Genome wide identification of genes required for bacterial plant infection by Tn-seq

Soft rot enterobacteria (Dickeya and Pectobacterium) are major pathogens that provoke diseases on plants of agricultural importance such as potato and ornamentals. Long term studies to identify virulence factors of these bacteria focused mostly on plant cell wall degrading enzymes secreted by the type II Out secretion system and the regulation of their expression. To identify new virulence factors we performed a Tn-seq genome-wide screen of a transposon mutant library during chicory infection followed by high-throughput sequencing. This allowed the detection of mutants with reduced but also increased fitness in the plant. Virulence factors identified differed from those previously known since diffusible ones (secreted enzymes, siderophores or metabolites) were not detected by this screen. In addition to genes encoding proteins of unknown function that could be new virulence factors, others could be assigned to known biological functions. The central role of the FlhDC regulatory cascade in the control of virulence was highlighted with the identification of new members of this pathway. Scarcity of the plant in certain amino acids and nucleic acids required presence of the corresponding biosynthetic genes in the bacteria. Their products could be targets for the development of antibacterial compounds. Among the genes required for full development in chicory we also identified six genes involved in the glycosylation of the flagellin FliC. We showed that absence of this modification reduces virulence of D. dadantii on celery. Since homologues of these genes are present in other Dickeya and Pectobacterium strains, this modification could be an important factor for soft rot enterobacteriale virulence. Author summary Identification of virulence factors of plant pathogenic bacteria has relied on the test of individual mutants on plants, a time-consuming method. New methods like transcriptomic or proteomic can now be used but they only allow the identification of genes induced during the infection process and non-induced genes may be missed. Tn-seq is a very powerful method to identify genes required for bacterial growth in their host. We used for the first time this method in a plant pathogenic bacteria to identify genes required for the multiplication of Dickeya dadantii in chicory. We identified about 100 genes with decreased or increased fitness in the plant. Most of them had no previously described role in bacterial virulence. We unveiled important metabolic genes and regulators of motility and virulence. We showed that D. dadantii flagellin is glycosylated and that this modification confers fitness to the bacteria during plant infection. Our work opens the way to the use of Tn-seq with bacterial phytopathogens. Test by this method of large collections of environmental pathogenic strains now available will allow an easy and rapid identification of new virulence factors.

the plant. Virulence factors identified differed from those previously known since diffusible 23 ones (secreted enzymes, siderophores or metabolites) were not detected by this screen. In  Table 1). The average number of reads per TA is 88 and 75, respectively. The results were 115 reproducible with a Pearson correlation coefficient of 72% (Fig. S1) The location of the 116 unique insertions showed an even distribution around the chromosome (Fig. 1A). For each 117 gene, we calculated a log 2 FC corresponding to a ratio between the measured number of reads 118 and the expected number of reads. The density plot (Fig. 1B) indicates that essential and non-119 essential genes are easily distinguishable, confirming the good quality of our Tn-seq libraries.

120
Indeed non essential genes are expected to have a positive FC, whereas essential genes are 121 expected to have none or few insertions.

122
Then, gene essentiality of the Tn-seq input libraries was determined by using the TRANSIT 123 software [29]. We decided to use the Hidden Markov Model (HMM) method which predicts 124 essentiality and non-essentiality for individual insertion sites since it has been shown to give 125 good prediction in datasets with density as low as 20% [29]. The HMM analysis led to the 126 identification of 665 genes essential for growth in LB (ES), representing 14% of the genes of 127 D. dadantii 3937, a number in the range of those found for this type of analysis with bacteria.
128 552 genes were categorized as Growth Defect genes (GD, i.e. mutations in these genes lead 129 to loss of fitness), 125 as growth advantage genes (GA, i.e mutations in these genes lead to 130 gain of fitness) and 3320 as non-essential genes (NE) (Table S2 and Fig. 1B)). Genes necessary for chicory leaf maceration. We used chicory leaf infection as a model to 138 identify D. dadantii genes required for growth in plant tissues. Biological duplicates were 139 performed to insure the reproducibility of the results. Each chicory was inoculated with 10 7 140 bacteria from the mutant pool and after 2 days more than 10 10 bacteria were collected from 141 the rotten tissue. Sequencing transposon insertion sites in these bacteria followed by the TPP 142 analysis indicated a density of unique insertion in TAs comparable to that of the input 143 datasets (23-24%). Surprisingly, the results were more highly reproducible than in LB with a 144 very high Pearson correlation coefficient of 98% (Fig. S1).

145
In order to test the statistical significance of the identified genes conferring to D. dadantii a 146 loss or a gain of fitness in planta, we performed the RESAMPLING (permutation test) 147 analysis of the TRANSIT software. Applied to our Tn-seq datasets and selecting only genes 148 achieving a FDR adjusted p-value (q-value) ≤ 0.05, we identified 122 genes out of 4666 149 required for fitness in planta, as shown with the volcano plot of RESAMPLING results 150 comparing replicates grown in LB versus in planta (Fig. S2). From these 122 genes, we 151 applied an additional cutoff by removing 20 genes with a mean read count in LB <5 (less 152 than 5 reads in average per TA). These genes were categorized as ES or GD in LB. We also 153 removed from the analysis 6 genes with a log 2 FC comprised between -2 and 2. By applying 154 all these criteria, we retained only 96 genes for a further analysis (Fig. 2). 92 of them were 155 identified as GD genes in the chicory (log 2 FC ≤2), the 4 left as GA genes in the chicory 156 (log 2 FC ≥2). Some of these genes, in bold in Fig. 2, were already known to play a role in D.  (Table S3). We   (Table S3). In the purine metabolism 166 pathway, the inosine monophosphate (IMP) biosynthesis pathway that produces IMP from L-167 glutamine and 5-phosphoribosyl diphosphate is particularly important for D. dadantii in 168 planta since 5 out of the 10 genes of this pathway were significantly GD genes in planta ( Fig.   169 3). IMP is the precursor of adenine and guanine. Next, IMP can be converted in xanthosine 170 5'-phosphate (XMP) by the IMP dehydrogenase GuaB. guaB gene was also a GD gene in 171 planta, with a strong log 2 FC of -10.06 (Fig. 3). In the pyrimidine synthesis, the uridine 172 monophosphate (UMP) biosynthesis pathway that converts L-glutamine to UMP, a precursor 173 of uracyl, is very important in planta since carAB, pyrB, pyrC and pyrE, involved in this 174 enzymatic pathway, were all required for growth in planta (Fig. 3). This pyrimidine 175 biosynthesis pathway is specific to bacteria. It is noteworthy that in the human pathogen S.

213
Their exact role is unknown but their absence leads to avirulence in certain bacteria such as 214 D. dadantii [34]. This absence induces a membrane stress that is sensed and transduced by 215 the Rcs envelope stress response system. This system controls the expression of many genes, 216 including those involved in motility, and those encoding plant cell wall degrading enzymes 217 through the RsmA-RsmB system [35][36][37]. Thus, mutants defective in OPG synthesis are 218 expected to have a reduced virulence. Indeed, in our experiment, mutants in the two genes 219 involved in OPG synthesis, opgG and opgH were non competitive in chicory (Fig. 2). unable to acquire iron and thus are unable to grow in the plant. In accordance, tonB was 228 essential in chicory while the genes coding for siderophore synthesis or secretion were not.

229
Similarly a mutant devoid of the iron-loaded chrysobactin transport gene (fct) was non-   The GGDEF proteins are c-di-GMP synthase. Their genes are often located next to their 247 cognate EAL diguanylate phosphodiesterase gene. ecpC (yhjH) encodes an EAL protein that 248 was shown to activate virulence factor production in D. dadantii [45]. gcpA, which is located 249 next to ecpC encodes a GGDEF protein. However, a gcpA mutant could not be constructed 250 and analysed in this previous study. We observed that gcpA mutants (Dda_03858) were 251 overrepresented in chicory (Fig. 2). This increased virulence is a phenotype opposite to that 252 described for the ecpC mutants, indicating that overproduction of c-di-GMP could reduce D. 253 dadantii virulence.

254
Of the eighteen regulators of the LacI family present in D. dadantii, four were found to be 255 involved in plant infection [46]. One of those, LfcR, which has been found important for 256 infection of chicory, Saintpaulia and Arabidopsis, was identified as important for chicory 257 infection in our experiment. LfcR is a repressor of adjacent genes [46]. Surprisingly none of 258 these genes appeared to play a role for chicory infection suggesting that other targets of LfcR 259 probably remain to be discovered.

260
Finally, it is noteworthy to mention that the ackA and pta genes were GD in planta. These

Additional genes could be involved in virulence 306
Several genes have a log 2 FC >4 or <-4 but do not satisfy the statistical permutation test 307 (Table S4). However, most of them belong to the categories described above and could be 308 required for growth in planta. Among those with a log 2 FC< -4 can be found genes involved 309 in amino acid and nucleic acid synthesis (cysH, ilvC, pyrF, pyrD, purC, thrC, metA, cysK,

325
To validate the Tn-seq results, we performed coinoculation experiments in chicory leaves 326 with the wild type strain and various mutants in genes conferring a growth advantage, a 327 growth defect or no fitness benefit in a 1/1 ratio. We calculated a competitivity index (CI) by 328 counting the number of each type of bacteria in the rotten tissue after 24 h. We found a 329 correlation between the competitivity index of mutants with the log 2 FC of the corresponding 330 genes with a Pearson coefficient of 0.50 (Fig. 6), indicating that Tn-seq is a reliable technique 331 to identify genes involved in plant colonization and virulence. Additional validations were 332 performed by inoculation of uncharacterized leucine and cytosine auxotroph mutants in 333 chicory leaves which confirmed that these mutants are unable to grow in plant and are thus 334 avirulent ( Fig S4).

341
This includes all the proteins secreted by the type II secretion system and small molecules 342 such as siderophores and butanediol. Other categories of genes were not found: for example, 343 no genes involved in response to acidic or oxidative stresses were identified. However, 344 chicory has been described as an inadequate model to study the response of D. dadantii to 345 oxidative stress [54]. Similarly, the type III hrp genes were not identified in our study. The

364
Bacterial strains and growth conditions. Bacterial strains, phages, plasmids and 365 oligonucleotides used in this study are described in Table S1. D. dadantii and E. coli cells 366 were grown at 30 and 37°C respectively in LB medium or M63 minimal medium 367 supplemented with a carbon source (2 g/L). When required antibiotics were added at the 368 following concentration: ampicillin, 100 µg/L, kanamycin and chloramphenicol, 25 µg/L.

369
Media were solidified with 1.5 g/L agar. Transduction with phage PhiEC2 was performed 370 according to [56]. The motility of each mutant was compared with that of the wild-type strain 371 on semisolid (0.4%) LB agar plates as previously described [57]. medium. 10 µL of the bacterial suspension were inoculated into leaves in a hole made with a 460 pipet tip. The wound was covered with mineral oil and the leaves were incubated at 30°C at 461 high humidity for 2 days (celery) or 24 h (for chicory). Length or rotten tissue was measured.

499
In red are indicated the growth defect genes in chicory that pass the permutation test (q-value 500 ≤ 0.05). The log 2 FC of read numbers between chicory and LB for each gene is indicated in 501 bracket. Some genes do not pass the permutation test (in black) but have a strong negative

551
Each leaf was inoculated with 10 6 bacteria. Length of rotten tissue was observed after 24h.