Plant-microbe co-evolution: allicin resistance in a Pseudomonas fluorescens strain (PfAR-1) isolated from garlic

The antibiotic defense substance allicin (diallylthiosulfinate) is produced by garlic (Allium sativum L.) after tissue damage, giving garlic its characteristic odor. Allicin is a redox-toxin that oxidizes thiols in glutathione and cellular proteins. A highly allicin-resistant Pseudomonas fluorescens strain (PfAR-1) was isolated from garlic, and genomic clones were shotgun electroporated into an allicin-susceptible P. syringae strain (Ps4612). Recipients showing allicin-resistance had all inherited a group of genes from one of three similar genomic islands (GI), that had been identified in an in silico analysis of the PfAR-1 genome. A core fragment of 8-10 congruent genes with redox-related functions, present in each GI, was shown to confer allicin-specific resistance to P. syringae, and even to an unrelated E. coli strain. Transposon mutagenesis and overexpression analyses revealed the contribution of individual candidate genes to allicin-resistance. Moreover, PfAR-1 was unusual in having 3 glutathione reductase (glr) genes, two copies in two of the GIs, but outside of the core group, and one copy in the PfAR-1 genome. Glr activity was approximately 2-fold higher in PfAR-1 than in related susceptible Pf0-1, with only a single glr gene. Moreover, an E. coli Δglr mutant showed increased susceptibility to allicin, which was complemented by PfAR-1 glr1. Taken together, our data support a multi-component resistance mechanism against allicin, achieved through horizontal gene transfer during coevolution, and allowing exploitation of the garlic ecological niche. GI regions syntenic with PfAR-1 GIs are present in other plant-associated bacterial species, perhaps suggesting a wider role in adaptation to plants per se.

was complemented by PfAR-1 glr1. Taken together, our data support a multi-component The reaction proceeds rapidly and alliin conversion to allicin is approximately 97 % complete 48 after 30 sec. at 23 °C (Lawson & Hughes, 1992). The evolutionary investment of garlic in this 49 mechanism is further emphasized by the fact that a single clove of approximately 10 g fresh 50 weight can liberate up to 5 mg of allicin (Lawson et al., 1991;Block, 2010). 51 Allicin is an electrophilic oxidant that oxidizes thiols, or more precisely the thiolate ion, 52 in a modified thiol-disulfide exchange reaction (Scheme 1), producing S-allylmercapto 53 disulfides (Rabinkov et al., 2000;Müller et al., 2016). Cellular targets are accessible cysteines 54 in proteins, and the cellular redox buffer glutathione (GSH). In this way, allicin can inhibit 55 essential enzymes (Wills, 1956) and shift cell redox balance (Gruhlke et al., 2010). Allicin 56 causes oxidative stress and was shown to directly activate the Yap1 transcription factor which 57 coordinates the oxidative stress response in yeast (Gruhlke et al., 2017). Inded, allicin has been 58 described as a 'redox toxin' (Gruhlke et al., 2010). competitive environments (Soucy et al. 2015). 80 In the work reported here, we isolated a highly allicin-resistant bacterium from its 81 ecological niche on garlic, an environment hostile to non-adapted microorganisms, and we used 82 a shotgun genomic cloning strategy to functionally identify genes conferring allicin resistance.  Additional information about bacterial strains, plasmids, primers, chemical allicin synthesis, 94 and allicin analysis are given in the supplementary material (SM). 95 PfAR-1 genomic library construction. Genomic DNA of PfAR-1 was extracted and partially 96 digested with Sau3AI to obtain approx. 10 kbp fragments which were subcloned in BamHI 97 digested pRU1097 (SM7, SM8). 98 Transposon mutagenesis. Genomic clone 1 plasmid was transposon mutagenized in Ps4612 99 using the transposon IS-Ω-km/hah on plasmid pSCR001. Plasmid DNA was transferred from 100 Ps4612 to E. coli MegaX to seperate chromosomal and plasmid Tn integrations as described in 101 SM9. 102 Overexpression of putative allicin resistance genes. Genes were amplified from genomic 103 clone 1 plasmid DNA with overhangs for NotI/XbaI subcloning in pJABO (SM10).

109
Plates were then incubated over night.

110
To investigate the effects of different oxidants, bacterial cultures were freshly inoculated from 111 over-night cultures and grown from an OD600 of 0.05 to an OD600 of 1.0 to 2.0. OD600 of all 112 strains was adjusted to an OD600 of 1.0 and 125 µl were applied on top of 20 ml solid medium 113 in round Petri dishes (Ø = 9 cm). Bacteria were spread with glass beads (Ø = 3 mm) by soft 114 shaking to avoid heat stress compared to seeded agar. Holes (Ø = 0.6 cm) were punched out to 115 apply the test solution.

116
Streak tests. A bacterial colony was harvested from agar plates in liquid medium, mixed, and 117 streaked on 20 ml prepared solid media with holes in the center (Ø = 0.6 cm) to apply test 118 solutions.

119
Drop tests. Liquid bacterial cultures grew over night and were subsequently diluted ten-fold 120 from OD600 = 1.0 to OD600 = 10 -5 . 5 µl of each bacterial dilution was dropped on solid media 121 (LB) containing different amounts of allicin. Plates were incubated at 37 °C over night.

122
Protein extraction and glutathione reductase activity assay. Pseudomonads were grown 123 over night in liquid M9JB medium (SM2) to decrease slime production. Crude bacterial cell 124 lysate was prepared from bacteria by vortexing with glass beads. Glutathione reductase activity 125 assay was performed as described in SM 12. was performed by GATC. The resulting two datasets, combined with the Illumina datasets 145 described above, were then assembled, using SPAdes 3.5.0, yielding a single contig sequence 146 of ~6.26Mbp.

147
Self-alignment of this contig revealed that 9,642 bp sequence was duplicated on each end 148 which was then removed from one end. In order to simplify cross-genome comparisons, this 149 sequence was aligned against the Pf0-1 reference sequence, and oriented to match, resulting in  In-silico analysis of the PfAR-1 Genome. The low-GC regions identified in the PfAR-1 154 genome were initially compared manually by cross-referencing the functional annotation of 155 genes. This revealed a list of genes from each region which have a potentially common origin.

156
After removing low confidence protein annotations, which were both unique to a single region 157 and lacking a definitive functional annotation (PfAR1.peg.1058 and PfAR1.peg.1070), the 158 remaining genes were manually reconciled into a putative ancestral arrangement of 26 genes.

159
Comparison of Putative HGT Regions across the Pseudomonas Genus. A set of bait genes 160 was created based on the putative 26-gene ancestral arrangement described above. Since these 161 26 groups were generally represented in more than one region, the set comprised 57 sequences

216
An allicin-resistant Pseudomonas fluorescens isolated from garlic. We reasoned that 217 if allicin-resistant bacteria were to be found in nature, then it would likely be in association with 218 garlic cloves. Therefore, bacteria isolated from garlic bulbs were tested for allicin resistance in 219 a Petri plate agar diffusion test with bacteria-seeded agar. An extremely resistant isolate was 220 detected that was able to grow right up to the allicin solution in the agar, whereas allicin-  Table 1). The overall 260 arrangement of the genes was highly conserved among the clones. Clones 1-7 contained two 261 sets of genes conserved in the direction of transcription: osmC, sdr, tetR, dsbA, and trx and the 262 second set being ahpD, oye, 4-ot, kefF, and kefC ( Fig. 3 C). Clone 8, which conferred slightly 263 less allicin resistance than the other clones ( Fig. 2), lacked the ahpD and oye genes. The kefF 264 and kefC genes are part of a glutathione-regulated K + efflux/H + influx system and are classified 265 as transporters, although they too are regulated by cellular glutathione, and thus are also redox-266 dependent.
267   sequenced. Various oxidants were tested (Fig. 4 A), and it was found that the genomic clones conferred allicin-specific resistance in both E. coli and Ps1448A, as evidenced by a reduction 286 in inhibition zone area against allicin, but not the other oxidants tested (Fig. 4 B). The degree 287 of allicin-resistance conferred by genomic clones 1 and 5 was similar, but as previously 288 observed in the streak test with Ps4612 (Fig. 2), clone 8 was less effective than the other clones 289 (Fig. 4 B). screening Tn-mutants for loss-of-function. In addition, subcloning and over-expression of 301 individual genes in Ps4612 was undertaken, with screening to assess for gain-of-function.  tests. All cultures were diluted to OD600 = 1 (= 10 0 ) and 5 µL of a 10 n dilution series down to 10 -5 was 308 dropped onto LB medium supplemented with different allicin concentrations.

309
In total, 86 out of the 132 Tn-mutants investigated showed a decrease in allicin resistance found in the osmC, sdr or tetR genes, but for the majority of the remaining genes, several 313 independent Tn insertion sites were found and these showed a tendency towards decrease in 314 resistance SM13). Tn-mutants for each gene were selected for testing in a more sensitive drop 315 test (Fig. 5 B). In the absence of allicin stress, all Tn-mutants grew less well than controls (wt the data suggest that this protein plays a major role in being able to confer allicin resistance to 324 PfAR-1. The contributions of the dsbA and trx genes to allicin resistance were more than those 325 of the kefC and oye genes, but all of these Tn-mutants showed a clear allicin phenotype, 326 especially at the 200 µM allicin level (Fig. 5 B). 327 Overexpression of ahpD and dsbA conferred high allicin-resistance to Ps4612. The set of

342
In-silico analysis of the PfAR-1 Genome. As stated above, Pseudomonas flourescens Pf0-1 is (<55%), that were absent in Pf0-1 (Fig. 7 A, B). The combination of low GC content and 347 absence from a near-relative genome suggests that these regions might have arisen by horizontal 348 gene transfer. Further analysis revealed that each of the 3 genomic islands (GI1, GI2, and GI3) 349 contained a highly similar region, which we labeled RE1, RE2 and RE3, respectively. These genes within these regions had many annotations in common and a syntenic organization 351 (SM14), suggesting a shared origin.

352
The maintenance of high-similarity regions is generally rare in prokaryotes, and typically 353 requires that such regions offer a substantial evolutionary benefit. Intriguingly, the allicin-354 resistance-conferring clones found in the functional analysis originated within these 3 regions 355 (Fig. 7 B, C, D and E), suggesting that the evolutionary benefit may be, in fact, increased allicin 356 resistance. Interestingly, each genomic clone covered almost the complete corresponding repeat 357 region, and thus the genes shared between the RE regions (Fig. 8) matches closely with those 358 shared between the clones (Fig. 3 B). Given   Inter-Species Codon Analysis. Expecting the codon usage of a horizontally transferred gene 395 region to resemble the donor species rather than the current host, we performed a codon usage 396 analysis to complement the bait-sequence analysis described above. For this, we compared for 397 the full PfAR-1 genome, the 3 RE regions, the 3347 other available Pseudomonas genomes, 398 and 8 representative non-Pseudomonas Gammaproteobacteria. The results were plotted using 399 Principal Component Analysis (PCA), and are shown in Fig. 9.

404
The first principle component, which accounts for almost 78% of the variation, is

412
The resulting plot loosely clusters the 3 GI regions with 4 sequenced Pseudomonas 413 species, namely P. luteola, P. lutea, P. zeshuii and P. sp. HPB0071. Unfortunately, none of 414 these 4 species were found to contain matches for the bait sequences in the cross-species 415 comparison above, and thus they are unlikely to be the origin of the putative HGT regions.

416
In addition to the genome-wide analysis, we also did a gene-window analysis of PfAR-1, clade D is shown in Fig. 10 A, while the corresponding whole-genome tree is shown in Fig. 10

477
Coordinates of syntenic regions are given in SM19.

478
When PfAR-1, P. salomonii ICMP14252, Pst DC3000 and Ps1448A were tested in a 479 simple streak assay, we observed that PfAR-1 and P. salomonii are most resistant against in Pst DC3000 (Fig. 13). with PfAR-1 genomic clone 1 or pRU1097 (empty vector control), while the other strains were not 488 genetically modified. PfAR-1 has three copies of a set of ten genes that were identified on genomic clones 489 (e.g. genomic clone 1) that confer resistance to allicin in P. syringae strain 4612. P. salomonii ICMP 14252 490 has two copies of this set of genes in its genome while P. syringae pv. tomato DC3000 has one and P.
The role of glutathione reductase (Glr) in PfAR-1 allicin resistance. Both GI1 and GI2 have 494 a glr gene (glr2, glr3) outside of the allicin-resistance-conferring genomic clone regions in RE1 495 and RE2, respectively, and a further glr gene (glr1) is present on the PfAR-1 chromosome.

496
Furthermore, PfAR-1 also had a two-fold higher basal Glr activity than Pf0-1 (Fig. 14). Since 497 allicin targets -SH groups in proteins and GSH metabolism is critical for resistance to allicin, 498 we investigated the potential contribution of Glr to allicin-resistance in PfAR-1.

503
The importance of GSH metabolism and Glr for allicin-resistance is shown in Fig. 15.

504
The agar-diffusion test showed that deleting the glr gene from E. coli BW25113 increased its 505 susceptibility to allicin compared to the wt (Fig. 15 A and B). Allicin-resistance was restored 506 by complementing the BW25113 Δglr strain with the chromosomal glr1 gene from PfAR-1 507 (Fig. 15 C). bacterial isolates and, although the genetic basis for this variation was unknown, we reasoned 525 that we might find organisms with high allicin-resistance associated with the garlic bulb itself 526 as a niche-habitat. This was indeed the case, and we were able to isolate a highly allicin-resistant 527 bacterium, which we named P. fluorescens Allicin Resistant-1 (PfAR-1), from a garlic bulb. In 528 inhibition zone tests, comparison with E. coli K12 DH5α or P. syringae 4612, PfAR-1 showed 529 an exceptionally high degree of allicin-resistance (Fig. 1). We employed parallel approaches of 530 functional testing of random genomic clones, and whole genome sequencing, to characterize 531 genes conferring allicin-resistance to PfAR-1 (Figs 2 & 3). Interestingly, positive clones 532 conferred allicin-resistance not only to closely-related pseudomonads, but also to distantly 533 related bacteria such as E.coli (Figs 4 & 5).

534
Resistance-conferring PfAR-1 genomic clones contained 16 unique genes in total, with a 535 congruent set of 8 genes shared by all clones (Fig. 3). Since allicin is a redox toxin causing 536 oxidative stress, it was interesting to observe that half of these genes were annotated with redox-537 related functions (Fig. 3). Moreover, these genes were linked to reports in the literature in the 538 context of oxidative and disulfide stress (Table 1).

539
Transposon mutagenesis indicated that the dsbA, trx, kefC, and oye genes worked conferring almost as much resistance as the complete genomic clone (Fig 6). However, we 553 observed no effect for trx or oye overexpression; although transposon insertions in the complete 554 genomic clone led to a decrease in resistance (Fig. 5). This might indicate that the function of KefC is normally tightly regulated by KefF and GSH, and an imbalance could lead to a toxic 559 decrease in cellular pH and loss of potassium, which is needed to maintain turgor and enable 560 cell growth and division (Epstein, 2003). Overexpression of sdr and oye showed no phenotype 561 (not shown).

562
In parallel to the gain-of-function approach, genome analysis revealed unique features of 563 the PfAR-1 genome compared to the Pf0-1 reference strain. Thus, three large genomic regions 564 (GI1 to GI3, Fig. 7) were identified, with sizes between 79 to 98 kbp, having a lower GC content 565 (Δ%GC approximately 5-10 %). These GI's contained repeat regions RE1, RE2 and RE3 with 566 the resistance-conferring clones identified in the functional analysis being contained within 567 them (Fig. 3 B, 8).

568
A codon usage analysis showed differences in RE1, RE2 and RE3 compared to the core 569 PfAR-1 genome (Fig. 9); which was a strong indication that these regions were obtained by It is unusual for multiple copies of genes to be maintained in bacteria because of the genomic 578 instability that arises through homologous recombination leading to genome rearrangements 579 and loss of essential interim sequences (Rocha, 2003). The presence of such large, widely 580 spaced REs in the PfAR-1 genome, suggested that there was a high selection pressure to 581 maintain them. The latter is presumably associated with the allicin-resistance-conferring 582 functions of many of the genes and this fits with the competitive advantage they offer over other 583 bacteria in occupying the garlic niche.

584
Although the GI donor remains unknown, phylogenetic analysis identified similar 585 syntenic regions to the REs from PfAR-1 in other bacterial genomes (Fig. 12). Thus, the garlic 586 pathogen P. salomonii ICMP14252 has two syntenic regions, and the well-described model 587 pathogen P. syringae pv. tomato DC3000 has one syntenic region. In P. salomonii ICMP 14252 588 and Pst. DC3000 the syntenic regions have the set of ten core genes identified in genomic clones 589 1-7 of PfAR-1 (Fig.12). Furthermore, we observed that the species with multiple copies of the 590 syntenic regions, 3 for PfAR-1 and 2 for P. salomonii, showed higher allicin resistance than 591 those with only one or zero copies, namely Pst DC3000 and Ps 1448A, respectively (Fig 13).

592
This further supports our findings that these regions are putative allicin resistance factors in 593 PfAR-1. P. salomonii causes the café-au-lait disease on garlic (Gardan et al., 2002) and its high 594 degree of allicin resistance corresponds well to its niche as a pathogen of garlic. One might 595 expect that a pathogen like P. salomonii could be the origin of allicin resistance genes in 596 PfAR-1, but according to our codon usage analysis, the allicin resistance regions in P. salomonii 597 ICMP14252 are quite distinct from the remainder of the genome, and therefore were also likely 598 obtained by horizontal gene transfer (Fig 9, 10). Pst DC3000 is a model pathogen with a fully 599 sequenced genome (Buell et al., 2003) and is pathogenic on tomato and on the model organism 600 Arabidopsis thaliana (Xin and He, 2013). To the best of our knowledge, the genes and their 601 function in allicin resistance have not been described before in this well studied strain. Although 602 our experiments suggest that the resistance conferred by the core-region is allicin-specific ( Fig. 4) with only one copy of glr (Fig. 14). The importance of Glr activity for tolerance to allicin is 642 clear from the enhanced sensitivity to allicin of a Δglr knockout in E. coli. Moreover, we were 643 able to complement this phenotype by introducing glr1 from PfAR-1 (Fig. 15). Glr recycles 644 oxidized glutathione (GSSG) to GSH using NADPH as a reductant. GSH protects cells from 645 oxidative stress, either by direct reaction with pro-oxidants, thus scavenging their oxidative presumably because the lowered pH works against thiolate ion formation (Ferguson 1993(Ferguson , 664 1995(Ferguson , 1996(Ferguson , 1997Poole, 2015). KefC activation would be expected to protect against oxidative 665 stress caused by the electrophile allicin in the same way (Fig. 16).

666
Because PfAR-1 lacks the 6-phospho-fructokinase gene necessary for glycolysis, it uses 667 the Entner Dodouroff Pathway (EDP) to metabolize glucose to pyruvate. During oxidative 668 stress, the EDP is advantageous over glycolysis because, in addition to NADH, NADPH is 669 produced (Conway, 1992). For example, it was shown for P. putida that key enzymes of the 670 EDP are upregulated upon oxidative stress (Kim et al., 2008). NADPH is used as reducing 671 equivalents for antioxidative enzymes like glutathione reductase and Oye-dehydrogenases.    receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. 697 Nikolaus Schlaich ist thanked for helpful discussions and Ulrike Noll for proof-reading the MS. 698 699