Escherichia coli strains were grown in Luria-Bertani (LB) broth at 37°C. P. fluorescens was grown in King’s medium B (KB) broth or mannitol-glutamate (MG) minimal medium at 28°C. Pathogenic fungi were grown on Potato Dextrose Agar (PDA) at 28°C, except Phytophthora, which was grown on V8 juice agar. Pathogenic bacteria Ralstonia solanacearum and Acidovorax avenae were grown in Nutrient Broth (NB) at 28°C (Table 1). The following concentrations of antibiotics were used: ampicillin at 50 μg/mL and kanamycin at 50 μg/mL. Nicotiana benthamiana was grown in a greenhouse with 16 h light/8 h dark, 65% humidity, and temperatures of 24°C during daylight and 22°C at night.

To locate the SPI-1 T3SS gene cluster, we performed a BlastP search in the complete genome of strain 2P24 (GenBank accession: CP025542.1), using protein sequences of T3SS conserved components that were identified previously in Salmonella enterica (InvA: QBG67165.1, SpaP: EBB2042894.1, and PrgK: NUD93968.1). DNAMAN Version 9 was used to analyze the identity of T3SS components between P. fluorescens 2P24, P. kilonensis F113, and S. enterica. To produce a phylogenetic analysis of SPI-1, we searched the National Center of Biotechnology Information (NCBI) database for InvA protein sequences and 16S rRNA sequences in five genera: Pseudomonas, Salmonella, Shigella, Yersinia, and Burkholderia. Phylogenetic trees were created with MEGA7 (Kumar et al., 2016) using the Maximum Likelihood method.

For scanning InvF binding sites from intergenic regions of strain 2P24, Geneious Prime^® (Biomatters Ltd) was used to generate intergenic regions by retrieving the genome sequence of strain 2P24. Then, we input the sequence matrix of the InvF binding site ‘‘ANNGGNCNTTTTTTNAANGTT’’ and the intergenic region file to the Find Individual Motif Occurrences (FIMO)¹ (Grant et al., 2011); we set the p value ≤ 10^–5.

To make a hilA mutant of P. fluorescens 2P24, a 2.8 kb fragment that covered the full length of the hilA gene was amplified from strain 2P24 using primer pairs WHL140 (5′-TTCTAAGCTTGAGCAACAGCAGCG-3′) and WHL141 (5′-GGACACGCCACTTCTAGAAGTTGG-3′) and digested with HindIII and XbaI. The fragment was cloned into the same sites of pK303SacB. The resulting pK303SacB derivative was digested with NarI to eliminate a 0.9 kb fragment of the hilA gene from the insertion. The parental fragment was recycled and self-ligated. The final pK303SacB derivative pK303SacBΔhilA was constructed and transformed into strain 2P24. Colonies were selected from KB medium with 10% sucrose. Then kanamycin-sensitive mutants were selected and screened by PCR.

To make an invF mutant of strain 2P24, two 1.0 kb fragments that carried the left and right flanking regions of invF were amplified by PCR using primer pairs WHL92 (5′-CCAAAGCTTCAGGACTGGTCACGCC-3′) and WHL93 (5′-GGCATCGCTCATGTCGAAAAAAGC-3′), and primer pairs WHL94 (5′-CTCGCTGACCGATGTGGCACT-3′) and WHL95 (5′-TCGAGGATCCACAACGACAACTC-3′), respectively. The left and right regions were digested with HindIII and BamHI, respectively, and were cloned into the relevant sites of pK303SacB. The resulting pK303SacB derivative pK303SacBΔinvF was transformed into strain 2P24. The correct mutant was screened using the same method as hilA mutant, and the same protocol was used to make an invE-C mutant. The left and right flanking regions of invE-C were amplified by PCR using primer pairs WHL15 (5′- TGGCGAATTCATCCTCGGAACAAG-3′) and WHL16 (5′- CTGACAATTCGTCGCTGATCTGC-3′), and primer pairs WHL17 (5′-GCGTCTCGAGCAACTGCAGGTCT-3′) and WHL18 (5′-ATCAAGCTTGTATCGGCAGCGGT-3′), respectively. The left and right regions were digested with EcoRI and HindIII, respectively, and then cloned into the relevant sites of pK303SacB.

Strains 2P24, 2P24ΔinvF, and 2P24ΔhilA were grown on KB plates with 50 μg/mL ampicillin at 28°C. Fresh lawns were suspended in 5 mL of KB and incubated at 28°C overnight with 200 rpm shaking. Bacterial suspension was diluted in 50 mL of MG medium in flasks to a final OD₆₀₀ of 0.1. The cultures were incubated at 28°C at 200 rpm on a rotary shaker for 6 h/12 h. The cell fractions were separated by centrifugation at 10, 000 rpm for 5 min at 4°C. Total RNA was isolated using a E.Z.N.A.^® Bacterial RNA Kit (Omega Bio-tek, United States) as described by the manufacturer for RNA-seq and qRT-PCR.

A prokaryotic chain specific sequencing library was constructed using a NEBNext Ultra Directional RNA Library Prep Kit and then sequenced on an Illumina HiSeq platform. Gene expression was calculated using HTseq software (V 0.6.1) based on the FPKM (Expected number of Fragments Per Kilobase of transcript sequence Per Millions base pairs sequenced) method. Genetic differences were analyzed using DESEQ2 (V1.6.3) in the Bioconductor software package. For qRT-PCR, a FastKing RT Kit (With gDNase) (Tiangen, China) was used to synthesize cDNA from 50 ng-2 μg of total RNA. Luna Universal qPCR Master Mix (New England Biolabs, United States) was used for a real time PCR reaction to quantify the cDNA level of target genes in different samples. Reactions were run and data were collected using the ABI QuantStudio6 Flex real-time PCR system (Applied Biosystems, United States).

Antagonistic tests were performed as described by Gu et al. (2020). For antagonism of pathogenic fungi, a 1 cm diameter agar plug with mycelium was placed in the center of an agar plate, and 10 μL cultures of strains 2P24, 2P24ΔhilA, 2P24ΔinvF, and 2P24ΔinvE-C were dropped on the plate at four directions, approximately 3 cm from the center. The plates were incubated at 28°C and checked for inhibition zones of mycelial growth. For antagonism of pathogenic bacteria, we incubated pathogenic bacteria in 5 mL of NB overnight and diluted in melting Nutrient Agar (NA) at a ratio of 1:100 for preparing plates. The cultures of strains 2P24, 2P24ΔhilA, 2P24ΔinvF, and 2P24ΔinvE-C were inoculated and observed as above.

We conducted motility tests as described by Rashid and Kornberg (2000), with slight modification. For swimming tests, we used KB medium that contained 0.2% agar, and for swarming tests, we used KB medium that contained 0.5% agar and 0.5% glucose. Freshly cultured strains were dipped using pipette tips and inoculated on the surface of the plates’ center. For twitching tests, we used KB medium that contained 1% agar. Freshly cultured strains were dipped using pipette tips and inoculated into the bottom of the plates’ center. Then, plates were placed stably in the incubator and cultured at 28°C.

Chemotactic responses of strain 2P24 and SPI-1 related mutants to glucose were performed according to a simplified capillary assay method (Mazumder et al., 1999). Bacteria grew to late log phase in KB liquid medium. Pellets were collected by centrifugation and washed twice with BHA (NH₄NO₃, 1 g/L; FeCl₃, 0.05 g/L; KH₂PO₄, 1 g/L; K₂HPO₄, 1 g/L; MgSO₄, 0.2 g/L; CaCl₂, 0.02 g/L; pH 7.0) (Ugochukwu et al., 2013). Finally, bacteria were suspended into BHA (OD₆₀₀ = 1.0). One hundred microliters of bacterial suspension were sucked into a 200 μL pipette tip, and 100 μL BHA that contained 0.1% glucose was drawn up through the needle into a 1 mL syringe, and 100 μL BHA without carbon source was drawn up through the needle into a 1 mL syringe as negative control. The needle-syringe capillary was inserted into the pipette tip that contained the bacterial suspension. After 45 min of incubation at room temperature, the contents from the needle-syringe were removed and diluted in 25 mM PBS buffer (pH 7.0) and plated onto LB medium. Accumulation in the capillary was calculated from the CFUs on the plates.

The assay of biofilm formation was performed according to the classic approach (Christensen et al., 1982). One hundred microliters cultures of bacterial strains that were incubated overnight were added to 2 mL centrifuge tubes, and 900 μL KB was added to each tube. Bacterial strains were incubated at 28°C for 48 h statically. Then, the contents of each tube were emptied and washed three times with sterile water. The tubes were stained for 15 min with 200 μL of 1% crystal violet and then washed with sterile water. After the tubes were air dried, the dye bound to the adherent cells was resolubilized with 2 mL of 95% ethyl alcohol per tube. The OD of each tube was measured at 570 nm by using a L3S visible spectrophotometer (INESA, China).

For ROS assays, bacterial strains were injected into 5-wk-old N. benthamiana leaves at 5 × 10⁸ cfu/mL. Leaf disks were excised 12 h later with a 0.5 cm diameter cork borer and placed into wells of 96-well plates to which 10 μL ddH₂O had been pre-added. Then, 100 μL of 0.5 mM L-012 (Wako, Kyoto, Japan) in 10 mM morpholinepropanesulfonic acid-KOH buffer (pH 7.4) was added. Chemiluminescence was monitored immediately for 10 h using a Veritas luminometer (GENios Pro, TECAN, Switzerland).

All tests in this study were performed at least triplicate, and standard deviations were calculated. One-way ANOVA was performed to determine significant changes.

Through genome mining and comparison, a 23 kb SPI-1 T3SS gene cluster comprised of 25 predicted open reading frames (ORFs) was identified in P. fluorescens 2P24, and the gene names were assigned according to the SPI-1 T3SS of S. enterica. Comparing SPI-1 T3SS gene clusters of PGPR and mammalian pathogens indicated that the organization and orientation of SPI-1 T3SS in strain 2P24 were very close to those in S. enterica and P. kilonensis F113 (Figure 1). The SPI-1 T3SS of strain 2P24 retained all assembly proteins except InvH, which indicated that the structure of SPI-1 T3SS in strain 2P24 was conserved. Two transcriptional regulator homologs, HilA and InvF, were retained in strain 2P24. However, compared with the SPI-1 T3SS in S. enterica, effector proteins such as SipA, SptP, and OrgC were lost, and only the chaperone protein SicA was retained in strains 2P24 and F113. Amino acid sequence analysis of the SPI-1 T3SS components in strain 2P24 showed higher identity to strain F113 than to Salmonella (Table 2). The export apparatus (i.e., InvA, SpaP, SpaQ, SpaR, and SpaS), basal body proteins (InvG and PrgK), accessory protein (IagB), needle (PrgI), and chaperone protein (SicA) displayed >50% identity between S. enterica and strain 2P24, although the remaining accessory proteins (i.e., InvE, InvC, InvI, SpaO, PrgH, PrgJ, OrgA, and OrgB) showed less identity (between 25 and 50%). In addition, less conserved regulators and translocons suggested a changed function of SPI-1 T3SS in PGPR.

A few reports showed that the SPI-1 T3SS was present in some plant beneficial bacteria, especially in fluorescent Pseudomonas (Chauhan et al., 2011; Loper et al., 2012). To investigate the relationship of SPI-1 T3SS among beneficial and pathogenic bacteria, the 16S rRNA sequences and InvA amino acid sequences were compared among Pseudomonas, Salmonella, Shigella, Yersinia, and Burkholderia spp. Based on alignment of the 16S rRNA, the maximum likelihood method was used to construct a phylogenetic tree that depicted evolutionary distance among these T3SS-containing bacteria (Figure 2A). There were two clades in the tree, and all of the fluorescent Pseudomonas strains were located on a separate subbranch of one clade, which indicated a distant evolutionary relationship with mammalian pathogens. Then, the SPI-1 T3SS phylogenetic tree was constructed based on InvA amino acid sequences using the same method (Figure 2B). The phylogenetic tree contained two clades, and fluorescent Pseudomonas strains were located in a subbranch together with Shigella spp. and Salmonella spp. The difference in phylogenetic relationship between 16S rRNA and InvA suggested a horizontal transfer of SPI-1 T3SS from mammalian pathogens, such as Salmonella and Shigella, to plant beneficial fluorescent Pseudomonas.

HilA is a transcriptional activator, which regulates transcription of the SPI-1 T3SS apparatus genes directly (Ellermeier and Slauch, 2007). To verify whether HilA had a similar function in strain 2P24, the hilA deficient mutant 2P24ΔhilA was constructed, and the expression of five apparatus genes (invG, sipB, sipD, prgI, and prgK) of the SPI-1 T3SS complex in strains 2P24 and 2P24ΔhilA was detected by qRT-PCR. Compared to strain 2P24, the expression of these five apparatus genes was decreased significantly in 2P24ΔhilA (Figure 3). HilA of strain 2P24 had a transcriptional activation function similar to mammalian pathogens and regulated SPI-1 apparatus genes positively.

InvF, an AraC-type transcriptional activator, regulated the expression of genes that encoded the secreted effector molecules SipABCD, SigD, SptP, and SopE (Darwin and Miller, 1999; 2001). However, it seems that almost all effector homologs were lost in strain 2P24 (Figure 1). To identify the InvF regulon in strain 2P24, the invF deficient mutant 2P24ΔinvF was constructed, and we implemented transcriptome sequencing (GEO accession: GSE181272). A total of 127 significantly differentially expressed genes (DEGs) were obtained by comparing the transcriptional level of strain 2P24ΔinvF with 2P24 at 6 h and 12 h [Fold change ≥ 1.5; p value ≤ 0.05] (Figure 4A and Supplementary Table 1). KEGG analyses indicated that DEGs at 6 h and 12 h involved biosynthesis of secondary metabolites, biosynthesis of amino acids, and ABC transporters (Supplementary Figure 1). Fifty-two down-regulated genes were regarded as InvF regulon candidates. Among the InvF regulon candidates, 39 genes were selected for qRT-PCR verification and eight genes were confirmed to be regulated positively by InvF (Figure 4B).

Previous study found an InvF binding site in the promoter regions of sicA, sigD, and sopE that was recognized by InvF to activate downstream expression of genes (Darwin and Miller, 2001). To identify the InvF regulon more comprehensively, the intergenic regions of the genome of strain 2P24 were scanned for the InvF binding site using the FIMO website (see text footnote 1) (Grant et al., 2011). Eleven InvF binding sites were obtained at p value ≤ 1 × 10^–5 (Supplementary Table 2). The adjacent downstream genes of these binding sites were searched, and their expression in strains 2P24 and 2P24ΔinvF was verified by qRT-PCR (Supplementary Table 2). Six genes among them were regulated positively by InvF (Figure 4C). Finally, 14 genes were identified as InvF regulons through transcriptional and promoter analysis (Table 3). However, all of the InvF regulons are not homologous to any effector proteins in pathogenic bacteria.

A previous study found that strain 2P24 antagonized multiple plant pathogens (Wei et al., 2004a). To investigate whether SPI-1 T3SS was involved in the antagonism, the T3SS deficient mutant 2P24ΔinvE-C was constructed. Then, the antagonistic ability of strains 2P24, 2P24ΔhilA, 2P24ΔinvF, and 2P24ΔinvE-C against various plant pathogens was detected. 2P24ΔinvE-C and 2P24ΔhilA exhibited no significant difference with strain 2P24 in antagonism (Figure 5). 2P24ΔinvF also showed no significant difference with strain 2P24 in antagonism to a variety of pathogens, but it reduced antagonistic ability against F. graminearum significantly. We speculated that InvF might influence antagonism to F. graminearum indirectly by regulating the expression of some genes rather than the T3SS.

Bacteria can carry out single bacterial swimming and multiple bacterial swarming by the rotational movement of flagella. In addition, the expansion and contraction of type IV secretion system (T4SS) can drive twitching (Henrichsen, 1972). To explore the influence of SPI-1 T3SS in the motility of strain 2P24, we detected swimming, swarming, and twitching of strain 2P24 and the SPI-1 mutants. Strains 2P24, 2P24ΔhilA, 2P24ΔinvF, and 2P24ΔinvE-C exhibited similar swimming, swarming, and twitching motility under different concentrations of agar (Figure 6A). Statistical analysis showed that no significant difference was found between the strains (Figure 6B).

Bacteria perceive the change of chemical concentration in the environment to produce an approach or a retreat response. This chemotactic behavior usually depends on the motility of flagella and can help bacteria survive better (Chet and Mitchell, 1976). In this study, we tested the influence of SPI-1 T3SS in perceiving exogenous glucose. Strain 2P24 had a strong chemotactic response to glucose. However, 2P24ΔhilA, 2P24ΔinvF, and 2P24ΔinvE-C reduced the chemotactic response to glucose significantly (Figure 7). These results suggest that the SPI-1 T3SS of strain 2P24 is involved in chemotaxis independent on motility.

Biofilm formation of bacteria is related to many factors. Ding and Wang (2009) reported that the T3SS mutant of Yersinia pseudotuberculosis reduced its ability to form biofilm significantly. To verify whether SPI-1 T3SS affected the ability of strain 2P24 to form biofilm, we detected the biofilm of strain 2P24 and related mutants of SPI-1 after static incubation at 28°C for 12, 24, and 48 h in 2 mL centrifuge tubes. Compared with negative control, strain 2P24 formed obvious biofilm at 48 h (Figure 8). 2P24ΔinvF and 2P24ΔinvE-C showed no significant difference with wild type, but 2P24ΔhilA enhanced biofilm formation significantly (Figure 8) which suggest that HilA plays important role in biofilm formation via a transcriptional regulation on some genes. However, the mechanism remains to be explored further.

Previous study found that strain 2P24 triggered a ROS burst in N. benthamiana leaves (Liu et al., 2016). In this study, we detected whether SPI-1 T3SS was involved in triggering a ROS burst in N. benthamiana. After injection of N. benthamiana with wild type and SPI-1 mutants of strain 2P24 for 12 h, ROS was measured every 5 min for 10 h. Consistent with P. fluorescens Pf0-1, strain 2P24 triggered a strong ROS burst, and deficient mutants of hilA and invF did not influence this immunity (Figure 9). However, 2P24ΔinvE-C reduced the accumulation of ROS significantly (Figure 9). This indicated that the SPI-1 T3SS was involved in triggering a ROS burst in N. benthamiana.