A highly polymorphic effector protein promotes fungal virulence through suppression of plant-associated Actinobacteria

Plant pathogens secrete effector proteins to support host colonization through a wide range of molecular mechanisms, while plant immune systems evolved receptors to recognize effectors or their activities to mount immune responses to halt pathogens. Importantly, plants do not act as single organisms, but rather as holobionts that actively shape their microbiota as a determinant of health, and may thus be targeted by pathogen effectors as such. The soil-borne fungal pathogen Verticillium dahliae was recently demonstrated to exploit the VdAve1 effector to manipulate the host microbiota to promote vascular wilt disease in absence of the corresponding immune receptor Ve1. We now identified a multiallelic V. dahliae gene displaying ~65% sequence similarity to VdAve1, named VdAve1-like (VdAve1L). Interestingly, VdAve1L shows extreme sequence variation, including alleles that encode dysfunctional proteins, indicative of selection pressure to overcome host recognition. We show that the orphan cell surface receptor Ve2, encoded at the Ve1 locus, does not recognize VdAve1L. Furthermore, we show that the full-length variant VdAve1L2 possesses antimicrobial activity, like VdAve1, yet with a divergent activity spectrum. Altogether, VdAve1L2 is exploited by V. dahliae to mediate tomato colonization through the direct suppression of antagonistic Actinobacteria in the host microbiota. Our findings open up strategies for more targeted biocontrol against microbial plant pathogens.

The plethora of microbes a plant associates with, its so-called microbiota, encompasses a 52 diversity of microbes that establish a spectrum of symbiotic relationships with their host, 53 ranging from commensalistic through endophytic to mutualistic and pathogenic (Hassani et  Verticillium dahliae is a soil-borne fungal pathogen that causes vascular wilt disease in 77 hundreds of plant species (Fradin & Thomma, 2006). The presumed asexual fungus generates 78 genomic diversity through extensive chromosomal rearrangements and segmental duplications 79 that gave rise to dynamic so-called lineage-specific (LS) regions, more recently referred to as Only few genetic resistance sources to V. dahliae have been identified. In tomato, the 95 Ve locus provides resistance against V. dahliae and has been introgressed into most commercial 96 tomato cultivars (Schaible et al., 1951;Diwan et al., 1999). The Ve locus contains two closely 97 linked genes, SlVe1 and SlVe2, that both encode extracellular leucine-rich repeat receptor-like 98 proteins (eLRR-RLPs), of which only SlVe1 was confirmed to confer V. dahliae resistance 99 2010). Then, raw read coverage values were transformed to a binary matrix by applying a cut-145 off of 10 reads for short-read data; >=10 reads indicate presence (1) and <10 reads indicate 146 absence (0) of the respective genomic region. In the case of long-read data, a cut-off of 1 read 147 was applied; >=1 reads indicate presence (1) and <1 reads indicate absence (0). This matrix 148 was further summarized to obtain the total number of presence/absence counts in each 100 bp 149 genomic window within 50 kb upstream and downstream of the VdAve1L locus.  promoter, the coding sequence of VdAve1L2 was amplified and cloned into pFBT005. V. 180 dahliae transformations were performed as described previously (Santhanam, 2012). 181

Disease assays 182
Inoculation of tomato plants to determine the virulence of the V. dahliae was performed as 183  incubated at 22°C and 19°C during 16-h day and 8-h night periods, respectively, with 70% 210 relative humidity. Leaves were inspected for HR at 5 dpi. 211

Protein production, purification and refolding 212
Heterologous production and purification of VdAve1L2 was performed as described 213 previously for VdAve1 (Snelders et al., 2020). 214

Oxford Nanopore Technology sequencing 215
Library preparation of the PCR fragment was performed according to the protocol of Oxford 216 Nanopore, skipping the DNA fragmentation step. The library was loaded on a Nanopore flow 217 cell. The run yielded about 870 high quality long reads with an average length of 3,869 bp with 218 the longest read being ~14 kb. Using Nanocorrect, we used all the obtained reads to correct the 219 50 longest reads, of which 28 were corrected to generate a consensus. Finally, reads were used 220 for BLAST analysis to confirm the presence of VdAve1L3 fragments at both ends. 221

Root microbiota analysis 222
Tomato inoculations were performed as described previously (Fradin et al., 2009). After 223 10 days, plants were carefully uprooted and gently shaken to remove loosely adhering soil from 224 the roots. Next, roots with rhizosphere soil from two tomato plants were pooled to form a single 225 biological replicate. Alternatively, the tomato plants that received the water-treated and 226 vancomycin-treated microbial communities where uprooted 18 days post inoculation with V. 227 dahliae and a single root system with adhering river sand was collected as a biological control. 228 All samples were flash-frozen in liquid nitrogen and ground using mortar and pestle. Genomic 229 DNA isolation was performed using the DNeasy PowerSoil Kit (Qiagen, Venlo, The 230 Netherlands. Sequence libraries were prepared following amplification of the V4 region of the 231 bacterial 16S rDNA (341F and 785R) and paired ends (300 bp) were sequenced using the 232 MiSeq sequencing platform (Illumina) at Baseclear (Leiden, The Netherlands). Data analyses 233 were performed as described previously (Snelders et al., 2020). Institute (Utrecht, the Netherlands). 243

Antimicrobial activity assays 244
In vitro antimicrobial activity assays were performed as described previously using 0.2x 245 tryptone soy broth as growth medium for the bacteria (Snelders et al., 2020). 246

In vitro competition assays 247
Following eleven days of cultivation on PDA, the conidiospores of V. dahliae strains DVDS26 248 ΔVdAve1L2, DVD-S26 ΔVdAve1L2 + pVdAve1::VdAve1L2 #1 and DVD-S26 ΔVdAve1L2 + 249 pVdAve1::VdAve1L2 #2 were harvested from plate and stored at -80° C at a concentration of 250 4*10 5 spores/mL in low salt TSB (17 g/L tryptone, 3 g/L soy peptone, 0.5 g/L NaCl, 2.5 g/L 251 K2HPO4 and 2.5 g/L glucose) supplemented with 10% glycerol until use. Next, bacterial 252 isolates were grown on low salt TSA at 28°C. Single colonies were selected and grown 253 overnight at 28°C while shaking at 150 rpm. Overnight cultures were resuspended to an 254 OD600=0.02 in fresh low salt TSB, while the fungal spore suspensions were allowed to thaw at 255 room temperature. Finally, the bacterial and fungal spore suspensions were mixed in 500 µl of 256 low salt TSB to a final concentration of OD600=0.01 and 10 3 spores/mL, respectively. 257 Following six days of incubation at 22°C, the microbial suspensions were transferred to clear 258 24-well flat-bottom polystyrene tissue culture plates to allow imaging of fungal growth using 259 an SZX10 stereo microscope (Olympus) equipped with a EP50 camera (Olympus). with considerable allelic variation. In total we identified six allelic variants (VdAve1L1 to 271 VdAve1L6) that share 93-99% sequence similarity (Fig. 1a,b). Like VdAve1L1, also VdAve1L3 272 and VdAve1L4 encode truncated 24 amino acid proteins (Fig. 1a). In contrast, and similar to 273 VdAve1, VdAve1L2 and VdAve1L5 encode 134 amino acid proteins including an 18 amino acid 274 N-terminal signal peptide (Fig. 1a). Finally, VdAve1L6 only differs by one amino acid when 275 compared with VdAve1L5 and is truncated after 120 amino acids (Fig. 1a). As expected based on the observed PAV among some V. dahliae strains (Fig. 1b), 297 VdAve1L is localized in an AGR (Fig 1c,d) Supplementary Fig. 3), and that all of the identified SNPs, 30 302 in total ( Supplementary Fig. 4), cause protein sequence variation. Thus, VdAve1L is a highly 303 polymorphic gene that displays accelerated evolution by PAV, transposon-mediated sequence 304 disruption, and sequence variation. 305

VdAve1L proteins are not recognized by SlVe2 306
The extreme sequence variation of VdAve1L is likely the result of selection pressure to 307 overcome recognition by a plant immune receptor. The only R gene known to confer resistance 308 to V. dahliae is tomato SlVe1 (Kawchuk et al., 2001; Fradin et al., 2009), which resides in the 309 Ve locus together with SlVe2 that similarly encodes a receptor-like protein (Fradin et al., 2009). VdAve1L1*, VdAve1L3* and VdAve1L4* (Fig. 2a,b). Similarly, we replaced stop codons in 316 VdAve1L3* and removed the retrotransposon (Fig. 2a,b). Additionally, based on an alignment 317 consensus sequence we constructed the predicted common VdAve1L ancestor VdAve1L** 318 (Fig.2a,b). Subsequently, we co-expressed the various genes with SlVe2 in Nicotiana tabacum, 319 and with SlVe1 as a negative control, but no hypersensitive response (HR) could be observed 320 except upon co-expression of SlAve1 (Fig. 2c). Consequently, it is unlikely that SlVe2 321 recognized VdAve1L and drove its diversification. 322 Cladosporium fulvum effector Avr9 serves as a negative control for recognition by SlVe1 and 338 SlVe2. 339 340

VdAve1L2 is a virulence factor that functionally diverged from VdAve1 341
Only the alleles VdAve1L2 and VdAve1L5 encode full length proteins (Fig. 1), yet none of the 342 strains that encodes VdAve1L5 is pathogenic on tomato (Li, 2019). To determine if 343 VdAve1L2, like VdAve1, contributes to virulence on tomato, we assessed its expression in V. 344 dahliae race 2 strain DVD-S26 during host colonization, showing clear expression in planta 345 ( Supplementary Fig. 5). Interestingly, while we previously also detected expression of VdAve1 346 during cultivation in vitro on PDA and in soil extract (Snelders et al., 2020), we did not detect 347 VdAve1L2 expression under these conditions ( Supplementary Fig. 5). 348 To determine the importance of VdAve1L2 for tomato colonization, we generated 349 deletion mutants in the race 2 strain DVD-S26 and in the race 1 strains ST14.01 and CBS38166 350 ( Supplementary Fig. 6). Inoculation of tomato plants with the deletion mutants of the race 1 351 strains revealed no virulence contribution of VdAve1L2 (Fig 3a,b). Strikingly, however, we 352 detected strongly compromised tomato colonization of the VdAve1L2 deletion mutant in strain 353 DVD-S26 (Fig 3c), which was restored in a complementation mutant (Fig. 3c)  VdAve1L2 or VdAve1 deletion mutants, and the mutants expressing VdAve1L2 under control of 372 its native or VdAve1 promoter. V. dahliae biomass in tomato stems was quantified by real-time 373 PCR. Letters represent non-significant biomass differences (one-way ANOVA and Tukey's 374 post hoc test; p<0.05; N≥17). 375

VdAve1L2 promotes V. dahliae virulence through suppression of Actinobacteria 376
While most effector proteins functionally characterized to date act in manipulation of host 377 physiology, we recently showed that VdAve1 is an antibacterial effector protein that is secreted 378 by V. dahliae to suppress microbial antagonists in the microbiomes of its hosts (Snelders et al.,  379   2020). Thus, we hypothesized that VdAve1L2 may similarly exert antibacterial activity. In 380 vitro assays previously revealed a strong activity of VdAve1 on the Gram positive bacterium 381 Bacillus subtilis (Snelders et al., 2020). Interestingly, VdAve1L2 affected B. subtilis growth as 382 well, albeit markedly less effectively (Fig. 4a). Furthermore, similar to VdAve1 (Snelders et 383 al., 2020), VdAve1L2 inhibited the growth of plant-associated Novosphingobium sp. and 384

Staphylococcus xylosus, but not of Agrobacterium tumefaciens, Pseudomonas corrugata and 385
Ralstonia sp.. However, in contrast to VdAve1, VdAve1L2 did not inhibit growth of 386 Sphingobacterium sp. (Supplementary Fig. 7). Collectively, we conclude that VdAve1L2 is an 387 antibacterial effector with a diverged activity spectrum when compared with VdAve1. 388 To determine if VdAve1L2 secretion by V. dahliae impacts host microbiota, we 389 performed bacterial community analysis based on 16S ribosomal DNA profiling on tomato 390 roots colonized by V. dahliae strain DVD-S26 and the VdAve1L2 deletion mutant. Furthermore, 391 the VdAve1 deletion mutant of V. dahliae strain JR2 and the corresponding transformant 392 expressing VdAve1L2 were included. In correspondence with previous observations (Snelders 393 et al., 2020), colonization by V. dahliae did not dramatically impact the overall composition of 394 bacterial phyla in tomato root microbiota, and also not their α-diversities (Fig. 4b,c). 395 Importantly, however, a principal coordinate analysis based on Bray-Curtis dissimilarities (β-396 diversity) revealed separation of the bacterial communities based on V. dahliae genotype (Fig.  397 4d), suggesting that secretion of VdAve1L2 impacts root microbiota compositions. Based on 398 pairwise comparisons between the abundances of the bacterial phyla detected in the microbiota 399 in the presence and the absence of VdAve1L2, we identified Actinobacteria as the sole phylum 400 that was significantly suppressed in the microbiota colonized by V. dahliae strains secreting 401 VdAve1L2 (Fig. 4e). tested Actinobacteria displayed higher sensitivity to the effector than most of the other bacteria 424 tested thus far (Fig. 5a), suggesting that Actinobacteria are genuine and direct targets of 425 of the VdAve1 promotor that is highly active during in vitro growth, in contrast to the 438 VdAve1L2 promoter (Supplementary Fig. 5; Supplementary Fig. 8). As anticipated, secretion 439 of VdAve1L2 failed to counter the antagonistic activity of A. tumefaciens and did not promote 440 V. dahliae growth when confronted with this bacterium (Fig. 5b). However, V. dahliae clearly 441 benefited from VdAve1L2 secretion in competition with both Actinobacteria, as it mediated 442 enhanced fungal growth and development of larger colonies (Fig. 5b). Collectively, our 443 findings suggest that V. dahliae secretes VdAve1L2 to antagonize Actinobacteria in the host 444 assess the virulence contribution of VdAve1L2. As determined using 16S ribosomal DNA 465 profiling, the tomato plants exposed to the vancomycin-treated microbiota did not harbor a 466 dramatically altered community of bacterial phyla when compared with plants exposed to the 467 water-treated microbiota ( Supplementary Fig. 9a). Moreover, the vancomycin treatment did 468 not affect the α-diversity or total abundance of bacteria in the plant microbiota (Supplementary 469 Fig. 9b,c). However, as anticipated, we detected a severe impact of vancomycin treatment on 470 the Actinobacteria community structure (Supplementary Fig. 9d) in the water-treated community (Fig. 6a). Quantification of V. dahliae biomass in the root 475 microbiomes using real-time PCR confirmed significantly increased colonization by the 476 VdAve1L2 deletion mutant in the presence of the vancomycin-treated microbiota (Fig. 6b). 477 Importantly, while on plants that were treated with the water-treated community VdAve1L2 478 markedly contributes to virulence (Fig. 6a), this virulence contribution is not observed on plants 479 that were treated with the vancomycin-treated community, in line with the hypothesis that the 480 Actinobacteria that are targeted by VdAve1L2 are no longer present in the host microbiota 481 (Fig. 6a). Accordingly, in contrast to the microbiomes with the water-treated microbial 482 community, principal coordinate analysis based on Bray-Curtis dissimilarities (β-diversity) 483 failed to reveal a clear separation of the root microbiomes with the vancomycin-treated 484 community based on their colonization by the different V. dahliae strains or mock treatment 485 (Fig. 6c). Moreover, we only detected a VdAve1L2-mediated repression of Actinobacteria 486 genera in plants that received the water-treated communities (Fig. 6d). Likely, treatment with 487 vancomycin limited the abundance of antagonistic Actinobacteria such that interference by 488 VdAve1L2 is no longer required for optimal V. dahliae colonization. In conclusion, our 489 findings suggest that Actinobacteria in the tomato root microbiota antagonize host colonization 490 by V. dahliae, and that the fungus exploits VdAve1L2 in turn to suppress these antagonists and 491 Functional characterization of the full-length effector variant VdAve1L2 uncovered 517 that this effector, like its homolog VdAve1, exerts antibacterial activity to directly suppress 518 microbial competitors in planta. Remarkably, however, VdAve1L2 is exclusively expressed 519 during plant colonization and, in contrast to VdAve1, not expressed in vitro or in soil. 520 Moreover, the two effectors display distinct antibacterial activities in vitro. Accordingly, we 521 observed that secretion of VdAve1L2 by the race 2 strain DVD-S26, unlike VdAve1 secreted 522 by the race 1 strain JR2, impacts the abundance of Actinobacteria and not of Sphingomonadales 523 in tomato. VdAve1L2 and VdAve1 thus functionally diverged from each other. So far, the 524 mode of action of VdAve1 remains unclear and, accordingly, it is unclear how VdAve1L2 525 functionally diverged from VdAve1 to target Actinobacteria rather than Sphingomonadales. 526 Actinobacteria represent a core phylum that is found in virtually any plant grown in any Considering these beneficial traits, it is not surprising that members of this phylum are targeted 532 by microbial plant pathogens to weaken plant holobionts. Interestingly, the oomycete 533 Arabidopsis pathogen Albugo candida was recently reported to deposit several antibacterial 534 effector proteins in the leaf apoplast (Gómez-Pérez et al., 2022). Interestingly, some of these 535 effectors impact growth of Actinobacterial keystone taxa of the Arabidopsis phyllosphere in 536 vitro, suggesting that the suppression of Actinobacteria in host microbiota might be a strategy 537 adopted by diverse microbial plant pathogens. These findings furthermore support the 538 hypothesis that effector-mediated manipulation of host microbiota communities may be a 539 widely deployed strategy of plant pathogens to support host colonization (Snelders et al., 540 2022). 541 VdAve1 is recognized by the tomato immune receptor SlVe1, encoded by a gene in a 542 locus that also encodes the highly similar orphan receptor SlVe2 (Fradin et al., 2014). 543 Considering the similarity between VdAve1 and VdAve1L, we tested if SlVe2 was able to 544 recognize any of the current VdAve1L alleles or their putative progenitors, but none of those 545 evoked a detectable hypersensitive response upon overexpression in combination with SlVe2. 546 However, if recognition took place in tomato, it may equally well have been mediated by any 547 other putative immune receptor encoded in the tomato genome. Perhaps even more likely, 548 recognition may also have occurred in any of the hundreds of other V. dahliae hosts. 549 Importantly, Actinobacteria are ubiquitously present in a wide diversity of plants, and thus V. fashion as effector proteins that interfere with intrinsic immune components are perceived. As 568 a consequence of recognition, microbial plant pathogens need to mutate, purge or inactivate 569 their microbiota-manipulating effector proteins to escape host recognition, which leads to 570 pathogen races with divergent suites of antimicrobial effectors. A possibility for the more 571 effective use of microbial biocontrol agents could be to base their selection on the genotype of 572 a plant pathogen, for instance by selecting antagonists that are insensitive to the activity of a 573 specific (lineage-specific) effector. Conversely, in case a resistance gene has been described to 574 recognize a microbiota-manipulating effector protein, the application of a strong antagonistic 575 biocontrol agent that is sensitive towards the activity of the corresponding effector can be 576 considered. In this manner, a strong selection pressure is exerted to retain that particular 577 effector gene in the pathogen, which may contribute to enhanced durability of the resistance in 578 turn. In this manner, the further identification and characterization of microbiota-manipulating 579 effectors secreted by microbial plant pathogens may aid in the development of more 580 sophisticated, and perhaps more successful, biocontrol strategies. 581