Soil Inoculation and Blocker-Mediated Sequencing Show Effects of the Antibacterial T6SS on Agrobacterial Tumorigenesis and Gallobiome

ABSTRACT The type VI secretion system (T6SS) is deployed by many proteobacteria to secrete effector proteins into bacterial competitors for competition or eukaryotic cells for pathogenesis. Agrobacteria, a group of soilborne phytopathogens causing crown gall disease on various plant species, deploy the T6SS to attack closely and distantly related bacterial species in vitro and in planta. Current evidence suggests that the T6SS is not essential for pathogenesis under direct inoculation, but it remains unknown whether the T6SS influences natural disease incidence or the microbial community within crown galls (i.e., the gallobiome). To address these two key questions, we established a soil inoculation method on wounded tomato seedlings that mimics natural infections and developed a bacterial 16S rRNA gene amplicon enrichment sequencing platform. By comparing the Agrobacterium wild-type strain C58 with two T6SS mutants, we demonstrate that the T6SS influences both disease occurrence and gallobiome composition. Based on multiple inoculation trials across seasons, all three strains induced tumors, but the mutants had significantly lower disease incidences. The season of inoculation played a more important role than the T6SS in shaping the gallobiome. The influence of the T6SS was evident in summer, during which two Sphingomonadaceae species and the family Burkholderiaceae were enriched in the gallobiome induced by the mutants. Further in vitro competition and colonization assays demonstrated the T6SS-mediated antagonism to a Sphingomonas sp. R1 strain isolated from tomato rhizosphere in this study. In conclusion, this work demonstrates that the Agrobacterium T6SS promotes tumorigenesis in infection processes and provides competitive advantages in gall-associated microbiota.

M any proteobacteria, including pathogens and commensals, deploy the type VI secretion system (T6SS) for antagonism or pathogenesis (1,2) . The T6SS is a protein translocation apparatus used to inject effectors into target cells, mainly in a contact-dependent manner. Based on the destination and biological function of known effectors (3), the T6SS mainly functions as an antibacterial weapon used by bacteria to inhibit or kill the competing bacterial species, thus providing a competitive advantage and shaping microbiota in their ecological niche (4).
Previous studies of the microbiome associated with animal guts or plants indicated that the T6SS genes are enriched in these communities, suggesting the T6SS may be important for niche competition (5)(6)(7)(8). Metagenomic analysis of human gut microbiota revealed a role of the T6SS in the domination of the gut symbiont Bacteroides fragilis by targeting other members of the microbiome in vitro and in vivo (6,7). A recent study further showed that a murine pathogen, Citrobacter rodentium, and resident commensal Enterobacteriaceae share the same strategy by using the T6SS for niche competition in the murine gastrointestinal tract (9). The T6SS is also deployed by plant pathogens to gain a competitive growth advantage in planta as well as for beneficial bacteria to prevent or reduce disease symptoms caused by competing pathogens (10)(11)(12). Comparative metagenomic analysis of microbiota between T6SS 1 and T6SSbacterial strains were also carried out. The gut microbiota of pests infected by Pseudomonas protegens, a plant-beneficial bacterium capable of invading insect pests, showed that the T6SS has no significant impact on microbiota diversity at phylum/class level but affects the abundance of Enterobacteriaceae (13). These studies collectively suggest that the T6SS is a potent antibacterial weapon used by invading pathogens or resident bacteria to gain competitive advantage in their ecological niches. However, knowledge regarding the degree to which the T6SS shapes the microbiota and the molecular mechanisms of interbacterial competition in complex microbial communities are limited.
Agrobacteria are a diverse group of bacteria that include members from several genera (14). These plant pathogens are capable of inducing crown gall or hairy root disease on plants by transferring a piece of DNA named transfer DNA (T-DNA) from bacteria into plants via the type IV secretion system (T4SS) (15). Another secretion system, the T6SS is highly conserved in several Agrobacterium species and plays a role in interbacterial competition (16)(17)(18). Among these Agrobacterium species, the T6SS is encoded by a gene cluster consisting of an imp operon encoding the main T6SS components and an hcp operon encoding the puncturing device and effectors (10,18,19). Using a key Agrobacterium reference strain, C58, which is commonly known as a member of A. tumefaciens but recently reclassified as A. fabrum (20), we previously discovered that it uses a T6SS DNase effector to gain competitive growth advantage in vitro and in planta (10). Interestingly, agrobacteria with incompatible effector-immunity (EI) pairs exhibit strong antagonism between species, while only weak or nondetectable effects within species (16). Moreover, a higher T6SS-mediated killing outcome was observed when nutrients were scarce, as opposed to nutrient-rich conditions (21). Thus, genetic and environmental factors beyond EI pairs also contribute to interbacterial competition.
To date, agrobacterial T6SS has been only demonstrated as an antibacterial weapon (16)(17)(18). There is no evidence for its role in promoting virulence when tumor assays were conducted in sterile conditions or when it was directly inoculated on the stems of various plant species, including tomato plants (19). Considering that agrobacteria may need to compete with other bacteria in bulk soil or rhizosphere to gain access to plant wounds for inciting crown galls, we reasoned that agrobacterial T6SS may influence the microbial community and pathogenesis under a more natural setting. To address these questions, we developed a soil inoculation protocol to mimic natural infection on wounded tomato seedlings across seasons for evaluating the impact of agrobacterial T6SS on disease occurrence and crown gall microbiota (termed gallobiome). Moreover, to overcome the challenge of interference of host organelles in the study of plant-associated microbiota, we optimized a blocker-based method for enriching true bacterial reads in 16S rRNA gene amplicon sequencing. Based on results from our infection assays, gallobiome composition analysis, as well as in vitro competition and colonization assays, this work indicates that the agrobacterial T6SS may provide competitive advantages on the plant surface for effective infection, leading to a higher disease incidence.

RESULTS
Environmental factors and the T6SS affect crown gall disease incidence. The wild-type (WT) strain C58 and two C58-derived T6SS mutants with deletion of essential T6SS genes, DtssL and DtssB, were used to study the effects of the agrobacterial T6SS in tumorigenesis using a soil inoculation method. Nine batches of inoculation experiments across different seasons were conducted in this study (Table 1). In total, 139 crown galls were collected, including 70 induced by the WT, 41 induced by DtssL, and 28 induced by DtssB. The results showed that all three strains are capable of inducing tumors, but the disease incidences of DtssL and DtssB were significantly lower than that of WT (Table 1, Fig. 1A). There was no significant weight difference among the crown galls induced by different strains (Fig. 1B). It is notable that there is an inverse correlation between disease incidence and temperature across seasons for all three strains throughout the year ( Fig. 1C and D). These results indicated that the presence of a functional T6SS and the month of inoculation both affected the disease incidence under soil inoculation conditions.
Round I of 16S rRNA gene amplicon sequencing: initial trial. To determine the gallobiome composition, crown galls with similar weights across seasons were selected for 16S rRNA gene amplicon sequencing. For round I, six crown gall DNA samples, three induced by the WT and three by DtssL, were amplified with two commonly used 16S rRNA gene primer sets, V3 to V4 and V5 to V7 (see Table S1A in the supplemental material). The Illumina sequencing produced 140,830 and 601,299 reads from the V3 to V4 and V5 to V7 sets, respectively. However, the majority (99.1% in V3 to V4 and 93.6% for V5 to V7) of these reads were derived from plant chloroplast and mitochondria. After removing these host contaminants, only 1,230 and 38,320 bacterial reads remained (Table 2).
Round II of 16S rRNA gene amplicon sequencing: optimization. Due to the severe host contamination observed in round I, that used the universal 16S rRNA gene primers, we evaluated the performance of different 16S rRNA gene primers with blockers (39 modified oligonucleotides with C3 spacer) that could prevent the amplification of tomato chloroplast and mitochondrial rRNA genes (22)(23)(24).
Based on the alignment of bacterial, tomato chloroplast, and mitochondrial 16S rRNA gene sequences (Fig. S1), we designed the primer sets and cognate blockers against different variable regions in 16S rRNA genes (Table S1A and B). The new primer sets and blockers were tested using the three WT-induced crown gall DNA samples from round I. After sequencing and data processing, the result indicated that the blockers have variable effectiveness in reducing host contamination (Table 3, Fig. 2A). For the V3 to V4 and V5 to V7 primer sets, adding blockers increased the bacterial reads to 5.3 to 41.5% of total reads, compared to only 0.2 to 1.9% of bacterial reads per sample without blockers. For the V1 to  T6SS Effects on Tumorigenesis and Gallobiome mBio V3 and V6 to V8 primer sets, adding blockers increased the bacterial reads to only 1.6 to 5.6% of total reads per sample. The reads were grouped into species-level operational taxonomic units (OTUs) with 99% sequence identity. The alpha rarefaction curves based on the OTU counts indicated that, for those crown gall samples, the curves would approach saturation when the sequencing depth was over about 10,000 reads per sample (Fig. 2B).
According to the bar plot of bacterial composition at the family level, we observed a higher number of families identified from the same crown galls amplified by adding blockers (Fig. 3A). The result of principal-coordinate analysis (PCoA) indicated that adding blockers did not cause significant biases in the inferred microbiota composition (Fig. 3B). However, the use of different primer sets resulted in a significant difference in the inferred microbiota composition ( Fig. 3B and C), which has been reported previously (25). Overall, adding blockers at the PCR step increased bacterial read counts and the resolution of the gallobiome. Because the use of V5 to V7 primers and cognate blockers obtained the highest number and percentage of bacterial reads, the V5 to V7 setup with blockers was selected for 16 rRNA gene amplicon sequencing of crown galls induced by WT and T6SS mutants.
Round III of 16S rRNA gene amplicon sequencing: impact of the T6SS on the gallobiome. Among all of the 139 crown galls collected, 53 tumors in the range of 0.06 to 0.56 g (24 induced by WT, 16 by DtssL, and 13 by DtssB) were used for the analysis. We obtained an average of 69,980 6 26,452 (mean 6 standard deviation [SD]) reads per sample, ranging from 9,814 to 116,072 reads per sample. Before the analysis, the amplicon sequence  variants (ASVs) were clustered into species-level OTUs, and the singletons were removed. Based on the alpha rarefaction curves and the minimal read counts of the samples (Fig. S2), diversity analysis was conducted with data set subsampling at 9,800 reads per sample. The result of PCoA showed that the gallobiomes induced by the WT and two T6SS mutant strains were not significantly different (Fig. 4A). Most of the variation between samples was contributed by different seasons (July versus October/November) (Fig. 4B). After splitting and reanalyzing the data set based on the month of inoculation, we observed the difference between the gallobiomes associated with the WT and two T6SS mutants in July (Fig. 4C). In contrast, no difference was detected in those galls induced in October or November ( Fig. 4D and E). The alpha diversity indices, including observed OTUs, Shannon index, and Pielou's evenness index had no significant difference between the gallobiomes associated with the WT and T6SS mutants ( Fig. 5A to C). Interestingly, we found that in galls induced in July, the WT gallobiome showed significantly higher Faith's phylogenic diversity than the gallobiome associated with DtssL; and in galls induced in November, the DtssL gallobiome showed higher diversity than the WT gallobiome (Fig. 5D). The gallobiome associated with DtssB also exhibited higher diversity than the WT gallobiome, but the difference was not statistically significant.

T6SS Effects on Tumorigenesis and Gallobiome mBio
Rhizobiaceae, which Agrobacterium belongs to, was found to be highly variable and accounted for 5 to 85% of the entire community. The most abundant OTU in this Rhizobiaceae data set was 100% matched to 16S rRNA genes of C58, suggesting that the abundance of this OTU could be referred to the relative abundance of WT or mutant  T6SS Effects on Tumorigenesis and Gallobiome mBio inoculum in crown galls. When the weight of crown galls was plotted against the relative abundance of this OTU in gallobiomes, we found that the abundance of Rhizobiaceae in July was dramatically lower than those in October and November (Fig. 5F), but there was no significant difference observed among different strains. Furthermore, a positive correlation between the weight of crown galls and the relative abundance of this agrobacterial OTU was observed (Fig. 5G).
Sphingomonadaceae and Burkholderiaceae were more abundant in the gallobiomes induced by the T6SS mutants in July. Next, we analyzed the species-level OTUs and families with differential abundance (DA) between gallobiomes associated with the WT and T6SS mutants in July. Two OTUs belonged to Sphingomonadaceae, here named SphinOTU1 and SphinOTU2, and Burkholderiaceae family members were significantly enriched in the gallobiomes induced by T6SS mutants (Fig. 6A). SphinOTU1 was present in gallobiomes only in July but not in October/November, and SpinOTU2 was present in both July and October but not in November (Fig. 6B). Neither SpinOTU1 nor SpinOTU2 was identified in gallobiomes induced by the WT in July. Burkholderiaceae was enriched in gallobiomes induced by the T6SS mutants in July, but there was no consistent enrichment in October and November (Fig. 6C). Although SphinOTU1 and SphinOTU2 were only present in gallobiomes induced by T6SS mutants in July, at the family level, Sphingomonadaceae was not enriched in a T6SS-dependent manner in any months (Fig. 6C). T6SS-dependent antagonism between agrobacteria and a Sphingomonas sp. isolate. The T6SS-dependent differential abundance of two Sphingomonas OTUs in gallobiomes motivated us to investigate whether C58 exhibits T6SS antibacterial activity to Sphingomonas. A Sphingomonas sp. strain, R1, isolated from the tomato rhizosphere in one soil inoculation experiment was used for interbacterial competition assays in planta and in vitro. Each of the agrobacterial strains was mixed at a 1:1 ratio with Sphingomonas sp. R1 as inoculum, and soil inoculation on wounded tomato seedlings was performed. The colonization efficiency of those strains was determined by counting the CFU recovered from wounded stem segments at 10 days postinoculation (dpi). The results showed an ;0.5-log reduced CFU of Sphingomonas sp. R1 when it was coinoculated with the WT compared to that with DtssL and DtssB or R1 only (Fig. 7A). The recovered CFU of agrobacterial WT and T6SS mutants inoculated alone or with Sphingomonas sp. R1 were 1 to 1.5 log higher than that of Sphingomonas sp. R1, but there was no difference among agrobacterial WT and T6SS mutants inoculated alone or with Sphingomonas sp. R1. These results suggest that agrobacteria have higher competitive colonization efficiency than Sphingomonas sp. R1, in part dependent on a functional T6SS. Accordingly, an interbacterial competition assay on agar plates also showed T6SS-dependent antibacterial activity to Sphingomonas sp. R1 (Fig. 7B). However, when interbacterial competition was carried out in vitro on agar plates, only weak antibacterial activity was observed at 1:1 ratio, but ;0.5-log reduced CFU of Sphingomonas sp. R1 was detected when competition was carried out at a 10:1 ratio of agrobacteria to Sphingomonas sp. R1. The results show that the T6SS provides agrobacteria with a higher competitive advantage against Sphingomonas sp. R1 on plant wounding sites than in vitro. We further evaluated whether the number of CFU of Sphingomonas sp. R1 was enriched in crown galls induced by T6SS mutants compared to those induced by the WT via the soil inoculation method used for 16S rRNA gene amplicon sequencing. Surprisingly, Sphingomonas sp. R1 was not always present in crown galls at 28 dpi, and R1 was recovered from only one out of three independent experiments by direct induction of crown galls on tomato stem (Fig. S3). The abundance of C58 ranged from 10 4 to 10 7 CFU per gall, and no consistent difference could be observed between the WT and the T6SS mutants. Together, these results suggest that C58 exhibits antibacterial activity to inhibit the in vitro growth and plant colonization of Sphingomonas sp. R1. The T6SS may help agrobacteria to gain competitive growth over other competing rhizobacteria on the plant surface for effective infection, leading to higher disease incidence.

DISCUSSION
In this study, we designed experiments to address whether the T6SS affects agrobacterial tumorigenesis and gallobiome composition. By establishing a soil inoculation platform to evaluate the disease incidence of wounded tomato seedlings across seasons, we revealed that a functional T6SS is positively correlated with disease occurrence but not gall weight. Furthermore, by developing an effective protocol for enrichment of bacterial 16S rRNA genes by utilizing blocker to prevent host contamination, we found that seasons or environmental factors are the major drivers in shaping gallobiome composition, while the T6SS could also influence microbiota in a more specific manner.
The evidence that the agrobacterial T6SS promotes tumorigenesis in soil inoculation (Table 1, Fig. 1) but not direct inoculation (19) suggests that the T6SS is not directly involved in virulence. Instead, agrobacteria may deploy the antibacterial weapon to increase the occurrence of infection in the presence of relatively complex microbial communities in rhizosphere (Table S2) (26). Once galls are induced, the T6SS is not required for gall development since no significant difference in weight was observed between the WT and T6SS mutants (Fig. 1B). The presence of a functional T6SS has no impact on the population size of agrobacteria, which is consistent with recent findings that T6SS genes were not identified as fitness genes by a transposon insertion sequencing (Tn-seq) screen in T6SS Effects on Tumorigenesis and Gallobiome mBio crown galls (27). The T6SS of C58 also appears not to be the key factor in shaping gallobiome composition, suggesting that agrobacteria may also use other methods of niche competition in galls. It is generally believed that the agrobacteria inciting crown galls and later becoming residents have privilege to access the specific opines synthesized by transformed plant cells containing the agrobacterial T-DNA (28). This opine concept has not been experimentally validated but is supported by a study showing that the ability of agrobacteria to trap opines can be a competitive advantage over siblings which cannot utilize opines (29). Therefore, it would be interesting to investigate whether opine synthesis and other fitness genes would impact the crown gall microbiota in which the T6SS may synergistically enhance the niche occupation. The findings that disease incidence and crown gall weight are inversely correlated with temperature (Fig. 1C) are expected. Early studies demonstrated that crown gall sizes were FIG 7 Competition between agrobacteria and Sphingomonas sp. R1 in vitro and on wound sites of tomato stems. Each of the three Agrobacterium strains (i.e., C58 WT, DtssL, and DtssB) was individually tested for competition against Sphingomonas sp. R1 isolated from rhizosphere of tomato. (A) Agrobacterium only or coinfected with Sphingomonas sp. R1 at 1:1 ratio in soil was recovered from the surface of wounded stem segments at 10 dpi, and the CFU of Agrobacterium and Sphingomonas sp. R1 were plotted. A t test indicated the CFU differences of Sphingomonas sp. R1 inoculation alone or coinfected with T6SS mutants compared to WT with indicated P values. Data are the mean 6 SD of three independent experiments, each with five seedlings inoculated for each strain. The ANOVA test indicated no significant difference among means (P = 0.0945 and 0.0551 for Agrobacterium and Sphingomonas sp. R1 CFU, respectively). (B) The in vitro competition results based on Agrobacterium: Sphingomonas sp. R1 at 1:1 and 10:1 ratios and the survival of Sphingomonas sp. R1 were plotted; lines and error bars indicate the mean 6 SD of four biological replicates from two independent experiments. Based on ANOVA followed by Tukey's multiple-comparison test, the C58 WT significantly reduced Sphingomonas sp. R1 CFU when mixed in a 10:1 ratio (P = 0.0153) but not in a 1:1 ratio (P = 0.2753).

T6SS Effects on Tumorigenesis and Gallobiome mBio
dramatically decreased when the host plants were inoculated at 28 to 30°C, while there was no gall formation at 31°C or above (30,31). In addition, the T4SS complex and its associated T-pilus are more stable or produced at higher levels at 19°C than 28°C, whereas T4SS-mediated plasmid conjugation is deficient at 28°C (32-36). The increased tumor weight when inoculation occurred in fall/winter is also correlated with a higher relative abundance of agrobacterial OTU (Fig. 5G), suggesting the active proliferation of the agrobacterial population inside crown galls. Furthermore, our data show that crown galls produced in summer have distinct microbiota compared to those from fall/winter (Fig. 4B), which echoes previous studies of Allorhizobium vitis showing that the crown gall microbiota of grapevine is significantly different in summer (37). Interestingly, while there is no global difference in the gallobiome induced by the WT or T6SS mutants inoculated in fall/ winter, we found that the two Sphingomonadaceae OTUs and the family Burkholderiaceae were only present or significantly enriched in the gallobiome induced by T6SS mutants in summer (Fig. 6). By counting viable agrobacterial inoculum, we showed that C58 exhibits a T6SS-mediated antibacterial activity to Sphingomonas sp. R1 in vitro and in rhizosphere ( Fig. 7A and B), but there was no significant difference in R1 abundance between the galls coinfected with C58 WT or T6SS mutants (Fig. S3). Interestingly, we noticed that the colonization efficiency of those T6SS mutants was more variable compared to that of those of the WT from the combined data set (Fig. S4). To this end, it remains to be determined whether the absence or reduced abundance of two Sphingomonadaceae OTUs and Burkholderiaceae in gallobiome induced by the WT compared to that induced by T6SS mutants is a direct consequence of antibacterial activity conferred by the T6SS. However, current evidence suggests that agrobacteria may use the T6SS to gain a competitive advantage by its antibacterial activity against other bacterial species residing in the soil or rhizosphere and therefore increase its success in inciting tumors on infected plants.
Besides the Rhizobiaceae, which agrobacteria belong to, other bacterial families also occupied considerable proportions of crown gall microbiota, such as Comamonadaceae, members of which are commonly found in soil and water, as well as the plant-associated bacterium Variovorax, shown to be important for root growth of Arabidopsis (38,39). Bacteria belong to the Xanthobacteraceae and Pseudomonadaceae, the two families highly associated with plants as either pathogens or commensals, also had high relative abundance in the crown galls. Some Pseudomonas species, such as P. fluorescens and P. putida are capable of utilizing opines (40,41), which may explain why bacterial species other than the agrobacterial inoculum could also colonize and propagate in high relative abundance in crown galls.
We found that the predominance of plant DNA in crown gall samples led to the low resolution of the endophytic bacterial microbiota via standard 16S rRNA gene amplicon sequencing due to the coamplification of chloroplast and mitochondrial sequences. Although the use of 16S rRNA gene primers containing mismatches to the 16S rRNA gene of plant chloroplast or mitochondria could potentially reduce host contamination in endophytic microbiota studies (25,26), the bacteria-specific primers resulted in .90% of host reads of our initial attempt (round I) ( Table 2), which is similar to results of a previous study (22). When adding blockers that could specifically anneal to chloroplast or mitochondrial DNA but not be extended by DNA polymerase during PCR, we successfully reduced the host contamination and obtained sufficient sequencing depth for this study. Our results support the effectiveness of blockers in increasing bacterial reads of the endophyte community (22)(23)(24). Our newly developed soil inoculation method coupled with blocker-mediated enrichment of bacterial 16S rRNA gene amplicon sequencing (SI-BBacSeq) protocol for gallobiomes may be applicable to microbiota associated with tomato or even other plants after modifying the blocker sequences.
To date, collective plant microbiota studies suggest the enrichment of T6SS genes in microbiota as a trait for niche competition in microbiota (5,8). However, the role and impact of the T6SS in different plant and microbiome contexts may be different. Our study suggests that the T6SS may provide agrobacteria with competitive advantages at the initial stage of infection but is likely not critical for proliferation once they established their niche inside the crown galls. Future studies to investigate the spatial distribution of agrobacterial cells and other endophytes inside crown galls or plant wounding sites, together with temporal and spatial visualization of the T6SS activity, are important for dissecting the roles of T6SS in agrobacterial disease ecology.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The bacterial strains and plasmids used in this study are listed in Table S3A. Agrobacterium strain C58 wild type and two T6SS mutants (i.e., DtssB and DtssL) (42) were first streaked on 523 medium agar plates (43) and incubated at 25°C for 48 h. Sphingomonas sp. R1 was grown on R2A medium without soluble starch and pyruvate (44). Escherichia coli was grown on lysogeny broth (LB) medium at 37°C. Freshly grown colonies were inoculated into corresponding broth for overnight culture. The concentrations and antibiotics used are 10 mg/mL gentamicin for E. coli, 50 mg/mL gentamicin for Agrobacterium, and 20 mg/mL kanamycin and 50 mg/mL gentamicin for Sphingomonas sp. R1.
Construction of strains enables antibiotic selection. The primers used for construction are listed in Table S3B. The DNA fragment containing the gentamicin resistance (Gm R ) gene and green fluorescence protein gfp (S65T) were amplified from pRL662::GFP(S65T) with the primer pair BclI-GFP-GmR-F-25 and BclI-GFP-GmR-R-25 and then cloned into XbaI-digested pJQ-COM. The newly constructed plasmid was named pJQ-com-GmR-GFP, and it enabled the generation of the Gm R gene and gfp (S65T) knocked in upstream of actC (atu5330) of the Agrobacterium genome after double crossover. The knock-in strains were generated in an Agrobacterium C58, DtssB, and DtssL background. The knock-in event and excision of the pJQ200ks backbone were confirmed by green fluorescent protein (GFP) signal, gentamicin resistance, and PCR with the primer pair 39sacB-59GmR pJQ200 F and 39sacB-59GmR pJQ200 R, which target the fragment between sacB and the Gm R gene on pJQ200ks. The plasmids pRL662::GFP(S65T) and pBBR::mCherry, conferring kanamycin resistance, were each transformed into Sphingomonas sp. R1 through electroporation. The procedure of electroporation and double crossover via pJQ200ks was performed as described previously (45).
Soil inoculation. The soil inoculation method was optimized based on the method used for agrobacterial infection of pea and walnut (46,47). Overnight (14 to 16 h) culture of the agrobacterial strain in 12 mL of 523 was were centrifuged at 6,000 Â g for 10 min. The pellets were washed once using 10 mL of sterile saline (0.9% NaCl in H 2 O). The washed pellets were centrifuged and resuspended in sterile saline with the optical density at 600 nm (OD 600 ) adjusted to 1. The bacterial suspension was mixed into unsterilized soil (1 mL suspension per 100 g of soil), which is expected to result in ;10 7 CFU/g of soil, the optimized concentration that could incite crown galls but not at 100%. Tomato seedlings with two true leaves (2 to 3 weeks old) were wounded at the site between the primary root and cotyledons using a fire-sterilized sewing needle, planted in the inoculated soil, and grown in the greenhouse. Each strain was inoculated with 14 to 20 tomato seedlings, and the inoculation with saline was used as a negative control.
Harvest, surface sterilization, and storage of crown gall samples. Two months after the inoculation, tomato plants were harvested to examine gall formation on the wounded site. The disease incidence was calculated by dividing the number of plants with visible gall formation by the total number of inoculated plants. The sections with crown gall were cut out and sterilized in 35 mL of sterilization solution (3% NaOCl, 0.01% Tween 20) for 30 s, and then transferred to 35 mL of 70% ethanol and washed in 35 mL of sterile H 2 O three times (48). The sterilization was evaluated by spreading 100 mL of liquid from the third wash onto 523 agar plates and observing for 2 days to make sure there was no growth of bacteria. After sterilization, the crown galls were dissected from each of the plant segments, placed in a sterile petri dish, wrapped in sterile aluminum foil, frozen by liquid nitrogen, and stored at 280°C prior to DNA extraction.
DNA extraction from crown galls. Crown gall samples were homogenized with a pestle and mortar with liquid nitrogen. Then, 0.25 g of homogenized tissue was transferred to the beating tube of a DNeasy PowerSoil kit (Qiagen, Germany) following the manufacturer's instructions. Concentrations of extracted DNA were determined using NanoDrop 1000 (Thermo Fisher Scientific, USA) and Qubit double-stranded DNA (dsDNA) high-sensitivity (HS) (Invitrogen, USA) systems.
16S rRNA gene amplification and sequencing. Three rounds of amplicon sequencing were conducted in this study. Round I was conducted according to established protocols of soil microbiota studies using 16S rRNA gene variable regions (25). Each primer with a partial adaptor sequence was compatible with the Illumina TruSeq combinatorial dual index system. The first PCR amplifications were carried out in the following 25-mL reaction: 5 ng DNA template, 2Â KAPA HiFi HotStart DNA polymerase ReadyMix (Roche, Switzerland), and 0.2 mM for each of the forward and reverse primers. The first PCR program and reaction details for V3 to V4 and V5-to V7 are listed in Table S4. The PCR products were purified using Ampure XP beads (Thermo Fisher Scientific, Inc., Sweden). The V5 to V7 PCR products underwent BluePippin (Sage Science) size-selection to remove tomato amplicon. To attach dual indices and full-length Illumina sequencing adapters, a second round of PCR amplifications was employed for 50-mL reactions with purified or size-selection PCR products, 2Â KAPA HiFi HotStart DNA polymerase ReadyMix (Roche), and 0.5 mM for each of the forward and reverse primers. All of the library products of the second PCR were purified and pooled. The pooled libraries were loaded onto an Illumina MiSeq V3 flow cell (Illumina, USA) for 2 Â 300-bp paired-end sequencing.
Round II of amplicon sequencing was conducted to test the effect of adding PCR blockers on reducing plant chloroplast and mitochondrial sequences. Three gall samples (1115-W21, 1115-W22, and 1115-W25) were used as the DNA templates to provide biological replicates. For each DNA sample, four T6SS Effects on Tumorigenesis and Gallobiome mBio primer pairs targeting different variable regions of 16S rRNA gene were used: V1 to V3, V3 to V4, V5 to V7, and V6 to V8 (Table S1A). Each primer contained a partial adaptor sequence of the Illumina TruSeq combinatorial dual index system. Cognate 39 C 3 spacer-modified oligonucleotides (blockers) which would block the amplification of tomato chloroplast and mitochondrial 16S rRNA genes were designed according to a previous study (22) and applied to one of the two PCRs for each sample-primer combination (Table S1B). The procedures for PCR and Illumina sequencing were based on those used for round I, but the amount of DNA template was increased to 25 ng to reduce PCR cycle time, and the annealing temperature was adjusted for blockers. More detailed technical information is provided in Table S4. Based on results from the first two rounds of amplicon sequencing, an optimized protocol that utilizes blockers and targets the V5 to V7 region was used for round III. All 53 crown gall samples were included in this final round.
Raw read processing and microbiota analysis. The Illumina raw reads in FASTQ format were imported into QIIME 2 release 2020.8 (49). The primer region of imported sequences was trimmed using DADA2 (50). The denoised and merged paired-end reads were used to construct output feature tables containing amplicon sequence variant (ASVs) IDs and counts in different samples. For amplicon sequencing round III, the ASVs were further clustered into species-level OTUs based on 99% sequence identity using the qiime vsearch cluster-features-de-novo function (51), and the singletons were removed using the qiime feature-table filter-features function prior to the downstream analysis. The taxonomic information was inferred for representative OTU sequences via the pretrained q2-feature-classifier (52) and classify-sklearn naive Bayes classifier based on the SILVA version 138 non-redundant small subunit rRNA reference sequences dataset (SSU Ref NR 99) (53). The sequences assigned as chloroplasts, mitochondria, eukaryotes, archaea, or unknown were removed. After the filtering, a phylogenetic tree of eubacterial sequences was inferred using the q2-fragment-insertion plugin (54-57) based on the SILVA 128 SATé-Enabled Phylogenetic Placement (SEPP) reference database. Weighted UniFrac (58) analysis of those data sets was performed based on an insertion tree; the distance matrix and principal-coordinate analysis (PCoA) plot were generated using QIIME 2. For round III, phylogenic and diversity analysis was conducted using qiime diversity core-metrics-phylogenetic, and the subsampling depth was set to 9,800 reads/sample. Differential abundance analysis of OTU. Analysis of the composition of microbiomes (ANCOM) was applied to identify OTUs that had significantly different relative abundance in the gallobiome associated with the C58 WT and T6SS mutants using the function qiime composition ancom in QIIME 2. The feature table collapsed into the family level was imported and transformed by centered log-ratio transformation. The null hypothesis of ANCOM is that the abundance of an OTU is not different between two study groups. After performing all comparisons between each OTU in two study groups, the times that the null hypothesis was rejected is called w. The OTUs having w values at the 70th percentile or higher are considered significant (59,60).
Isolation of Sphingomonas sp. R1. Sphingomonas sp. R1 was isolated from a wounded tomato stem during a soil inoculation assay. After 10 dpi of soil inoculation as mentioned above, the 1-cm wounded segments of tomato seedlings were harvested and rinsed with sterile water and then transferred to 1 mL 0.9% NaCl. After vortexing at maximum speed for 3 min, the supernatant was transferred to a new microcentrifuge tube and spread on 523 agar plates via the exponential mode of the easySpiral Dilute system (reference [ref.] no. 414 000, Interscience, France). After growth at 25°C for 2 days, ;20 colonies were selected and streaked on 523 agar plates for pure culture. The partial 16S rRNA gene of the isolates was amplified and sequenced using the V5 to V7 primer set, which was used to conduct a BLAST search of the 16S rRNA gene sequence database in NCBI for identification of 16 bacterial species (Table S2).
Interbacterial competition assay in vitro and in planta. For interbacterial competition in vitro, the bacterial suspension was adjusted to an OD 600 of 3.0 in 0.9% NaCl (wt/vol) after overnight culture. For the in vitro competition, the attacker strains (C58 WT, DtssL, and DtssB) were further diluted to an OD 600 of 1, and the prey Sphingomonas sp. R1 containing pRL662::GFP was further diluted to an OD 600 of 0.3 or 0.1 for mixing with the attacker in equal volumes for different density ratios. Two spots of 10 mL of mixture were spotted on Agrobacterium kill-triggering medium (21). After being coincubated for 16 h at 25°C, the colonies were scraped and resuspended in 1 mL 0.9% NaCl (wt/vol) and spread via the exponential mode of the easySpiral Dilute system (ref. no. 414 000, Interscience, France) on 523 medium containing 30 mg/mL gentamicin to recover the prey.
For agrobacterial colonization with or without Sphingomonas sp. R1, each of the agrobacterial strains (C58 WT, DtssL, and DtssB) with Gm R -GFP knock-in was mixed with Sphingomonas sp. R1 containing pBBR::mCherry at an OD 600 of 1.0. The 5-mL mixture was inoculated on the needle-wounded stem of tomato seedlings with first pairs of true leaves. Tomato seedlings were harvested at 10 days postinoculation (dpi). For each plant, the 1-cm wounded segment was cut and rinsed with sterile water before being transferred into a sterilized Eppendorf tube containing 1 mL of 0.9% NaCl. After being vortexed at maximum speed for 3 min, the liquid was considered the rhizosphere sample. The samples were diluted and plated on medium with gentamicin for CFU calculation.
For counting the CFU of C58 and Sphingomonas sp. R1 in the tumor, the crown galls induced by direct inoculation (19) were collected after growing in the greenhouse for 3 weeks. The 1-cm tomato stem section containing the gall in middle was cut and homogenized with 1 mL 0.9% NaCl (wt/vol) using a sterilized mortar and pestle. The homogenized tissue was diluted 10-fold and spread via the exponential mode of the easySpiral Dilute system on 523 medium containing 50 ppm gentamicin to recover the attacker and prey. After incubation for 2 days at 25°C, the colonies were counted by using the Scan 500 automatic colony counter (ref. no. 436 000, Interscience, France; software version 8.6.1) to calculate CFU.

T6SS Effects on Tumorigenesis and Gallobiome mBio
Statistics and visualization. GraphPad Prism 8 was used to perform t tests on sequencing read counts, one-way analysis of variance (ANOVA) on CFU counts, two-way ANOVA followed by Turkey's multiplecomparison test on crown gall disease incidence, the alpha diversity index and Kruskal-Wallis test on tumor weight, and Spearman correlation on gall weight and the relative abundance of agrobacterial OTU. The alphararefaction curves were plotted using phyloseq (61) in R version 3.6.2. ADONIS in the qiime diversity betagroup-significance function of QIIME 2 (62,63) was used to test if groups of samples had significantly different microbiota compositions. The taxonomic composition plots were generated using qiime2R (version 0.99.35, https://github.com/jbisanz/qiime2R) in R studio 2.
Data availability. All raw data sets are available in the National Center for Biotechnology Information (NCBI) under BioProject accession number PRJNA894311.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank Teng-Kuei Huang and Mao-Sen Liu for careful reading of this manuscript and the members of Lai and Kuo labs for helpful discussion. We also acknowledge Lay-Sun Ma for constructing pRL662GFP(S65T) and Stanton Gelvin for providing pBBR::mCherry.
The Illumina sequencing library preparation was carried out in the Genomic Technology Core (Institute of Plant and Microbial Biology, Academia Sinica). We are grateful for the Illumina MiSeq sequencing service provided by the Genomics Core (Institute of Molecular Biology, Academia Sinica). Conceptualization