Unveiling the microbiome of hydroponically cultivated lettuce: impact of Phytophthora cryptogea infection on plant-associated microorganisms

Abstract Understanding the complex interactions between plants and their associated microorganisms is crucial for optimizing plant health and productivity. While microbiomes of soil-bound cultivated crops are extensively studied, microbiomes of hydroponically cultivated crops have received limited attention. To address this knowledge gap, we investigated the rhizosphere and root endosphere of hydroponically cultivated lettuce. Additionally, we sought to explore the potential impact of the oomycete pathogen Phytophthora cryptogea on these microbiomes. Root samples were collected from symptomatic and nonsymptomatic plants in three different greenhouses. Amplicon sequencing of the bacterial 16S rRNA gene revealed significant alterations in the bacterial community upon P. cryptogea infection, particularly in the rhizosphere. Permutational multivariate analysis of variance (perMANOVA) revealed significant differences in microbial communities between plants from the three greenhouses, and between symptomatic and nonsymptomatic plants. Further analysis uncovered differentially abundant zero-radius operational taxonomic units (zOTUs) between symptomatic and nonsymptomatic plants. Interestingly, members of Pseudomonas and Flavobacterium were positively associated with symptomatic plants. Overall, this study provides valuable insights into the microbiome of hydroponically cultivated plants and highlights the influence of pathogen invasion on plant-associated microbial communities. Further research is required to elucidate the potential role of Pseudomonas and Flavobacterium spp. in controlling P. cryptogea infections within hydroponically cultivated lettuce greenhouses.


Introduction
In r ecent years, ther e has been a gr owing r ecognition of the crucial role played by plant-associated micr oor ganisms in enhancing plant health and pr oductivity, especiall y in the context of sustainable crop production (Berg et al. 2014a ,b ).Plant-associated micr oor ganisms can contribute to disease suppression either dir ectl y by antagonizing the pathogen, or indirectly by priming the plant's immune system (also known as induced resistance; IR) (Berendsen et al. 2012, Raaijmakers and Mazzola 2012, Zamioudis and Pieterse 2012, Köhl et al. 2019, De Kesel et al. 2021 ).In this r egard, a gr owing body of r esearc h is highlighting the impact of the r oot micr obiome in pr omoting plant health.Specificall y, the r oot rhizospher e, defined as the narr ow r egion of soil (or substr ate) surrounding the roots plays a vital role, as it is harboring a diverse range of microorganisms, including bacteria, fungi, oomycetes, arc haea, pr otozoa, and algae .T hese micr oor ganisms ar e attr acted by the plant through root exudates giving rise to complex microbial root communities that may suppress soil-borne pathogens or provide other benefits to the plant (Costa et al. 2007, Hartmann et al. 2008, Badri and Vivanco 2009, Lahlali et al. 2022 ).Although m uc h r esearc h has focused on the rhizospher e, it is important to note that micr oor ganisms can colonize other parts of the plant as well, which may also contribute to plant growth and plant health.The r oot endospher e, for instance, whic h encompasses all micr oor ganisms r esiding within the plant r oots, is incr easingl y r ecognized for its unique inter action with the plant and its role in promoting plant health (Ryan et al. 2008 , Reinhold-Hurek andHur ek 2011 ).Mor eov er, in addition to micr obial individuals both direct and indirect interactions between microorganisms can have significant impacts on plant health.Ther efor e, understanding the ov er all micr obiome, its functioning and its inter actions with the plant and environment is crucial to promote plant health (Berendsen et al. 2012 ).
In addition to the beneficial micr oor ganisms that pr omote plant health and growth, the microbial communities associated with plants also include pathogens that can infect plants and alter the plant microbiome.For example, when plants are infected with pathogens, the signaling of defense-related hormones, such as salicylic acid and jasmonic acid, can change, leading to altered root exudates and, in turn, to changes in microbial community composition (Badri et al. 2008, Pieterse et al. 2009, Berendsen et al. 2012 ).Studies have also shown that substantial differences occur in the microbiomes of natur all y infected plants and their healthy counterparts originating from the same field, indicating that pathogens can alter the plant microbiome (Suhaimi et al. 2017, Wei et al. 2018, Shi et al. 2021 ).Also in the case of artificial pathogen inoculation, significant differences in microbial comm unities wer e found following pathogen infection (Cha pelle et al. 2016, Snelders et al. 2020 ).Ho w e v er, r esults seem to be contextdependent, and show either a higher r elativ e abundance of plantbeneficial microbes in noninfected plants (Suhaimi et al. 2017, Wei et al. 2018 ) or an increase in their r elativ e abundance in infected plants (Chapelle et al. 2016 ).The latter supports the "cry-for-help" str ategy, pr oposing that plants recruit beneficial microorganisms when exposed to str ess, whic h can r esult in diminished disease se v erity (Bakker et al. 2018 ).For example, in sugar beet plants artificially infected by the soilborne fungal pathogen Rhizoctonia solani , a higher r elativ e abundance of bacterial families whic h ar e known for their antifungal potential was observ ed compar ed to noninfected plants (Chapelle et al. 2016 ).By contrast, a significantly lo w er r elativ e abundance of plant-beneficial micr obes was found in Ralstonia solanacearum -infected tomato plants compared to the noninfected plants (Wei et al. 2018 ).Similarl y, higher r elative abundance of the plant beneficials Pseudomonas spp.and Bacillus spp. was found in healthy banana endospher es compar ed to the endosphere of bacterial-wilt infected banana plants (Suhaimi et al. 2017 ).To date, most attention has been given to microbiome responses to fungal and bacterial pathogens.By contrast, only little is known on how microbial communities respond to oomycete plant pathogens such as Phytophthora spp.and Pythium spp., despite the fact that this group represents one of the most economically important and widespread categories of plant pathogens (Sullam and Musa 2021 ).In a recent study by Gómez-Pérez et al. ( 2023 ), it has been demonstrated that the oomycete pathogen Albugo candida releases proteins into the host plant apoplast repressing plant-associated bacteria.The antimicrobial activity of these proteins was found to enhance host colonization by the pathogen (Gómez-Pérez et al. 2023 , Rovenich andThomma 2023 ).
Although plant microbiomes have received extensive attention in recent times, primarily focusing on soil-cultivated crops, there is still limited information available on the microbiomes of hydr oponicall y cultiv ated cr ops and how they ar e affected by plant pathogens.To the best of our knowledge, so far only one study has been conducted on the influence of a pathogen on the microbiome of hydr oponicall y gr own plants .T his study demonstrated that the root microbiome of tomato plants significantly differs between healthy plants and plants infected with rhizogenic a gr obacteria (Vargas et al. 2022 ).The root microbiome of healthy plants sho w ed a higher r elativ e abundance of Paenibacillus spp.(Vargas et al. 2022 ), fr om whic h specific linea ges hav e anta gonistic activity a gainst rhizogenic a gr obacteria (Bosmans et al. 2017 ).Ne v ertheless, further studies are needed to increase our understanding of the microbiomes of hydroponically cultivated crops.
The main goal of this study was to investigate the impact of oomycete plant pathogens on micr obial comm unities associated with hydr oponicall y gr own cr ops, focusing on both rhizosphere and endosphere bacterial communities.More specifically, we focused on the interaction between hydroponic lettuce ( Lactuca sativa L.) and Phytophthora cryptogea as our study system.Curr entl y, lettuce is widel y gr own in hydroponic systems as a r a pidl y gr owing leafy vegetable (Safaei et al. 2015 ), whereas in the past, lettuce was pr edominantl y cultiv ated in soil-bound systems .T he choice of the cultivation method is largely determined by the cultivar, environ-mental conditions, potential disease pr essur e, and the ca pital of the gro w er (Parkell et al. 2015 ).In soil-bound cultivation, lettuce is often grown in an intensive monocropping system, through which soil-borne diseases are likely to build up.In Flanders, a switch tow ar d hydroponic lettuce cultivation has been implemented due to the high disease pr essur e caused by soil-borne pathogens such as Fusarium oxysporum f .sp .lactucae , coupled with the phasing out of soil fumigation chemicals (e.g.methyl bromide) (Vandevelde 2019 , Claerbout 2020 ).Other major adv anta ges of hydr oponics ov er soilbound cultiv ation ar e shorter pr oduction times, r educed labor r equirements as some agricultural practices, such as weeding, tilling, and spraying can be eliminated, more efficient utilization of a vailable space , and reduced water usage through the recycling of nutrient solutions .T he major disadv anta ge, ho w e v er, is the easy spread of water-borne diseases through the recirculation system (Sharma et al. 2018, Vande v elde 2019 ).Fr om 2017 onw ar d, the devastating r oot r ot-causing oomycete P. cryptogea is se v er el y thr eatening hydroponic cultivation of lettuce.Notably, it was found that P. cryptogea tends to pr e v ail during warm periods (e .g. heatwa ves), which will most probably become worse in the future due to climate c hange (Berc kmoes and Van Cleemput 2021 ).The economic repercussions of this pathogen were estimated to lead to economic losses up to €50 000/ha/year (I.Vande v elde, personal communication).Phytophthora cryptogea produces flagellated, asexual zoospor es, whic h can attac h to the r oot surface and germinate to form a germination tube or a ppr essorium on the host tissue, after which hyphae invade the plant tissue (Hubrechts and De Marez 2014, Pettitt 2015, Berckmoes and Van Cleemput 2021 ).First, a slimy coat around the roots is formed, after which the roots start to show rotting symptoms followed by complete rotting of the lettuce cr op ( Figur e S1 , Supporting Information ).Because of the motile zoospor es, whic h can easil y spr ead thr ough the r ecirculated nutrient solution, contamination of the entire system can occur r a pidl y (Hubr ec hts and De Mar ez 2014 , Pettitt 2015, Sharma et al. 2018 ), leading to major economic losses.To date, there are no plant protection products available to control P. cryptogea in hydr oponicall y cultiv ated lettuce.
The aim of this study was to investigate the rhizosphere and r oot endospher e of hydr oponicall y cultiv ated lettuce.Additionall y, we sought to explore the hypothesis that P. cryptogea affects the bacterial communities of the lettuce rhizosphere and root endosphere.To test this hypothesis, both symptomatic and nonsymptomatic plants were sampled from three greenhouses naturally infested with P. cryptogea .Bacterial comm unities wer e examined by deep sequencing of partial 16S ribosomal RNA (rRNA) gene amplicons, and differences between healthy and symptomatic plants were described.By studying naturally infested greenhouses, we aimed to gain a better understanding of the impact of this important pathogen on plant-associated microbial communities under real-world conditions.

Sampling and DNA extraction
Our study was done in collaboration with three commercial hydr oponic lettuce gr o w ers in Flanders (Sint-Katelijne-Waver, Belgium), r eferr ed to as greenhouse 1, 2, and 3.In all greenhouses, lettuce was cultivated in a nutrient film technique system, more particularly a mobile gutter system, in which plant density can be adapted according to the plant growth stage.Climatic conditions, EC, and pH of the nutrient solution were similar for eac h gr eenhouse .T he plants in eac h gr eenhouse wer e sourced from the same plant nursery and were grown in substrate cubes containing 85% black peat and 15% wood fiber.Greenhouses 1 and 3 were cultivating butterhead lettuce, cv.Emeldia (Rijk Zwaan) and cv.Finley (Enza Zaden), r espectiv el y, while gr eenhouse 2 was cultiv ating m ulticolor lettuce, consisting of cv.Lugano, Satine, and Xodos (Rijk Zwaan).In eac h gr eenhouse, symptomatic and nonsymptomatic plants were collected including their substrate cubes, whic h wer e selected based on the visual pr esence or absence of typical P. cryptogea symptoms (i.e.brown or necrotic roots).Sampling was conducted without a predefined time point, as diseased plants were not alwa ys a vailable .Instead, samples were taken when P. cryptogea infection was observed (i.e. during warmer periods in the growth season).To confirm P. cryptogea as the causing agent of the symptoms, a qPCR assay was performed (see below), and for some plants the pathogen was isolated as described pr e viousl y (Dr enth and Sendall 2001 ).Plants wer e sampled after 21 (greenhouses 2 and 3) or 28 days (greenhouse 1) of growth on the gutters in June 2021 (greenhouses 2 and 3) and August 2021 (greenhouse 1), meaning that plants were almost fully grown.In general, lettuce has a cycle length of 21-35 days in summer (depending on the variety).On av er a ge in the inv estigated gr eenhouses, 220 plants/m 2 /year were cultivated during 8-12 c ycles/y ear.To av oid effects of age, both symptomatic and nonsymptomatic plants were sampled at the same time in the same and/or adjacent gutters.
Following cooled transport of the plants to the laboratory, samples wer e pr ocessed within 24 h after sampling based on the protocols described by Lakshmanan et al. ( 2017 ) and Bergna et al. ( 2018 ) ( Figure S2 ,Supporting Information ).Briefly, all roots outside the substrate cube of a plant were collected, as these roots came first into contact with the pathogen, and cut into 2 cm pieces.Subsequently, 20 ml of sterile phosphate buffered saline (PBS) was added to 2 g of r andoml y pooled root pieces.Next, root pieces were vortexed for 30 s and rinsed with another 10 ml of PBS, followed by sonication at 14.5 kHz for 30 s in 15 ml of PBS for r emov al of remaining external microbes.Finally, roots were rinsed again with 5 ml of PBS.After each step, the PBS solution was collected and pooled per plant.The pooled solution was then centrifuged at 4 • C at 5000 × g for 15 min.Next, the supernatant was discarded and the pellet was dissolved in 1 ml sterile 0.9% NaCl.The samples collected in this way r epr esented micr obiome samples of the rhizosphere .T he corr esponding r oots wer e further pr ocessed for sampling of the endosphere microbiome .T herefore , the roots were rinsed with 25 ml of sterile demineralized w ater, follo w ed by surface sterilization with 3% bleach for 5 min and three times rinsing with 25 ml sterile demineralized water.Surface-sterilized r oots wer e then used for further processing.A total of eight samples from nonsymptomatic plants and eight samples from symptomatic plants were analyzed for both greenhouses 1 and 2. Additionall y, se v en samples fr om nonsymptomatic plants and se v en samples from symptomatic plants were analyzed for greenhouse 3.For all greenhouse, both the rhizosphere and endosphere were analyzed ( Table S1 , Supporting Information ).From each sample, genomic DN A w as extracted using the DNeasy Po w erSoil Pro Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions with one modification: in the second step the use of a vortex adapter was replaced by two cycles of 30 s (with a 10 s break in between) in the Pr ecell ys ® 24 Tissue Homogenizer at a speed of 6000 r/m.To this end, 500 μl of the cell pellet was used for rhizosphere samples and ∼250 mg of the roots for endosphere samples.A negativ e contr ol in whic h the sample material was r eplaced by sterile, DNA-free water was included to confirm absence of r ea gent contamination.

Microbiome analysis
For micr obiome anal ysis, all DN A samples w ere subjected to PCR amplification of the bacterial hypervariable V4 region of the 16S rRNA gene using the Illumina barcoded primer pair 515F/806R (Ca por aso et al. 2011 ), designed according to Kozich et al. ( 2013 ) (dual-index sequencing strategy; Table S1 , Supporting Information ).A negativ e contr ol for PCR amplification, in whic h sterile, DN A-free w ater w as used instead of the DNA template, was included in each PCR run.Additionally, a DNA sample from a mock community composed of diverse bacteria was included as a r efer ence ( Table S2 , Supporting information ).PCRs were performed in a 40 μl reaction volume, containing 1x Titanium Taq PCR Buffer (Takara Bio), 0.15 mM of each dNTP (Invitrogen TM ), 1x Titanium Taq DNA Pol ymer ase (Takar a Bio), 0.5 μM of each primer, and 2 μl DNA template.Amplification started with an initial denaturation for 2 min at 94 • C, follo w ed b y 34 c ycles of denaturation, annealing and extension for 45 s each at 94 • C, 59 • C, and 72 • C, respectiv el y, and ended with a final elongation for 10 min at 72 • C. Next, amplicons were purified using Agencourt AMPure XP magnetic beads (Beckman Coulter Genomics GmbH, South Plainfield, UK) following the manufacturer's instructions .T he concentration of purified DNA fr a gments was then measured using a Qubit high sensitivity fluorometer (Invitrogen TM , Carlsbad, USA) and diluted to a concentration of 20 nM.Next, samples were pooled into a library and subjected to ethanol precipitation, after which the library was loaded on a 1.5% a gar ose gel.Subsequentl y, the target band r epr esenting fr a gments of ar ound 400 bp was excised from the gel and purified using the QIAquick Gel Extraction Kit (Qia gen).Finall y, the DNA libr ary was diluted to 2 nM and sent for sequencing at the Center for Medical Genetics (University of Antw erp, Antw erp, Belgium), using an Illumina MiSeq sequencer with a v2 500-cycle r ea gent kit (Illumina, San Diego, USA).
Sequences were received in the form of a demultiplexed FASTQ file with r emov ed barcodes and primer sequences.To mer ge pair ed-end r eads, USEARCH (v11.0.667) was used to form consensus sequences (Edgar 2013 ), allowing for no more than 10 mismatches in the overlap region.The resulting sequences were trimmed at the 250th base, and any reads with a length shorter than 250 bp or a total expected error threshold above 0.1 were discarded using USEARCH (v11.0.667).Next, Mothur (v1.39.5) was used with the SILVA database (v1.38) to identify and r emov e mitoc hondrial, c hlor oplast, or other nontarget sequences using the commands "classify.seqs"and "r emov e.linea ge" or "get.lineage".Bacterial sequences were classified into zer o-r adius oper ational taxonomic units [zO TUs , also known as amplicon sequence variants (ASVs)] (Edgar 2016, Callahan et al. 2017 ) using the UNOISE3 algorithm implemented in USEARCH (Edgar and Fl yvbjer g 2015 ).
The microDecon (v1.2.0) package in R (v3.5.2) was used to correct the data set for potential contaminants based on zOTU pr e v alence in the samples compared to the mean of the PCR negative control samples (Davis et al. 2018, R Core Team 2018 ).The DNA extr action contr ol was r emov ed fr om the data set as it yielded low sequence numbers and no additional zOTUs compared to the PCR controls.To further eliminate potential contaminants, zOTUs with a r elativ e abundance below 0.1% per sample were removed from the data set (Gloder et al. 2021 ).In this way, all members of the mock community were detected, while contaminating reads w ere discar ded ( Table S3 , Supporting Information ), indicating that the experimental conditions were met to obtain robust data.Finally, in order to correct for uneven sequencing depth among samples, samples were rarefied to an equal number of sequences.Specificall y, endospher e samples were rarefied to 750 reads, while rhizospher e samples wer e r arified to 25 000 r eads.Samples that did not yield enough sequence reads w ere discar ded from further data analysis (12 samples in total).For the rhizosphere, two samples from symptomatic plants of greenhouse 1 and one sample from symptomatic plants of greenhouses 2 and 3 were discarded.For the endosphere, one sample from nonsymptomatic plants of greenhouse 1 and six samples of nonsymptomatic plants of greenhouse 2 w ere discar ded, as w ell as one sample from symptomatic plans of greenhouse 2. The taxonomic origin of each zOTU was determined with the SINTAX algorithm as implemented in USE-ARCH based on the SILVA Living Tr ee Pr oject v123.Further, the identity of the most important zOTUs was verified with a BLAST search in GenBank against type materials .T he BLAST search was extended to the entire database when no significant similarity was found with type materials ( < 97% identity).Sequences obtained in this study were deposited in the Sequence Read Arc hiv e (SRA) at NCBI under Bioproject PRJNA947809.

Quantification of P. cryptogea and bacterial densities
Prior to subjecting the samples to microbiome analysis, P. cryptogea was quantified in all samples investigated using a probe-based qPCR assay with the primers Pcry-F (5 -TGA CGTTGCTGGTTGTGGA GG-3 ) and Pcry-R (5 -GA CA CCCTA CTTCGCA CCA CA-3 ) and the FAM-labeled, double quenched probe Pcry-P (5 -/56-FAM/A TT AAACGC/ZEN/ CGC AGC AGAC AAACC/3IABkFQ/-3 ).T hese primers amplified a section of the internal transcribed spacer (ITS) region of P. cryptogea .T he qPCR assa y was performed in a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.) in a 20 μl reaction volume, containing 1x PrimeTime Gene Expression Master Mix (Integrated DNA Technologies, Inc.), 0.25 nM of each primer, 0.25 μM of the probe, and 2 μl DNA template (10 ng/ μl).The qPCR assay was initiated with an initial denaturation for 3 min at 95 • C, follo w ed b y 40 c ycles of denaturation and annealing/extension for 5 s and 30 s at 95 • C and 60 • C, r espectiv el y, according to the manufacturer's instructions.As a negativ e contr ol, DNA-fr ee w ater w as used instead of DN A template.C T values lo w er than 40 wer e consider ed to be positiv e, whic h was consistentl y lo w er than C T values obtained for blank samples.Evaluation of the specificity of the assay against various fungi and oomycetes, including the target species as well as a number of close r elativ es, r e v ealed that the assay was highly specific under these conditions.Quantification of P. cryptogea DN A w as performed based on a standard curve generated with a 10-fold dilution series of ITS amplicons from P. cryptogea (E = 91.1%;R 2 = 0.999).
In the rhizosphere samples, also the bacterial density was assessed through a SYBR Green-based qPCR assay using unmodified 515F/806R primers to determine the bacterial 16S rRNA gene copy numbers (for details, see Borremans et al. 2019 ).T he qPCR assa y was performed in a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.) in a 20 μl reaction volume, containing 1x SsoAdvanced univ ersal SYBR ® Gr een supermix (Bio-Rad Laboratories, Inc.), 0.20 μM of each primer and 2 μl DNA template (10 ng/ μl).DNA concentrations were determined based on a standard curve established from the analysis of a 10-fold dilution series of 16S rRNA gene amplicons from Bacillus amyloliquefaciens QST713 (E = 87.24%;R 2 = 0.999).A negative control in which template DN A w as replaced b y sterile, DN A free w ater w as included in e v ery qPCR run performed, and a C T value of 40 was taken as the detection thr eshold, whic h w as belo w the C T value obtained for any negati ve control.Gi ven that the 16S rRNA gene primers used may also amplify mitochondrial and chloroplast DNA in plants (King et al. 2012 ; Table S4 , Supporting Information ), it was not possible to determine the bacterial density in the endosphere samples.

Sta tistical anal yses
Statistical anal yses wer e performed using Rstudio 2022.02.2 + 485 (R Core Team 2022) .As the number of r ar efied r eads for rhizospher e and endospher e samples was too different, the data set was split in two subsets, r epr esenting either rhizospher e or endosphere samples, and were analyzed separately.To determine whether the sequencing depth was sufficient to estimate micr obial div ersity, r ar efaction curv es wer e made using the vegan pac ka ge in R, plotting the number of observed zOTUs in function of the number of sequences.Next, the alpha diversity metrics observed zOTU richness (i.e. the number of zOTUs present in a sample) and the Simpson's diversity index, taking into account the number of zOTUs present as well as the r elativ e abundance of each zOTU, were calculated.The alpha diversity metrics and qPCR data wer e statisticall y anal yzed using a Wilcoxon (Mann-Whitney U) test, which can be used for small and unequal sample sizes (Smalheiser 2017, Zhu 2021 ).Further, based on the Hellinger tr ansformed r elativ e abundance data, the bacterial community composition (beta diversity) was visualized by nonmetric multidimensional scaling (NMDS) using the Bray-Curtis coefficient as distance measure in the R software package vegan .To test the hypothesis that infection with P. cryptogea altered the bacterial community composition, i.e. that there were significant differences in bacterial communities between samples from symptomatic and nonsymptomatic plants, a permutational multiv ariate anal ysis of v ariance (perMANOVA) was performed on the same data set.In order to identify zOTUs that were differentially abundant in samples from nonsymptomatic plants compared to symptomatic plants a combination of three methods was used.These included linear discriminant analysis effect size (LEfSe), DESEq2, and EdgeR.Although the latter two methods are commonly used to analyze differential gene expression with RNAseq data, they are also increasingly used in microbiome studies to identify zOTUs that are differentially abundant among conditions (Halfvarson et al. 2017, Jiang et al. 2022 ).Analyses were carried out using the phyloseq pac ka ge, followed by the functions run_lefse, run_deseq2 and run_edger ( microbiomeMarker packa ge).Differ entiall y abundant zOTUs wer e pr esented using volcano plots, with (1) a cutoff of 4 for the fold change, (2) a linear discriminant analysis score cutoff of 2 for LEfSe analysis, and (3) an adjusted P -value of .05.

Results
In order to assess whether bacterial communities in the rhizospher e and r oot endospher e of hydr oponicall y gr own lettuce plants differ between nonsymptomatic and symptomatic plants, both nonsymptomatic and symptomatic lettuce plants were sim ultaneousl y collected fr om differ ent gr eenhouses infested with P. cryptogea .First, using a qPCR assay targeting P. cryptogea DN A, w e confirmed that plants that were classified as symptomatic plants, based on visible symptoms, were infected with P. cryptogea.Gener all y, a significantl y higher number of P. cryptogea ITS copies was detected in samples of symptomatic plants compared to the nonsymptomatic plants for both the rhizosphere (Fig. 1 A) and the endosphere (Fig. 1 B) in each greenhouse, except for the rhizosphere samples of greenhouse 3 and the endosphere samples of greenhouse 2. Ho w ever, also in most nonsymptomatic plants very low amounts of P. cryptogea DNA were detected.
Next, DN A samples w ere subjected to a micr obial comm unity analysis.In total, 2 405 622 r eads wer e obtained for the rhizosphere samples, while 2 712 629 reads were retrieved for the endosphere samples.After quality control, the number of obtained r eads was r educed to 2 267 428 and 1 056 933, r espectiv el y, particularly due to the presence of chloroplast and mitochondrial DNA sequences in the endosphere samples ( Table S4 , Supporting Information ).After decontamination and r ar efying of the data, sequence analysis revealed a total of 633 bacterial zOTUs for the rhizosphere data set and 734 zOTUs for the endosphere data set.Rar efaction curv es r eac hed satur ation (especiall y for rhizospher e samples) or tended to a ppr oac h satur ation ( Figur e S3 , Supporting Information ), indicating that the sequencing depths were sufficient to cover microbial diversity.With regard to alpha diversity, the number of zOTUs per sample ranged from 95 to 159 for the rhizosphere samples and from 23 to 167 zOTUs for the endosphere samples (Fig. 2 A and B).For eac h gr eenhouse, a significantl y higher zOTU richness was found in the rhizosphere samples compared to the endosphere samples ( Figure S4 , Supporting Information ).With regard to the rhizosphere samples, a significantly higher richness was found in the samples of symptomatic compared to the nonsymptomatic plants for the plants from greenhouses 2 and 3, while no difference was found for greenhouse 1 (Fig. 2 A).Ho w e v er, for the endospher e samples, no significant differ ence in zOTU richness was found between the symptomatic and nonsymptomatic plants for none of the greenhouses (Fig. 2 B).In terms of the Simpson Diversity Index, no significant differences were found between symptomatic and nonsymptomatic plants for both the rhizosphere and endosphere, except for a significantly higher Simpson Diversity Index for the rhizosphere samples of symptomatic compared to the nonsymptomatic plants fr om gr eenhouse 3 ( Figur e S5A and B , Supporting Information ).
NMDS plots for rhizosphere (stress = 0.07) and endosphere samples (stress = 0.11) sho w ed a clear clustering between samples from the different greenhouses, especially for the rhizosphere samples (Fig. 3 A).With regard to the rhizosphere, samples from greenhouse 3 were separated from the other greenhouses by NMDS1, while NMDS2 separated samples from greenhouse 1 and greenhouse 2. Additionally, NMDS2 clearly separated samples from nonsymptomatic and symptomatic plants for eac h gr eenhouse.Similar observ ations can be made for the endosphere samples, but separation here is less pronounced (Fig. 3 B).PerMANOVA r e v ealed significant differences in bacterial community composition between plants of the three lettuce gr eenhouses (rhizospher e: F = 36.16,P < .001and endosphere: F = 13.16,P < .001)and across health status (rhizosphere: F = 3.70, P < .01 and endosphere: F = 1.98,P < .05).To estimate the absolute abundance of bacterial cells, a qPCR assay was performed to determine the number of 16S rRNA gene copies (Fig. 4 ).Gener all y, no significant differ ences in gene copy numbers were found between rhizosphere samples from symptomatic and nonsymptomatic plants (Fig. 4 ).Ho w e v er for gr eenhouse 3, a significantly higher number of 16S rRNA gene copies was observed in the nonsymptomatic plants for the rhizosphere, although a lo w er alpha diversity was observed (Figs 2 and 4 ).
When zooming in at the bacterial community composition, a high variability can be observed between the three greenhouses.Ov er all, the bacterial comm unity is dominated by the phylum of Pseudomonadota (Proteobacteria) in both rhizosphere and endosphere samples and both nonsymptomatic and symptomatic plants for all greenhouses (Fig. 5 A and B).Other abundant phyla in the rhizosphere, for both symptomatic and nonsymptomatic plants, ar e Bacter oidota (Bacter oidetes), Actinomycetota (Actinobacteria), Bacillota (Firmicutes), and V errucomicrobiota (V errucomicrobia) (Fig. 5 A).Ho w ever, at greenhouse level, a significantly higher relative abundance of Bacteroidota was found in the symptomatic compared to the nonsymptomatic plants of greenhouse 1 (Fig. 5 A; Figure S6 , Supporting Information ), while a significantl y higher r elativ e abundance of Pseudomonadota and Verrucomicrobiota was observed in the nonsymptomatic compared to the symptomatic plants.For greenhouse 2, a significantly higher relative abundance of Bacteroidota and Bacillota and a significantly lo w er r elativ e abundance of Verrucomicrobiota was observed in the nonsymptomatic compared to the symptomatic plants.For greenhouse 3, a significantly higher relative abundance of Actinomycetota was found in the symptomatic compared to the nonsymptomatic plants, while a significantly lo w er r elativ e abundance of Pseudomonadota and Verrucomicrobiota was found in these plants.Regarding the endosphere, the bacterial communities were also dominated by Pseudomonadota, Figur e 2. zO TU of bacterial communities of the rhizosphere (A) and root endosphere (B) of lettuce plants collected at three hydroponic lettuce greenhouses for both nonsymptomatic (NS) and symptomatic (S) plants.Results are presented for the three greenhouses separately.The lo w er, middle and upper lines of the boxplots correspond to the first quartile, median and third quartile, r espectiv el y, while the whiskers r epr esent the range from the minimum to the maximum.Data points represent the different replicates (number provided between brackets).Significant differences between symptomatic and nonsymptomatic plants are indicated by one or more asterisks ( P > .05(ns), P ≤ .05( * ), and P ≤ .01 ( * * )).
F igure 3. NMDS or dination plots based on r elativ e abundance data of the bacterial comm unities in the rhizospher e (str ess = 0.07) (A) and endospher e (stress = 0.11) (B) of lettuce root samples taken at three hydroponic lettuce greenhouses.Samples were taken from both nonsymptomatic and symptomatic plants .T he closer two data points , the mor e similar the bacterial comm unities ar e. Bacteroidota, Bacillota, and Actinomycetota, although mor e v ariability was present between the greenhouses (Fig. 5 B).In greenhouse 2, one of the most abundant phyla was Planctomycetota (Planctomycetes), which was mainly absent in greenhouses 1 and 3. Significant differences between symptomatic and nonsymptomatic plants at gr eenhouse le v el wer e onl y found in greenhouse 1, wher e a significantl y higher r elativ e abundance of Bacter oidota was found in the symptomatic compared to the nonsymptomatic plants.
To identify bacterial zOTUs linked to nonsymptomatic and/or symptomatic plants, a differential abundancy analysis was performed using DESeq2, EdgeR, and LEfSe anal ysis ( Figur es S7 -S9 , Supporting Information ).The differential abundant zOTUs shared between the thr ee anal ysis methods and with an ov er all r elativ e abundance lar ger than 5% for at least one greenhouse and nonsymptomatic/symptomatic samples are summarized in a relativ e abundance-pr e v alence matrix (Fig. 6 ).This matrix shows for eac h differ entiall y abundant zOTU its r elativ e abundance (%) and its pr e v alence (color shade).Gener all y, when considering the rhizospher e, differ entiall y abundant zOTUs sho w ed higher relative abundance in nonsymptomatic compared to symptomatic plants (Fig. 6 A).In greenhouses 2 and 3, a member of the Comamonadaceae family (zOTU1) and a zOTU identified as Aeromonas sp. ( zOTU18) wer e significantl y mor e abundant in nonsymptomatic compared to symptomatic plants (4.6% vs. 2.5% and 2.1% vs. 0.4%, r espectiv el y, in gr eenhouse 2; 30.6% vs. 18.5% and 6.6% vs. 4.1%, r espectiv el y, in gr eenhouse 3).Similarl y, in gr eenhouses 1 and 2, a Sphingobium sp.(zOTU2) and Flavobacterium sp.(zOTU11) sho w ed significantly higher r elativ e abundance in nonsymptomatic plants (6.1% vs. 3.4% and 6.0% vs. 1.9%, r espectiv el y, in greenhouse 1; 14.4% vs. 6.5% and 1.6% vs. 0.8%, r espectiv el y, in greenhouse 2).An unknown bacterium (zOTU15) was significantly more abundant in nonsymptomatic compared to symptomatic plants in greenhouses 1 and 3 (6.9% vs. 0.8% in greenhouse 1; 2.7% vs. 1.3% in greenhouse 3, respectively).On the contrary, two zO TUs (zO TU4 and zO TU10) wer e significantl y mor e abundant in the symptomatic compared to the nonsymptomatic plants and were identified as Pseudomonas sp. and Flavobacterium sp., respectiv el y.T he zO TU identified as Flavobacterium sp.(zO TU10) was significantl y mor e abundant in symptomatic compared to nonsymptomatic plants in all three greenhouses (16.1% vs. 1.1% in greenhouse 1; 1.9% vs. 0.1% in greenhouse 2; and 0.4% vs. 0.0% in The lo w er, mid dle and upper lines of the bo xplots correspond to the first quartile, median and third quartile, r espectiv el y, while the whiskers r epr esent the range from the minimum to the maximum.Data points r epr esent the different replicates (number provided between brackets).Significant differences are indicated with one or more asterisks ( P > .05(ns), P ≤ .05( * ), P ≤ .01 ( * * ), and P ≤ .001( * * * )).gr eenhouse 3).Like wise , the Pseudomonas sp.(zO TU4) sho w ed significantl y higher r elativ e abundance in symptomatic plants fr om greenhouses 1 and 2 compared to nonsymptomatic plants (1% vs. 0.6% in greenhouse 1; 19.9% vs. 8.6% in greenhouse 2).BLAST analysis r e v ealed that this Pseudomonas species most likel y belongs to the Pseudomonas fluorescens or Pseudomonas putida group.
Inter estingl y, most zOTUs differ entiall y abundant in the rhizospher e wer e also found to be differ entiall y abundant in the endosphere (Fig. 6 A and B).With regard to the endosphere, the zOTU belonging to the Comamonadaceae family (zOTU1) was significantl y mor e abundant in the nonsymptomatic compared to the symptomatic plants in both greenhouse 1 (5.0% vs. 1.5%, respectiv el y) and gr eenhouse 3 (3.3% vs. 0.1%, r espectiv el y).Also zOTU2, identified as Sphingobium sp., sho w ed a significantly higher relative abundance in nonsymptomatic compared to symptomatic plants in greenhouse 2 (6.9% vs. 1.8%, r espectiv el y), as well as an unknown bacterium (zOTU15) in greenhouse 3 (7.6% vs. 1.0%, respectiv el y).On the contrary, the other differentially abundant zO-TUs wer e significantl y mor e abundant in the symptomatic compared to the nonsymptomatic plants.Two of them were identified as Flavobacterium sp.(zOTU10 and zOTU30), of which zOTU10 was differ entiall y abundant in both greenhouse 1 (11.2% vs. 0.7%, r espectiv el y) and gr eenhouse 2 (7.1% vs. 0.3%, r espectiv el y), while zOTU30 was onl y differ entiall y abundant in greenhouse 2 (20.1% vs. 1.4%, r espectiv el y).Inter estingl y, zOTU10 sho w ed also significant higher r elativ e abundance in the rhizosphere of the symptomatic plants in all three greenhouses.Also a Duganella sp.(zOTU19) was significantly more abundant in the symptomatic compared to nonsymptomatic plants in greenhouse 1 (15.4 vs. 0.5%, r espectiv el y).

Discussion
Although micr obiome r esearc h has gained consider able attention in recent years, studies that examine the root microbiome of hydr oponicall y gr own cr ops ar e still limited, despite its importance in modern a gricultur e. Specificall y, ther e is limited knowledge re-garding the impact of pathogens on the microbial community in hydroponic systems.To fill this r esearc h ga p, we conducted a study on the interaction between hydroponically grown lettuce and P. cryptogea .Our objective was to investigate how P. cryptogea influences the bacterial community in the rhizosphere and endosphere of lettuce plants grown in hydroponic systems.We sampled both symptomatic and nonsymptomatic plants from differ ent gr eenhouses to ensur e a compr ehensiv e r epr esentation of the real-w orld scenario.Ho w ever, it is important to note that, although plants without typical symptoms of P. cryptogea were considered nonsymptomatic plants, they could have been infected just before sampling.This is supported by the low le v els of P. cryptogea DNA detected in most of the nonsymptomatic plants.We noticed that nonsymptomatic plants wer e pr edominantl y located at the beginning of the gutters, while symptomatic plants were mor e commonl y found to w ar ds the end.It is highl y pr obable that a disparity in oxygen concentration along the gutters exist, with higher le v els at the beginning and lo w er le v els to w ar d the end.Higher oxygen le v els hav e a beneficial impact on plant growth, but can also influence bacterial communities and pathogen de v elopment (Chérif et al. 1997, Suyantohadi et al. 2010, Martínez-Arias et al. 2022 ).
Although microbial communities in the endosphere have long been ov erlooked r egarding their r ole in plant health, endophytes hav e r ecentl y r eceiv ed incr easing attention because of their intimate interaction with plants (Compant et al. 2019(Compant et al. , 2021 ) ).Our results show that P. cryptogea can reach high densities in the endosphere following root infection.Nevertheless, its impact on the micr obial comm unity was less pr onounced in the endospher e compared to the rhizosphere.With regard to alpha diversity, no significant differences were found between the symptomatic and nonsymptomatic plants for the endosphere samples, although it w as sho wn earlier that the endophytic microbiome can be affected by pathogen infection (Proença et al. 2017, Suhaimi et al. 2017, Kaushal et al. 2020 ).For example, a lo w er OTU richness and diversity have been detected in the endosphere of healthy banana plants compared to those infected with F. oxysporum f. sp.cubense in the same field (Kaushal et al. 2020 ).Ho w e v er, it is important to highlight that in our study the r ar efaction curv es for the endosphere samples did not reach saturation, suggesting that alpha diversity may be underestimated and/or the impact of the pathogen might not be fully captured.This was particularly caused by the lack of specificity of the 16S rRNA gene primers, leading to an ov er amplification of mitoc hondrial and c hlor oplast plant DNA.Likewise, it was not possible to determine the bacterial densities in the endosphere samples by qPCR due to the nonspecificity of the primers .T his can possibly be circumvented by using more specific primers or by adding peptide-nucleic acid PCR clamps, although the efficacy of both methods depends on the plant species (Lundberg et al. 2013, Mori et al. 2014, Thijs et al. 2017, Fitzpatrick et al. 2018 ).Regarding the rhizosphere samples, an increased alpha diversity was found for samples of symptomatic compared to nonsymptomatic plants, indicated by a significantly higher number of bacterial zOTUs in samples from P. cryptogea infected plants .Although we did not c hec k for differences in root exudate composition and/or fatty acids of healthy and diseased plants in this study, it is highly probable that infection by P. cryptogea leads to a significant alteration in the composition of root exudates .T his phenomenon is well-documented in pr e vious studies exploring pathogen-induced c hanges in r oot exudation.Consequentl y, these alter ed r oot exudates can act as signals and attract plant-associated micr oor ganisms in an attempt to overcome the infection, also known as the so-called  for at least one greenhouse and symptomatic/nonsymptomatic plants.Results are presented for the three greenhouses separately for both the rhizosphere (A) and the endosphere (B).For each zOTU, the relative abundance is given as a number (%), while the color represents prevalence (i.e.fraction of samples in which the zOTU was present; white is absent).The number of replicates included is given between brackets .zO TUs are identified by a BLAST search against type materials in GenBank.When no significant similarity was found with type materials ( < 97%), the BLAST analysis was performed against the entire GenBank (indicated with an asterisk).Identifications were performed at genus level; when identical scores were obtained for different genera, identifications were performed at family level.
"cry-for-help" strategy (Lombardi et al. 2018, Yuan et al. 2018, Rolfe et al. 2019 ).Accordingl y, se v er al studies hav e shown that pathogen inv asion incr eases micr obial div ersity in the rhizosphere.For example, the rhizosphere of hydroponically grown tomato plants infected with rhizogenic a gr obacteria exhibited higher zOTU ric hness compared to their healthy counterparts (Vargas et al. 2022 ).Also a higher diversity of Gamma pr oteobacteria was found in the rhizospher e and phyllospher e of pot-cultiv ated lettuce plants infested by R. solani compared to healthy lettuce plants (Erlacher et al. 2014 ).Ho w e v er, se v er al studies hav e also r eported the opposite phenomenon, wher ein ther e is a decr ease in micr obial div ersity in the rhizosphere following invasion of a pathogen (Trivedi et al. 2012, Zhang et al. 2017, Wei et al. 2018 ).
In terms of beta diversity, significant differences were found between samples of nonsymptomatic and symptomatic plants for eac h gr eenhouse.Similarl y, Var gas et al. ( 2022) observ ed significant differences in bacterial communities between healthy and rhizogenic a gr obacterium-infected tomato plants in hydr oponics.In soil-bound cultivation, a distinct gamma pr oteobacterial com-munity was found in nonsymptomatic lettuce plants compared to plants infected by R. solani (Erlacher et al. 2014 ).Infestation of R. solanacearum in soil-grown tomato plants has also led to changes in the rhizosphere microbiome composition (Wei et al. 2018 ).Also upon artificial inoculation of pathogens, changes in the microbiome have been observed.For example, infection with Verticillium dahliae clearl y c hanged the tomato r oot micr obiome compar ed to mock-inoculated tomato plants (Snelders et al. 2020 ).
Apart fr om differ ences between symptomatic and nonsymptomatic plants, major significant differences in bacterial communities were also found between greenhouses for both the rhizospher e and endospher e samples.Considering the v ariability in cultiv ation pr actices, differ ences in beta div ersity wer e not unexpected.For instance, different lettuce varieties (butterhead lettuce vs. multicolor lettuce) and different cultivars were grown.Previous r esearc h suggests that cultivars may have a significant impact on micr obial comm unities in the rhizospher e, as e videnced for example in Brassica napus , where two different cultivars sho w ed r emarkable differ ences in the endophytic bacterial populations and total microbial load, even in the seed stage (Granér et al. 2003 ).Although the substrate, fertigation and climatic conditions were the same among the gr eenhouses, differ ences in the use of plant pr otection pr oducts, water disinfection str ategies , or en vironmental conditions might have affected the composition of the microbial r oot comm unity as well, as demonstr ated in pr e vious studies (Vallance et al. 2011, Sang iorg io et al. 2022 ).For example, it was shown that the use of Teldor WG50 ® (activ e ingr edient fenhexamid) against Botrytis cinerea in the nutrient solution of an hydroponic tomato cultivation system had an influence on the microbial community composition (Alsanius et al. 2013 ).While no information is available on the use of plant protection products in greenhouses 2 and 3, chemical pesticides against downy mildew wer e spr ayed in greenhouse 1.Also, the microbial load and/or microbial composition of the nutrient solution ma y ha ve been different, as samples were taken at different moments during the growing season (June and August, 2021) and each greenhouse utilized r ainwater fr om their gr eenhouse basin for the nutrient solution.
Despite the variability among gr eenhouses, a differ ential abundance analysis revealed several zO TUs , which were differentially abundant between the rhizosphere of nonsymptomatic and symptomatic plants in at least two out of three greenhouses .T hese included members of Pseudomonas and Flavobacterium , whic h wer e particularl y associated with symptomatic plants.Although certain Pseudomonas spp.could be pathogenic for plants, BLAST analysis of the zOTU associated with symptomatic plants identified as Pseudomonas sp.sho w ed that it most likely belongs to the P. fluorescens or P. putida group.These Pseudomonas spp.are w ell-kno wn for their biocontrol potential and are often r eferr ed to as plant gr owth-pr omoting rhizobacteria (PGPR).They can exhibit both direct and indirect effects to promote plant health.As direct effect, Pseudomonas spp.can produce bacterial alleloc hemicals, suc h as antibiotics and sider ophor es, while indir ectl y they can promote plant health through IR.The production of bacterial allelochemicals by Pseudomonas spp.and other PGPR can inhibit the growth of plant pathogens and provide a competiti ve ad vantage to the plant (Santo y o et al. 2012, Dorjey et al. 2017 ).Furthermor e, the induction of systemic r esistance by PGPR can activate the plant's defense mechanisms and provide long-term protection against pathogen attacks (Choudhary et al. 2007 ).For instance, Pseudomonas strains isolated from both the phyllospher e and rhizospher e of potato plants hav e been found to exhibit antagonistic potential against Phytophthora infestans and other potato pathogens when cocultivated in the lab (Guyer et al. 2015, Hunziker et al. 2015 ).These Pseudomonas strains produce volatiles such as 1-undecene, which inhibit mycelial growth, as well as zoospore germination and release (Hunziker et al. 2015 ).Also, it w as sho wn that some Pseudomonas corrugata strains exhibit biocontr ol activity a gainst Phytophthora blight of pepper (caused b y Ph ytophthora capsici ) through successful colonization of plant roots (Sang and Kim 2014 ).BLAST analysis of the zOTUs identified as Flavobacterium sp. was less conclusive about the species identity and consequently their characteristics.Ho w ever, to our knowledge, no Flavobacterium spp. with plant pathogenic c har acteristics are known, while their plant-beneficial effects (e.g.plant gr owth-pr omoting c har acteristics and antimicr obial activity) ar e well-documented (Alexander and Stewart 2001, Sang et al. 2008, Soltani et al. 2010, Kolton et al. 2014, Kwak et al. 2018, Carrión et al. 2019 ).The endophytic volatile-producing strain Flavobacterium johnsoniae GSE09, isolated and c har acterized fr om surfacesterilized roots of pepper plants, has demonstrated the ability to colonize pepper roots as well as the rhizospher e, r esulting in r educed colonization by Phytophthora capsici , and consequently a di-minished disease se v erity in pepper plants (Sang et al. 2008 , Sang andKim 2012 ).Mor eov er, it was found that Flavobacterium spp.are often highly abundant in the plant rhizosphere, where they can colonize plant roots and have a role in increasing the plant imm une r esponse (K olton et al. 2014 ).T hese c har acteristics make Pseudomonas and Flavobacterium spp.promising candidates for biocontr ol str ategies in a gricultur e, including biocontr ol of P. cryptogea .Inter estingl y, we found that these bacteria were particularly associated with P. cryptogea -infected plants, in line with the "cryfor-help" hypothesis (Bakker et al. 2018, Rolfe et al. 2019, Rizaludin et al. 2021 ).Ther efor e, giv en their known biocontrol potential and natur al occurr ence in the hydr oponic cultiv ation of lettuce, we hypothesize that the Pseudomonas sp. and Flavobacterium spp.identified in this study could be explored as biocontr ol or ganism (BCO) a gainst P. cryptogea.Pr e vious studies hav e demonstr ated that both rhizospher e micr oor ganisms and endophytes can r estrict infection b y Ph ytophthora spp.This illustrates the po w er of isolating plant beneficial microbes from the plant-associated microbiome as a potential source of BCOs (Arnold et al. 2003, Abraham et al. 2013, Acebo-Guerr er o et al. 2015, Islam et al. 2016, de Vries et al. 2018, Xi et al. 2022 ).For instance, pr etr eatment of cacao tree ( Theobroma cacao ) roots with Pseudomonas c hlororaphis , whic h was isolated from the rhizosphere of the tree, was found to reduce symptom se v erity follo wing Ph ytophthor a palmivor a inoculation (Acebo-Guerr er o et al. 2015 ).Mor eov er, inoculation of cacao leaves with fungal endophytes, isolated from healthy cacao trees, sho w ed a significant decrease in leaf mortality and necrosis after inoculation with Phytophthora sp.(Arnold et al. 2003 ).Ho w e v er, further r esearch is needed to explore the potential of the identified species as BCO against P. cryptogea infection in hydroponically grown lettuce crops.
Altogether, our study has clearly shown that the bacterial community composition of hydroponically grown lettuce is largely dependent on the investigated greenhouse.Further, our study indicates that P. cryptogea infection has a strong impact on the composition of the bacterial community in the rhizosphere and endospher e. Differ ential abundance anal ysis r e v ealed Pseudomonas spp.and Flavobacterium spp. as potentially important genera associated with symptomatic plants.Further r esearc h is needed to further explore their potential as BCOs.

Figure 1 .
Figure 1.Number of P. cryptogea ITS copies per μl DNA (log scale) present in roots of nonsymptomatic (NS) and symptomatic (S) plants, for both the rhizosphere (A) and root endosphere (B).Results are presented for the three greenhouses separately.The lo w er, middle and upper lines of the boxplots correspond to the first quartile, median and third quartile, r espectiv el y, while the whiskers r epr esent the range from the minimum to the maximum.Data points r epr esent the different replicates (number provided between brackets).Significant differences between symptomatic and nonsymptomatic plants are shown by one or more asterisks ( P > .05(ns), P ≤ .05( * ), P ≤ .01 ( * * ), and P ≤ .001( * * * )).

Figure 4 .
Figure 4. Number of 16S rRNA gene copies per μl DNA (log scale) present in roots of nonsymptomatic (NS) and symptomatic (S) lettuce plants, collected at three hydroponic lettuce greenhouses, for the rhizospher e. Results ar e pr esented for the thr ee gr eenhouses separ atel y.The lo w er, mid dle and upper lines of the bo xplots correspond to the first quartile, median and third quartile, r espectiv el y, while the whiskers r epr esent the range from the minimum to the maximum.Data points r epr esent the different replicates (number provided between brackets).Significant differences are indicated with one or more asterisks ( P > .05(ns), P ≤ .05( * ), P ≤ .01 ( * * ), and P ≤ .001( * * * )).

Figure 5 .
Figure 5. Phylum composition in the rhizosphere (A) and endosphere (B) of nonsymptomatic (NS) and symptomatic (S) lettuce plants collected at three hydroponic lettuce greenhouses.Results are presented for the three greenhouses separately.The number of replicates included is given between br ac kets.

Figure 6 .
Figure 6.Summary of the r elativ e abundance and pr e v alence of the zOTUs, which are differentially abundant between nonsymptomatic (NS) and symptomatic (S) plants according to the combined results of the DESeq2, EdgeR, and LEfSe analyses and with an overall relative abundance of ≥ 5% for at least one greenhouse and symptomatic/nonsymptomatic plants.Results are presented for the three greenhouses separately for both the rhizosphere (A) and the endosphere (B).For each zOTU, the relative abundance is given as a number (%), while the color represents prevalence (i.e.fraction of samples in which the zOTU was present; white is absent).The number of replicates included is given between brackets .zO TUs are identified by a BLAST search against type materials in GenBank.When no significant similarity was found with type materials ( < 97%), the BLAST analysis was performed against the entire GenBank (indicated with an asterisk).Identifications were performed at genus level; when identical scores were obtained for different genera, identifications were performed at family level.