Abstract
Knowledge of adaptive processes encompasses understanding of the emergence of new genes. Computational analyses of genomes suggest that new genes can arise by domain swapping, however, empirical evidence has been lacking. Here we describe a set of nine independent deletion mutations that arose during the course of selection experiments with the bacterium Pseudomonas fluorescens in which the membrane-spanning domain of a fatty acid desaturase became translationally fused to a cytosolic di-guanylate cyclase (DGC) generating an adaptive phenotype. Detailed genetic analysis of one chimeric fusion protein showed that the DGC domain had become membrane-localised resulting in a new biological function. The relative ease by which this new gene arose along with its profound functional and regulatory effects provides a glimpse of mutational events and their consequences that are likely to play a significant role in the evolution of new genes.
Introduction
The emergence of new genes by mutation - readily identified through comparative genomics - provides an obvious and important source of adaptive phenotypes (Chen et al. 2013; Long et al. 2013; Zhang and Long 2014). Mutational mechanisms involve divergence of duplicate genes (Ohno 1970; Lynch and Conery 2000; Bergthorsson et al. 2007; Nasvall et al. 2012), exon shuffling, the domestication of transposable elements, retrotransposition, gene fusion, and the de novo evolution of new reading frames (Long et al. 2003; Kaessmann 2010; Tautz and Domazet-Loso 2011; Ding et al. 2012; Long et al. 2013; Zhao et al. 2014).
Mutations that result in chimeric genes - by recombination of parental genes into a single open reading frame, or by retrotransposition of a gene into an alternate reading frame - are likely to generate new genes with relative ease (Rogers et al. 2010; Ranz and Parsch 2012). This is because fusions stand to produce novel combinations of functional domains and regulatory elements with few mutational steps. For example, promoter capture, whereby a fusion event couples an existing gene to a new promoter, can cause abrupt changes in temporal and spatial patterns of expression (Blount et al. 2012; Rogers and Hartl 2012; Annala et al. 2013). Additionally, novel combination of domains may result in a range of post-translational outcomes, ranging from relocalisation of domains, to novel inter-protein associations, regulation of enzymatic activity and possibly even the formation of novel protein functions (Patthy 2003; Bashton and Chothia 2007; Jin et al. 2009; Peisajovich et al. 2010; Rogers and Hartl 2012; Singh et al. 2012)
Comparative computational analyses provide evidence for the evolutionary importance of gene fusion events generating chimeric proteins. Compelling data comes a diverse range of organisms, including bacterial species (Pasek et al. 2006) Oryza sativa (Wang et al. 2006), Drosophila spp. (Long and Langley 1993; Wang et al. 2000; Wang et al. 2002; Jones et al. 2005; Rogers et al. 2009; Rogers and Hartl 2012), Danio rerio (Fu et al. 2010), Caenorhabditis elegans (Katju and Lynch 2006) and humans (Courseaux and Nahon 2001; Zhang et al. 2009). Despite the apparent importance of gene fusion events, the functional effects of presumed instances of domain swapping have received little attention. Notable exceptions are studies of “new” Drosophila genes: sphinx, jingwei and Sdic, which affect courting behaviour, alcohol-dehydrogenase substrate specificity and sperm-motility, respectively (Zhang et al. 2004; Dai et al. 2008; Zhang et al. 2010; Yeh et al. 2012). These three genes have complex mutational histories, but originate from gene fusion events: retrotransposition of distal genes (Long et al. 1999; Wang et al. 2002) or duplication and fusion of neighbouring genes (Ranz et al. 2003).
Beyond studies of new genes inferred from computational analyses, synthetic biology has shown that genes created by domain swapping can have important phenotypic effects. For example, in vitro recombination of genes involved in yeast mating was shown to generate greater diversity compared with control manipulations in which the same genes were duplicated in the absence of recombination (Peisajovich et al, 2010). While an elegant demonstration of the potential importance of gene fusion events in evolution, observation of such events in natural populations is desirable.
Evolution experiments with microbes provide as yet unrealised opportunities to understand evolutionary process, including mechanistic details. During the course of studies designed to elucidate the range of mutational pathways to a particular adaptive “wrinkly spreader” (WS) phenotype (Spiers et al. 2002; Goymer et al. 2006; Bantinaki et al. 2007; McDonald et al. 2009), we discovered a number of gene fusion events (Lind et al. 2015). Two classes bore the hallmarks of promoter capture whereby deletions caused a focal gene to come under control of a more highly expressed promoter eliciting the adaptive phenotype (Lind et al. 2015). A third class of mutation appeared not to conform to expectations of promoter capture. Intriguingly, this class was defined by eight independent deletions each of which caused a translational fusion between two adjacent genes. As we show here, these fusions define new genes that are chimeras between a membrane-localised fatty acid desaturase and a cytosolic di-guanylate cyclase (DGC).
Results
The WS model system and FWS types
When propagated in a spatially structured microcosm P. fluorescens SBW25 undergoes rapid diversification, producing a range of niche specialist types (Rainey and Travisano 1998). Among the best studied is the wrinkly spreader (WS) (Rainey and Travisano 1998; Spiers et al. 2002; Spiers et al. 2003; Goymer et al. 2006; Bantinaki et al. 2007; McDonald et al. 2009; Lind et al. 2015). WS types are caused by mutations in one of numerous genes that regulate or encode di-guanylate cyclases (DGCs). These genes catalyse the synthesis of 3', 5' -cyclic-di-guanosine monophosphate (c-di-GMP), a signalling molecule that allosterically regulates a complex of enzymes that produce an acetylated cellulosic polymer (ACP) (Ross et al. 1987; Tal et al. 1998; Amikam and Galperin 2006; De et al. 2008). Over-production of the polymer is the proximate cause of the WS phenotype (Spiers et al. 2002; Spiers et al. 2003).
Mutations that cause the WS phenotype in P. fluorescens SBW25 typically reside in one of three DGC encoding pathways: Wsp, Aws or Mws (McDonald et al. 2009). These pathways encode post-translational negative regulators that, when the targeted by loss-of-function mutations, result in constitutive DGC activity (Goymer et al. 2006). Elimination of these three pathways by deletion led to the discovery of a further 13 mutational routes to the WS phenotype, all involving pathways encoding DGC domains (Lind et al. 2015). Within this set of 12 new mutational routes were three loci where deletion events led to gene fusions. In two instances, the phenotypic effects were explained by increased transcription of DGCs resulting from promoter capture (Lind et al. 2015). However, for the third, which involved fusion between two open reading frames (pflu0183 and pflu0184), there was little to suggest that the phenotypic effects derived from promoter capture (Lind et al. 2015).
All mutations involving pflu0183 described in Lind et al. (2015) involved deletions spanning parts of pflu0183 and pflu0184. Each deletion resulted in a predicted single open reading frame, transcribed from the promoter of pflu0184 (Figure 1A), thus maintaining the reading frame of pflu0183. Of the eight independent mutations at this locus, four consisted of an identical 467bp deletion (see Figure 1B and 1C). A ninth fusion, “L2-M5”, obtained from an independent experiment (see below) is also depicted in Figure 1. Reconstruction of the 467bp deletion in the ancestral SM genotype was sufficient to cause the WS colony morphology, niche specialisation and over-production of ACP (Figure 1 - figure supplement 1). All fusion mutations involving pflu0183 produced the characteristic WS phenotype: hereafter these are referred to as “FWS” types and the focal FWS type studied here is termed “FWS2” (Figure 1 - figure supplement 2).
The fadA-fwsR fusion
Pflu0183 encodes a predicted di-guanylate cyclase (DGC) henceforth referred to as ‘fwsR’. FwsR forms a predicted protein of 335 residues in length with a predicted PAS fold and a GGDEF domain at the C-terminus (Figure 1A). This GGDEF domain features a GGEEF motif, which is indicative of DGC activity (Chan et al. 2004; Galperin 2004; Malone et al. 2007; Wassmann et al. 2007). The neighbouring gene pflu0184 encodes a predicted protein 394 residues in length, including a predicted N-terminal fatty acid desaturase (residues 11 and 246) and a predicted transposase element (residues 261 to 289). Pflu0184 is hereafter referred to as 'fadA’. FadA contains two predicted transmembrane domains (TMDs) between residues 10 to 32 and 135 to 157. The gene arising following deletion is termed 'fadA-fwsR'.
Little is known about the functions of proteins encoded by either fadA or fwsR. FadA shares 83% identity with DesA (PA0286) from Pseudomonas aeruginosa (Winsor et al. 2011), which is a fatty acid desaturase (Zhu et al. 2006). Fatty acid desaturases modify phospholipids in the cell membrane in order to modify membrane fluidity in response to environmental change (Zhang and Rock 2008). In P. aeruginosa, transcription of desA is promoted during anaerobic conditions resulting in increased membrane fluidity as a consequence of double bonds introduced at the 9-position of fatty acid acyl chains (Zhu et al. 2006). Increased production of short-chained fatty acids has recently been shown to increase production of cyclic di-GMP by the Wsp pathway (Blanka et al. 2015). fwsR has orthologs across members of the genus Pseudomonas, however, none of these have been the subject of study.
It is not known whether the proximal relationship of fadA and fwsR orthologs across many Pseudomonas spp. reflects a functional or regulatory relationship between the two genes (or their protein products), however the operon prediction tool DOOR suggests they are separate transcriptional units (Mao et al. 2014). In species such as P. putida and P. stutzeri orthologs of fadA and fwsR are adjacent to each other (Figure 1 - figure supplement 3). However, in P. aeruginosa the locus of the GGDEF domain-encoding gene (PA0260) is separated from desA by three open reading frames (ORFs).
DGC activity of FadA-FwsR is necessary to generate FWS2
How do deletions between fadA and fwsR (and generation of fadA-fwsR) produce FWS phenotypes? Like other WS-causing mutations (Goymer et al. 2006; Bantinaki et al. 2007; McDonald et al. 2009) the causal fadA-fwsR fusion is likely to alter the activity of the DGC domain of fwsR resulting in over-production of c-di-GMP. The constrained length of the spontaneous deletion mutations causing FWS types (all fadA-fwsR fusions arose by deletion of no more than 90 bp of fwsR - see Figure 1C) suggests that activity of the fwsR-encoded DGC is required for the FWS phenotype. To test this hypothesis the conserved GGEEF motif that defines the active site of the DGC was replaced by a mutant allele (GGAAF) expected to eliminate DGC function (Malone et al. 2007). The phenotype of FWS2 carrying the mutant fadA-fwsR allele was smooth, non-cellulose (ACP) producing and unable to colonise the airliquid interface of static broth microcosms (Figure 2). This demonstrates that diguanylate cyclase activity of fadA-fwsR is necessary for the FWS2 phenotype.
FWS2 does not require fatty acid desaturase activity
A fadA-fwsR fusion mutant isolated during the course of an independent experiment, termed ‘L2-M5’, contains a fadA-fwsR deletion that removed 347 bp of the predicted FadA domain (Figure 1B and 1C). This deletion results in a fusion of nucleotide 340 of fadA to nucleotide 83 of fwsR, while preserving the reading frame of fwsR. This ‘L2-M5’ mutation resulted in a characteristic WS phenotype upon substitution for fadA and fwsR in the ancestral SM genotype (Figure 3). That a fadA-fwsR fusion causes FWS, despite missing most of the enzymatic domain shows that fatty acid desaturase activity is not required for the FWS phenotype. It also suggests that structural componentry (such as non-enzymatic domain folds) of FadA is not required to cause FWS, unless that structure is proximal to the N- terminus.
Altered transcription of fwsR is insufficient to cause FWS
One hypothesis to account for the effects of the fadA-fwsR fusion predicts effects mediated via transcription. Promoter mutations at several GGEEF domain-encoding genes have been associated with transcriptional increases and the evolution of WS types (Lind et al. 2015). In the case of the fadA-fwsR fusion, this possibility is given credence by the presence of a rhoindependent transcriptional terminator downstream of the stop-codon of fadA (predicted by the WebGeSTer transcription terminator database (Mitra et al. 2011)). Removal of this terminator by the spontaneous deletion mutation could lead to increased transcription of fwsR, which may in turn elevate intracellular levels of the DGC, resulting in increased production of c-di-GMP and ultimately the FWS phenotype.
To test this hypothesis the promoter of fadA was fused to fwsR (resulting in PfadA-fwsR, Figure 4A) and integrated into the genome such that it replaced the fadA fwsR locus in the ancestral SM genotype. No effect of the fadA promoter fusion was observed: the genotype retained its smooth colony morphology, did not colonize the air-liquid interface, or produce ACP (Figure 4B). This demonstrates that changes in fwsR transcription caused by the fadA-fwsR fusion are not sufficient to cause WS.
Translational fusion of FadA to FwsR is necessary for FWS2
The alternate hypothesis, namely, that the translation fusion is itself necessary to generate FWS2 was next tested. This was achieved by construction of a fadA-fwsR fusion in which translational coupling was eliminated: 182 bp between the stop codon of fadA and the ATG start codon of fwsR were deleted and replaced by stop codons (one in each reading frame). The in situ ribosome-binding site (RBS) of fwsR was left intact to allow translation initiation of an independent protein product for fwsR. The allele was introduced into the FWS2 background (where it replaced the fadA-fwsR fusion of FWS2) generating fadA-3X-fwsR (Figure 5A). The phenotype of this mutant resembled the ancestral SM type in all respects (Figure 5B). This indicates that translational fusion between fadA and fwsR is necessary to generate FWS2.
The FadA-FwsR fusion relocates the DGC domain
To test the hypothesis that fadA-fwsR results in localisation of FwsR to the membrane, the cellular locations of proteins encoded by fadA, fwsR and the fadA-fwsR fusion were visualised by creating a green fluorescent protein (GFP) translational fusion to the C-terminal region encoded by each gene. All fusions were cloned into the multiple cloning site (MCS) of mini-Tn7-lac and modified to contain a Pseudomonas-specific RBS (attempts to visualise foci using native gene promoters proved unsuccessful). Two controls were used during microscopy: a negative control of the ancestral SM genotype without a gfp insert in the mini-Tn7-lac cassette; and a positive control of the ancestral SM type containing gfp expressed in the integrated mini-Tn7-lac cassette. A cassette expressing fadA-3X-fwsR-gfp was also made to confirm that the fadA-3X-fwsR construct allowed independent translation of FwsR.
The fluorescent signal from each genotype was quantified to show that non-specific fluorescent signals from FwsR-GFP and FadA-3X-FwsR-GFP are significantly greater than the negative control (Welch t-test, p-value = 4.8 x 10−10 and 5.0 x 10−8), indicating that the diffuse expression of fluorescence results from the induced expression of GFP fusions (Figure 6).
Fluorescence microscopy demonstrates that the fadA-fwsR fusion alters the location of the FwsR DGC from the cytosol to the membrane (Figure 7). The location of the fluorescent signal of cells expressing fwsR-gfp is visually diffuse and is located within the cytoplasm (Figure 7C). Similarly, the signal from fadA-3X-fwsR-gfp is dispersed throughout the cytoplasm (Figure 7D). In contrast, cells expressing fadA-gfp have clear foci localised to the edge of the cell and, by inference, near the membrane (Figure 7E). The foci are distributed predominately at the lateral edge of cells in visual reference to the phase contrast images, similar to observations of the localisation of WspA in P. aeruginosa (O'Connor et al. 2012). The protein expressed by fadA-fwsR-gfp (Figure 7F) is localized in a manner identical to that observed in cells expressing fadA-gfp. The visual co-localisation of foci with the edge of the cells in genotypes expressing fadA-gfp and fadA-fwsR-gfp was confirmed by co-localisation analysis by using Van Steensel’s approach (van Steensel et al. 1996) (Figure 7 - figure supplement 1). Together, the location of fluorescent foci seen in cells with induced fadA-gfp and fadA-fwsR-gfp demonstrates the fadA-fwsR fusion event relocated FwsR from the native cytosol to the membrane location of FadA.
Localisation of FwsR to the membrane is sufficient to generate FWS
If membrane localisation of FwsR is all that is required to activate synthesis of c-di-GMP, then fusion of FwsR to any membrane spanning domain-containing protein should suffice to generate the FWS phenotype. To test this hypothesis, the membrane-spanning domain of mwsR (PFLU5329) was fused to fwsR. Details of the construct are shown in Figure 8A. Replacement of the native fadA fwsR locus with the mwsR-fwsR fusion in the ancestral genotype lacking mwsR (to avoid unwanted allelic exchange at this gene) resulted in the FWS phenotype (Figure 8B).
Discussion
The origin of new genes is a subject of fundamental importance and longstanding debate (Sturtevant 1925; Haldane 1932; Bridges 1936). Gene duplication and divergence, once seen as the primary source (Ohno 1970; Lynch and Conery 2000; Nasvall et al. 2012), is now recognized as just one of a number of routes by which new genes are born. Studies in comparative genomics indicate that new genes have arisen from retroduplication (in which mRNA is reverse transcribed into a complementary DNA copy and inserted into the chromosome (Brosius 1991; Long et al. 2003) from retrotransposon-mediated transduction (Moran et al. 1999; Cordaux and Batzer 2009), from deletion and recombination events that generate chimeric gene fusions (Marsh and Teichmann 2010; Ranz and Parsch 2012) from genomic parasites (Volff 2006; Feschotte and Pritham 2007), and even from previously non-coding DNA (Tautz & Domazet-Loso 2011, Neme & Tautz 2013).
Despite the apparent range of opportunities for the birth of new genes, there are few examples in which the evolution of a new gene has been captured in real time, the selective events leading to its formation understood, and mechanistic details underpinning function of the new gene revealed. Those thus far reported have involved recombination or deletion events leading to promoter capture. For example, in the Lenski long-term evolution experiment, Blount et al (2012) described a rare promoter capture event that underpinned evolution of citrate utilization in E. coli. Similarly, Lind et al (2015), using experimental Pseudomonas populations reported promoter capture events caused by deletions that increased transcription of genes encoding active DGCs necessary for over production of c-di-GMP and the evolution of WS types.
Here, provided with opportunities afforded by experimental evolution, we have observed, in real time, multiple independent deletion events, each of which caused a translational fusion between two genes: the membrane-spanning domain of fadA and the DGC domain of fwsR. Each fusion resulted in the birth of a new gene with the resulting fusion altering the cellular location of the DGC domain. This alteration conferred new biological properties: activation of the DGC, the synthesis of c-di-GMP, over-production of cellulose and generation of the adaptive wrinkly spreader phenotype.
In many respects, these events mirror those inferred from comparative studies responsible for new gene function, for example, the human gene Kua-UEV is thought to have originated from fusion of Kua and UEV resulting in localization of UEV to endomembranes by virtue of the N-terminal localization domain of Kua (Thomson et al. 2000). That similar, albeit simpler, events were detected in experimental bacterial populations is remarkable given that prokaryotic cell structure presents few opportunities for protein re-localization to occur; it thus adds weight to the suggestion that the birth of new genes via fusions that re-localize proteins is likely to be more common than recognized (Buljan et al. 2010). Indeed, inferences from comparative genomics support this prediction (Byun-McKay and Geeta 2007): approximately - 37% of duplicated gene pairs in Saccharomyces cerevisiae encode proteins that locate to separate cellular compartments (Marques et al. 2008). Of human multi-gene families known to encode proteins predicted to locate to the mitochondria, approximately 64% contain a gene predicted to relocate to an alternative subcellular location (Wang et al. 2009). Additionally, studies on the fate of duplicated genes in C. elegans suggest that approximately a third of new genes caused by duplication mutations are chimeric (Katju and Lynch 2003; Katju and Lynch 2006). Such chimeric mutations can introduce spatial encoding motifs to the duplicated protein, providing ample opportunity for the relocalisation of protein domains.
While the de novo fusion events reported here occurred between two adjacent loci, the fact that the DGC domain of fwsR could be fused (in vitro) to the membrane spanning domain of a gene more than one million nucleotides away (mwsR) and with similar effect reveals the considerable potential to generate new gene functions via fusion of distal genetic elements. This realization warrants consideration in context of standard models for the origin of new genes via duplication (amplification) and divergence (Ohno 1970, Bergthorsson et al 2007, Lynch and Conery 2000, Nävsall et al 2012). Typically divergence is considered a gradual process occurring via point mutations and aided by selection for original gene function plus selection for some promiscuous capacity (Bergthorsson et al 2007, Nävsall et al 2012). However, formation of chimeric proteins by gene fusion (following duplication) provides opportunity for divergence to occur rapidly and in a single - or small number - of steps. Such distal-fusion events have been reported from genomic studies (Rogers et al. 2009, Rogers and Hartl 2012) and shows how the modular nature of spatial localising domains can facilitate rapid generation of new functions (Pawson & Nash 2003).
With few exceptions (see for example studies on WspR (Bantinaki et al. 2007; Malone et al. 2007; De et al. 2008; O'Connor et al. 2012; Huangyutitham et al. 2013) and PleD (Aldridge et al. 2003; Paul et al. 2007) biochemical details underpinning the mechanisms of DGC activation and factors determining the specificity of c-di-GMP signaling are yet to be fully understood (Schirmer and Jenal 2009; Massie et al. 2012; Chou and Galperin 2016). The mechanism by which the fadA-fwsR fusion leads to DGC activation is unknown. The requirement for an active FwsR DGC domain and for that active domain to localize to the membrane makes clear that membrane localization is a necessary condition. One possibility is that membrane localization is alone sufficient to activate DGC activity (a suggestion bolstered by the fact that the DGC domain of fwsR could be fused to the membrane-spanning domain of mwsR with the same effect). For example, membrane localization may promote homodimerization, which is necessary for DGC activity (Wassmann et al. 2007). Activation may also be affected by fatty acid composition of the cell membrane as recently shown in P. aeruginosa (Blanka et al. 2015). Alternatively, localization of FwsR to the membrane may serve to facilitate spatial sequestration of c-di-GMP and thus molecular interactions between the DGC and the membrane localized cellulose biosynthetic machinery (Hengge 2009). Biochemical studies are required to explore these possibilities, but discovery that the DGC domain of FwsR becomes active upon localization opens new opportunities for understanding mechanisms of DGC activation and signaling, and points to the importance of connections between DGCs and the cell membrane. In this regard it is significant that DGCs are invariably membrane associated, or components of regulatory pathways that are - like Wsp (Bantinaki et al. 2006)) - membrane localized.
One notable feature of genes that encode DGC domains is the diverse array of domains with which they associate. Prior bioinformatic analysis of DGCs shows connections to numerous domains, ranging from CheY-like domains to kinases and phosphodiesterases (Galperin 2004; Jenal and Malone 2006; Romling et al. 2013). Gene fusion events such as we have described here are likely to contribute to the origin of these diverse spectra suggesting capacity of DGC domains to be readily accommodated within proteins of diverse function. If, as the evidence suggests, this is true, then questions arise as to why the underpinning evolutionary events have not previously been observed in experimental studies of evolution using bacteria. Our recent work provides a clue: of all single-step mutational routes to the wrinkly spreader phenotype, fusions generating chimeric proteins constitute almost 10 % of such events - but only when the genome is devoid of readily achievable loss-of-function routes to the adaptive phenotype (if loss-of-function routes are intact then fusions described here constitute ~0.1% of the total) (Lind et al. 2015). While such loss-of-function routes are readily realised in laboratory populations where functional redundancy among DGC domains is observed, the very real possibility is that in populations in natural environments such redundancy does not exist, making loss-of-function routes less achievable and mutations of the kind described here more likely. Taking experimental evolution into the wild - or at least scenarios more reflective of the complex challenges faced by natural populations (Hammerschmidt et al. 2014; Bailey and Bataillon 2016; Lind et al. 2016) is an important future goal likely to shed new light on the origins of genes.
Materials and Methods
Media and strains
All P. fluorescens strains are derivatives of P. fluorescens SBW25 (Silby et al. 2009) and were cultured at 28°C in King’s medium B (KB) (King et al. 1954). Escherichia coli DH5-a âpir (Hanahan 1983) was used to replicate all plasmids used to construct mutations. E. coli was cultured at 37°C in Lysogeny Broth (LB) (Bertani 1951). Solid media were prepared by the addition of 1.5% bacteriological agar. Strains intended for fluorescence microscopy were cultured in M9 minimal media (Sambrook and Russell 2001) containing 0.4% w/v glycerol and omitting thiamine. All bacterial overnight cultures were grown shaking at 160 rpm for approximately 16 hours in 30 mL glass microcosms containing 6 mL of medium. Antibiotics for the maintenance of plasmids were used at the following concentrations: gentamycin (Gm) mg L−1, kanamycin (Km) 100 mg L−1, tetracycline (Tc) 15 mg L−1, nitrofurantoin (NF) 100 mg L−1, and cycloserine 800 mg L−1. X-gal (5-bromo-4-chloro-3-indolyl-p-D- galactopyranoside) was used at a concentration of 60 mg L−1 in agar plates, IPTG at 1 mM in liquid culture and Calcofluor (fluorescent brightener 28) at a concentration of 35 mg L−1 in agar plates.
Mutation reconstruction
Strand overlap extension (SOE-PCR) was employed to construct all site-directed genomic mutations in P. fluorescens as previously described (Ho et al. 1989; Rainey 1999; Bantinaki et al. 2007). Briefly, Phusion High-Fidelity DNA polymerase (New England Biolabs) was used to generate an amplicon product approximately 1000 bp either side of the mutation, using overlapping primer sets with the required mutation within the 5’ region of the internal primers. A second round of PCR using only the external primers was performed to make a single amplicon with the required mutation. The amplicon was then cloned into pCR8/GW/TOPO (Invitrogen) using TA cloning, transformed into an E. coli host and the plasmid was sequenced to confirm the presence of the mutation and no additional mutations within the amplified region. The DNA regions were then excised from pCR8/GW/TOPO by restriction digest and ligated into the pUIC3 suicide vector (Rainey 1999), and was then introduced into the P. fluorescens host by two-step allelic exchange (Kitten et al. 1998). This involved transconjugation of pUIC3 into the P. fluorescens host by tri-parental mating, and selection for transconjugants with the pUIC3 vector homologously recombined at the target sequence. Overnight non-selective cultures of the transconjugates were then treated with tetracycline in order to inhibit growth of cells that had lost the pUIC3 vector. The culture was then treated with cycloserine to kill growing cells featuring the pUIC3 vector and thus enrich the fraction of cells that were either isogenic with the host, or which now featured the mutation following a second recombination event. The culture was serially diluted and plated on KB plates containing X-gal to enable visual screening for genotypes that had lost the pUIC3 vector. Several resulting white colonies were isolated and Sanger sequencing confirmed the presence of the mutant allele and the absence of mutations that may have been introduced during the cloning process. The sequence of each genomic mutation is accessible through the following accession numbers: fadA-fwsR GGAAF (KU248756), fadA-fwsR L2M5 (KU248757), PfadA-fwsR (KU248758), fadA-3X-fwsR (KU248759) and PfadA-mwsR1218-fwsR (KU248760).
The visualization of proteins required translational fusions of fadA, fwsR and fusions of these genes with gfp variant gfpmut3* (Andersen et al. 1998). The open reading frame of genes to be tagged were cloned into pCR8 and then moved (by digestion with restriction endonucleases and ligation) into the multiple cloning site (MCS) of mini-Tn7-LAC (Choi and Schweizer 2006). This plasmid was modified to contain both a Pseudomonas-specific ribosome-binding site (RBS) as well as a gfp encoding region 3’ of the MCS. This resulted in the ability to induce expression of our protein of interest, tagged at the C-terminus with GFP. These plasmids were sequenced to identify possible mutations introduced during cloning. Plasmids were then taken up by electrocompetent P. fluorescens SWB25 cells and the integration of the construct at the attB site was confirmed by PCR and electrophoresis. The sequence of each gfp-tagged construct cloned in the mini-Tn7-LAC MCS is accessible through the following accession numbers: gfp positive control (KU248761), fwsR-gfp (KU248762), fadA-gfp (KU248763), fadA-fwsR-gfp (KU248764) and fadA-3X-fwsR-gfp (KU248765).
Microscopy techniques
Colony-level photography was performed using a Canon Powershot A640 camera in conjunction with a Zeiss Axiostar Plus light microscope using a 10x objective. Microscopy of cells was visualised using an Olympus BX61 upright microscope and images were recorded using a F-view II monochrome camera. The production of acetylated cellulosic polymer (ACP) was detected by the in vivo staining of colonies with Calcofluor (fluorescent brightener 28, Sigma) and this stain was visualised with fluorescence microscopy. Colonies of P. fluorescens were grown for 48 hours on KB plates containing Calcofluor, and several colonies were resuspended in 100 μL of distilled water. 10 μL of this solution was pipetted onto a microscope slide and used for microscopy. Stained cells were visualised using a 60x objective following DAPI excitation.
The localisation of fluorescently tagged proteins was visualised in vivo by fluorescent microscopy. Strains were then prepared for microscopy by inoculation of single colonies in 6 mL of sterile M9 media supplemented with 0.4% glycerol (with appropriate antibiotics to prevent loss of the mini-Tn7 vector) and grown overnight (160 rpm, 16 h). The culture was then subcultured in fresh M9-glycerol media (without antibiotics) resulting in an OD600 ~0.05 culture. This was then incubated with shaking for 60 minutes, after which 1 mM Isopropyl p-D-1-thiogalactopyranoside (IPTG) was added to induce expression of the tagged genes. The culture was then returned to the incubator for 2 hours until an OD600 of approximately 0.3 was reached. In order to increase the density of cells for microscopy, a 1 mL aliquot of culture was centrifuged and resuspended in 100 μL of M9-glycerol media. Agarose pads were prepared consisting of M9-glycerol media with 1% w/v agarose. A 3 μL aliquot of resuspended culture of induced cells was pipetted onto the pads, which were left to dry and then covered with a coverslip. Cells were visualised using an Olympus BX61 upright microscope and both phase contrast and fluorescence images were recorded using a F-view II monochrome camera. Fluorescence images were observed using a constant 3500 ms exposure time across all strains. All images were processed using FIJI (Schindelin et al. 2012) to obtain measures of fluorescence and measures of co-localisation. Measures of fluorescence levels were obtained by calculating the mean gray values of cell transects observed under fluorescence microscopy. Four randomly picked cells were analyzed across two independent images for each biological replicate, with three biological replicates analyzed per genotype. Co-localisation analyses were prepared by generating the mean van Steensel’s distribution of two fields of view for each biological replicate. At least three biological replicates were used over two separate experiments in co-localisation analyses.
Competing interests
The authors declare no competing interests
Acknowledgements
This work was supported in part by Marsden Fund Council from New Zealand Government funding (administered by the Royal Society of New Zealand) and the New Zealand Institute for Advanced Study. We thank Jenna Gallie and Peter Lind for discussion and comments on the manuscript and Heather Hendrickson for assistance with microscopy.