Identification of novel virulence factors in Erwinia amylovora through temporal transcriptomic analysis of infected apple flowers under field conditions.

The enterobacterial pathogen Erwinia amylovora uses multiple virulence-associated traits to cause fire blight, a devastating disease of apple and pear trees. Many virulence-associated phenotypes have been studied that are critical for virulence and pathogenicity. Despite the in vitro testing that has revealed how these systems are transcriptionally regulated, information on when and where in infected tissues these genes are being expressed is lacking. Here, we used a high-throughput sequencing approach to characterize the transcriptome of E. amylovora during disease progression on apple flowers under field infection conditions. We report that type III secretion system genes and flagellar genes are strongly co-expressed. Likewise, genes involved in biosynthesis of the exopolysaccharide amylovoran and sorbitol utilization had similar expression patterns. We further identified a group of 16 genes whose expression is increased and maintained at high levels throughout disease progression across time and tissues. We chose five of these genes for mutational analysis and observed that deletion mutants lacking these genes all display reduced symptom development on apple shoots. Furthermore, these induced genes were over-represented for genes involved in sulphur metabolism and cycling, suggesting the possibility of an important role for maintenance of oxidative homeostasis during apple flower infection.

INTRODUCTION

Progression of fire blight disease of apple and pear trees requires the causative agent Erwinia amylovora to coordinately express virulence traits and spread through host tissues (Piqué et al., 2015; Zhao et al., 2005). During infection, E. amylovora cells encounter several differing and hostile environments presented by specific host cell and tissue types that present distinct physical and biochemical challenges. Precise expression of genes involved in overcoming host defences and barriers under appropriate conditions leads to successful infection for E. amylovora.

Multiple specific virulence factors are important for successful infection of apple and pear trees by E. amylovora (Kharadi et al., 2021), including flagellar motility (Bayot & Ries, 1986; Raymundo & Ries, 1980), production of exopolysaccharides (Bellemann et al., 1994; Geider, 2000; Geier & Geider, 1993; Gross et al., 1992; Nimtz et al., 1996), biofilm formation (Koczan et al., 2009, 2011), catalase activity (Santander et al., 2018), and the type III secretion system (T3SS; Baker et al., 1993; Barny, 1995; Wei et al., 1992). When E. amylovora cells are deposited on the stigma of an apple flower, populations grow to densities of 106 to 107 colony‐forming units (cfu) on and between papillae cells (Malnoy et al., 2012; Pusey, 2000; Slack et al., 2022). Moisture from rain or dew enables the bacterial cells to migrate down to the nectaries, where they use flagella to swim through nectar to reach natural openings in the hypanthium (Bayot & Ries, 1986). On invasion at the base of the floral cup, E. amylovora cells must use the T3SS to suppress host defences and initiate pathogenesis (Bogdanove et al., 1998). During leaf infection at shoot tips, E. amylovora cells will also invade vascular tissues and produce the exopolysaccharides amylovoran, levan, and cellulose to structurally promote biofilm formation and population growth (Castiblanco & Sundin, 2018; Koczan et al., 2009, 2011). From all infected tissues during the disease cycle, sufficient internal populations will emerge in ooze droplets, ready to be disseminated to new hosts (Slack et al., 2017). In addition to these virulence traits that are important for navigation of and survival in host tissues, the bacteria must also be able to appropriately acquire and utilize the nutrients that are available in the occupied environment. Studies have found that iron acquisition via siderophores (Dellagi et al., 1998) as well as nutrient acquisition and metabolism are essential for full virulence (Ramos et al., 2014).

Although several studies have been conducted to understand E. amylovora transcriptional regulatory networks using in vitro growth conditions (Ancona et al., 2016; Li et al., 2014; McNally et al., 2012; Wang et al., 2009), studies of gene expression during infection are limited. Expression of T3SS genes was assessed by reverse transcription quantitative real‐time PCR (RT‐qPCR) during infection of apple flowers, and it was found that these genes were rapidly induced, and expression peaked between 24 and 48 h postinoculation (hpi) (Pester et al., 2012). A recent study conducted transcriptomic analysis of E. amylovora in shoots of two apple cultivars, one with low susceptibility and the other highly susceptible (Puławska et al., 2017). This study found that the greatest difference in virulence gene expression was manifest at 24 hpi, but little difference between cultivars was observed at 6 days postinoculation (dpi). Another study used an in vivo expression technology (IVET) approach to identify genes rapidly induced on inoculation to immature pears (Zhao et al., 2005). This approach identified novel virulence factors important for disease progression in this model. These studies provide a foundation on which further experiments can build to provide a transcriptome level view of E. amylovora virulence trait expression.

While these in vivo expression studies provide important groundwork, the studies were conducted under controlled greenhouse (Pester et al., 2012; Puławska et al., 2017) or laboratory conditions (Zhao et al., 2005). In this study, we inoculated a native strain of E. amylovora to apple flowers on trees in a research orchard, and analysed E. amylovora gene expression at the transcriptome level. Because of the complexity of different host tissues and environments, rather than sample whole flowers we sampled and sequenced RNA from the stigma, flower base, or pedicel of inoculated flowers at four sampling time points postinoculation. Our sequencing results matched previously reported expression patterns (Pester et al., 2012) for T3SS genes. We furthermore confirmed several hypothesized gene expression patterns across time as infection progressed. We identified novel genes induced on inoculation and chromosomal deletion of these genes resulted in reduced virulence. Our results suggest an important role for sulphur cycling, metabolism, and oxidative state during fire blight disease progression on flowers.

RESULTS

Sequencing and gene expression

Our RNA sequencing of infected flowers resulted in a total of 424.5 million reads across all samples. The attempt to deplete plant‐derived ribosomal and poly(A) transcripts was only moderately successful at early time points. In samples from apple flower stigmas at 12 hpi, only 0.1% of sequenced reads mapped to the E. amylovora ATCC 49946 genome (Figure S1). By 48 hpi, 1% of sequenced reads successfully mapped to the E. amylovora genome, and at 96 hpi, about 30% of sequenced reads mapped to the E. amylovora genome. Pure inoculum was used as evidenced by 98% of sequenced reads from the inoculum mapping to the E. amylovora genome. Principal component analysis of all biological replicates showed that all in planta samples separated clearly from inoculum samples and replicates clustered by tissue and time point (Figure 1a). Similarly, Pearson's correlation coefficients were high between biological replicates and hierarchical clustering also showed that replicates clustered by tissue and time point (data not shown).

FIGURE 1

Multidimensional analysis of gene expression during infection under field conditions. (a) Principal component plot demonstrating clustering by tissue and time point. Squares represent Erwinia amylovora strain Ea110 grown in Luria‐Bertani (LB) medium (inoculum); dash, 12 h stigma sample; circles, 24 h nectary samples; diamonds, 48 h nectary samples; triangles, 96 h pedicel samples. (b) Heat map of genes differentially regulated between cells grown in LB medium (inoculum) and cells in 96 h postinoculation (hpi) pedicel samples. Hierarchical clustering of genes divided genes into two broad groups, designated I and II, based on whether the gene of interest had increased or decreased mRNA abundance in inoculated samples compared to inoculum

Because of the low E. amylovora read counts at early time points, we were unable to determine differentially expressed genes using traditional analysis, but we were able to do so for later sampling time points. We generated a heat map across all sample replicates using the 394 genes differentially expressed between pedicel 96 hpi samples and the source inoculum (Figure 1b). We found that in our data set a majority of the differentially expressed genes were down‐regulated during flower infection, represented as group II in Figure 1b. Among the genes with greatest decrease in expression in flower pedicels were a sugar efflux transporter, setA, and several putative two‐component signalling system proteins. Because of the low sequencing depth in the inoculated flower samples, this group of down‐regulated genes may have biological relevance, but also could be an artefactual observation, and further experimentation is needed to resolve this possibility. However, there are many genes with low expression levels in the inoculum (growth in Luria‐Bertani [LB] medium) that increased at some point during flower infection, represented as group I in Figure 1b. The genes with the greatest increases in expression levels were primarily flagellar and T3SS genes. We selected 10 genes and conducted RT‐qPCR from independent RNA samples to validate the expression levels observed in our RNA sequencing (RNA‐Seq) experiments and found similar expression patterns between RNA‐Seq and RT‐qPCR (Figure S2).

T3SS and hrpL gene expression profile

Because expression patterns for some T3SS genes were previously determined during flower infection (Pester et al., 2012), we analysed the expression patterns for all T3SS genes in our sequencing data set. We found that our results mirrored those previously found using RT‐qPCR approaches, in which T3SS genes are rapidly induced on flower inoculation and expression starts to diminish as disease progresses to later stages (Figure 2a,b). Also, we found that the average induction of T3SS genes was about 10 times greater than that of the alternative sigma factor, hrpL, which is essential for expression of T3SS genes. This is consistent with the role of HrpL as a transcriptional regulator that is capable of signal amplification.

FIGURE 2

Temporal expression patterns of flagellar and type III secretion system (T3SS) genes during flower infection, relative to inoculum. (a) Average expression patterns for each T3SS cluster operon, labelled by the first gene in the operon. (b) Average expression pattern across all T3SS genes, solid line with error bars showing standard error, and hrpL expression pattern, dashed line. (c) Average expression pattern of each of the six flagellar gene clusters in the ATCC 49946 chromosome. (d) Average expression pattern of all flagellar genes with error bars showing standard error

Flagellar gene expression profiles

Similar to T3SS genes, we found that expression of flagellar genes was also rapidly and strongly induced on inoculation on flowers. Like the expression of T3SS genes, we found that expression of flagellar genes began to diminish at 96 hpi in the pedicel (Figure 2c). Although T3SS genes are organized in a single cluster in the genome, the flagellar genes of E. amylovora are spread across six gene clusters in the genome. Some of these clusters represent duplications of several flagellar genes (Koczan et al., 2011). By separating the average expression level of each flagellar cluster individually, we see that flagellar clusters 1 (EAM_1436–EAM_1449; ATCC 49946 genome), 2 (EAM_1484–EAM_1498), and 4 (EAM_2062–EAM_2071) showed increased expression during flower infection, but clusters 3 (EAM_2019–EAM_2034), 5 (EAM_2537–EAM_2561), and 6 (EAM_2568–EAM_2588) only showed small changes in expression throughout infection (Figure 2d). Interestingly, flagellar clusters 5 and 6 are duplicates of clusters 2 and 3, respectively, and are not involved in flagellar motility (Zhao et al., 2010) but may have roles in surface attachment (Koczan et al., 2011).

Because it has been reported that flagellar and T3SS genes are co‐expressed during shoot infection (Puławska et al., 2017), we conducted a linear regression to test for correlation between expression of the T3SS genes and each of the flagellar gene clusters (Figure S3). We found that expression of flagellar cluster 1 had the strongest correlation to T3SS expression (R 2 = 0.588), and that clusters 2 and 4 had weak correlations (R 2 = 0.338 and R 2 = 0.109, respectively). In addition, flagellar clusters 3, 5, and 6 had essentially no correlation (R 2 < 0.05).

Exopolysaccharide and sugar utilization gene expression profiles

To succeed in building populations that will facilitate infection and overcome host defences, E. amylovora must efficiently utilize the available carbon sources. In apple tissues, the primary storage and transport sugar is sorbitol (Raese et al., 1977). In our expression data, we found that sorbitol utilization genes were induced on the flower stigma, then shut off when the cells were in the nectary area, and then expressed again in flower pedicels (Figure 3a,b). The down‐regulation of sorbitol utilization genes in the flower base and nectary of the flower corresponds to an abundance of sucrose, but not sorbitol, in nectar (Tóth et al., 2003).

FIGURE 3

Temporal expression patterns of select sugar‐related genes during flower infection, relative to inoculum. (a) Expression pattern for all sorbitol utilization genes. (b) Average expression pattern for all sorbitol utilization genes. (c) Expression pattern of all amylovoran biosynthesis genes. (d) Average expression patterns of amylovoran biosynthesis genes. Error bars represent standard error

Production of the exopolysaccharide amylovoran is essential for E. amylovora pathogenicity (Bernhard et al., 1993). In our expression data, we found that the amylovoran biosynthetic genes were induced on the stigma, shut off early in the flower base and then induced again after 48 h in the flower base and also in the pedicel (Figure 3c,d). This expression pattern is similar to that of the sorbitol utilization genes, and linear regression showed a positive correlation between expression profiles of sorbitol utilization and amylovoran biosynthesis genes (R 2 = 0.426; Figure S4a). To show that these correlations are specific, we observed no correlation between expression of amylovoran biosynthesis genes and T3SS genes (R 2 = 0.066; Figure S4b). Interestingly, genes involved in the regulation and biosynthesis of the exopolysaccharide levan, a homopolymer of fructose synthesized from sucrose, showed increased expression at 24 hpi in the flower base at the time when both amylovoran and sorbitol genes were down‐regulated (data not shown). Furthermore, genes involved in metabolism of sucrose were not up‐regulated in the flower base (data not shown), where sucrose would be abundant, suggesting that during infection of the flower base and nectary, E. amylovora is primarily converting sucrose to levan, but is not largely using the sucrose as a carbon source, consistent with the observation that under osmotic stress in nectar E. amylovora cells do not typically divide (Ivanoff & Keitt, 1941).

High expression under infection co‐expressors

We hypothesized that genes with expression that increased on apple flower infection, and where that increased expression was maintained across time points and tissues, were likely to represent genes that are important for E. amylovora during disease. Such genes are likely to be essential for successful infection by virtue of function as virulence factors and not simply due to major growth defects. To identify genes that match these criteria, we generated a numerical vector representing this expression pattern of interest and searched across all genes for genes with a statistically significant (p < 0.05) Pearson's correlation to the sample vector. We identified 16 genes with significant similarity to the target expression pattern, and corresponding expression profiles are represented in Figure 4. The identities of these genes and the putative functions of their products are specified in Table 1. Among these genes with induced and maintained expression we found waaL, an O‐antigen ligase gene that has previously been demonstrated to be a virulence factor for E. amylovora (Berry et al., 2009). Interestingly, a significant number of these genes (iscS, tpx, cysG, masA, and dsbA) are all involved in sulphur metabolism or maintenance of sulphur oxidation status (p < 0.0001 by Fisher's exact test). We selected five of the 16 genes (iscS, tpx, masA, puuE, and dsbA) to generate knockout mutants and characterize their impacts on virulence and virulence‐associated traits.

FIGURE 4

Expression pattern of 16 identified genes in which expression is induced and sustained during flower infection across time and tissues

TABLE 1

Erwinia amylovora genes with expression induced and maintained on apple flowers under field conditions

Gene name	Locus TAG	Putative function	Fold induction a
puuE	EAM_2305	Putrescine utilization	31.2
cysG	EAM_2691	Cysteine biosynthesis	14.0
tpx	EAM_1833	Thiol‐peroxidase	6.8
waaL	EAM_0086	O‐antigen ligase	5.8
rpsJ	EAM_3199	Ribosome small subunit protein	5.5
rpsF	EAM_0448	Ribosome small subunit protein	5.2
cspC	EAM_1967	Cold shock protein	5.0
iscS	EAM_2492	Iron‐sulphur cluster formation and repair	4.7
–	EAM_1467	Putative nucleotide‐binding protein	4.1
moaB	EAM_1219	Molybdenum cofactor	3.0
–	EAM_1468	Putative lipoprotein	2.7
dapD	EAM_0816	Lysine biosynthesis	2.3
aroQ	EAM_3140	Shikimate biosynthesis	2.3
dsbA	EAM_0025	Disulphide bond formation	2.1
masA	EAM_0886	Methionine salvage pathway	1.4
ppiB	EAM_1062	Proline isomerization, protein folding	1.4

Average fold induction across all tissues and time points.

Growth and virulence of knockout mutants

We hypothesized that genes with low expression in E. amylovora cells grown in LB medium but highly expressed during flower infection may be important for virulence, but dispensable to growth. Therefore, we tested the deletion mutant strains for effects on growth and virulence. When grown in LB medium, only the ∆iscS deletion mutant displayed slower growth than wild‐type E. amylovora strain Ea1189 (Figure S5). The ∆iscS mutant had a longer lag phase and slower entry to exponential growth, but ultimately reached the same OD600 as the wild‐type cells.

When inoculated onto immature pear fruits, the ∆dsbA mutant exhibited reduced development of disease symptoms at 3 dpi (Figure 5a). When inoculated onto apple shoots, all of the deletion mutants tested had reduced symptom development relative to wild‐type E. amylovora strain Ea1189 at 7 dpi (Figure 5b). In apple shoots, the ∆iscS deletion mutant had nearly no symptom development, even 7 dpi. The remaining mutants displayed initial symptom progression similar to wild‐type E. amylovora strain Ea1189, but later disease progression was delayed. The ∆masA and ∆dsbA mutants started to show reduced symptom development 4 dpi, and the ∆puuE and ∆tpx mutants did not manifest reduced symptom development until 6 and 7 dpi (Figure 5b). The virulence of all complemented mutant strains was restored to wild‐type levels (Figure S9).

FIGURE 5

Virulence of indicated deletion mutant strains when inoculated onto (a) immature pears or (b) apple shoots. Asterisks indicate significant difference (p < 0.05) from wild‐type Erwinia amylovora Ea1189 by Student's t test

Mutants and oxidative stress

Because all of the deletion mutants exhibited reduced symptom development when inoculated on apple shoots, and several of the deletion mutants are predicted to function in sulphur cycling or maintenance of sulphur oxidation state, we tested whether these mutants have altered catalase activity or altered susceptibility to exogenous hydrogen peroxide. We observed significantly reduced catalase activity in the ∆puuE, ∆tpx, and ∆iscS mutants, but only the ∆tpx and ∆iscS mutants were more susceptible to exogenous hydrogen peroxide than wild‐type E. amylovora strain Ea1189 (Figure 6). Susceptibility to exogenous hydrogen peroxide was diminished, and catalase activity was partially restored in complemented strains (Figure S10).

FIGURE 6

Deletion mutants have reduced catalase activity and ΔiscS has increased susceptibility to exogenous hydrogen peroxide. (a) Relative intracellular catalase activity of cells grown overnight in Luria‐Bertani (LB) medium. (b) Zone of inhibition of indicated strains around a disk containing hydrogen peroxide on solid LB medium. Asterisks indicate significant (p < 0.05) difference compared to wild‐type Erwinia amylovora Ea1189 by Student's t test

Mutants and other virulence‐associated traits

We further assessed the effects of mutations in these genes of interest by evaluating the mutant strains for a variety of virulence‐associated traits. The ∆masA deletion mutant produced significantly more (p < 0.05) amylovoran and the ∆iscS and ∆dsbA deletion mutants produced significantly less amylovoran than wild‐type E. amylovora strain Ea1189 (Figure 7a). The ∆tpx mutant exhibited significantly increased levansucrase activity, but the ∆iscS and ∆puuE mutants exhibited significantly reduced levansucrase activity relative to wild‐type E. amylovora strain Ea1189 (Figure 7b). When assessed for biofilm formation, growth of the ∆masA and ∆dsbA mutants resulted in significantly greater crystal violet staining of adherent cells (Figure 7c). Only the ∆dsbA mutant had a significant reduction in swimming motility relative to wild‐type E. amylovora strain Ea1189 (Figure 7d). Functions of virulence‐associated traits returned to levels equivalent to Ea1189 in complemented mutant strains, except for ∆puuE+pBBR1MCS5 puuE::puuE (Figure S11).

FIGURE 7

Deletion of genes affecting virulence also affects other virulence‐associated traits. (a) Amylovoran production by cells grown in modified basal medium A (MBMA). (b) Levansucrase activity secreted into supernatants of cultures grown in Luria‐Bertani medium. (c) Biofilm formation as assessed by crystal violet staining following 48 h of static growth in MBMA. (d) Swimming motility of indicated strains stab‐inoculated into soft agar (0.25% wt/vol) and incubated for 24 h. Asterisks indicate significant (p < 0.05) difference from wild‐type E. amylovora Ea1189 by Student's t test

DISCUSSION

We report novel E. amylovora single‐gene deletion mutants impacting virulence and virulence factors that were identified through a temporal transcriptomic analysis of E. amylovora gene expression during disease progression on apple flowers in the field. This analysis also agrees with previously reported T3SS gene expression patterns during flower infection, as well as hypotheses for the roles of virulence traits during early stages of infection, summarized in Figure 8.

FIGURE 8

Representation of apple flower tissues and heat map of average expression across time of virulence‐associated traits. The type III secretion and flagellar systems are strongly induced early during infection on flower stigmas and in the flower base, but expression decreases later during infection in the pedicel. Amylovoran production and sorbitol utilization genes are initially expressed on the stigma, but decreased in expression in the flower base and increased again in flower pedicels. We identified a group of novel virulence genes, which includes several genes involved in sulphur cycling and maintenance of sulphur‐oxidation status, probably through helping Erwinia amylovora to cope with oxidative stresses. These genes are induced by E. amylovora during flower infection and high expression is sustained across the tissues and time points sampled

Genes encoding the T3SS and flagellar systems were induced during early stages of infection on flower stigmas and in the flower base, but expression decreased during later infection stages in the pedicel. This expression pattern, along with the correlation between expression of flagellar genes and T3SS genes, supports the infection model in which E. amylovora must actively migrate down the flower stigma, swim through nectar, and then suppress host defences to invade the base of the flower through natural openings. Although the T3SS and flagella may continue to be expressed during later stages of infection, their highest expression occurred during early‐stage infection, suggesting that they are of greatest importance during these infection stages.

RNA‐Seq expression patterns indicate that amylovoran production and sorbitol utilization genes were initially expressed on the stigma, but then expression decreased in the flower base and increased again in flower pedicels. Because of the sucrose levels in the nectary, which can range from 10% to 50% (Tóth et al., 2003), we hypothesize that E. amylovora has little need for sorbitol utilization genes in this environment and this is the reason for decreased expression. A correlation between sorbitol utilization and amylovoran production was previously observed (Geider et al., 1996; Wang et al., 2010) and our data support this observation. We hypothesize that in nectar, where sucrose levels are high, the bacteria do not need to produce amylovoran because they are not interacting directly with host cells nor actively forming biofilms. Whether this correlation is due to co‐regulation of sorbitol utilization and amylovoran production genes must still be determined.

Among genes with consistently increased expression in all tissues and time points sampled, we identified a group of novel virulence‐associated genes, which includes several genes involved in sulphur cycling and maintenance of sulphur‐oxidation status, possibly through helping E. amylovora to cope with oxidative stresses. These genes were induced by E. amylovora during flower infection and high expression was sustained across the tissues and time points sampled. Testing of these genes that were highly expressed throughout flower infection confirmed that deletion of these genes results in virulence defects.

We constructed single deletion mutants of five genes with consistently higher expression across tissues and time points during infection of apple flowers: ∆dsbA, ∆iscS, ∆masA, ∆puuE, and ∆tpx. Decreased symptom development was observed when each of these deletion mutants was inoculated on susceptible apple shoots, but only the ∆dsbA mutant showed virulence defects relative to wild‐type E. amylovora Ea1189 when inoculated on immature pears.

Disulphide bond formation protein A (DsbA) enzymatically carries thiol disulphide oxidoreductase activity and plays an important role in the appropriate formation of disulphide bonds in other proteins in the periplasm (Akiyama et al., 1992; Wunderlich & Glockshuber, 1993). Several secreted proteins rely on DsbA for disulphide bond formation to facilitate proper secretion (Schierle et al., 2003). Deletion of dsbA resulted in reduced virulence on immature pears and on apple shoots. The ∆dsbA mutant was also affected in flagellar motility, a known effect of losing dsbA function in Escherichia coli (Dailey & Berg, 1993). In E. amylovora, deletion of dsbA resulted in low amylovoran production and high crystal violet staining in a biofilm assay, suggesting this mutant may affect attachment structures because deletion of the RNA chaperone hfq also results in reduced amylovoran production but high crystal violet staining in biofilm assays due to a hyperattachment phenotype (Zeng et al., 2013).

The iron‐sulphur cluster synthesis protein IscS is a cysteine desulphurase and a critical enzyme in formation of iron‐sulphur clusters for loading into many other cellular enzymes (Schwartz et al., 2000; Takahashi & Nakamura, 1999; Urbina et al., 2001). With its enzymatic function, IscS plays an important role in repair of oxidatively damaged iron‐sulphur clusters, and E. coli iscS deletion mutants exhibit increased susceptibility to hydrogen peroxide (Rogers et al., 2003; Yang et al., 2002). We also observed increased susceptibility to hydrogen peroxide in the E. amylovora ∆iscS mutant as well as decreased catalase activity. IscS is also required for thiamin biosynthesis in E. coli, specifically involved in biosynthesis of the thiazole moiety of thiamin via sulphur transfer to the C‐terminal carboxylate of ThiS, a small sulphur‐carrier protein (Begley et al., 1999; Lauhon & Kambampati, 2000). We have previously shown that thiamin biosynthesis in E. amylovora stimulates the function of the tricarboxylic acid cycle, thereby providing energy requirements for amylovoran biosynthesis (Yuan et al., 2021). The ∆iscS mutant exhibited decreased production of the exopolysaccharides amylovoran and levan, a slight growth defect in LB medium, and reduced catalase activity relative to wild‐type cells. Although it did not display reduced symptom development when inoculated on immature pears, the ∆iscS mutant was avirulent on apple shoots. Recent work indicates that oxidative stress due to hydrogen peroxide affects E. amylovora survival (Santander et al., 2018). Given this result, we hypothesize that the loss of virulence in the ∆iscS mutant in apple shoots is due to inability to cope with oxidative stress from host defences.

As determined for DsbA and IscS, thiol peroxidase Tpx is involved in cycling of the sulphur oxidative state (Baker & Poole, 2003; Kim et al., 1999). The E. amylovora tpx deletion mutant did not have a virulence defect in immature pears, but it was reduced in virulence in apple shoots. Deletion of tpx also had a small but significant effect on catalase activity and susceptibility to hydrogen peroxide. This is consistent with Tpx roles reported in Salmonella typhimurium, where deletion of tpx increases susceptibility to hydrogen peroxide and decreases bacterial survival in macrophages (Horst et al., 2010).

In addition to DsbA, IscS, and Tpx, which have direct roles in sulphur redox chemistry, MasA participates in sulphur recycling through the methionine salvage pathway (Albers, 2009). Sulphur utilization and sulphur stress are linked to ability to cope with oxidative stresses. We hypothesize that these genes that are up‐regulated during host infection and are involved in sulphur cycling play critical roles during disease development by enabling E. amylovora to cope with host‐derived reactive oxygen species. This hypothesis is supported by the observed decreases in virulence as well as the increased susceptibility to hydrogen peroxide in the ∆tpx and ∆iscS deletion mutants.

The ∆puuE deletion mutant does not have direct links to sulphur metabolism, and it is possible that the PuuE protein is playing another important role in disease progression on apple flowers and shoots. It was recently demonstrated that putrescine utilization is important for disease progression in Ralstonia solanacearum during xylem colonization in tomatoes (Lowe‐Power et al., 2018). It is possible that putrescine uptake and utilization in E. amylovora is similarly important during infection because both E. amylovora and R. solanacearum colonize host xylem.

In summary, we identified five new deletion mutants with reduced virulence on apple shoots in this work through gene expression pattern analysis of E. amylovora RNA‐Seq data during apple flower infection under field conditions. We specifically identified genes based on an expression pattern in which RNA abundance was low in cells grown in LB medium and used for plant inoculations, but high and sustained across apple flower tissues and time points. Future efforts to characterize the specific functions that affect the virulence of the proteins encoded by these genes will provide greater understanding of the roles they play that are important during infection but dispensable during growth in rich medium. Because we only identified novel factors based on a single expression pattern, our gene expression data set can be further used to identify additional novel genes and traits that are important during specific stages of fire blight disease progression.

EXPERIMENTAL PROCEDURES

Culture conditions, growth, and plasmids

E. amylovora strains were grown routinely using LB medium at 28°C. For field inoculations, the Michigan strain E. amylovora Ea110 (Zhao et al., 2005) was used. For generation of genetic mutants and virulence trait analysis, E. amylovora strain Ea1189 (McNally et al., 2012) was used. Gene deletion mutants were generated using a Red‐recombinase approach as described (Datsenko & Wanner, 2000). Bacterial strains and oligonucleotides used in this work are included in Tables 2 and S2, respectively. Deletion mutants were complemented by constructing plasmids from the pBBR1MCS‐5 plasmid with gentamicin resistance backbone and gBlocks (IDT‐DNA) containing the gene of interest and the native promoter.

TABLE 2

Bacterial strains used in this work and their relevant characteristics

Bacterial strains	Relevant characteristics	Source or reference
Erwinia amylovora
Ea110	Michigan‐native wild‐type	Ritchie and Klos (1976)
Ea1189	Wild‐type strain for genetic studies	GSPB a
Ea1189 ∆dsbA	dsbA deletion mutant	This work
Ea1189 ∆iscS	iscS deletion mutant	This work
Ea1189 ∆masA	masA deletion mutant	This work
Ea1189 ∆puuE	puuE deletion mutant	This work
Ea1189 ∆tpx	tpx deletion mutant	This work
Ea1189 ΔdsbA + pBBR1MCS5 dsbA::dsbA	dsbA complemented mutant, native promoter	This work
Ea1189 ΔiscS + pBBR1MCS5 iscS::iscS	iscS complemented mutant, native promoter	This work
Ea1189 ΔmasA + pBBR1MCS5 masA::masA	masA complemented mutant, native promoter	This work
Ea1189 ΔpuuE + pBBR1MCS5 puuE::puuE	puuE complemented mutant, native promoter	This work
Ea1189 Δtpx + pBBR1MCS5 tpx::tpx	tpx complemented mutant, native promoter	This work
Ea1189 + pBBR1MCS5	Control, empty vector GmR	Castiblanco(unpublished)
Escherichia coli
DH5α	F−80dlacZ1M151(lacZYA‐argF)U169 endA1 recA1 hsdR17 (rk −mk +)deoR thi‐1 supE44 gyrA96 relA1λ−	Invitrogen
Plasmids
pBBR1MCS5	Broad‐host‐range cloning vector, GmR	Kovach et al. (1995)

GSPB, Göttinger Sammlung phytopathogener Bakterien, Göttingen, Germany.

Field inoculations, cultivar, and sampling

Apple flowers of the cultivar McIntosh were inoculated with a 1‐µl droplet of E. amylovora Ea110 that was previously grown overnight in liquid LB medium. All flowers were inoculated on the day immediately following flower opening. For RNA‐Seq samples, inoculations occurred on 6 May 2019 at 11:00 (0 hpi). Weather conditions corresponding to the time course of the experiment can be found in Table S2. Four biological replicates were collected for each tissue × time point combination and some samples were later discarded due to low RNA quality. For stigmas, each biological replicate consisted of inoculated stigmas pooled from 15 to 20 flowers. For flower base, each biological replicate was a pool from four inoculated flowers. Biological replicates of pedicel samples were pooled from two pedicels. Stigma inoculations were conducted using a cell density of 109 cfu/ml (106 cfu inoculated per flower) and stigma samples were collected 12 hpi. For all other samples, the flower nectary was inoculated with a 1‐μl droplet with cell density of 108 cfu/ml (105 cfu/flower). Of these samples, the base of the flower was sampled at 24 and 48 hpi. For these samples the base was considered to be the remaining portion of the flower following removal of petals, anthers, stigmas, and pedicels. Inoculated flowers were monitored for emergence of ooze droplets from flower pedicels. When ooze appeared, the ooze droplet was removed and the pedicel was sampled. This occurred at approximately 96 hpi.

RNA extractions, library preparation, and sequencing

Total RNA was isolated from infected plant samples using the EZNA Plant kit (Omega) according to the manufacturer's recommendations, including the on‐column DNase treatment, which was carried out using Turbo DNase I (Invitrogen). RNA was checked for quality and quantity using Caliper LabChip GX (Perkin Elmer) and Qubit RNA HS (Thermo Fisher Scientific) analysis, respectively. To enrich RNA samples for bacterial messenger RNA, multiple depletion reagents were used, including oligo(dT) beads to deplete eukaryotic polyadenylated mRNA, RiboZero Plant rRNA removal beads to deplete host rRNA, and RiboZero bacterial rRNA removal beads to remove prokaryotic rRNA (Illumina). The remaining RNA was used for library preparation using an Illumina TruSeq Stranded Total RNA Library Preparation Kit (Illumina) on a Sciclone G3 robot following the manufacturer's recommendations (Perkin Elmer). Library quality was assessed using Caliper LabChipGX HS DNA (Perkin Elmer) and Qubit dsDNA HS (Thermo Fisher Scientific) assays. Samples were pooled and run on three lanes using the Illumina HiSeq4000 platform with single‐end 50 bp format. Sequencing was conducted at the Michigan State University Research Technology Support Facility. Base calling was conducted by Illumina Real Time Analysis v. 2.7.6.

Sequence and computational analysis

RNA sequencing reads were trimmed using Trimmomatic (Bolger et al., 2014) and mapped to the E. amylovora ATCC 49946 genome (Sebaihia et al., 2010) using bowtie2 (Langmead & Salzberg, 2012). Mapped reads were counted in each annotated gene using HTseq (Anders et al., 2015). Read counts per gene by sample were normalized to total sample read counts, using total reads mapping to the E. amylovora genome as a proxy for bacterial population. Normalized read counts were statistically analysed for differential gene expression using the R package DESeq (Anders & Huber, 2012). For gene expression pattern analysis, Pearson's correlation coefficients and associated test statistics were calculated using R and the vector used to identify consistently up‐regulated genes of interest was [1,10,10,10,10]. Principal component analysis, clustering analyses, and heat map generation were conducted using ClustVis (Metsalu & Vilo, 2015). RNA‐Seq results were confirmed using RT‐qPCR of 10 genes (Table S2) following the methods of Slack et al. (2017).

Virulence assays

Immature pears were inoculated as described previously (Zhao et al., 2005). Immature pears were wounded and inoculated with 103 cfu in a 1‐µl droplet and incubated in high‐humidity chambers at 28°C. Necrotic or water‐soaked lesion diameters were measured 4 dpi. Pear experiments were repeated three times with a total of at least 15 replicates per strain. Actively growing shoots of 2‐year‐old potted apple trees cultivar Gala were inoculated as described (Koczan et al., 2009). Briefly, shoots were cut with sterile scissors dipped in a bacterial suspension of 108 cfu/ml. Necrotic symptom development was monitored every 24 h and the distance from inoculation to the end of visible symptoms was measured at the indicated time points. Each replicate of a strain was inoculated onto a different tree. The shoot assay was repeated twice with at least six shoots per strain per experiment.

Oxidative stress testing

Catalase activity was measured as described (Iwase et al., 2013). Susceptibility to exogenous hydrogen peroxide was assessed through a described zone of inhibition approach (Santander et al., 2018). Briefly, overnight cell cultures were adjusted to an OD600 of 0.2 and 100 µl of cells were plated onto solid LB medium. A filter paper disc was placed on each plate and treated with 10 µl of 8 M H2O2. Following 24 h of growth at 28°C, plates were imaged and the zone of inhibition around each disc was measured using ImageJ (Abràmoff et al., 2004). Each phenotype was tested four times with at least three replicates per strain.

Virulence‐associated trait analysis

Virulence‐associated traits were assessed using established methods. Virulence phenotypes were tested three times independently for a total of at least 12 replicates per strain. Production of the exopolysaccharide amylovoran was assessed following growth for 48 h in modified basal medium A (MBMA; 3 g/L KH2PO4, 7 g/L K2HPO4, 1 g/L (NH4)2SO4, 2 ml/L glycerol, 0.5 g/L citric acid, 0.03 g/L MgSO4, 10 g/L sorbitol) at 28°C as described (Bellemann et al., 1994; Edmunds et al., 2013; Zhao et al., 2009). To quantify levansucrase activity, culture supernatants were mixed in a 1:1 ratio with 2 M sucrose, incubated for 24 h, and the resulting turbidity was measured as described (Geier & Geider, 1993). Biofilm formation for each strain was determined using 96‐well microtitre plates as described (Santander & Biosca, 2017). Swimming motility was assessed following stab inoculation of cells into soft‐agar plates (0.25% wt/vol) using a described method (Zeng & Sundin, 2014).

CONFLICT OF INTEREST

The authors declare that they have no conflicts of interest.

AUTHOR CONTRIBUTIONS

J.K.S. and G.W.S. conceived of and designed the experiments. J.K.S., K.G., and I.P. conducted the experiments. J.K.S., K.G., I.P., and G.W.S. analysed the data, and wrote and edited the manuscript.

FIGURE S1 Schematic sampling of infected tissues across time. Inoculum was grown in Luria‐Bertani medium overnight. Stigmas were pooled from 15 to 20 flowers to represent each biological sample. Flower bases were pooled from three flowers to represent each biological sample. Following the 48‐h postinoculation (hpi) sampling, the remaining infected flowers were monitored each morning for extrusion of bacterial ooze. When bacterial ooze was observed on the pedicel, the pedicel was sampled. This occurred at approximately 96 hpi. The ooze droplet was not included in the pedicel sample. Pedicel samples were pooled from three flowers to constitute each biological sample

Click here for additional data file.

FIGURE S2 Bacterial RNA‐Seq read depth increases as disease progresses and bacterial populations grow. (a) Percentage of reads mapping to the Erwinia amylovora ATCC 49946 genome. (b) Total genomic depth of reads mapping to the E. amylovora genome

Click here for additional data file.

FIGURE S3 Heatmap of Pearson correlations between samples and replicates

Click here for additional data file.

FIGURE S4 Reverse transcription quantitative PCR confirmation of RNA‐Seq data from independent experiments

Click here for additional data file.

FIGURE S5 Co‐expression between flagellar gene clusters and type III secretion system genes. Linear regression of average expression levels in each sample between flagellar cluster genes and type III secretion system genes

Click here for additional data file.

FIGURE S6 Amylovoran co‐expresses with sorbitol utilization genes, but not type III secretion system genes. Linear regression of average expression levels in each sample between (a) sorbitol and amylovoran or (b) amylovoran and type III secretion system genes

Click here for additional data file.

FIGURE S7 Genomic contexts of genes targeted for deletion in study

Click here for additional data file.

FIGURE S8 The ∆iscS deletion mutant is the only deletion mutant generated that had a growth defect in Luria‐Bertani medium, and this was a mild growth defect

Click here for additional data file.

FIGURE S9 Virulence of complemented mutant strains when inoculated onto (a) immature pears or (b) apple shoots. Data are averages of all replicates and error bars represent standard deviation. dsbA::dsbA in panel A for strain Ea1189ΔdsbA + pBBR1MCS5 dsbA::dsbA

Click here for additional data file.

FIGURE S10 Phenotypes of complemented mutant strains. (a) Relative intracellular catalase activity of cells grown overnight in Luria‐Bertani (LB) medium. (b) Zone of inhibition of indicated strains around a disk containing hydrogen peroxide on solid LB medium. Data are mean of all replicates and error bars represent standard error. Provided p values from t test between mutant strain and wild‐type Erwinia amylovora Ea1189. iscS::iscS for strain Ea1189ΔiscS+ pBBR1MCS5iscS::iscS, puuE::puuE abbreviation of strain Ea1189ΔpuuE+ pBBR1MCS5puuE::puuE, tpx::tpx abbreviation of strain Ea1189Δtpx+ pBBR1MCS5tpx::tpx

Click here for additional data file.

FIGURE S11 Phenotypes of complemented mutant strains. (a) Amylovoran produced by cells grown in modified basal medium A (MBMA). (b) Levansucrase activity of supernatants of cultures grown in Luria‐Bertani medium. (c) Biofilm formation as assessed by crystal violet staining following growth in MBMA. (d) Swimming motility of indicated strains stab‐inoculated into soft agar. Data are mean of all replicates and error bars represent standard error. Provided p values from t test between mutant strain and wild‐type Erwinia amylovora Ea1189. dsbA::dsbA for strain Ea1189ΔdsbA+ pBBR1MCS5dsbA::dsbA, iscS::iscS for strain Ea1189ΔiscS+ pBBR1MCS5iscS::iscS, masa::masA for strain Ea1189ΔmasA+ pBBR1MCS5masA::masA, puuE::puuE abbreviation of strain Ea1189ΔpuuE+ pBBR1MCS5puuE::puuE, tpx::tpx abbreviation of strain Ea1189Δtpx+ pBBR1MCS5tpx::tpx

Click here for additional data file.

TABLE S1 List of genes differentially regulated in 96‐h pedicel samples compared to inoculum

Click here for additional data file.

TABLE S2 Oligonucleotides/primers used in this work

Click here for additional data file.

TABLE S3 Hourly weather conditions during RNA‐Seq experiments

Click here for additional data file.