is located close to residue E172 (Fig. 6B), which is conserved between members of the XylR/DmpR family and is thought to participate in effector recognition (Garmendia et al., 2001; Devos et al., 2002; see Fig. 6B). The second modification not present in the parental protein XylRv17 was I136T. This change falls in the carboxyl-terminal end of α-helix 7 (Fig. 6B), a region that has been suggested to modulate interdomain signalling (Devos et al., 2002). Although distant in the primary sequence, both S174R and I136T mutations are spatially positioned in the vicinity of F65 and M37, which are residues described as key determinants of effector recognition specificity (Garmendia et al., 2001; Sarand et al., 2001).

Figure 6.

Localization of mutations borne by XylRv17 and XylR28.
A. Schematic representation of the A and B domain of XylR and the position of the mutations present in XylRv17 (in black) and in XylR28 (in black and in red).
B. Localization of XylR28 mutations within a structural model of the XylR A domain of (Devos et al., 2002). The threading model shown is based on the crystallographic data of the catechol o-methyltransferase (COMT; PDB code 1vid), a typical a/b fold, consisting of eight α-helices and seven β-strands (Dietrich et al., 2010). The C-terminal F48 (I48 in XylRv17 and XylR28) shown in violet, adjacent to the two C-terminal β-sheets comprising amino acids 193–207. The specific XylR28 changes T136 and R174 (shown in red) are close to residues F65, M37, E140 and E172 (shown in blue) that belong the theoretical binding pocket. The XylR A domain model was visualized with Pymol.

As the changes carried by XylR28 seem to imply functionally different portions of the A domain we set out to separate individual mutations for determining their effects on XylR activity and specificity. To this end, we recreated the single mutations I136T and S174R of XylR28 into the wild-type regulator, passed them to P. putida TaPuLUX in the same plasmid vector as before, and measured the response of Pu → luxCDABE against 2,4-DNT and m-xylene, the results being shown in Fig. 7A. While pBBxylR[I136T] was indistinguishable from the wild-type and pBBxylR[S174] had a noticeable higher basal level of activity, it is noteworthy that their combination in pBBxylR[S174/I136T] reduced the basal level of Pu activity. Such very low transcriptional output was put back to normal levels when the cells were exposed to 2,4-DNT. These somewhat small effects, which were altogether reproducible with the sensitive reporter lux system used, suggested an epistatic connection between the two mutations. Their functional association became more evident when the same individual xylR mutants were exposed to the native inducer m-xylene and the responses of strains carrying combined mutations compared as shown in Fig. 7B. While the scale of Pu activity increases in all cases by at least two orders of magnitude when cells are exposed to the standard effector, it is to be noted that introduction of I136T and S174R in the XylR sequence very significantly decreased promoter output. Because this behaviour does not appear in individual mutants, the data of Fig. 7 indicate a substantial allelic cooperation in the double I136T/S174R mutant for weakening the response to m-xylene. Taken together, the results above put forward an evolutionary scenario in which acquisition of responsiveness to a new effector and suppression of a promiscuous state occurs through various accumulative changes that build cooperatively on each other.

Figure 7.

Bioluminescence production of individualized mutations of XylR28. The panels show specific bioluminiscence produced by P. putida TaPuLux carrying pBBxylR, pBBxylR[I136T], pBBxylR[S174R] and pBBxylR[I136T/S174R] and induced with either 1 mM 2,4-DNT (graph a) or m-xylene (graph b). The same samples without inducer are shown as a control (NI). Note the different scales of luminescence value in each case. The figures shown are the mean value of three separate experiments with duplicate samples.


This work stems from our long-standing interest to engineer reporter bacteria that can be spread in soil as whole cell-biosensors for explosives (Garmendia et al., 2008; de Las Heras et al., 2008). Because natural TFs responding to TNT or 2,4-DNT have not been found thus far and the attempts to forward engineering artificial regulators to the same end have not worked well (Looger et al., 2003; Smirnova et al., 2004; Lonneborg et al., 2007), we have resorted instead to experimental evolution approaches. XylR is a toluene/m-xylene responsive, σ54-dependent TF of P. putida that can be evolved to recognize non-natural but structurally related effectors (Delgado and Ramos, 1994; Galvão et al., 2007). Previous attempts to make XylR responsive to 2,4-DNT originated variants containing combined mutations in the A and B domain of the protein. This fact made difficult to relate discrete changes in XylR structure to the capability of responding to the new effector. To tackle this question we focused on the variant called XylRv17, the best of the 2,4-DNT responders originated in the round of error-prone PCR mutation/selection described in Galvão et al. (2007). This variant contains two changes, F48I and L222R, which were segregated into two individual proteins and further analysed separately by site-specific saturation mutagenesis. The results above (Fig. 2) clarify that the bulk of the behaviour of XylRv17 regarding inducibility by 2,4-DNT can be traced to the L222R change, which lies in the B region (Fig. 1A), with only a small contribution of F48I. As mutants of XylR bearing disruptive changes in the B linker domain both broaden the effector profile of the protein (Galvão et al., 2007) and increase its level of basal activity (Garmendia and de Lorenzo, 2000), it is possible that the L222R makes the protein promiscuous to a variety of effectors with similar chemical structures. However, we show in this work that individual mutations in the A domain also lead to XylR variants with a broad effector profile.

The ease by which TFs acquire a promiscuous state when selected for a new inducer specificity has been observed in several cases (Cebolla et al., 1997; Wikstrom et al., 2001; Ju et al., 2009). These data with XylR and other regulators support a view of the evolution of inducer-specificity maturation of TFs that involves: (i) first, the expansion (rather than switch) of their effector/substrate profile to include the new one at little fitness cost and (ii) the later surmounting of the lack of specificity until a new effector-specialization is achieved (Tokuriki and Tawfik, 2009). One key feature of this model is that a promiscuous state (which we propose to designate stem protein type; Galvão et al., 2007) can be caused at high frequency by different types of single mutations. In contrast, the itinerary towards regaining effector specificity might require multiple, stepwise and cooperative changes, not as dramatic as the other way around. Consistently with this view, when we subjected the stem variant XylRv17 to a round of mutagenesis and selection for increased response to 2,4-DNT (along with a smaller reaction to m-xylene), we found a mutant (XylR28) that appeared to represent one step towards switching specificities between the old and the new inducer. The new changes (I136T and S174R) were in regions of XylR close to, but not engaging directly, the predicted minimal effector-binding region deduced from previous deletion mutants and domain swapping data (Devos et al., 2002). It is revealing that these amino acid substitutions did not increase the response to 2,4-DNT found in the precursor variant (XylRv17), but they lowered instead the response of the protein to m-xylene (Fig. 4C). In contrast, individual changes merely altered the basal activity level of the corresponding proteins, which remained insensitive to 2,4-DNT, in any case. The four mutations borne by XylR28 (as compared with the wild-type protein, Fig. 6) seem therefore to record the process of first becoming promiscuous (F48I and L222R) and then keeping the responsiveness to the new effector while suppressing the others (I136T and S174R). Moreover, these results reveal a considerable epistatic connection between the mutations that surely reflect a mechanistic cooperativity among all residues involved. In particular, the results of the second round of mutagenesis indicate that two changes act synergistically to produce the final phenotype.

The picture that emerges from this work is that of a non-symmetrical transition between peaks of effector-specificity, which seems to require first the acquisition of a stem protein state followed by the stepwise and cooperative climbing of a valley of effector promiscuity. The intriguing aspect of this process is that acquisition of the stem form seems to be easily achievable, while development of a new specificity requires sequential and well-orchestrated changes in various amino acid residues. This is somewhat counterintuitive, as varying directly the interaction sphere for the inducer would in principle appear to be a more straightforward way to go. Perhaps the main constraints that transcription factors must overcome while developing a new specificity in vivo are more related to protein stability and competitive fitness along the intermediate states than to assembling an optimal effector-binding architecture in the corresponding protein pocket (Bloom and Arnold, 2009; Khersonsky and Tawfik, 2010; Soskine and Tawfik, 2010). An interesting possibility is that the inducer-promiscuity of stem regulators is due to metamorphic structures in the inducer binding domain (Yadid et al., 2010), an scenario that deserves further studies. In any case, the development of new computational remodelling (Looger et al., 2003; Murphy et al., 2009) together with high throughput screening methods will help to accelerate mutation/selection in vitro and to make sense of the corresponding evolutionary roadmap.

Experimental procedures

Strains and culture conditions

Pseudomonas putida KT2440 (Franklin et al., 1981), P. putida SF05 (Fernandez et al., 1994), P. putida TEC2 (Galvao and de Lorenzo, 2005), P. putida TaPuLux and E. coli DH5α 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 TEC2 strain, amended with uracil (20 µg ml−1, Sigma Aldrich). Antibiotics were used at the following concentrations: Kanamycin (Km) at 50 µg ml−1, piperacilin 40 µg ml−1, potassium tellurite (Tel) at 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 (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

Pseudomonas putida TaPuLux was constructed by inserting the transposon mini-Tn5[Tel-attPuLux] in the chromosome of P. putida KT2440. This mobile element, which carries a transcriptional fusion between the Pu promoter of the TOL plasmid pWW0 and the luxCDABE operon was constructed as follows. First, a XmaI/BamHI 309 bp fragment containing the Pu promoter was excised from the plasmid pMAD (Cases et al., 1996) and introduced in pattFRTLux (de Las Heras et al., 2008), thereby replacing the FRT sequence located upstream the luxCDABE operon. The resulting plasmid was digested with NotI and a fragment containing the Pu → luxCDABE fusion was ligated to pJMT6 (pUT/mini-Tn5Tel; Sanchez-Romero et al., 1998), generating pTn5Tel-attPuLux. This construct was transferred to P. putida KT2440 by tripartite mating with the helper strain E. coli HB101 (RK600) as above. Exconjugants were selected on M9 minimal media with MgSO4, citrate and kanamycin and were tested to be sensitive to piperacillin. One of these strains derived from KT2440 was designated P. putida TaPuLux and kept for further use. The starting plasmid used in site-directed mutagenesis experiments and construction of the XylR17 mutagenesis library was pBBxylR (de Las Heras et al., 2008). This plasmid carries the xylR gene under the control of its own promoter Pr, which efficiently expresses the full-length XylR protein. An equivalent plasmid encoding variant XylRv17 (pBBxylRv17[F48I, L222R]) was constructed by exchanging the 713 bp EcoRI/AvrII fragment of pBBxylR carrying the A and B domains of XylR by the same fragment of plasmid pURXv17 (Galvão et al., 2007). The rest of the plasmids carrying mutant versions of XylR were made by site-directed mutagenesis or by random mutagenesis using the pBBxylR or pBBxylRv17[F48I, L222R] as templates (see below).

Site-directed mutagenesis

In order to introduce specific changes in the A and B domains of XylR we used the QuickChange Site-directed mutagenesis kit (Stratagene), which is based in the procedure of Wang and Malcolm (1999). Specifically, we used primers F48IFw/Rv, L222RFw/Rv, I136TFw/Rv and S174RFw/Rv to generate pBBxylR[F48I], pBBxylR[L222R], pBBxylR[I136T] and pBBxylR[S174R] respectively (Table S1). On the other hand, in order to construct saturation mutagenesis libraries in specific XylR residues we subjected pBBxylR to the same procedure but using primers F48NNNFw and F48NNNRv for targeting position F48 and L222NNNFw and L222NNNRv for L222 position (Table S1). This method has two parts. In the first stage, 200 ng of pBBxylR was used as template in each of two PCR reactions performed in separate tubes, one containing 1 µM of the forward primer and the other 1 µM of the reverse, both containing the amplification cocktail described in the protocol of Quick Change Site-directed mutagenesis kit. The reaction was initiated by preheating to 95°C for 1 min and then allowed to proceed for 10 cycles of 95°C (30 s), 55°C (1 min) and 68°C (8 min). In the second step 25 µl of each reaction tubes was mixed in one tube along with 2.5U of the Herculase Enhanced DNA Polymerase (Stratagene) and subjected to a preheating of 95°C for 1 min and 18 cycles of 95°C (50 s), 55°C (50 s) and 68°C (16 min) and a final extension step of 68°C for 7 min. Following amplifications, 10 units of DpnI was added for 3 h in order to eliminate the template. 3 µl of the resulting PCR sample was then electroporated into 50 µl of E. coli DH5α.

Random mutagenic PCR

Error-prone amplification of target DNA sequences was performed using 25 ng of plasmid pURXv17 as template and primers NXylREco and AvrREV described previously (Galvão et al., 2007), according to the directions of the GeneMorph Random Mutagenesis Kit (Stratagene). The reaction was allowed to proceed for 30 cycles of 95°C (1 min), 55°C (1 min) and 72°C (8 min). The resulting PCR product (713 bp) was digested with EcoRI and AvrII, and ligated to pBBXylR predigested with the same enzymes (what removes the sequence of the wild-type AB domains of XylR). Finally, the ligation mixture was electroporated into E. coli DH5α creating a 106 library of xylR variants evolved from the precursor mutant xylRv17.

Screening of XylR mutants

The pools of xylR sequences obtained after either saturation mutagenesis of positions F48 and L222 of the wild-type or the error-prone PCR procedure (above) made on xylRv17 were captured in E. coli DH5α, transferred by conjugation to the test strain P. putida TEC2 and selected on solid M9 minimal media without uracil but amended with MgSO4, citrate, kanamycin and 1 mM 2,4-DNT (from a 1 M stock in DMSO). After 48 h at 30°C, colonies were replica-plated in the same medium without 2,4-DNT in order to identify and discard constitutive XylR variants. Where indicated, those clones whose growth in this medium was dependent on 2,4-DNT were by re-plated on M9 medium with uracil, MgSO4, citrate, Km and FOA (Galvao and de Lorenzo, 2005) and incubated in the presence of m-xylene to identify when required, clones with a reduced growth. After these two steps of selection/counter selection, the surviving clones were transferred first to the Pu → lacZ strain P. putida SF05 (Fernandez et al., 1994) and then to the Pu → luxCDABE equivalent P. putida TaPuLUX for further analyses.

XylR activity readouts with Pu → lacZ and Pu → luxCDABE reporters

Overnight cultures of P. putida SF05 and P. putida TEC2 carrying pBBxylR and its derivatives were diluted in 10 ml LB to OD600 of approximately 0.05 and grown at 30°C. When cultures reached OD600 = 1–1.2, the appropriate inducer (m-xylene or 2,4-DNT) was added to the concentration indicated in each case. β-galactosidase activity of cells permeabilized with chloroform and SDS was measured after 4 h, as described (Miller, 1972). To determine bioluminescence, 2 ml of the cultures of the strains under study was first pre-grown in 10 ml test tubes overnight in LB at 30°C. Next, they were diluted to an OD600 of 0.05 and grown to an OD600 ∼ 1.0 in 100 ml flasks. At this point they were exposed, where indicated, to m-xylene or 2,4-DNT. After 4 h, 200 µl aliquots was placed in 96 well plates (NUNC) and light emission was measured in a Victor II 1420 Multilabel Counter (Perkin Elmer). The bioluminescence data values reflect total light emission (in arbitrary units) divided by the optical density of the culture (OD600). The values of either β-galactosidase or luminescence reported throughout this work represent the average of at least three independent experiments.

Protein techniques

Protein analyses were carried out according to published protocols (Sambrook et al., 1989). For detection of the XylR (wild type and variants), 5 µg of protein extract of whole P. putida cells was denatured in a sample buffer containing 2% SDS and 5% β-mercapthoethanol and run in 10% polyacrylamide gels. These were subsequently blotted onto a polyvinylidene difluoride (PVDF) membrane (Immobilon-P; Millipore) using a semi-dry electrophoresis transfer apparatus (Bio-Rad). After protein transfer, membranes were blocked for 2 h at room temperature with MBT buffer (0.1% Tween and 5% skim milk in PBS). For detection of XylR, membranes were incubated with MBT buffer containing a dilution 1/2000 of anti-XylR Phab (Fraile et al., 2001). The membranes were subjected to 5 min washing steps in 40 ml of MBT buffer alone or MBT with 0.1% sodium deoxycolate in the case of the washing of the membranes hybridized with Phabs. To detect the anti-XylR Phab an anti-M13 peroxidase conjugate was utilized (1/5000 dilution in MBT). Blots were then incubated for 1 h at room temperature with secondary antibodies and washed in MBT buffer in 5 steps of 5 min. The protein bands corresponding to XylR, its derivatives and σ54 were developed, followed by reaction with a chemoluminescent substrate (ECL; Amersham Pharmacia Biotech). The structure of XylR A domain was analysed using The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC.


This work was supported by generous research grants of the Spanish Ministry of Science and Innovation (CONSOLIDER), by contracts of the Framework Program of the EU (BACSINE, MICROME) and by Funds of the Autonomous Community of Madrid.