The XylS/Pm regulator/promoter system and its use in fundamental studies of bacterial gene expression, recombinant protein production and metabolic engineering

Summary The XylS/Pm regulator/promoter system originating from the Pseudomonas putida TOL plasmid pWW0 is widely used for regulated low‐ and high‐level recombinant expression of genes and gene clusters in Escherichia coli and other bacteria. Induction of this system can be graded by using different cheap benzoic acid derivatives, which enter cells by passive diffusion, operate in a dose‐dependent manner and are typically not metabolized by the host cells. Combinatorial mutagenesis and selection using the bla gene encoding β‐lactamase as a reporter have demonstrated that the Pm promoter, the DNA sequence corresponding to the 5′ untranslated end of its cognate mRNA and the xylS coding region can be modified and improved relative to various types of applications. By combining such mutant genetic elements, altered and extended expression profiles were achieved. Due to their unique properties, obtained systems serve as a genetic toolbox valuable for heterologous protein production and metabolic engineering, as well as for basic studies aiming at understanding fundamental parameters affecting bacterial gene expression. The approaches used to modify XylS/Pm should be adaptable for similar improvements also of other microbial expression systems. In this review, we summarize constructions, characteristics, refinements and applications of expression tools using the XylS/Pm system.


Introduction
The Pm promoter and its cognate regulator gene xylS originate from the Pseudomonas putida TOL plasmid pWW0 and control expression of an operon encoding enzymes involved in the degradation of aromatic hydrocarbons (Worsey and Williams, 1975). The xylS gene encodes the AraC family positive transcriptional regulator XylS which upon binding to its effector becomes activated, binds as a dimer to its operator sequence and induces transcription from Pm. On pWW0, xylS is transcribed from two tandem promoters, Ps1 and Ps2. Ps1 is r 54 -dependent and inducible, while Ps2 is r 70 -dependent and provides constitutive, low-level expression of XylS (Gallegos et al., 1996) (Fig. 1). The XylS/Pm expression cassette including the Ps2 promoter was initially used for the construction of broad-host-range vectors, which were based on the RSF1010 replicon and contain xylE as a reporter protein. Regulated expression of XylE from vector pNM185 was demonstrated in 15 of totally 18 tested different Gram-negative bacterial species (Mermod et al., 1986). Further modifications of pNM185 resulted in construction of plasmids pERD20 and pERD21 (Ramos et al., 1988) carrying a mutant xylS gene, designated xylStr6, with altered effector specificity. These latter plasmids were tested in Escherichia coli by using b-galactosidase as a reporter gene and 3-8 fold elevated expression levels were obtained compared with the original vectors under Pm induction conditions. Moreover, pERD20 and pERD21 displayed a broad temperature range for inducible expression and functioned well in 7 different Gram-negative bacterial species tested. These early studies showed for the first time that XylS/Pm has a potential for regulated recombinant expression in many different Gram-negative species. Later, the system was used in combination with the broad-host-range minimal replication elements oriV (origin of vegetative replication) and trfA (encodes the TrfA protein required for initiation of replication at oriV and also controls plasmid copy number) from the naturally occurring RK2 plasmid, resulting in a set of versatile plasmid vectors with known nucleotide sequences (Blatny et al., 1997a,b). In these vectors, the copy number is adjustable by introducing desired point mutations in the trfA gene. The vectors also contain polylinkers facilitating cloning of target genes at an ATG site appropriately positioned relative to the DNA sequence encoding the native Pm ribosome binding site to ensure efficient translation. A common feature for expression vectors containing XylS/Pm is that several alternative, non-toxic and cheap benzoic acid derivatives can be used as inducers. They enter the cells by passive diffusion and function in a dose-dependent manner, usually without being metabolized. These characteristics together with the properties of the expression cassette itself explain why XylS/Pm is such a useful tool for a variety of applications in many different bacterial species.
Despite the many favourable properties of wild-type XylS/Pm, the system has been substantially improved by refinement of the Pm promoter sequence, the 5 0 -untranslated region of Pm-derived transcript, the xylS coding sequence and certain 5 0 -terminal fusion partners. This allowed for a full exploitation of the potential of this expression cassette both for high-level protein production and applications in which tightly controlled expression at more physiologically relevant levels is desired. The outcomes of such approaches have provided new genetic tools widening the range of further applications and have also contributed to new fundamental insights into bacterial gene expression. The methods used to achieve these results should be possible to apply for any expression cassette in principle in any host, Fig. 1. The upper scheme illustrates the catabolic operon for degradation of aromatic compounds in the Pseudomonas putida TOL plasmid pWW0. The upper-pathway operon is under transcriptional control of XylR/Pu and encodes enzymes that transform toluene into benzoate. Subsequently, benzoate is converted to tricarboxylic acid (TCA) intermediates by enzymes encoded in the meta-pathway operon, under transcriptional control of XylS/Pm. Circles, XylR; triangles, XylS; open symbols, regulator unable to activate transcription; solid symbols, regulator able to stimulate transcription; +, activating effect; À, repressing effect (Inouye et al., 1987;Greated et al., 2002). The lower part of the figure presents how XylS/Pm can be applied to express recombinant genes. Pm is activated by the XylS regulator when it forms a complex with a metapathway substrate entering passively into the cell. although the technologies have so far been used almost exclusively in E. coli. E. coli is among the most commonly used and best studied bacterial hosts for heterologous gene expression (for a review, see Rosano and Ceccarelli, 2014). Despite of this, functional heterologous protein production even in this host is still often a matter of trial and error, indicating that fundamental aspects of bacterial gene expression have not yet been fully understood. In this review, we describe many and, in several respects, unique properties of the XylS/Pm regulator/promoter system. We present its current applications in E. coli and many other bacterial species, most often for recombinant protein production and metabolic engineering types of experiments. Finally, we consider the potential of XylS/Pm as a valuable model system for basic studies on gene expression in bacteria.

The architecture and functioning of the XylS/Pm expression cassette
XylS is a positively regulating transcription factor belonging to the AraC-XylS family, and when activated by a benzoate-derived inducer, it binds to an operator sequence and initiates transcription from the Pm promoter (Fig. 2). The inducer molecules can appear in protonated or non-protonated states depending on the growth medium pH. Only the protonated forms can diffuse passively into the cells, meaning that the expression level when using this system may also be affected by pH (Winther-Larsen et al., 2000a,b). Transcription initiation from Pm is mediated by the XylS/inducer complex and r 32 -dependent RNA polymerase in early exponential growth phase or r 38 -dependent RNA polymerase in early stationary phase and thereafter Dom ınguez-Cuevas et al., 2005).
The origin and biology of the AraC-XylS family of transcription activators and their regulation have been extensively reviewed previously (Gallegos et al., 1997;Ramos et al., 1997;Martin and Rosner, 2001;Egan, 2002;Tobes and Ramos, 2002;Schleif, 2003Brautaset et al., 2009Santiago et al., 2016) and will therefore be only briefly outlined here. XylS is a 321 amino acid protein with a molecular mass of 36 kDa (Dom ınguez-Cuevas et al., 2008). To date, no experimental 3D structure of XylS exists as this protein is, like many AraC members, poorly soluble which so far has rendered its purification in active form unsuccessful (Aune et al., 2010;Dom ınguez-Cuevas et al., 2010). XylS appears to be composed of two separate and functionally independent domains: a conserved C-terminal domain (CTD) for DNA binding and interactions with RNA polymerase and an N-terminal domain (NTD) responsible for effector recognition and protein dimerization (Gallegos et al., 1997). Double mutants harbouring two specific single amino acid substitutions in the N-terminal domain of XylS have been constructed (Ru ız and Ramos, 2001, and the mutant protein displayed altered induction properties confirming the role of this domain for the effector binding. It has also been demonstrated that the C-terminal domain alone is able to activate transcription as efficiently as the full-length protein (Kaldalu et al., 2000), while the N-terminal domain represses DNA binding in the absence of benzoate effectors (Dom ınguez-Cuevas et al., 2010). XylS-CTD consists of seven a-helices folding into two helix-turn-helix (HTH) motifs. Two recognition helices (a-helix 3, a-helix 6) are critical for establishing contact with the Pm promoter in a region organized as two homologous 15-base pairs direct repeats, each consisting of a 5 0 -box A (TGCA) and a 3 0box B (GGTA) separated by 6 base pairs (Fig. 2). Recognition of these direct repeat sequences by XylS leads to the formation of a dimer acting in two consecutive steps. The first XylS monomer occupies the proximal binding site and facilitates binding of the second monomer to the distal site. Binding of the two monomers provokes a gradual bending of DNA and interaction of XylS with RNA polymerase as illustrated in Fig. 2 (Kessler et al., 1993;Gonz alez-P erez et al., 1999Gonz alez-P erez et al., , 2002Dom ınguez-Cuevas et al., 2008. Activation of XylS may be caused either by effector binding or by XylS hyperproduction leading to auto-induction independent of any inducer. One striking feature is that a high number of different benzoic acid-based compounds function as inducers for XylS (Silva-Rocha et al., 2011). To date, 59 different compounds have been tested and among these 31 compounds were reported to induce expression from Pm with different induction ratios (Table 1).
Combinatorial mutagenesis and selection to alter and improve the elements of the XylS/Pm expression system Directed evolution combining random mutagenesis protocols and selection has turned out to be more efficient than rational site-directed mutagenesis to improve enzymes, indicating that our current understanding of the correlation between protein structure and function is still limited. The same is, at least to some extent, true also for gene expression elements. Despite the substantial knowledge accumulated about DNA sequences and features of promoters, spacer regions, ribosome binding sites and 5 0 -untranslated mRNA regions, rational engineering of new and better expression systems remains challenging and highly unpredictable. Separate elements of the XylS/Pm system, i.e. the Pm promoter, the DNA sequence corresponding to the 5 0 -untranslated mRNA region (5 0 -UTR) of the Pm-derived transcript, the xylS coding region, as well as various types of 5 0 -terminal fusion partners that can stimulate expression of recombinant genes, have been modified and improved by using random mutagenesis and screening approaches. The common procedure has been to generate huge mutant libraries of the relevant genetic elements (i.e. by using doped oligonucleotides, error-prone PCR, DNA shuffling, or combinations thereof). Such elements were further cloned into plasmid vectors harbouring the bla gene (encodes b-lactamase), put under control of Pm and transformed to E. coli. Mutants were then directly selected by plating the heterogeneous recombinant cultures on solid medium containing different ampicillin concentrations, taking advantage of the fact that host ampicillin tolerance  The main focus has been on elevated expression levels, but the technology can be used to identify many other types of mutants too. The typical design and features of the selection vectors are illustrated in Fig. 3. These efforts resulted in construction of new and better expression tools expanding the properties and application range of XylS/Pm, and the experiments have also provided new basic understanding of genetic parameters affecting bacterial gene expression. Below, we summarize the most important findings from these experiments.
Generation of Pm promoter and 5 0 -UTR mutants displaying higher expression levels Plasmid pJT19bla has the bla gene under transcriptional and translational control of Pm and Pm-derived transcript including its rbs respectively (Winther-Larsen et al., 2000b). In this vector, the region upstream of the Pm transcriptional start site and the region corresponding to the 5 0 -UTR of the mRNA could be replaced in one-step cloning procedures by doped oligonucleotides carrying random nucleotide substitutions relative to the wild-type sequence (see Fig. 3A). By using this method, the desired regions could be mutagenized to an extent predetermined by the manner of synthesis of the oligonucleotides. Libraries of plasmids representing different Pm promoter and 5 0 -UTR mutants were constructed in E. coli and variants leading to altered expression properties were selected by plating the recombinant cells on solid media containing different ampicillin concentrations. Point mutations (in most cases several in each clone) leading to enhancements of induced expression levels and mutations causing reduced background expression The following compounds promoted no induction of the XylS/Pm or the induction ratios were not reported: 2,3-dichlorobenzoate (Liu et al., 2010), 3,5-dichlorobenzoate (Ramos et al., 1990a;Michan et al., 1992a;Liu et al., 2010), sodium benzoate (Purvanov and Fetzner, 2005), 5-chlorosalicylate, 4-chlorosalicylate, 3,5-dichlorosalicylate (Cebolla et al., 2002), 2,4-dimethylbenzoate (Ramos et al., 1990a;Zhou et al., 1990;Michan et al., 1992a), 2,6-dichlorobenzoate (Ramos et al., 1990a;Michan et al., 1992a), 3,5-dimethylbenzoate (Ramos et al., 1990a;Zhou et al., 1990), 2,6-difluorobenzoate, 2-iodobenzoate, 2,4-dichlorobenzoate, 3-hydroxybenzoate, 4-hydroxybenzoate, 2,5-dichlorobenzoate (Ramos et al., 1990a), 2,6dimethylbenzoate (Zhou et al., 1990), m-xylene, o-chlorotoluene, p-ethyltoluene, 1,2,3-trimethylbenzene, 1,3,4-trimethylbenzene, 2,5dichlorotoluene, 2,6-dichlorotoluene, benzyl alcohol, p-methylbenzyl alcohol, p-ethylbenzyl alcohol, m-chlorobenzyl alcohol (Abril et al., 1989). a. The ratio of the induced/basal expression. Induction was performed by using inducers at a concentration between 0, 1 and 5 mM (in most cases 1 or 2 mM). b. The activity of indicated inducer or presented induction ratio was reported for the mutagenized form of XylS protein. A. Vector tool for mutagenesis and selection using bla (encoding blactamase) as a reporter gene. The Pm promoter coding region, the Pm 5 0 -UTR coding region and the xylS coding region can individually be substituted by libraries of randomly mutagenized oligonucleotides and genes. The libraries were made by synthesizing one mutated strand for each of the three regions. During synthesis of these strands, the three alternative nucleotides were mixed at a varying percentage (for example, 4% each) with the nucleotide of the wild-type strand. The bases to be mutagenized were varied in different libraries. After synthesis, the DNA strands were annealed to their respective non-mutagenized complementary strands (Winther-Larsen et al., 2000b;Bakke et al., 2009). Cloning was then done by using the relevant restriction endonuclease sites indicated on the figure. Up and down mutants can be directly selected by growing the recombinant cells on different ampicillin concentrations. T is a transcriptional terminator. B. Vector tool for selection of 5 0 -UTR mutants based on translational re-initiation. The gene of interest (goi) is placed under control of Pm and the bla coding gene with its translational start codon overlaps with the goi stop codon (TGATG). Construction of mutant libraries of the Pm 5 0 -UTR coding region and screening for increased ampicillin resistance relative to wild-type was done as described in A above. were identified. Combination of such mutations in some cases resulted in generation of expression cassettes with strongly extended induction windows compared with the wild-type system. Pm promoter mutants with very high basal expression were also identified, indicating that background expression from both wild-type and altered Pm promoter sequences is independent of XylS (Winther-Larsen et al., 2000b). Later, these studies were extended by mutagenizing a larger region of the Pm promoter and by expansion of the mutant library . First Pm mutants with up to 10-fold elevated expression level of the bla reporter gene were selected, and the best mutant was then used as a template for a second round of random mutagenesis and selection. In this way, Pm promoter mutants with up to 14-fold elevated expression level of reporter genes were eventually selected. Interestingly, mapping of the mutations causing the improved phenotypes revealed that the DNA sequence alternations were apparently randomly distributed in the mutagenized region in a presumably unpredictable manner ). The 5 0 -UTR of Pm-derived transcripts contains the rbs consisting of the Shine-Dalgarno (SD) motif, the AUG initiation codon and a spacer region separating them, as well as upstream nucleotides. Initially, site-directed mutagenesis was used to generate 12 different mutants with 1-6 nucleotide substitutions in the DNA sequence corresponding to the rbs of Pm-derived transcripts (denoted as 'SD mutants') and the effects of these mutations on expression from Pm were tested by using the phosphoglucomutase gene celB as a reporter (Brautaset et al., 1998(Brautaset et al., , 2000. The obtained results demonstrated that the SD mutations caused 1.5-to 50-fold reduced expression level of CelB in E. coli. By using two alternative reporter genes, it was also shown that the effects of these SD mutants were highly gene specific (Winther-Larsen et al., 2000b). The experiments were later extended by constructing a 5 0 -UTR mutant library consisting of more than 25 000 variants and by selecting for up and down mutants using bla as a reporter gene, as described above . A number of 5 0 -UTR mutants, all carrying mutations located outside of the SD sequence and causing up to 20-fold elevated expression of b-lactamase were in this manner selected. This indicated that the native SD region is likely already close to optimal for high-level expression from Pm. Surprisingly, quantitative PCR analyses revealed that these 5 0 -UTR mutations caused up to 7-fold elevated bla transcript levels in the cells. By using alternative reporter genes, 5 0 -UTR mutants causing up to 15-fold increased transcript level from Pm were eventually isolated . For one selected 5 0 -UTR mutant, it was deduced that this effect was not caused by increased mRNA stability or any alternation within the Pm transcription start site. This was the first documentation that the 5 0 -UTR coding sequence can have a high impact on the transcription level of the cognate gene in E. coli.
By using celB as an alternative reporter gene, it was observed that the effects of the 5 0 -UTR mutations were context dependent. Therefore, the vector system was redesigned to enable selection of 5 0 -UTR mutants optimal for efficient expression of any recombinant gene (Berg et al., 2012). This time, celB and bla were cloned as a synthetic operon under control of Pm and with an overlap between the celB translational stop codon and the translational start codon of the bla gene. No rbs associated with the bla start codon was included, and any translation was thus dependent on translational reinitiation by ribosomes translating the upstream celB coding sequence of the mRNA (see Fig. 3B). In this way, any 5 0 -UTR mutations causing increased transcription and/or translation of celB should directly cause elevated b-lactamase production levels eventually affecting the ampicillin tolerance level of the recombinant cells. By using this novel selection tool, 5 0 -UTR mutants causing up to 3-fold elevated celB transcription level and 1.5-fold elevated CelB production level were selected (Brautaset et al., 1998(Brautaset et al., , 2000, demonstrating that this dual selection approach indeed functioned. Thus, because of its flexibility, the XylS/Pm cassette represents a valuable model system for basic studies aiming at expanding our understanding of genetic features affecting gene expression in bacteria. Mutagenesis of the xylS coding region and selection of mutants with altered functional properties As described above, XylS has a non-conserved N-terminal domain for effector binding and protein dimerization and a conserved C-terminal domain important for DNA binding and interactions with the host RNA polymerase. For identification of the specific regions involved in effector binding and determination of the models for XylS-mediated Pm activation, a large number of xylS mutations resulting in different amino acid substitutions within C-and N-terminal ends of XylS were constructed. The isolated XylS mutant proteins exhibited constitutively mediated transcription from Pm (Zhou et al., 1990), increased basal Pm activity and altered effector specificity and affinity (Ramos et al., 1990a,b;Michan et al., 1992a,b;Kessler et al., 1994). It was shown that XylS and its mutants can bind and respond differentially to many different chemical inducer molecules (Table 1) and when overproduced in the cells XylS can activate transcription from Pm in the absence of any inducer. Interestingly, amino acid substitution Phe291-Tyr in the second helix-turn-helix motif of XylS resulted in a mutant with a significantly higher activity than wild-type XylS (Manzanera et al., 2000). Aune et al. (2010) used a combination of error-prone PCR and DNA shuffling to randomly mutagenize the xylS coding region. XylS mutants with new effector profiles were then selected by using the bla reporter gene under the control of Pm (see above). Initially, a library of 430 000 xylS mutants was constructed in E. coli and screened for increased ampicillin tolerance level under induction conditions. The 40 most promising mutants contained totally 14 different amino acid substitutions within the xylS coding region and presented up to 3-fold stimulation of expression from Pm compared with wildtype XylS. Interestingly, all identified xylS mutations were located in the region encoding the N-terminal domain of XylS. Combination of two or more of the selected mutations in one xylS gene generated unpredictable results. Therefore, 28 of the best mutant xylS genes were used as templates for DNA shuffling to recombine random combinations of mutations, and then the resulting library was exposed to a second round of screening for high levels of expression from Pm. This approach identified xylS variants with several beneficial mutations causing up to 10-fold increased ampicillin host tolerance under induction conditions. One of these xylS mutant genes (StEP-13) used together with the wild-type Pm promoter allowed for 9-fold increased protein production level of a single-chain antibody variable fragment denoted scFv-phOx (Sletta et al., 2004) in comparison with the wildtype xylS (Aune et al., 2010). However, it was later shown that at high inducer and XylS concentrations the wild-type and the StEP-13 proteins resulted in similar maximum expression levels (Zwick et al., 2013). 5 0 -terminal DNA sequences that encode protein translocation signal sequences can strongly affect the recombinant gene expression level from Pm 5 0 -terminally fused DNA sequences encoding signal peptides for Sec pathway dependent protein translocation might confer an unexpectedly high impact on the expression level of heterologous proteins in E. coli (Sletta et al., 2007), and the choice of an optimal signal sequence is presumably protein dependent in an unpredictable manner (Li, 2015). A novel model system based on XylS/Pm was designed, enabling selection of improved signal sequences affecting the expression level and/or the translocation efficiency of recombinant proteins in E. coli. More specifically, a plasmid vector (pCSP1bla) was constructed in which the bla gene, with its native signal sequence, was replaced with a synthetic signal peptide denoted CSP (Sletta et al., 2004) and put under control of Pm (Heggeset et al., 2013). This approach takes advantage of the fact that b-lactamase must be both expressed and translocated into periplasm to confer its biological function, and any mutations in the CSP coding region that could improve expression and/or translocation of this protein should result in elevated ampicillin tolerance of the host cells. The CSP coding region was randomly mutagenized by using doped oligonucleotides and the resulting library consisting of ca 137 000 clones was screened for mutants with increased ampicillin tolerance level. In this way, CSP mutants causing up to 5.5-fold elevated expression and translocation of b-lactamase were identified. Bioinformatics-based analyses of around 20 different selected CSP variant DNA sequences could not rationally explain the obtained results. Interestingly, some of the CSP mutants could be also used for efficient expression and translocation of several alternative heterologous proteins in E. coli. These results highlighted the importance of optimizing the 5 0 -terminal region of coding genes for their efficient expression in E. coli and also demonstrated that rational approaches using available bioinformatics tools could apparently not be applied to predict the outcomes. To decouple any potential effects of translocation process itself, the 5 0 -terminal region of the celB gene (Brautaset et al., 1994) was tested as an alternative fusion partner to overexpress the human interferon alpha 2b (IFN-a2b) gene intracellularly . celB can be functionally expressed to very high levels, while IFN-a2b is poorly expressed, using XylS/Pm in E. coli (Brautaset et al., 1998(Brautaset et al., , 2000Winther-Larsen et al., 2000b;Bakke et al., 2009;Berg et al., 2012). Totally 13 different celB fusion partners of varying lengths were fused in frame with the 5 0 -end of the IFN-a2b coding sequence, and expressed under control of the Pm promoter. The celB fusion partners ranging from 24 to 207 nucleotides long caused between 7-fold and 60-fold stimulation of expression at the transcript and protein levels, respectively, in E. coli. Further mutagenesis of the selected celB fusion sequences allowed for additional improvements which were also shown to be useful for high-level heterologous production of IFN-a2b in E. coli under high-cell density cultivations ) (see below).

Combining optimized mutant genetic elements to expand the expression window of XylS/Pm
Overall, by using combinatorial engineering approaches, several different genetic elements of the XylS/Pm expression systems including the Pm promoter, the DNA sequence corresponding to its mRNA 5 0 -UTR, the xylS coding region as well as external fusion partners (celBderived regions and the CSP translocation signal sequence) have been modified to improve both transcription, translation and translocation of heterologous proteins in E. coli. Zwick et al. (2012) reported that the b-lactamase expression level could be up to 75-fold and 50-fold increased at the protein and transcript levels, respectively, by combining optimized Pm, 5 0 -UTR and xylS regions. Similar results were obtained when using alternative reporter proteins. It was shown that even a single copy of such a multisite modified XylS/Pm expression cassette integrated into the E. coli chromosome could confer higher recombinant b-lactamase production level than the analogous wild-type plasmid present in multiple copies per genome (Zwick et al., 2013).

Comparison of XylS/Pm performance with other relevant expression systems for regulated and high-level heterologous gene expression in E. coli
Several different expression systems have been developed and work well for heterologous protein production in E. coli (Terpe, 2006;Brautaset et al., 2009;Tegel et al., 2011). For example, the T7 promoter originating from bacteriophage T7 (Studier and Moffatt, 1986) is recognized by its strength associated with the affinity of the highly selective T7 RNA polymerase which provides effective transcription initiation and in vitro elongation rate of 250 nucleotides per second compared with 50 for E. coli RNA polymerase (Golomb and Chamberlin, 1974). Placing the T7 polymerase gene under control of the Plac promoter allows induction of transcription from the PT7 promoter by adding isopropyl-b-D-1-thiogalactopyranoside (IPTG) as an inducer. This system is, in contrast to XylS/Pm, based on negative transcriptional regulation mediated by the LacI repressor. A similar control mechanism is applied in case of the strong synthetic Ptrc promoter, which has been used to express heterologous proteins up to 15-30% of total cell protein in E. coli (Terpe, 2006). Another popular expression system is AraC/P BAD in which the positive regulator AraC stimulates transcription from P BAD upon induction with arabinose. Balzer et al. (2013) made an extensive comparative analysis of LacI/PT7lac, LacI/Ptrc, AraC/ P BAD , wild-type XylS/Pm and its high-level expression variant Pm ML1-17, in E. coli hosts. The main premise of the study was standardization of the design of the vectors to reduce influence of parameters unrelated to the features of the expression systems themselves. The most apparent observation following from these experiments was that the LacI/PT7lac system generates uniquely high amounts of transcripts. This property, however, typically correlates with an overload of the translational machinery eventually resulting in production of insoluble and inactive proteins. When considering protein functionality, weaker promoters sometimes allow to obtain higher yields of soluble and correctly folded proteins. In terms of low background expression, LacI/Ptrc turned out to be most leaky and it also displayed the smallest induction window. The most tightly regulated promoter system was AraC/P BAD . Calculation of translation initiation rate (TIR) values indicated that AraC/P BAD and LacI/PT7lac transcripts are theoretically characterized by the most efficient translation. LacI/PT7lac and XylS/Pm ML1-17 tended to produce the highest amount of recombinant proteins while XylS/Pm ML1-17 showed higher yields of active proteins per transcript. One general advantage of using XylS/Pm is that it does not require any host-mediated inducer uptake system, and most often the inducer is not consumed. In contrast, the PT7 promoter needs host strains expressing the T7 RNA polymerase, usually from the Plac promoter. In case of AraC/P BAD , bacterial hosts should preferably be unable to catabolize the L-arabinose inducer, but must be able to take up this compound. The XylS/Pm system is therefore easy to adapt to new bacterial hosts, what makes making it a very attractive candidate when the conditions of recombinant protein production have not previously been standardized.
Recently, a comparative microfluidic single-cell analysis of LacI/PT7lac, AraC/P BAD and XylS/Pm ML1-17 in the synthetic M9CA growth medium was reported (Binder et al., 2016). Such well-defined experimental set-up provided high environmental homogeneity. The focus was preliminary on investigating the influence of different inducer molecules including their concentrations, and uptake mechanisms, on phenotypic heterogeneity as well as other system specifications. It was demonstrated that IPTG induction of LacI/PT7lac analysed in the E. coli strain BL21 (DE3) with an active lactose uptake mechanism, and salicylate induction of XylS/Pm ML1-17 analysed in E. coli strain Tuner (DE3), led to the strongest initial expression and significant growth impairment. In contrast, analogous induction with IPTG using E. coli Tuner (DE3) with a passive lactose uptake mechanism and m-toluate induction of XylS/Pm ML1-17 showed intermediate responsiveness and hardly any interference with growth compared with respective non-inducing conditions. Analysis of leaky expression confirmed the results obtained by Balzer et al. (2013) (see above) indicating that AraC/P BAD is the most tightly regulated among the expression systems tested. Interestingly, XylS/Pm induced either by m-toluate or salicylate revealed significant leaky expression leading to the subsequent moderate dynamic ranges of induction. Authors concluded that observed high basal expression of the system was probably triggered by use of mutagenized high-level expression variant of Pm promoter (ML1-17). Finally, expression responses of XylS/Pm induced by mtoluate and LacI/PT7lac induced by IPTG in the absence of the active lactose uptake mechanism were characterized as the most homogenous. Summarized, it was proven that the type of inducer and the presence of inducer uptake systems can have a high impact on phenotypic heterogeneity, and this should be considered when choosing between different promoter systems.
In a separate study (Royo et al., 2005a), the performance of different promoter systems for expression of dioxygenase genes in E. coli was investigated. Comparison of the rate of indigo accumulation in the recombinant strains revealed that the induction level of Pm was slightly higher than in case of the PT7 and Ptac promoters. However, all of these multicopy plasmid-based systems were unstable when serially diluted batch experiments were performed without a selective pressure. The problem was solved by integrating the Pm expression module into the bacterial chromosome. Despite the gene dosage reduction and initially slower accumulation rate, the chromosomal system allowed for tightly controlled and stable production of indigo in amounts comparable to a multicopy plasmid, or a different plasmid system based on the tac promoter. In general, expression systems based on strong promoters like Ptac, Ptrc and PT7 are rather considered to be unstable and very little is known about their performance after single copy integration into the chromosome (Royo et al., 2005a).
Sometimes utility of the promoter system may be limited by its host specificity. Bi et al. (2013) demonstrated that among all tested promoter systems, PBAD and Pm provided the highest expression level of the red fluorescent protein (RFP) upon induction in Ralstonia eutropha, whereas the PlacUV5, Ptet and Ppro systems were not functional in this host. These results argue in favour of the broad-host-range properties of the XylS/Pm regulatory cassette.

Application of XylS/Pm for heterologous protein production under high-cell density cultivations in E. coli
In addition to the laboratory-scale experiments aiming at high-level protein production, the XylS/Pm expression system has also been tested under more industrially relevant conditions. The expression system was reported to be useful for high-level production of different human medical proteins, including granulocyte-macrophage colony-stimulating factor (GM-CSF), IFN-a2b and a singlechain antibody variable fragment (scFv-phOx) as well as recombinant fish vaccines under high-cell density cultivations (HCDC) of E. coli (Sletta et al., 2004(Sletta et al., , 2007Tøndervik et al., 2013). Under such conditions, low background expression and strong induction have proven to be critical to obtain a high-level of cell growth in the fermenter prior to Pm induction, leading to high volumetric production yields and preventing unwarranted loss of plasmids from the recombinant cells during the growth phase. Single-chain antibody fragment scFv-phOx must be translocated to the periplasm to fold into soluble and functional form. It is also regarded as host toxic in the sense that its overexpression and translocation eventually cause lysis of the recombinant cells. Thus, careful regulation of scFv-phOx expression is critical in particular under HCDC. In agreement with what was described above, it was demonstrated that fusion sequences optimized by combinatorial mutagenesis and selection could stimulate high-level expression also of other heterologous protein in E. coli under HCDC (Heggeset et al., 2013;Kucharova et al., 2013;Tøndervik et al., 2013).

Application of XylS/Pm for fine-tuned regulated expression of genes and gene clusters in many different bacterial species
XylS/Pm displays many favourable properties, which make it a valuable tool for metabolic engineering and other purposes which require fine-tuning of expression of genes or gene clusters, usually at physiologically relevant levels. The expression system has been shown to function well in a wide range of different Gram-negative organisms, and recently also in Gram-positive species. The possibility to use XylS/Pm for fine-tuning of expression is a consequence of the nearly proportional relation between the expression level and the concentration of the inducer. In some cases, even the uninduced expression level from Pm may be higher than desired, emphasizing the need for mutants that exhibit reduced background expression while still being inducible. A complete list of all reported bacterial species in which the XylS/Pm system has been applied is presented in Table 2 and selected examples are outlined below. Due to its applicability to a wide range of bacterial hosts, XylS/Pm is available in the SEVA-DB as one of the broad-host-range expression cargos formatted following the SEVA standard to allow combination of the system with other optimal plasmid elements (SEVA-DB, http://se va.cnb.csic.es; Silva-Rocha et al., 2013).

Broad-host-range applications of the wild-type XylS/Pm expression cassette
The xanA gene of Xanthomonas campestris encodes a bifunctional phosphogluco-mannomutase required for biosynthesis of the commercially important polysaccharide xanthan (K€ oplin et al., 1992). By expressing xanA from XylS/Pm in a xanA-deficient X. campestris host, the synthesis of xanthan could be monitored in induced and uninduced cells. There was virtually no xanthan synthesis in the absence of inducer, while polymer synthesis was activated to wild-type levels upon induction of the Table 2. Host organisms used for the XylS/Pm-mediated expression of heterologous proteins and examples of applications in these hosts.

Azotobacter vinelandii
Over-expression of NifH (Fe protein), AlgE3 expression, assay of b-galactosidase activities Pm promoter. The XanA production under these conditions was not limiting for xanthan biosynthesis in X. campestris (Winther-Larsen et al., 2000a). Mart ınez-Garc ıa and de Lorenzo (2011) employed a procedure involving two basic plasmid architectures aiming for multiple markerless gene replacements in a range of different of Gram-negative bacterial species including P. putida. One plasmid was responsible for introducing I-SceI site(s) within the target genome region through homologous recombination between plasmid-encoded DNA and the chromosome. The second plasmid provided conditional expression of the I-SceI endonuclease upon activation of XylS/Pm in a manner that does not depend on the cellular growth phase (Mart ınez-Garc ıa and de Lorenzo, 2011). The results obtained proved the effectiveness of the I-SceI methodology which later allowed for introduction of targeted deletions into 11 chromosomal regions (comprising 300 genes) of P. putida and significantly improved the growth properties of the resulting recombinant strain (Mart ınez-Garc ıa et al., 2014). Chromosome cleavage with unique I-SceI sites and XylS/Pm-controlled expression of the target enzyme were also applied to eliminate operons encoding anthranilate phosphoribosyltransferase, indole-3glycerol phosphate synthase and chorismate mutase and to establish anthranilate production in P. putida (Kuepper et al., 2015).
The XylS/Pm-based reporter system was used to control expression of a fluorescent protein denoted EcFbFp (E. colioptimized flavin mononucleotide-based fluorescent protein) and alkyl halide degradation operon from Pseudomonas pavonaceae responsible for organohalide metabolism. Successful expression and resulting activity of these proteins confirmed the capabilities of the recombinant strain to grow under anoxic conditions (Nikel and de Lorenzo, 2013).  reported the construction of a programmed self-disruptive P. putida BXHL strain that facilitates the release of polyhydroxyalkanoic acid (PHA) granules to the extracellular medium. This is biotechnologically important as an efficient PHA recovery process is essential to reduce the cost of microbial PHA production. The engineered system was based on two proteins from the pneumococcal bacteriophage EJ-1, Ejh holin and Ejl endolysin, and the corresponding ejh and ejl genes were inserted into the chromosome of a tolB mutant of P. putida KT2440 under control of the XylS/ Pm. The tolB gene encodes a periplasmic protein and a mutation in this gene causes alternations in the outer membrane stability. With this expression system, cell lysis could be controlled by using 3-methylbenzoate as inducer. Valls et al. (2000) described the construction of a R. eutropha strain with an enhanced ability to immobilize Cd 2+ ions from the external media. The effect was observed as a result of stable chromosomal integration of the minitransposon TnMTb-1 containing the mtb gene placed downstream of the Pm promoter. This cassette allowed for expression of the mouse metallothionein I (MT) protein fused to the autotransporter b-domain (MTb) of the IgA protease of Neisseria gonorrhoeae. Production of MTb was found to be strictly dependent on the presence of 3-methylbenzoate in the growth medium, thus demonstrating the tight control of the Pm promoter in R. eutropha.
In Myxococcus xanthus, the myxothiazol biosynthetic gene cluster mta originating from Stigmatella aurantiaca was placed under control of the XylS/Pm system and integrated into the M. xanthus chromosome . The resulting recombinant strain produced myxothiazol in yields comparable to the natural S. aurantiaca producer strain. XylS/Pm was also used for controlled expression of genes involved in biosynthesis of secondary metabolites that may be toxic for the host strain. Heterologous expression of the myxochromide S cluster from S. aurantiaca in a P. putida mutant strain resulted in high myxochromide production levels in the recombinant cells (Wenzel et al., 2005).
Interestingly, the XylS/Pm expression system was recently also demonstrated to function in Gram-positive bacteria (Dragset et al., 2015). By making some necessary modifications to the XylS/Pm regulated gene expression vector, robust time-and dose-dependent reversible induction accompanied by low background expression levels was obtained in both Mycobacterium smegmatis and Mycobacterium tuberculosis (Table 2). This result should open up opportunities for exploring the application of the XylS/Pm expression system also in other Gram-positive bacteria.
In the RK2-based broad-host-range expression vectors harbouring XylS/Pm (see above), plasmid replication relies on the initiation protein TrfA encoded by the trfA gene located on the vectors (Blatny et al., 1997a). A plasmid denoted pJBSD1 was constructed with a mutant version of the trfA gene placed under control of Pm, and this plasmid was demonstrated to be dependent on a Pm inducer to replicate in E. coli. This plasmid can be used as a conditional suicide vector system for targeted chromosomal integration via homologous recombination in E. coli and potentially also in other Gram-negative bacteria (Karunakaran et al., 1999).
XylS2 is a mutant derivative of XylS that can be activated by salicylic acid . The xylS2 gene together with Pm was used as key components in the construction of a novel regulatory circuit demonstrated to be useful in Salmonella bacteria with two modules operating in cascade (Royo et al., 2007). More specifically, the expression of xylS2 was coupled to the NahR-dependent Psal promoter, and the cassette was inserted into the bacterial chromosome. NahR is a transcriptional activator that can be induced by salicylate and then promotes transcription from Psal. This genetic background was then used in a host for the expression of plasmid-borne reporter genes placed under the control of the Pm promoter. In the absence of salicylate, XylS expression levels were low and XylS was not active, and accordingly expression of reporter genes from Pm was very low. In the presence of salicylic acid, NahR activated transcription from Psal and thus produced XylS2, which then subsequently bound the effector molecule salicylic acid, becoming activated and causing synergistically increased transcription from Pm. The system was demonstrated to be useful for studying bacterium-host interactions in vivo in both mouse and tumour cells by expressing the GFP protein from the Pm promoter (for a review, see Becker et al., 2010). Later the circuit has been modified and improved by using different replicons for the Pm expression module (Medina et al., 2011). This has been useful for in vivo studies of Salmonella upon infection of different eukaryotic cells (Mesa-Pereira et al., 2013).

Application of XylS/Pm mutant derivatives
Totally 12 different derivatives of the XylS/Pm expression cassette with alterations in the rbs coding region (denoted SD mutants; see above) were used for finetuned low-level expression of a heterologous phosphoglucomutase (Pgm) in a pgm-deficient mutant of E. coli growing in the presence of galactose (Brautaset et al., 1998(Brautaset et al., , 2000. Galactose enters the cells eventually as G-1-P and can be channelled into catabolism by the action of Pgm. In the absence of Pgm, G-1-P accumulation leads to biosynthesis of intracellular amylose. The recombinant cells were cultivated without any induction, and Pgm activity was downregulated up to 51-fold when using the SD mutants compared with the wild-type XylS/ Pm. In this way, amylose accumulation in the respective cells could be gradually varied demonstrating that very low expression levels may be needed to obtain full control of metabolic pathway activities. The induction ratios of mutant derivatives were also shown to be strongly affected compared with the wild-type XylS/Pm cassette (Winther-Larsen et al., 2000b).
XylS/Pm was also employed to modulate production, composition and localization of biosynthesis and export components of the important biopolymer alginate in P. fluorescens. This was achieved by controlled expression of the mannuronan C-5-epimerase gene algG (Gimmestad et al., 2003), the alginate lyase gene algL (Bakkevig et al., 2005) and the porin gene algE (Maleki et al., 2016) respectively. In the latter example, the density of alginate secretion components in the cell membrane could be modulated by using the unique properties of XylS/Pm for regulated low expression. In all these cases, a combination of chromosomal integration approach and a specific Pm promoter mutant denoted PmG5 was used to ensure physiological relevant low expression of the respective genes. PmG5 provides lower background expression in the absence of inducer than Pm wild type in P. fluorescens (Gimmestad et al., 2003).
XylS/Pm and its mutant derivatives were effectively used for controlled and functional expression of the biosynthetic gene cluster of the C50 carotenoid sarcinaxanthin originating from Micrococcus luteus, enabling efficient sarcinaxanthin production in E. coli (Netzer et al., 2010). The gene cluster includes totally seven protein coding sequences and substitution of single genes with heterologous genes allowed for production of unnatural C50 carotenoids in E. coli. By using certain Pm 5 0 -UTR down mutants, it was later demonstrated that the XylS/ Pm system could be used to control the sarcinaxanthin production level in recombinant E. coli cells and in this way metabolic bottlenecks in the sarcinaxanthin biosynthetic pathway could be identified (Lale et al., 2011). Recently, one of these 5 0 -UTR region modifications allowed to reduce the leakiness of the XylS/Pm system which was used in combination with PT7lac promoter to investigate potential of two Pseudomonas spp. strains for low-temperature expression of a red fluorescent reporter protein (mCherry) (Bjerga et al., 2016). Gemperlein et al. (2016) employed a specific 5 0 -UTR mutated version of the Pm promoter region for expression of the pfa biosynthetic gene cluster originating from Aetherobacter fasciculatus in a P. putida KT2440 host strain. The pfa gene cluster encodes a set of myxobacterial polyunsaturated fatty acid (PUFA) synthases which are polyketide synthase-like enzymes catalysing biosynthesis of long-chain (LC) PUFAs in A. fasciculatus. The recombinant strain was further engineered for co-expression of the afppt gene from A. fasciculatus encoding a phosphopantetheinyl transferase, proposed to catalyse phosphopantetheinylation of the PUFA synthases. The afppt gene was placed under the control of a separate XylS/ Pm cassette and integrated into the host chromosome. Induced recombinant expression in recombinant P. putida KT2440 strain resulted in 3-fold increased LC-PUFA production yield compared with the wild-type strain.

Conclusions
The inducible Pm promoter together with its cognate positive regulator XylS display many favourable properties that makes the XylS/Pm system highly valuable for different applications related to recombinant gene expression. The cassette can be used in a wide range of different Gram-negative bacteria and it was recently also demonstrated to function in Gram-positive organisms further extending the range of its potential applications. The system is characterized by the simple mode of regulation which can be achieved by using different inducers which are typically non-metabolized by a host and enter passively to the cells. These properties together with a dose-dependent induction response and low background expression in the absence of inducer make this expression system highly flexible for both high-level protein production and metabolic engineering in many host organisms. Synthetic biology continuously raises increasing need for useful expression systems, enabling fine-tuned expression of genes and gene cluster around physiologically relevant levels. Ongoing bioprospecting and advanced genetic engineering aim to generate synthetic gene clusters for microbial production of complex chemicals, such as antibiotics, biopolymers and terpenoids, for various medical and industrial applications. To fully explore the potential of such approaches, genetic tools enabling the functional expression of the desired genes in the preferred microbial host will be crucial. The combinatorial mutagenesis efforts made to improve and expand the properties of XylS/Pm have provided better tools for such purposes, and the technologies used to improve this particular system, as presented here, have great potentials to be used for alternative bacterial expression systems.