Differential dynamics of microbial community networks help identify microorganisms interacting with residue-borne pathogens: the case of Zymoseptoria tritici in wheat

Background Wheat residues are a crucial determinant of the epidemiology of Septoria tritici blotch, as they support the sexual reproduction of the causal agent Zymoseptoria tritici. We aimed to characterize the effect of infection with this fungal pathogen on the microbial communities present on wheat residues and to identify microorganisms interacting with it. We used metabarcoding to characterize the microbiome associated with wheat residues placed outdoors, with and without preliminary Z. tritici inoculation, comparing the first set of residues in contact with the soil and a second set without contact with the soil, on four sampling dates in two consecutive years. Results The diversity of the tested conditions, leading to the establishment of different microbial communities according to the origins of the constitutive taxa (plant only, or plant and soil), highlighted the effect of Z. tritici on the wheat residue microbiome. Several microorganisms were affected by Z. tritici infection, even after the disappearance of the pathogen. Linear discriminant analyses and ecological network analyses were combined to describe the communities affected by the infection. The number of fungi and bacteria promoted or inhibited by inoculation with Z. tritici decreased over time and was smaller for residues in contact with the soil. The interactions between the pathogen and other microorganisms appeared to be mostly indirect, despite the strong position of the pathogen as a keystone taxon in networks. Direct interactions with other members of the communities mostly involved fungi, including other wheat pathogens. Our results provide essential information about the alterations to the microbial community in wheat residues induced by the mere presence of a fungal pathogen, and vice versa. Species already described as beneficial or biocontrol agents were found to be affected by pathogen inoculation. Conclusions The strategy developed here can be viewed as a proof-of-concept focusing on crop residues as a particularly rich ecological compartment, with a high diversity of fungal and bacterial taxa originating from both the plant and soil compartments, and for Z. tritici-wheat as a model pathosystem. By revealing putative antagonistic interactions, this study paves the way for improving the biological control of residue-borne diseases. Electronic supplementary material The online version of this article (10.1186/s40168-019-0736-0) contains supplementary material, which is available to authorized users.

one left outdoors in contact with the soil, and the other left outside but not in contact with the soil, at different sampling dates during two consecutive years. The diversity of experimental conditions was expected to lead to the establishment of different microbial communities according to the origin of the constitutive taxa (plant or soil), thereby increasing the probability of detecting effects of Z. tritici on the residue microbiome, and of the residue microbiome on Z. tritici.

Results
Overall diversity of the bacterial and fungal communities on residues The response of the residue microbiome to Z. tritici inoculation was assessed by analyzing the composition of the fungal and bacterial communities of wheat residues, after inoculation with Z. tritici (n=240) or in the absence of inoculation (n=240). We also investigated the impact of cropping season (n=2), season (n=4), and soil contact (n=2) on the dynamics of these communities (see materials and methods for a detailed explanation of the experimental design; Figure 1).
We investigated the structure of the residue microbiome by analyzing the v4 region of the 16S rRNA gene and ITS1. Overall, 996 bacterial amplicon sequence variants (ASVs) and 520 fungal ASVs were obtained from 390 and 420 samples, respectively. Some samples (July 2016) were removed from the analysis due to the co-amplification of chloroplasts.
The high relative abundance (RA) of ASVs affiliated to Zymoseptoria in samples collected in July 2016 (21.59.8%) and 2017 (30.37.1%) highlights successful colonization of the wheat tissues by this pathogen following inoculation ( Figure 2). However, the RA of Zymoseptoria rapidly decreased to 21.64% and 1.40.9% on residues not in contact with the soil (above ground residues) collected in October 2016 and 2017, respectively, and this species was below the limit of detection in December and February. For residues in contact with soil, this decrease occurred more rapidly, with Zymoseptoria ASV already undetectable in samples collected in October.
Alpha diversity, estimated with the Shannon index, was low in July for both bacterial (2.700.75) and fungal communities (1.820.19; Suppl. Figure 1). A gradual increase was then observed during residue degradation. Z. tritici inoculation had no impact on bacterial alphadiversity, but decreased fungal diversity (Kruskal-Wallis: p = 0.008). More specifically, bacterial diversity was higher in inoculated residue samples in July 2017 (2.920.80 for inoculated samples versus 2.470.6 for non-inoculated samples; Wilcoxon: p = 0.022), but no such difference was detected for the other sampling dates. Conversely, for fungal communities, inoculation had no effect in July, but led to a significant decrease in diversity in subsequent months during the second cropping season (October and December 2017, for the two soil contact conditions).
Beta diversity analysis (Bray-Curtis index) showed large dissimilarities between bacterial community composition in July and at the other sampling dates, as illustrated in the hierarchical clustering of the samples, justifying separate analyses and MDS representations ( Figure 3).
Rhizobium, Nocardioides, Pseudomonas, and Sphingomonas were more abundant in residues in contact with the soil, whereas Cladosporium, Massilia, Paracoccus, Stagonospora and Cryptococcus were more abundant in above ground residues.

Impact of Z. tritici inoculation on microbial communities
The influence of Z. tritici inoculation on the RA of residue microbiome members was assessed, through LDA scores. In total, the RA of 115 ASVs (74 bacterial ASVs and 41 fungal ASVs) was significantly affected by Z. tritici inoculation, for at least one sampling date (listed in Suppl. Figure 3). The effect of inoculation on microbial communities persisted throughout the experiment, despite the absence of Zymoseptoria detection from December onwards ( Figure 2). ASVs with significant differences in RA decreased over time for residues in contact with the soil (Suppl. Table 1). By contrast, for above ground residues, the number of Inoculation with Z. tritici decreased the RA of fungal ASVs, including those affiliated to Sarocladium, Gibellulopsis and Blumeria, and increased the RA of bacterial ASVs affiliated to Curtobacterium and Brachybacterium (listed in Suppl. Figure 3). The ASVs affected by inoculation differed between above ground residues and residues in contact with soil. The pattern of change (i.e. promoted or inhibited by inoculation) was always the same within a given year, regardless of soil contact conditions. For example, Brachybacterium and Curtobacterium were promoted by inoculation, in both soil contact conditions, whereas Sarocladium was inhibited by inoculation, in both soil contact conditions. Impact of the actual presence of Z. tritici on microbial communities Ecological network analyses (ENA) combining bacterial and fungal datasets were performed to predict the potential interactions between Z. tritici and members of microbial communities associated with wheat residues.
Dynamics of ecological interaction networks -The dataset was split according to the effects previously described (cropping season, seasonality, soil contact conditions). Six ENA were performed per experimental year, corresponding to residue samples in contact with the soil and above ground residues, collected in October, December, and February ( Figure 5). The networks for July are presented in Suppl. Figure 4. The mean number of nodes in the network (205.347.5) increased over the season (Suppl . Table 1). Overall, networks were sparse, with a mean node degree of 2.760.43. For each network, the positive/negative edge ratio decreased over time, reaching 1.0-1.5 in February. Most nodes were common to October, December and February. Zymoseptoria ASV was one of the fungal ASV with the largest number of degrees and greatest betweenness (measurement of centrality in a graph based on the shortest paths) for above ground samples in October. By contrast, for samples in contact with soil, it was absent the first year and had low betweenness and degree values for the second year ( Figure 6). Microorganisms with the same differential pattern (i.e. "promoted by inoculation" or "promoted in the absence of inoculation") did not interact negatively with each other in networks. Conversely, microorganisms with opposite differential patterns systematically interacted negatively with each other. These results highlight the consistency of the LDA and ENA approaches.
The subnetworks generated with microorganisms presenting differential relative abundances and their adjacent nodes were strongly connected: each subnetwork consisted of a principal component and, in some cases, smaller components of less than four nodes ( Figure   7).

Discussion
By sequencing the microbial communities of 420 samples of wheat residues, we obtained a total of 996 bacterial ASVs and 520 fungal ASVs. Using this large dataset, we estimated the potential interactions occurring between a plant pathogen (Z. tritici) and the members of microbial communities associated with crop residues in field conditions. By combining two approaches -LDA and ENA -we were able to demonstrate an effect of pathogen infection, even after disappearance of the pathogen, on the structure and composition of the microbial communities during residue degradation.

Effect of soil contact on microbial communities
Our aim here was not to characterize the organisms colonizing wheat residues, but our findings nevertheless highlight major changes in the microbial community over time for residues in contact with soil. The taxa favored in above ground residues, such as Cladosporium, Alternaria, Pedobacter and Massilia, were already present on the plant. This is consistent with previous findings showing a decrease in the abundance of these plant-associated taxa during the degradation of residues in contact with soil and the colonization of these residues with soilborne competitors, such as Chaetomium, Torula, and Nocardioides [16]. Some fungal genera not present in July were favored by above ground conditions (e.g. Cryptococcus, Stagonospora, and Myrmecridium). This finding is consistent with our knowledge of fungal dispersal processes, mostly involving aerial spores.
Decline of Z. tritici during residue degradation Z. tritici rapidly decreased to below the limit of detection between October and December.
This result is surprising in light of the quantitative epidemiological data acquired for the same plot, which suggested that Z. tritici ascospores may be ejected from residues until March [3,37].
The observed decline of Z. tritici may be due to lower levels of contamination of adult wheat plants in residues than would be achieved in the field after natural infection. Indeed, in field conditions, Z. tritici establishes itself on all parts of the plant (leaves, but also sheaths and stems) through multiple secondary infections, driven by the repeated splash dispersal of asexual spores leading to an accumulation of contaminating raindrops at the points of insertion of the leaf sheaths. The single inoculation event in the greenhouse resulted in contamination principally of the leaves, the organs most exposed to spraying, with relatively little contamination of the stems and sheaths, the parts of the plant most resistant to degradation. Indeed, the results of a previous study [16] support this hypothesis: in the same field, during the same season, Z. tritici was detected in wheat residues originating from plants grown in natural conditions until February, and even May, with a similar metabarcoding approach.

Effect of Z. tritici on microbial communities
Endophytes and pathogens induce changes in plant tissues (e.g. necrosis), which may themselves modify the microbial communities inhabiting the plant (e.g. impact of secondary saprophytes or opportunistic pathogens [38]; selection of microorganisms by secondary metabolites produced by microorganisms or the plant [39,40]). This general phenomenon may explain the impact of Z. tritici on the microbial communities observed in both LDA and ENA.
The impact of Z. tritici on residues, even after its decline between October and December, persisted until February, particularly for fungal communities. Within microbial networks, Z.
tritici was one of the keystone taxa, despite its low abundance, in above ground residues in October (Suppl. Figure 5). The high levels of Zymoseptoria in July (between 10 and 40% of reads) account for its central position in the network. The number of microorganisms displaying changes in abundance due to Z. tritici inoculation decreased during residue degradation. This finding highlights the resilience of the community (i.e. its ability to return to its original composition after a disturbance, in this case, Z. tritici inoculation) [41].

Specific interactions with Z. tritici
Most of the predicted interactions with Zymoseptoria involved fungi, such as Fusarium, Blumeria or Cladosporium. Z. tritici infection has been shown to be associated with the accumulation of H2O2 [42]. This compound is known to inhibit biotroph fungal pathogens [43], such as Blumeria graminis [44,45]. This may explain the negative interaction between Z. tritici and B. graminis in July and October 2017-2018. In addition, Z. tritici infection induces leaf necrosis, potentially decreasing wheat susceptibility to B. graminis, due to a significant physiological interaction during the latent, endophytic period of Z. tritici development [45].
H2O2 is also known to promote necrotrophic agents, such as Fusarium. We detected both positive and negative interactions between Zymoseptoria and Fusarium, depending on the ASV considered. On adult wheat plants, such differential interactions have been demonstrated in loglinear analyses [46], with both species giving positive results on stem bases and negative results on the upper parts of stems. Positive interactions between Z. tritici and Cladosporium have also been demonstrated on adult plants [46], consistent with our findings for wheat residues.
Although the use of ENA based on bacterial and fungal data sets can introduce many biases (distortion of the microbial community composition due to analysis by separate PCRs, inherent limitations in terms of resolution of the taxonomic markers, etc.), these results lend a biological meaning to the interactions detected, confirming the relevance of network analyses for highlighting ecological interactions within crop residue communities.
Trichoderma was more abundant in residues from wheat plants inoculated with Z. tritici (July 2016), as shown by LDA (Suppl. Figure 4). Conversely, Epicoccum and Cryptococcus were more abundant in residues from non-inoculated wheat plants (October 2016). The overabundance of those taxa, described as biocontrol agents in previous studies [34][35][36]47], was influenced by the presence of the pathogen. However, no direct interactions between Z.
tritici and these species could be established. This exemplifies the difficulties highlighting beneficial species within complete microbial communities. These difficulties are not specific to the residue compartment and also apply to the spermosphere [48], phyllosphere [49] and rhizosphere compartments [14,50].

Other interactions
Other interactions between ASVs highlighted in the network analysis were examined in light of published results for fungal pathogens of cereals. For instance, it has already been shown that B. graminis growth on barley is inhibited by Trichoderma harzianum [51] and Stagonospora norodum [52], that Stenetrophomas maltophila attenuates the seedling blight of wheat caused by F. graminearum [53], that Acremonium zeae has antibiotic activity against Fusarium verticillioides [54], and that Chaetomium sp. produces compounds (e.g. chaetomin) active against Alternaria triticimaculans [34]. Conversely, certain non-pathogenic bacteria were shown to be associated with significantly more disease on wheat caused by B. graminis and Z. tritici and to "help" Phaeosphaeria nodorum to infect wheat tissues [55]. Newtoon et al.
[38] has proposed the hypothesis of "induced susceptibility" to explain such an interaction between bacteria and biotroph fungal pathogens.
ENA also suggested that intra-kingdom interactions were favoured over inter-kingdom interactions in certain conditions (Suppl. Table 2). This may reflect differences in ecological niches and dynamics, as illustrated by the temporal changes in microbial communities over a season, with a densification of the networks during residue degradation. Further investigations are required to determine whether inter-or intra-kingdom interactions are more intense, and thus more promising for use in biocontrol engineering. Should we preferentially focus on fungal communities to improve the management of a fungal disease, and on bacterial communities to improve the management of a bacterial disease? The ability to answer this question with the approach developed in this study should be nuanced. Indeed, the weakness associated with separate analysis of fungal and bacterial communities (see above) may have impacted our observation that intra-kingdom interactions were more difficult to discern that inter-kingdom interactions (see below), and may increase the difficulty of identifying actual biological interactions between bacteria and fungi.

Identification of beneficial species, and potential biocontrol agents
Network models provide new opportunities for enhancing disease management and can be helpful for biocontrol. Our study, combining LDA and ENA based on a metabarcoding approach and differential conditions (plants inoculated with a pathogen or left non-inoculated; plant residues in contact with soil vs. residues not in contact with the soil), fits into the framework described by Poudel et al. [56], which considers several types of network analyses, including pathogen-focused analyses, taking into account diseased and healthy plant hosts, with a view to elucidating direct and indirect pathogen-focused interactions within the pathobiome.
Network analyses revealed no significant direct interactions between Z. tritici and microorganisms reported to be useful biocontrol agents. However, pathogen infection had a strong effect on the entire microbial community present in residues during the course of their degradation. Most of the interactions were difficult to interpret. Several interactions appeared to be transient, changing over time with residue degradation, and their presence or absence depended on whether the residues were in contact with the soil. This suggests that interactions between microorganisms are not stable and can be modified by changes in the environment, for example, or by the arrival of a new microorganism.
Network models, although effective in characterizing putative interactions between ASVs within a microbial community and highlighting changes due to disturbance (e.g. presence of a pathogen, application of fungicides, introduction of a resistance gene in a host plant population, etc.), do not necessarily allow to identify the species concerned by these interactions: indeed, the taxonomic markers employed (16S v4 and ITS1) have inherent limitations in terms of resolution and difficulties for distinguishing microorganisms below the level of genus remain. This is the case for bacteria, but also for a number of fungi, such as those associated with the genus Alternaria: some Alternaria sp. are sometimes described as biocontrol agents and others as pathogens, while ITS1 sequences do not allow to distinguish them. Having said that, this type of work combining LDA and ENA based on a metabarcoding approach can be considered as a hypothesis generator or a guide for the targeted isolation of microorganisms that may have the desired biocontrol phenotypes.
The neglect of complex interactions between biocontrol agents and their biotic environment (the plant, the soil and their microbiomes), the physical and chemical properties of which change over time, may account for lower levels of efficacy in field conditions than in laboratory conditions (concerning the phyllosphere, e.g. [38], but also the residue compartment, e.g. [57]). Indeed, several studies have demonstrated the value of studying the effect of entire communities on biotic and abiotic stresses rather than the effects of single species. For example, resistance to B. cinerea in Arabidopsis thaliana was shown to be not due to a single species, but to the action of the microbiome as a whole [58]. By comparing the structure of microbial communities associated with Brassica rapa plants inoculated with the root pathogen Plasmodiophora brassicae, Lebreton et al. [14] showed significant shifts in the temporal dynamics of the root and rhizosphere microbiome communities during root infection.
Moreover, the rhizospheres of plants infected with P. brassicae were significantly more frequently colonized with a Chytridiomycota fungus, suggesting interactions between these two microorganisms.
The most frequently studied cases of microbial community effects include "suppressive soils", which provide defense against soil-borne pathogens, rendering them unable to establish themselves or to persist in the soil or the plant [59]. The basis and dynamics of this disease suppression vary, and suppression may be general or specific, under the control of antibioticproducing Pseudomonas or Streptomyces populations, for example [60]. Differences in the composition, structure and diversity of microbial communities on crop residues remain poorly understood, and further studies are required to determine the potential for use in biocontrol not of single agents, but of microbial communities, as for these suppressive soils. Despite this ecological reality, the current perception of biocontrol engineering is still too often limited to the action of a single species, even a single strain, with a direct, strong and durable effect against a plant pathogen.
Potential utility of the residue microbiome Improving our understanding of the relationship between biodiversity and ecosystem functioning will require the development of methods integrating microorganisms into the framework of ecological networks. Exhaustive descriptions of microbial diversity combined with ENA are particularly useful for identifying species within microbial communities of potential benefit for disease management [56]. By revealing antagonistic interactions between pathogen species (e.g. Z. tritici) and other microorganisms, our study suggests that this strategy could potentially improve the control of residue-borne diseases, as suggested by another recent study on Fusarium [17]. This strategy, which has been developed separately for the plant [61,62] and soil [14,50,63] compartments, would undoubtedly benefit from further development on crop residues. Indeed, decreasing the presence of pathogens on residues during the interepidemic period can decrease disease development on subsequent crops [21]. More generally, our case study highlights that an interesting way to use ENA is the definition and comparison of indicators, such as node degree and centrality, to characterize the impact of human-induced perturbations on the microbial component of agroecosystems.

Conclusion
This study provides one of the first example of research revealing alterations to the crop residue microbiome induced by the presence of a mere residue-borne fungal pathogen using highthroughput DNA sequencing techniques. The strategy developed here can be viewed as a proofof-concept focusing on crop residues as a particularly rich ecological compartment, with a high diversity of fungal and bacterial taxa originating from both the plant and soil compartments.
Our findings pave the way for deeper understanding of the complex interactions between a pathogen, crop residues and other microbial components in the shaping of a plant-protective microbiome, to improve the efficacy of biocontrol agents and to preserve existing beneficial equilibria through the adoption of appropriate agricultural practices.

Methods
We investigated the effect of Z. tritici on the diversity of the wheat microbiome and the effect of the wheat microbiome on Z. tritici, by characterizing the composition of the microbial communities of 420 residue samples (210 per year) from plants with and without preliminary Z. tritici inoculation. The residues were placed outdoors, either directly in contact with the soil in a field plot or "above ground" , i.e. not in contact with the soil, to assess the effect of their colonization by microorganisms originating from the soil, the plant and the air on the saprophytic development of Z. tritici. We investigated the persistence of interactions between the pathogen and the whole microbial community, and changes in those interactions over time, by sampling the residues before exposure to outdoor conditions (in July), and every two months thereafter (in October, December, and February) ( Figure 1).

Preparation of wheat residues
The 420 wheat residue samples were obtained from 60 winter wheat cv. Soissons plants grown in a greenhouse in each of the two years of the study, as described in [64]: two weeks after sowing, seedlings were vernalized for eight weeks in a growth chamber and then transplanted into pots. Three stems per plant were retained. Half the wheat plants were inoculated with a mixture of four Z. tritici isolates (two Mat1.1. isolates and two Mat1.2 isolates; [65]) to ensure that sexual reproduction occurred as in natural conditions. This equiproportional conidial suspension was prepared and adjusted to a concentration of 2  10 5 spores.mL -1 , as previously described [64]. Thirty plants were inoculated at the late heading stage in early May, by spraying with 10 mL of inoculum suspension. The other thirty plants were sprayed with water, as a control. Inoculated and non-inoculated plants were enclosed in transparent plastic bags for three days to ensure moist conditions favoring pathogen infection.
Septoria tritici blotch lesions appeared three to four weeks after inoculation ( Figure 1A). All plants were kept in the same greenhouse compartment until they reached complete maturity (mid-July).
For each "inoculated" and "non-inoculated" condition, stems and leaves were cut into 2 cm-long pieces and homogenized to generate the "wheat residues", which were then distributed in 105 nylon bags (1.4 g per bag; Figure 1B) for each set of inoculation conditions, in each year.

Exposure of residues to natural conditions
Ninety nylon bags were deposited in contact with the soil in a field plot (the "soil contact" treatment) or without contact with the soil ("above ground" residue treatment). Thirty batches of residues (15 inoculated and 15 non-inoculated) were used to characterize the communities present in July before the exposure of the residues in the nylon bags to natural conditions. The field plot ("OWO" in [16]; Grignon experimental station, Yvelines, France; 48°51′N, 1°58′E) was the same in both cropping seasons. It was sown with wheat in 2015-2016, with oilseed rape in 2016-2017, and with wheat in 2017-2018. The 90 bags for the "soil contact" treatment were deposited in the OWO field plot ( Figure 1C) in late July, at 15 sampling points 20 m apart (three "inoculated" and three "non-inoculated" bags at each sampling point). The 90 bags of the "above ground" treatment were placed on plastic grids exposed to outdoor conditions and located about 300 m from the OWO field plot ( Figure 1D).
We assessed the impact of seasonality on the fungal and bacterial communities on residues by collecting samples of each "inoculated" and "non-inoculated" treatment at three dates (October, December and February): 15 bags from plastic grids ("above ground" treatment) and one bag from each sampling point in the field ("soil contact" treatment) At each date, nylon bags were opened, the residues were rinsed with water and air-dried in laboratory conditions. Residues were then crushed with a Retsch™ Mixer Mill MM 400 for 60 seconds at 30Hz with liquid nitrogen in a Zirconium oxide blender.

PCR and Illumina sequencing
Fungal and bacterial communities profiles were analyzed by amplifying ITS1 and the v4 region of the 16S rRNA gene, respectively. Amplifications were performed with ITS1F/ITS2 [66] and 515f/806r [67] primers. All PCRs were run in a total volume of 50 µL, with 1x Qiagen

Sequence processing
Runs were analyzed separately. Primer sequences were first cut off in the fastq files with Cutadapt [68]. Files were then processed with DADA2 v.1.8.0 [69] according to the recommendations for the "DADA2 Pipeline Tutorial (1.8)" workflow [70], with quality trimming adapted for each run (Suppl. Table 3).
A mock sample consisting of equimolar amounts of DNA from known microorganisms was included in each run (see Suppl. Figure 6) to establish a detection threshold for spurious haplotypes. At a threshold of ≤ 0.3 ‰ of the size of the library, amplicon sequence variants (ASVs) were considered spurious and were removed from the sample. We used the naive Bayesian classifier on RDP trainset 14 [71] and the UNITE 7.1 database [72] to assign ASVs.
ASVs assigned to chloroplasts (for bacteria) or unclassified at the phylum level (for bacteria and fungi) were also removed from each sample. Due to the larger proportion of chloroplast sequences among the 16S rRNA gene products obtained from living plant tissues compared to dead tissues, all samples from July 2017 were removed from the analysis.  [75]). Since the July samples were derived from living plant tissues (greenhouse), we carried out a PERMANOVA to test the effects of inoculation (for fungi and bacteria) and season (for fungi only; Table 1), and a PERMANOVA for the other sampling dates together to test the effects of inoculation, season and contact with soil.

Differential community analysis
A linear discriminant analysis (LDA) implemented in Galaxy [76] (LefSe, http://huttenhower.org/galaxy) was used to characterize the differential abundances of fungal and bacterial taxa between each soil contact condition and each Z. tritici inoculation condition.
In this analysis, differences in the relative abundance of taxa between treatments were evaluated with a Kruskal-Wallis test; a Wilcoxon test was used to check, by pairwise comparisons, whether all subclasses agreed with the trend identified in the Kruskal-Wallis test. The results were used to construct an LDA model, to discriminate between taxa in the different conditions. For the comparison between "soil contact" and "above ground" treatments, inoculation condition was used as a subclass, with the Wilcoxon test alpha value set at 0.05, and the alpha value of the Kruskal-Wallis test set at 0.01. For the comparison between "inoculated" and "noninoculated" treatments, the alpha value of the Kruskal Wallis test was set at 0.01 (no subclasses). For both analyses, the threshold for the LDA analysis score was set at 2.0.

Ecological interaction network analyses
For characterization of interactions within the different wheat residue microbial communities, we performed ecological network analyses (ENA) with SPIEC-EASI [77] for combined bacterial and fungal datasets [78]. The same parameters were used for all networks.         Figure 4A and composed of differential bacterial and fungal ASVs identified in residue samples (originating from wheat plants inoculated and non-inoculated with Zymoseptoria tritici) and of the first adjacent nodes. Node color corresponds to the results of LefSe differential analysis between inoculated (orange) and noninoculated (blue) treatments. Only genera with p-values < 0.01 for the Kruskal-Wallis tests and LDA scores > 2 were retained for the plot. The first adjacent nodes of each differential ASV are not named, except for ASVs interacting with Z. tritici. Edges represent positive (green) or negative (red) interactions. Differential ASVs are plotted with genus name abbreviations: