Promoter Activation in Δhfq Mutants as an Efficient Tool for Specialized Metabolite Production Enabling Direct Bioactivity Testing

Abstract Natural products (NPs) from microorganisms have been important sources for discovering new therapeutic and chemical entities. While their corresponding biosynthetic gene clusters (BGCs) can be easily identified by gene‐sequence‐similarity‐based bioinformatics strategies, the actual access to these NPs for structure elucidation and bioactivity testing remains difficult. Deletion of the gene encoding the RNA chaperone, Hfq, results in strains losing the production of most NPs. By exchanging the native promoter of a desired BGC against an inducible promoter in Δhfq mutants, almost exclusive production of the corresponding NP from the targeted BGC in Photorhabdus, Xenorhabdus and Pseudomonas was observed including the production of several new NPs derived from previously uncharacterized non‐ribosomal peptide synthetases (NRPS). This easyPACId approach (easy Promoter Activated Compound Identification) facilitates NP identification due to low interference from other NPs. Moreover, it allows direct bioactivity testing of supernatants containing secreted NPs, without laborious purification.


Introduction
Natural products (NPs), also known as secondary or specialized metabolites,a re produced by almost all bacteria, archaea and fungi. They fulfill numerous functions as part of their ecology acting for example as antibiotics,s iderophores, toxins or signals mediating all aspects of organismic interaction between the microbes and their environment. [1,2] NPs and chemical derivatives thereof are also central to our health and agriculture,b eing applied as clinically-relevant antibiotics,immunosuppressants,anticancer,antiviral drugs or as pesticides. [3] Their biological properties are ar esult of their chemical structures that have been optimized during evolution towards as pecific target. Hence,t hey represent ar ich source of promising leads for new drugs capable of overcoming microbial resistances and to fight emerging diseases.
Thee ver-increasing number of sequenced microbial genomes has created an umber of resources and repositories for mining the data, with aparticular emphasis on the identification of biosynthetic gene clusters (BGCs) involved in NP production. [4,5] In most cases,the number of these BGCs encoded in the genomes far outnumbers the quantity of NPs produced under laboratory conditions.How to exploit the potential of this hidden chemical diversity and consequently deliver pure NPs in as imple,r apid and cost-efficient method, gaining sufficient amounts of NPs for broad bioactivity testing is am ajor scientific challenge.
Different strategies have been implemented for the activation of these BGCs that sometimes might not be expressed under laboratory conditions and therefore are considered "silent". Methods for BGC activation range from varying cultivation conditions (also called the OSMAC approach) to co-cultivation approaches. [2,6,7] An individual BGC can also be activated using deletion/overexpression of global (or specific) transcription factors, [7,8] application of transcription factor decoys [9] or promoter exchange approaches activating these BGCs using inducible promoters. [10,11] Heterologous expression of acomplete BGC has also been applied successfully for NP production. [12][13][14] However,m any challenges remain with all of these methods and particularly with heterologous hosts that may lack required building blocks (e.g.f atty acids,a mino acids) for proper biosynthesis of the original NP.F urthermore,expression levels may be low due to toxicity against the heterologous producer. [15] Thed rawback of all described approaches is that the NP of interest is generated in addition to undesired NPs that are also produced under any given condition. Ther esulting complex NP mixture might be very difficult to separate.I deally,t he activation of as ingle BGC would result in the production of as ingle corresponding NP and its derivatives.I nt he prolific NP producing bacterial genus Photorhabdus,werecently showed adependence of NP production on the RNAchaperone,Hfq, that modulates BGC expression through sRNA/mRNAi nteractions. [16] In a Dhfq strain, the biosynthesis of NPs is almost completely lost. Here we show that activation of desired BGCs in a Dhfq background led to the nearly exclusive production of the corresponding NPs in several proteobacteria following targeted BGC activation using the inducible promoter P BAD . Compared to BGC activation in wild type strains,orapproach termed easyPACId (easy Promoter Activated Compound Identification) leads to culture supernatants lacking most undesired NPs,t hereby enabling not only simplified identification and purification of the desired NP,b ut also direct bioactivity testing of culture extracts or supernatants against different target organisms,w ithout time-consuming NP purification ( Figure 1).

Results and Discussion
Compared to the wild type Photorhabdus laumondii TTO1 and Xenorhabdus szentirmaii, Dhfq mutants appear colorless ( Figure 2a)d ue to the absence of their main pigments,a nthraquinones (1)a nd phenazines,r espectively   Table 1f or all NP structures). HPLC-MS analysis of culture supernatants confirmed the absence of all NPs in the Dhfq strains compared to WT when extracted ion chromatograms (EICs) of NPs were analyzed (Figure 2b,c). Although in some Proteobacteria like Pseudomonas aeruginosa deletion of hfq results in ag rowth defect compared to the respective wild type, [17] this was hardly observed for Photorhabdus and Xenorhabdus (Supplementary Figure 2).

Targeted NP production in Xenorhabdus and Photorhabdus Dhfq mutants
In P. laumondii TTO1-Dhfq, the known NPs [18,19] Game-XPeptide A( 4a), glidobactin A( 6), and ririwpeptide A( 7) and in X. szentirmaii-Dhfq GameXPeptides (4), rhabdopeptides (8)a nd pyrrolizixenamides (10)w ere individually produced upon promoter activation following genomic integration of the non-replicating pCEP plasmids carrying the first 600 bp of the first gene in the BGC of interest behind the inducible P BAD promoter ( Figure 1). ForN Ps that are also produced in wild type strains (e.g. 4, 8)t heir activation in a Dhfq-mutant often leads to as trong increase in the production titer [10] but with very little undesired NPs from other BGCs being produced. Furthermore,i ta llows the activation of BGCs that seem silent under the cultivation conditions used (6, 7, 10, 14, 15 a-c, 27). Theindependence of the induced promoter from (often unknown) intracellular regulation mechanisms may be areason for this overproduction as well as the increased availability of building blocks due to all other NP pathways being inactive.
Promoter exchange of Xsze_03460 and Xsze_03680 in X. szentirmaii-Dhfq led to the activation of the two BGC for the known xenobactin (11) [20] and szentiamide (13) [21] for which the BGCs had not been identified yet (Supplementary Figure 3). Promoter exchange of Xsze_03663 and Xsze_0377 in the Dhfq mutant resulted in the production of an oxidized diketopiperazine named szentirazine (14)and three lipopeptides (15 a-c)t hat represent shortened PA X-peptides (Figure 2d), [22] none of which are detected in the wild type strain. Thes tructures of 15 a-c were solved by detailed MS-MS analysis (Supplementary Figure 4). Szentirazine (14)w as isolated from al arge-scale culture and its structure was solved by NMR spectroscopy (Supplementary Figures 5-9, Supplementary Table 2). Compared to standard non-ribosomal peptide synthetases (NRPS), the bimodular NRPS involved in the production of 14 encodes an additional Nterminal acyl-CoA dehydrogenase (ACAD) domain [23] that might introduce the double bond (Supplementary Figure 3).

Bioactivity testing of single-NP-enriched Dhfq culture supernatants
Culture supernatants of induced Dhfq-P BAD _xy mutants grown in Luria Bertani medium enriched only with the desired NP suggested the possibility of direct testing for bioactivity.W et herefore tested 38 supernatants from different strains,i ncluding the corresponding wild type and Dhfq controls,i nm ultiple bioassays (Supplementary Table 5). These included antibiotic activity against Gram-negative and Gram-positive bacteria, quorum quenching (QQ) activity against Vibrio campbellii and Chromobacterium violaceum, activity against different human and plant pathogenic fungi, oomycetes,t oxicity against higher organisms (zebrafish, nematodes,insects,mites) and biochemical assays ( Figure 3).
In these assays we observed aloss of activity for all Dhfq mutants compared to most wild type strains ( Figure 3). One exception was an unknown quorum quenching activity in all Dhfq mutants of both Photorhabdus strains.S everal known bioactivities were confirmed with our method, including quorum quenching activity of the phenylethylamides (17)and tryptamides against C. violaceum [24] and apoptosis-inducing activity of the proteasome inhibitor glidobactin A( 6). [25,26] Glidobactin A(6)additionally showed antifungal activity and inhibited the production of NO,b ut not of prostaglandin E 2 (PGE 2 )i nv itro.X enocoumacins (16 a and/or 16 b) [27] and fabclavine (27) [28] appeared to be the main bioactive contributors in both X. doucetiae and X. nematophila. While xenocoumacins show ab road-spectrum bioactivity in most assays including inhibition of NO and PGE 2 production, fabclavines show as imilar broad-spectrum activity without inhibiting the production of NO and PGE 2 .I tm ust be mentioned that from the activation of some BGCs,m ultiple NP derivatives are produced (e.g.f or the rhabdopeptides, xenoamicins or GameXPeptides) [19] and that the corresponding bioactivity data cannot identify the active derivative. However,o nce ad esired bioactivity is observed, the most active derivative can be identified following isolation of these derivatives and repeating the target assay(s) with the pure NPs.D ifferences in the amount or structure of these derivatives might also account for bioactivity differences as it was observed for activation of the xenocoumacin producing BGC in X. doucetiae and X. nematophila (compare the production of 16 in Supplementary Figure 10 and 11). While both xenocoumacin I( 16 a)a nd xenocoumacin II (16 b)a re produced in wild type and promoter exchange mutant of X. nematophila,i nX. doucetiae only 16 b was observed but at ah igher amount. Wild type supernatants of P. laumondii TTO1 showed ag ood antibiotic activity against V. campbelli that could not be repeated in any of the promoter exchange mutants suggesting that the responsible BGC was not activated. Activation of different BGCs in the same parental strain showed high bioactivity against higher eukaryotes like zebrafish and nematodes exemplified by 16/18 and 16/19 in X. doucetiae and X. nematophila. In general, the bioactivity of two NPs in the same assay might point towards an important ecological role of these NPs to act synergistically in aw elldefined (or concerted) mixture.T he free-living stage of the nematodes carrying Photorhabdus or Xenorhabdus in their gut, infect insect larvae in the soil that are used as af ood source and shelter for nematode development. To guarantee an undisturbed propagation, the insect cadavers must be protected by NPs delivered by the nematode symbiont against potential food competitors including also invertebrates and vertebrates. [19] Since many NPs are only produced in low amounts in the natural environment, [29] synergism might be an efficient way to potentiate the overall activity as previously shown also for clinically used drugs [30] while at the same time using less resources for NP production. Subsequently,activa-tion of two BGCs together or mixing of the individual supernatants might help to elucidate such synergistic pairs.

easyPACId in Pseudomonas entomophila
Since Hfq has been shown to influence the production of some NPs in other proteobacteria like Pseudomonas, [17,31,32] Serratia [33][34][35] and Burkholderia, [36] we applied our method to Pseudomonas (Ps) entomophila since it contained known and unknown BGCs.The Dhfq mutant in Ps.entomophila showed as trong reduction of pyoverdines [37] and labradorins (30) [38] and ac omplete loss of the lipopeptide entolysin (31). [39] Ps. entomophila-Dhfq lost its swarming ability and antibiotic activity against Micrococcus luteus and Saccharomyces cerevisiae (Supplementary Figure 17). Activation of the BGC (PSEEN_RS10885) for the recently described pyreudiones (29) [40] indeed resulted in the products that were not produced in the wild type strain under the cultivation conditions tested (Supplementary Figure 17). Additionally,an ew tetrapeptide named pseudotetratide A(33) (Figure 2d)was obtained from activation of PSEEN_RS12600 in the Dhfq mutant showing the potential of this approach in other NP-producing proteobacteria encoding Hfq. Thes tructure of 33 was confirmed after isolation from alarge-scale culture followed by detailed NMR analysis (Supplementary Figure 18-23, Supplementary  Table 4).

Chances and limitations of easyPACId
If BGCs are composed of multiple separated transcription units,promoter activation of only one of these would result in only partial BGC activation and production of either none or not the complete NP.T oa chieve full BGC activation for complete NP production, multiple promoters must be activated. This limitation also applies to promoter activation in Dhfq mutants as it was evident for the BGC responsible for entolysin (31)biosynthesis in Ps.entomophila that is split into two loci etlA and etlBC. [39] Activation of etlA encoding two NRPS modules only produced the starter fragments (32 a-b) of entolysin that were not detected in the wild type under the same conditions (Supplementary Figure 17). In cases where the biosynthesis of an unusual building block (e.g. amino acid, iso-fatty acid) is also Hfq-dependent, either no NP or nonnative NP derivatives will be produced which can also be advantageous due to the production of novel derivatives that the wild type does not produce.This is exemplarily shown for the production of xenorhabdins (18)i np romoter exchange mutants of the X. doucetiae wild type and Dhfq strains: neither an iso-fatty acid nor an N-methyl group was found in the xenorhabdins produced in the Dhfq mutant in contrast to the derivatives produced in the wild type strain (Supplementary Figure 24). [10] This suggests that enzymes responsible for these pathways/modifications are not encoded in the activated operon but encoded elsewhere in the genome and therefore are not produced in the Dhfq mutant. However,i n general the BGC structure in proteobacteria is often rather simple compared to other prolific NP producers like actinobacteria, making them ideal targets for this approach that we termed easyPACId.
Even if Dhfq mutants would have ag rowth defect compared to the parental wild type strain as observed for Ps.a eruginosa [17] and Ps.e ntomophila ( Supplementary Figure 2), the cleaner background of the Dhfq strain would still be advantageously for NP detection and isolation.

Conclusion
Although we applied easyPACId mainly to NRPS and NRPS/PKS-derived NPs as they often represent the major NP classes,w ea ssume that it also works for other BGC classes that are controlled by asingle promoter as it is often the case in proteobacteria. [42] Since the generation of Dhfq mutants as well as the activation of BGCs of interest can easily be performed in high-throughput in these (and other) strains,i t should be possible to obtain multiple new NPs in the future. This will accelerate the identification of bioactive NPs for various applications from direct testing of supernatants or crude extracts without time-consuming isolation ( Figure 3). In more well-established NP producers like Actinobacteria, in the future maybe other global regulatory mechanisms could be used. As an example,i th as been shown in Streptomyces that N-acetylglucosamine acts as as ignal for the onset of development and as ag lobal elicitor molecule for antibiotic production. [43] This said, there might also be ag lobal suppressor mechanism for NP production in these bacteria that would be equivalent to Dhfq mutants in proteobacteria and that can be used for promoter exchange approaches even in BGCs with multiple transcriptional units applying CRISPR/ Cas [24] or similar technologies.Ingeneral, the isolation of NPs from such mutants with areduced NP-background or no NPs at all would be greatly simplified allowing the future illumination of the biosynthetic "dark matter" present in most microbes. [10,44]