Presence of the Hmq System and Production of 4-Hydroxy-3-Methyl-2-Alkylquinolines Are Heterogeneously Distributed between Burkholderia cepacia Complex Species and More Prevalent among Environmental than Clinical Isolates

ABSTRACT The Burkholderia cepacia complex (Bcc) comprises several species of closely related, versatile bacteria. Some Bcc strains produce 4-hydroxy-3-methyl-2-alkylquinolines (HMAQs), analogous to the 4-hydroxy-2-alkylquinolines of Pseudomonas aeruginosa. Using in silico analyses, we previously estimated that the hmqABCDEFG operon, which encodes enzymes involved in the biosynthesis of HMAQs, is carried by about one-third of Bcc strains, with considerable inter- and intraspecies variability. In the present study, we investigated by PCR, using consensus primers, the distribution of hmqABCDEFG in a collection of 312 Bcc strains (222 of clinical and 90 of environmental origins) belonging to 18 Bcc species. We confirmed that this operon is not distributed evenly among Bcc species. Among the 30% of strains bearing the hmqABCDEFG operon, we found that 92% of environmental isolates and 82% of clinically isolated Bcc strains produce levels of HMAQs detectable by liquid chromatography-mass spectrometry in at least one of the tested culture conditions. Among the hmqABCDEFG-positive but HMAQ-negative strains, none expressed the hmqA gene under the specified culture conditions. Interestingly, the hmqABCDEFG operon is more prevalent among plant root environment species (e.g., Burkholderia ambifaria and Burkholderia cepacia) and absent in species commonly found in chronically colonized individuals with cystic fibrosis (e.g., Burkholderia cenocepacia and Burkholderia multivorans), suggesting a role for the Hmq system in niche adaptation. We investigated the impact of the Hmq system on plant growth promotion and found that Pisum sativum root development by B. ambifaria required a functional HMAQ system. IMPORTANCE Environmental bacteria belonging to the various closely related species forming the Burkholderia cepacia complex (Bcc) can infect plants and animals, including humans. Their pathogenicity is regulated by intercellular communication, or quorum sensing, allowing them to collaborate instead of acting individually. Bcc organisms generally exploit interacting quorum sensing systems based on N-acyl-homoserine lactones as signaling molecules. Several Bcc strains also carry an hmqABCDEFG operon responsible for the biosynthesis of 4-hydroxy-3-methyl-2-alkylquinolines (HMAQs), molecules analogous to the Pseudomonas quinolone signal (PQS) system of P. aeruginosa. Our finding that the prevalences of the Hmq system and HMAQ production are very different between various Bcc species suggests a key role in niche adaptation or pathogenicity. This is supported by a significant reduction in plant growth promotion in the absence of HMAQ production for a beneficial Bcc strain.

similarly to a set of environmental B. ambifaria isolates (36,43). Therefore, to better understand HMAQs, we postulated that (i) the hmqABCDEFG operon is more frequently found among clinical Bcc strains and (ii) among HMAQ-producing Bcc isolates, clinical strains produce higher concentrations than environmental ones. Our previous bioinformatic study on the distribution of the hmqABCDEFG operon in the Bcc, based on available whole-genome sequences for 1,257 strains belonging to 21 Bcc species, showed that B. ambifaria, B. cepacia, B. contaminans, B. pyrrocinia, B. stagnalis, B. territorii, and B. ubonensis strains carry the hmqABCDEFG operon, although not all strains within a species do (55). On the other hand, all sequenced genomes of B. anthina, B. arboris, B. cenocepacia, B. diffusa, B. latens, B. metallica, B. multivorans, B. pseudomultivorans, B. seminalis, B. stabilis, and B. vietnamiensis-species mainly isolates from clinical cases-that have been investigated are lacking the hmqABCDEFG operon (55). To experimentally validate our in silico study and verify the ability of Bcc isolates carrying the hmqABCDEFG operon to actually produce HMAQs, a collection of 312 Bcc strains, comprising 222 clinical and 90 environmental isolates belonging to 18 different Bcc species, was analyzed. We first assessed the presence of the hmqABCDEFG operon in their genomes and then directly determined, by liquid chromatography coupled to mass spectrometry (LC-MS), the ability of all the strains bearing the hmqABCDEFG operon to produce HMAQs. Our data show that the Hmq system is heterogeneously distributed between Bcc species, with high prevalence in some species (e.g., B. cepacia) and near absence in others (e.g., B. cenocepacia and B. multivorans). Finally, we investigated the impact of the Hmq system on plant growth using hmqA and hmqG mutants and found that Pisum sativum root growth promotion by a B. ambifaria strain was lost in the absence of HMAQ production.

RESULTS
The hmqABCDEFG operon is heterogeneously distributed across and within Bcc species. We previously examined 1,257 whole-genome sequences belonging to 21 different Bcc species to assess the distribution of the hmqABCDEFG operon by orthology and homology analyses (55). We found that at least one sequenced strain belonging to 7 of the 21 Bcc species carries the hmqABCDEFG operon (B. ambifaria, B. cepacia, B. contaminans B. pyrrocinia, B. stagnalis. B. territorii, and B. ubonensis); one striking initial finding was that prevalence of the hmqABCDEFG operon within a species appeared highly variable (55). Here, to validate our in silico analyses (55) and to globally determine the ability of Bcc to actually produce HMAQs, we screened a collection of 312 Bcc strains (222 of clinical and 90 of environmental origins; listed in Table S1 Fig. S1]). We had previously shown that presence of a hmqG orthologue correlates with the presence of a complete hmqABCDEFG operon (55). Here, we found that 30% of the 312 tested strains possess an hmqABCDEFG operon, including 53% of the environmental strains but only 21% of clinical strains (Fig. 1A). A Fisher test with a P value of 2.13 Â 10 210 indicates that the prevalence of hmqABCDEFG is significantly different depending on the strains' sampling origin (environmental versus clinical). Among the 18 different Bcc species investigated, 14 species comprise at least one strain carrying the operon ( Fig. 1B and Fig. S2). However, B. cenocepacia, B. multivorans, and B. vietnamiensis are overrepresented among clinical strains compared to the other species. We thus decided to subsample these three species with only 20 clinical strains, increasing the prevalence of the hmqABCDEFG operon in clinical Bcc strains to 30% (46 strains carrying the operon among the 151 subsampled clinical strains), which is still statistically significantly lower than the prevalence of the hmqABCDEFG operon within environmental strains (Fisher test with a P value of 1.324 Â 10 26 ).
We found isolates of B. anthina, B. dolosa, and B. vietnamiensis carrying the hmqABCDEFG operon, which was not predicted in our previous analysis of available genomic data (55); we confirmed these PCR results by sequencing the amplicons using primers listed in Table S2.
Considering that horizontal gene transfer and pc3 chromosomal rearrangement could explain the heterogeneous distribution of the hmqABCDEFG operon in the Bcc. Since not all Bcc species carry hmqABCDEFG, we asked whether the distribution of the operon could be related to a loss of the third chromosome in Bcc or to horizontal gene transfer. Bcc bacteria can lose their third chromosome, also described as a virulence megaplasmid (pc3) containing a few core genes (56,57). The hmqABCDEFG operon being generally located on the pc3 replicon (except in B. ubonensis, which carries this operon on its second chromosome like B. pseudomallei and B. thailandensis), and because the synteny of the hmqABCDEFG operon is conserved within a species (55), we determined whether the strains missing the operon were also missing their pc3 replicon. Our analysis showed that the absence of the hmqABCDEFG operon does not correlate with the loss of pc3 (Table S3).
By comparing the phylogenetic tree based on multilocus sequence typing (MLST) sequences and the distribution of strains possessing the hmqABCDEFG operon, we found that the resulting MLST and hmqABCDEFG phylogenies mostly match, except for two groups, which are inverted (Fig. 2). Hence, the Hmq system originates from a common ancestor, with possible chromosomic rearrangement for some strains.
Most Bcc strains carrying the hmqABCDEFG operon produce HMAQs. We then verified whether the presence of the biosynthetic genes is indicative of known HMAQ production. We cultured the 94 strains that were PCR-positive for hmqA and hmqG in tryptic soy broth (TSB) medium at 30°C and 60 rpm overnight and detected the production of HMAQ in 72% of the PCR-positive strains-65% clinical and 79% environmental-carrying the hmqABCDEFG operon (Table S4). None of the PCR-positive B. anthina, B. diffusa, B. dolosa, B. metallica, B. seminalis, and B. ubonensis strains produced detectable HMAQ under these culture conditions.
Assuming that the absence of HMAQ production in 26 of the 94 PCR-positive strains could simply be the result of unfavorable culture conditions, we assayed the production of these metabolites in TSB but at 37°C, in artificial sputum medium (ASM; at 30°C at 60 rpm, overnight), and on tryptic soy agar plates (TSA; incubated at 30°C for 4 days). As shown in Table 1, 4 of 26 previously negative strains produced HMAQs when grown in TSB at 37°C, while ASM growth induced production levels allowing detection of HMAQs in 10 of the 26 Bcc strains-seven environmental and three clinical strains. Surface growth on TSA plates induced the detectable production of HMAQs for five environmental and four clinical strains. Among the strains already producing HMAQs in TSB at 30°C, most of them also produced HMAQs in ASM and TSA (Table S4). Interestingly, two B. contaminans strains that were negative in both TSB and ASM produced detectable HMAQ when cultured on TSA plates. These additional culture conditions reduced the number of HMAQ-negative strains from 26 to 15, increasing the number of Bcc isolates able to produce HMAQs to a total of 82, that is, 87% of strains carrying the hmqABCDEFG operon, including 92% environmental and 82% clinical isolates ( Fig. 3 and Fig. S3).
We also found that the concentration of HMAQs produced was variable among the various species when grown in TSB at 30°C. A Kruskal-Wallis test confirmed the absence of statistical difference in the concentrations produced between the clinical and environmental strains (P value of 0.19 for HMAQ-C 7 :29 and 0.22 for HMAQ-C 9 :29). However, clinical strains of B. ambifaria produced more HMAQs than environmental ones. The opposite was observed for B. cepacia and B. contaminans strains (Fig. 4).
Presence of the hmqABCDEFG operon and ability to produce HMAQs do not correlate with P. aeruginosa coisolation or the origin of the strain. Since the hmqABCDEFG operon is homologous to the pqsABCDE operon in P. aeruginosa, and  HMAQs might interact with the PQS system, we asked if HMAQ production of clinical Bcc is correlated with a coisolation or a colocalization with P. aeruginosa from the patient at some point as well as with the origin of the sample (sputum, throat, sinus, etc.). Information was available for 53 of 222 clinical strains (Table S7). Using Fisher's exact test for count data, we did not find a correlation between the presence of the hmqABCDEFG operon and the presence of P. aeruginosa at the sampling time or within the previous year or with the origin of the sample ( Table 2). Lack of expression of the hmqABCDEFG operon explains the absence of HMAQs detection in some Bcc cultures. Our screening revealed that 26 strains carrying the hmqABCDEFG operon do not produce HMAQs in TSB at 30°C. To investigate the possibility that too low transcription of the biosynthetic genes would explain this absence of detectable production, which is compatible with the induction seen when the culture conditions were changed, we measured the transcription of the hmqABCDEFG operon for one HMAQ-negative and one HMAQ-positive strain per species of B. ambifaria, B. cepacia, B. contaminans, and B. vietnamiensis by targeting the hmqA gene by reverse transcription-PCR (RT-PCR).
Results show that the hmqABCDEFG-positive but HMAQ-negative strains B. ambifaria AMMD, B. cepacia ATCC 25416, B. contaminans VC15406, and B. vietnamiensis VC9237 do not express the hmqA gene when grown in TSB at 30°C, while B. ambifaria HSJ1, B. cepacia VC13394, B. contaminans FFH2055, and B. vietnamiensis VC8245, which produce HMAQs under these conditions, produce a clear hmqA transcript (Fig. S5). Thus, we decided to investigate if a mutation in the promoter region could explain the lack of expression of the hmqABCDEFG operon, but there was no difference between B. ambifaria HSJ1 and AMMD promoter region sequences. The lack of expression of hmqABCDEFG is a matter of regulation, especially since strain AMMD produced HMAQs in ASM and TSA. Extending these results, we hypothesize that the other strains which carry the hmqABCDEFG operon and do not produce HMAQs do not express the hmqA gene under the specified culture conditions. This is supported by the finding that several of these strains could eventually produce measurable levels of HMAQs when culture conditions were changed (as described above) ( Table 1).
The capacity to produce HMAQs is required for root growth promotion. Since B. ambifaria is a good plant growth promoter (15, 16; reviewed in reference 19) and the Hmq system is more prevalent among environmental strains, we investigated the impact of the Hmq system on growth of the common pea (P. sativum). While B.

FIG 4
Concentration of the main produced HMAQs among Bcc species when cultured in tryptic soy broth. HMAQ-C 7 :29 and HMAQ-C 9 :29 were quantified by LC-MS with a lower limit of detection of 50 mg/liter for each molecule in the total culture. Tested strains are grouped by species. The production of HMAQs was quantified in three biological replicate cultures for each strain, and each dot represents the average HMAQ production for each strain, in milligrams per liter. Lines represent the production average for the clinical and environmental strains.

DISCUSSION
This study aimed at understanding the prevalence of the Hmq system and corresponding HMAQ production among the various species belonging to the Bcc to experimentally validate and extend our previous in silico analyses (55). Besides the non-Bcc species B. pseudomallei and B. thailandensis, only a few strains of B. ambifaria and B. cepacia were already known to carry the hmqABCDEFG operon and to produce HMAQs (16,36,(38)(39)(40)(41)(42)(43)(44). Indeed, the presence of the hmqABCDEFG operon is well conserved in B. pseudomallei and B. thailandensis, but its prevalence within the Bcc remains unclear (55). To understand the ecological role of the Hmq system, we first needed to evaluate its distribution and prevalence. Since available genomes are not equally distributed between Bcc species, we screened a panel of 312 Bcc strains for the presence of the hmqABCDEFG operon by PCR. Since the hmqG gene is detectable only when the complete hmqABCDEFG operon is present and no significant variation of the hmqABCDEFG operon was observed in the 447 positive genomes analyzed by bioinformatics (55), we  considered the operon to be present when both hmqA and hmqG genes were amplified. While the possibility that that some strains carry variations in the hmqABCDEFG operon cannot be excluded, we did not detect any in our previous bioinformatic analysis comparing whole genomes (55). The prevalence of the hmqABCDEFG operon found here for the Bcc followed the same distribution as that found by in silico analysis. Our PCR-based screening could miss some strains carrying the operon due to the limited availability of whole-genome sequences for some Bcc species (e.g., B. arboris, B. metallica, and B. stabilis). Nevertheless, our primers could amplify hmqA and hmqG targets in species not previously known to carry a hmqABCDEFG operon (e.g., B. vietnamiensis), which was expected due to the high level of identity (.88%) of hmqA and hmqG among identified and sequenced Bcc species (55). Furthermore, testing the production of HMAQs in a group of 31 PCR-negative strains confirmed the absence of production of these metabolites. Taking all the results together, we are confident that we identified the strains carrying the genetic ability to synthesize HMAQs. Based on our previous results obtained with a few B. ambifaria isolates, we had hypothesized that the Hmq system would be more prevalent among clinical isolates and produce more HMAQs than environmental ones (36,43,46). Unexpectedly, we discovered that B. cenocepacia, B. multivorans, and B. vietnamiensis (the prominent Bcc species colonizing immunosuppressed individuals and those with CF) transmitted between patients (22, 23, 58) never or rarely (0 of 72 B. cenocepacia, 0 of 35 B. multivorans, and 5 of 35 B. vietnamiensis strains) carry the hmqABCDEFG operon. As this operon in Bcc results from a common ancestor, its absence could result from chromosomic rearrangement (by deletion or replacement) (55). In contrast, our data suggest that the Hmq system could play a beneficial role in niche adaptation to the rhizosphere microbial community due to the large prevalence of the hmqABCDEFG operon among environmental strains of B. ambifaria, B. cepacia, B. contaminans, B. pyrrocinia, and B. ubonensis, species known for their preference for the plant root environment (19,(59)(60)(61)(62). The presence of the Hmq system in clinical strains of these five common environmental species is compatible with a recent or direct environmental acquisition in CF patients (60,63,64).
Among the strains carrying the hmqABCDEFG operon, not all produced HMAQs under the tested conditions. Since the limit of detection for these molecules was 50 mg/liter, we cannot exclude the possibility that some of these strains actually produced lower levels. For those strains producing HMAQs in TSB, growth in ASM did not inhibit HMAQ production and stimulated a detectable production in 10 additional strains, which could be explained by differences in regulation of the biosynthetic operon. Further, an increase in the number of HMAQ-positive strains when a third culture condition (4 days on TSA plates) was tested highlights the need to use nutritionally appropriate media for investigating the production of secondary metabolites (53). For this purpose, we tried to optimize a medium for HMAQ production. Nevertheless, our results show that the production of HMAQs in the Bcc is dependent on the conditions of growth and that hmqABCDEFG-positive strains not producing detectable HMAQs under the conditions tested here could produce enough HMAQs under different ones.
Lack of transcription of the hmqABCDEFG operon seemed to explain the observation that 28% of Bcc strains carrying the genes did not produce HMAQs when grown in TSB at 30°C. This result may be observed due to nonpromoter effects (e.g., posttranscriptional or posttranslational control). Indeed, production could be detected simply by changing the culture conditions, suggesting that the promoter of the hmqABCDEFG was not expressed, although we cannot exclude the possibility that it could be nonfunctional in some strains. We thus believe the same explanation (poor hmqABCDEFG transcription) could apply to the remaining 13% strains for which we could not yet identify appropriate culture conditions. Future work will involve developing a better understanding of the nutritional and regulatory elements (e.g., posttranscriptional or posttranslational regulation) controlling the expression of the hmqABCDEFG operon and production of HMAQs.
Because the hmqABCDEFG operon is more prevalent in environmental strains, which are often isolated from the rhizosphere, we investigated the impact of the Hmq system on the growth promotion ability of a B. ambifaria strain of a model plant, P. sativum. The inactivation of hmqA or hmqG prevented root growth promotion, presumably because of the inability to produce HMAQs. As is the case for cepacin and pyrrolnitrin, HMAQs could influence the rhizosphere microbial community, for instance, due to their antimicrobial activity (16, 36, 39-41, 45, 46, 50, 54, 65) or other properties. As we have previously reported that the HMAQs influence the Cep regulatory system, the molecules could indirectly impact the production of other plant-beneficial metabolites and functions regulated by quorum sensing (e.g., enacyloxin, pyrrolnitrin, cepacin, and antifungal cluster (afc) lipopeptide) (16, 30, 36, 39-41, 45, 46, 50, 54, 65-67). Further experiments with other HMAQ producers and pure HMAQs will provide exciting insights.

MATERIALS AND METHODS
Strains and culture conditions. A total of 312 strains isolated from either clinical or environmental settings were used in this study (Table S1). Uncertain identification was confirmed by amplifying and sequencing the recA and gyrB genes at the IRCM Sequencing Platform (Montreal) following the directions on the PubMLST database (https://pubmlst.org/organisms/burkholderia-cepacia-complex) ( Table S8).
Detection of the presence of hmqABCDEFG operon by PCR. Genomic DNA was extracted following a previously described method (68). Briefly, cells were resuspended in lysis buffer (50 mM Tris-HCl [pH 8], 5 mM EDTA-2Na [pH 8], 3% SDS) and transferred to a tube containing glass beads. The cells were lysed in a Fast-Prep-24 instrument (MP Biomedicals). Then the lysates were centrifuged at 8,000 Â g for 5 min, the supernatants were transferred to a new tube, and 2.5 N ammonium acetate was added. The supernatant was transferred to a new tube, and 1 volume of isopropanol was added. The pellets were washed with 75% ethanol and dried before resuspension in 50 ml water.
For the strains acquired from the Burkholderia cepacia Research Laboratory and Repository and from the Canadian Burkholderia cepacia complex Research and Referral Repository, genomic DNA was extracted using a 96-well plate gDNA extraction kit (Favorgen, Canada).
The hmqA and hmqG genes were amplified by PCR using EasyTaq polymerase (Transgen, Canada). Primers were designed based on a consensus sequence of 11 complete Bcc sequences available in 2017 (Table S8). A strain was considered to have a complete hmqABCDEFG operon when amplification of both hmqA and hmqG targets was attained, based on our previous analyses (55).
Detection of the presence of the pc3 chromosome and the hmqABCDEFG operon in Bcc strains. Assembled genomes are available on DB Burkholderia (http://www.Burkholderia.com). We selected only complete assembled genomes and searched for the presence of the hmqABCDEFG operon by orthology and homology directly on DB Burkholderia. Results are listed in Table S3.
Phylogeny of the Bcc based on MLST and hmqABCDEFG sequences. First, atpD, gltB, gyrB, recA, lepA, phaC, and trpB gene sequences were used to generate a Bcc phylogeny based on MLST. Second, the hmqABCDEFG operon sequence was used to determine the coevolution or horizontal transfer of the Hmq system within Bcc species. Sequences were found on DB Burkholderia (http://www.Burkholderia .com) and concatenated in the order atpD, gyrB, recA, gltB, lepA, phaC, and trpB, with hmqABCDEFG in parallel. The resulting concatenated sequences were aligned using Clustal Omega (69), and trees were generated by RAxML using the GTRGAMMA model and 1,000 bootstraps. The branches are labeled where bootstrap values are .50.
Quantification of HMAQ production by LC-MS/MS. Bcc strains were cultured in 5 ml TSB at a starting optical density at 600 nm (OD 600 ) of 0.05 and incubated at 30°C with shaking for an overnight. 5,6,7,8-Tetradeutero-4-hydroxy-2-heptylquinoline (HHQ-d4) was used as an internal standard (70). The total HMAQs were extracted from 4 ml culture with 1 volume of ethyl acetate. After nitrogen evaporation, the residues were dissolved in 400 ml high-performance liquid chromatography (HPLC)-grade acetonitrile. Samples were analyzed by liquid chromatography coupled with mass spectrometry (LC-MS) in positive electrospray ionization using a Kinetex 5-mm EVO C 18 100-Å 100-by 3-mm reverse-phase column as previously described (70). A Quattro Premier XE triple quadrupole was used as a detector (Waters). A multiple reaction monitoring (MRM) program was used to detect HMAQ families (36). This experiment was conducted with three independent biological replicates.
ASM, TSB, and TSA medium assays. The strains were grown in ASM liquid medium (71) and TSA agar plates out from overnight cultures and incubated at 30°C for 24 h and 4 days, respectively, or in grown TSB and incubated at 37°C for 24 h.
One milliliter of ASM culture was extracted as described for the HMAQ production method. For each TSA plate, 5 ml water was added to extract the HMAQs from the agar. For each sample, 1 ml was extracted with 1 volume ethyl acetate containing 4 ppm HHQ-D4, concentrated 10 times, dissolved in HPLC-grade acetonitrile, and analyzed as described above for the HMAQ extraction method.
The experiments were performed in two independent biological replicates.