Seedlings were grown in the chamber under controlled temperature (25˚C), humidity (90%), and light conditions (16 hours of light and 8 hours of darkness). When the third leaves were fully expanded, seedlings were inoculated with powdery mildew (race E20). Seven days later, infected and control tissue samples from seedling roots, stems and leaves were collected. Each sample has three biological replications. Each replication was mixed with the tissues from 5 individual plants. A total of 9 samples were used for further processing.

The wheat roots, stems and leaves samples were cleaned with tap water for remove soil particles. For surface sterilization, wheat roots, stems and leaves tissues were put in sterile falcon tubes and washed with sterilized water for 1 min, soaked in 1% sodium hypochlorite solution for 5 min, washed with 75% ethanol for 1 min, and then washed with sterilized water for five times. The surface sterilized tissues were then crushed in sterile mortar and suspended in sterile saline solution (2% NaCl). Suspensions were incubated for 2 h at 28˚C under shaking saline plated in duplicates onto Nutrient agar (NA). Peptone sucrose agar (PSA) was supplemented with 0.01% cycloheximide. Besides the washed wheat roots, stems and leaves samples, the last washed water was also spread onto different media plates for confirming the surface sterilization effect.

The genomic DNA of wheat roots, stems, and leaves in each group were extracted using E.Z.N.A.HP Plant DNA Kit (OMEGA Bio-Tek, USA, https://www.omegabiotek.com/). Total DNA concentration and purity were monitored on 1% agarose gels (Biowest Agarose, http://www.genehk.com).

The primers 335F (5'-CADACTCCTACGGGAGGC-3') and 769R (5'-ATCCTGTTTGMTMCCCVCRC-3'), which targeting the V3 and V4 hyper variable regions of bacterial 16S rDNA genes, were selected for bacterial taxa analysis [14] . Both forward and reverse primers were tagged with adapter, pad, and linker sequencing. The suitable system of PCR was 50 μL in which containing 0.2 μL Q5 High-Fidelity DNA Polymerase (NEB, USA), 10 μL High GC Enhancer (NEB, USA), 10 μL Buffer, 1 μL dNTP, 10 μM of forward and reverse primers, and 40 ng template DNA. The procedures were as follows: an initial denaturation at 95˚C for 5 min, each of 15 cycles at 95˚C for 1 min, 50˚C for 1 min, and 72˚C for 1 min, with a final extension at 72˚C for 7 min. Finally, the library was sequenced on an Illumina Hiseq2500 platform at Biomarker Technologies Co, LTD, Beijing, China.

Following amplification, 2 μL of PCR product was used to verify successful amplification by 2% agarose gel electrophoresis. The products of triplicate PCR reaction from each group were combined and the pooled mixtures were purified with Gel Extraction Kit (OMEGA Bio-Tek, USA, www.omegabiotek.com) and analyzed on an Agilent 2100 Bioanalyzer using High Sensitivity DNA Chips (Agilent Technologies, Germany ) for size distribution. The sequencing libraries were generated using NEB Next Ultra DNA Library Prep Kit for Illumina (NEB, USA) following manufacture’s recommendations and index codes were added. The library quality was assessed on the Qubit@ 2.0 Fluorometer (Thermo Scientific, USA) and Agilent Bioanalyzer 2100 system (Agilent Technologies, Germany).

The forward and reverse sequences were merged by overlapping paired-end reads using FLASH (V1.2.11, http://ccb.jhu.edu/software/FLASH/) [15] . All sequence reads with the same tag were assigned to the same sample according to the unique barcodes (raw tags). The raw tags were further strictly filtered by previous methods (clean tags) [16] , and the quality of clean tags were detected by Qiime (V2.1.0, http://qiime.org/index.html) [17] . The low quality tags were removed. The tags with chimera were detected and removed using UCHIME Algorithm (V4.2, http://www.drive5.com/usearch/manual/) [18] . The effective sequences were then clustered into operational taxonomic units (OTUs) at 97% sequence similarity using the UPARSE-OTU and UPARSE-OTUref algorithms of UPARSE software package (Uparse v7.0.1001, http://drive5.com/uparse/). The indices of alpha diversity were calculated based on the above algorithms [19] . Finally, the RDP (Ribosomal Database Project) Classifier (V2.2, https://sourceforge.net/directory/windows/) was used to assign representative sequence to the microbial taxa [20] . Sequence data have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under the accession number SRP056016 (16S).

Cluster analysis was preceded with principal component analysis (PCA) using the QIIME (V1.8.0) software package [21] . QIIME calculates both weighted and unweighted unifrac distances, which are phylogenetic measures of beta diversity [22] . Phylogenetic relations among different microbial taxa were further displayed by KRONA [23] . Alpha diversity indices of Chao1, ACE, Shannon, Simpson, and coverage were calculated to reflect the diversity and richness of the endophytic community in different samples [24] .

To fully understand the bacterial community, we performed 16S rDNA sequencing to wheat roots, stems, and leaves tissues under Bgt (Blumeria graminis f. sp. tritici) infection and un-infection conditions (Table 1, samples designated as PM-Root, CK-Root, PM-StemPM-Stem, CK-Stem, PM-Leaf, and CK-Leaf, respectively). Totally, we obtained 1,424,920 sequencing reads and, among them, 1,317,962 reads were raw tags. After splicing the original data, quality filtering the spliced sequences, and removing the chimeras, finally we obtained 1,121,671 high quality tag sequences, with a range from 150,591 to 217,010 sequences per sample. The average length of effective bacterial sequences was 437 bp (Table 1). Overall, at a similar level of 97%, these sequence reads were classified into 1279 Operational Taxonomic Units (OTUs), with a range from 422 to 943 OTUs per sample (Table 1). Generally speaking, these OTUs were divided into 20 Phylum, 58 Classes, 91 Orders, 165 Families, 277 Genus and 197 Species. And wheat samples were dominated by those OTUs taxonomy into Proteobacteria (60%), Bacteroidetes (20%), Acidobacteria (16%), Firmicutes (13%), and Actinobacteria (10%). Rarefaction curves, showing the increasing number of detected OTUs with the increasing number of obtained sequences per sample, were obtained for the three tissues upon Bgt infected and uninfected conditions, as shown in Figure 1.

Further analysis revealed that, in leaves, we detected the most OTUs (1701), and close number of OTUs were detected in powdery mildew infected and uninfected leaves (854 and 847 respectively). The detected number of OTUs in root samples was 1406, with 726 and 680 OTUs in infected and uninfected roots respectively. OTUs in the stem samples was the lowest (1365) and obvious difference of OTUs number was observed between infected and uninfected powdery

CK represents uninfected wheat powdery mildew, PM represents infected wheat powdery mildew.

mildew (422 and 943 respectively). Comparingly, it seemed the disease happening had different effect on the distribution of OTUs in root and stem tissues (Figure 1). The overlap and unique OTUs between infected and uninfected samples of three tissues were further analyzed. The results showed that 122 OTUs were overlapped by the six treatments (Figure 2). Meanwhile, 15, 7, 0, 0, 5, and 6 unique OTUs were detected in PM-Root, PM-StemPM-Stem, PM-Leaf, CK-Root, CK-Stem, and CK-Leaf samples, suggesting that the OTUs diversity are some differences among samples.

At the Phylum level, the heat map was drawn according to the species annotation and abundance information (Figure 3). It showed that samples from same tissues were clustered together, suggesting that OTUs have some tissue specific manners. However, as can be seen from Figure 3, the abundance was fair difference between infected and uninfected samples, suggesting the impact of disease happening on OTUs community.

Consist with the highest Chao1 index and ACE index (992.2143 and 972.3016 respectively), the most OTUs were detected in CK Stem sample (943), indicating that the endophytic bacteria community in leaf was more abundant and diversity. However, there are a minimum of 422 OTU clusters of endophytic bacteria in PM-Stem. From the perspective of community richness index (Chao1 index and ACE index), the Simpson index and Shannon index of the sample are 0.0641 and 3.3244, respectively (Figure 4).

This may be due to the changes in endophytic bacterial species affected by wheat powdery mildew, thereby regulating the migration of its internal mechanisms. Among the number of O T Uin the leaf tissue samples, the number of OTU infecting wheat powdery mildew was more than the number of OTU in the endophytic bacteria clustering of uninfected wheat powdery mildew. In brief, our results suggested that the pathogens that infect wheat powdery mildew and uninfected wheat powdery mildew have the highest bacterial abundance and uniformity.

Due to the large difference in the community structure of endophytic bacteria between different parts of the sample, when the Rank-Abundance curve was drawn, the roots, stems and leaves of the infected and uninfected pathogens were drawn according to the classification criteria. The Rank-Abundance curve of endophytic bacterial diversity, in which the CK-Stem sample curve is the most gradual, and the curve has the largest span on the X-axis, indicating that the endophytic bacteria have the highest abundance and the endophytic bacteria distribution is the most. Evenly. The PM-Stem sample curve has the narrowest span and the steepest trend, which indicates that the endophytic bacteria in the stem tissue is the lowest after infestation of wheat powdery mildew. The CK-Leaf and PM-Leaf tissue samples were relatively flat, and the span on the X-axis was relatively close. The CK-Root and PM-Root were also close, indicating that the leaf tissue samples had less difference in richness (Figure 5, Table 2).

Beta diversity analysis was performed using QIIME software to compare the similarity of species diversity in different samples. Beta diversity analysis mainly uses binary jaccard, brays curtis, weighted unifrac (bacteria limited), unweighted unifrac (limited bacteria) and other four algorithms to calculate the distance between samples to obtain the beta value between the samples. Principal coordinates analysis (PCoA [25] ), is a similar dimensionality reduction method to PCA, which can be used to classify multiple samples and reveal the diversity of species diversity among the samples.

The PCoA map was drawn based on the endophytic OTUs of different tissues to reflect the endophytic Beat diversity of all samples. As a result, it was found that the stem (CK-StemCK-Stem, PM-Stem) sample was on the PC1 (70.89%) axis in Figure 6(a) and the PC1 (50.86%) axis in Figure 6(b), and the PC1 (66.41%) axis in Figure 6(c) and Figure 6(d). PC1 (61.73%) was well separated from the other two sets of samples on the axis, while the other two samples were basically inseparable. The number of reads of CK-StemCK-Stem and PM-Stem is quite different, while the difference between CK-Leaf and PM-Leaf, CK-Root and PM-Root is small (Table 1), so it is inevitable for stems. For roots and leaves, the difference between difference and similarity is a lower level, CK-Root and PM-Root are separated on the PC1 (61.73%) axis in Figure 6(d), but the distance is closer. This indicates that wheat powdery mildew has an effect on the distribution of endophytic bacteria in roots, stems and leaves of wheat, and has a great influence on stem tissue parts.

The significant difference between groups was mainly used to find Biomarker with statistical differences between different groups. According to the set Biomarker screening criteria (LDA score > 4), the eligible Biomarker was found and displayed as an icon, LEFSe (Line Discriminant Analysis (LDA) Effect Size [26] ) is able to find statistically significant Biomarker between different groups (Figure 7, Figure 8). In the uninfected wheat powdery mildew mycorrhizal was enriched with Devosia, Cytophagia, Cytophagaceae, Cytophafales, Streptomyces and Streptomycetales (6 groups in totsl), the stems were enriched with Proteobacterio, Burkhoideriales and Comamonadaceae (3 group in total), the leaves enriched with Blastocatellales, Nitrospirae, Desulfurellaceae and Latescibacteria (4 groups in total), the fine lineages had an LDA value of 4 or higher in above groups (Figure 5). In infected wheat powdery mildew, the leaves were enriched with Acidobacteria, Chloroflexi, Gemmatimonadetes, Rhodospirillales, Nitrospirales, Sphingomonadales, Acidimicrobiales and Xanthomonadales, the roots were enriched with Myxococcales, Actinobacteria, Streptomycetac, Xanthomonadales, Hyphomicrobiaceae, etc. The stems were enriched with Gammaproteobacteria, Pseudomonas, Enterobacteriales, Staphylococcaceae, Brucellaceae, Ochrobactrum, Candidatus, etc. (Figure 6).