Non-pathogenic leaf-colonising bacteria elicit pathogen-like responses in a 1 colonisation density-dependent manner

by diverse

H3K4me3, H3K36me3 and lysine acetylation were found to contribute to the induction of 91 genes following pathogen exposure (Berr et al., 2012;Ding & Wang, 2015). By contrast, Plant growth: Plants were grown as previously described in (Miebach et al., 2020). Briefly, 116 sterilised seeds were germinated on ½ MS (Murashige and Skoog medium, including 117 vitamins, Duchefa, Haarlem, Netherlands) 1% phytoagar (Duchefa) filled pipette tips. 118 Healthy looking seedlings were aseptically transferred, without removal from the pipette 119 tip, aseptically into Magenta boxes (Magenta vessel GA-7, Magenta LLC, Lockport, IL, USA) 120 filled with ground Zeolite (sourced from cat litter -Vitapet, Purrfit Clay Litter, Masterpet 121 New Zealand, Lower Hutt, New Zealand) and watered with 60 mL ½ MS. Each box received 122 four seedlings. The boxes were closed with lids that allowed for gas exchange and placed 123 into a climate cabinet (85% relative humidity, 11 h light, 13 dark, 21 °C, 150-200 µmol light 124 intensity). Plants were grown for four weeks for the time course and six weeks for the 125 bacterial density experiment before they were treated with bacteria or mock controls. 126 127 Plant inoculation: Bacterial suspensions were prepared as previously described in 128 (Miebach et al., 2020). Briefly, bacteria were cultivated at 30 °C on minimal media agar 129 plates containing 0.1% pyruvate as a carbon source. Bacterial suspensions were prepared 130 from bacterial colonies suspended in phosphate-buffered saline (PBS, 0.2 g L −1 NaCl, 1.44 g 131 L −1 Na2HPO4 and 0.24 g L −1 KH2PO4) and washed twice via centrifugation at 4000 × g for 5 132 min followed by discarding the supernatant and again adding PBS. Table 1 contains the list 133 of bacteria used in this study. For the time course experiment the optical density (OD600 nm) 134 was adjusted so that the suspension contained 2 × 10 7 colony forming units (CFU) ml -1 . To 135 explore the influence of bacterial load on plant responses, the bacterial suspensions were 136 adjusted to 10 5 , 10 6 , 10 7 and 10 8 CFU ml -1 . Next, 200 µL (time course experiment) or 1 ml 137 (bacterial load experiment) of bacterial solution was sprayed per plant tissue culture box 138 using an airbrush spray gun (0.2 mm nozzle diameter, Pro Dual Action 3 #83406). To obtain 139 a homogeneous coverage, the distance between the airbrush spray gun and the plants was 140 increased by stacking a plant tissue culture box, with the bottom cut off, onto the plant 141 tissue culture box containing the plants being spray-inoculated. Primers that were first used in this study were designed using 'primer-blast' (NCBI,177 Bethesda, MD, USA). Primer efficiencies were determined via serial template dilutions 178 (Nolan et al., 2013). The mRNA concentration of each target gene was then normalised 179 against the mean mRNA concentration of two stably expressed, previously described 180 reference genes ( Table 2, (Czechowski et al., 2005)). Next, the normalised mRNA   Discovery Rate (FDR; Benjamini-Hochberg p correction) cut-off of 5% were kept as 210 differentially expressed genes (DEGs). The FC threshold was determined using elbow plots 211 ( Fig. S3a, Fig. S4a). MDS plots were generated from trimmed mean of M values method 212 normalised gene counts. K-means clusters were calculated from log2-transformed counts 213 per million, that were centred around the mean for each gene. A prior count of two was 214 added to each gene count to prevent taking the logarithm of zero. Elbow plots were used to 215 determine the optimal number of Ks (Fig. S3b, Fig. S4b). Heatmaps were generated from 216 log2-transformed counts per million (cpm), that were centred around the mean for each 217 gene. GO term enrichment analysis was performed using the PANTHER classification

235
The aim of this study was to broaden our knowledge of the intricate relationship between 236 the plant and its bacterial colonisers. The focus lay on assessing plant transcriptional 237 responses to a diverse array of microbial colonisers. Therefore, six microbial leaf 238 colonisers representing all major phyla of the core leaf microbiota were selected (Vorholt,

254
The temporal course of bacterial densities post inoculation confirmed that bacteria were 255 sprayed close to carrying capacity. Within 4 dpi the bacterial densities for Sphingo34 and 256 Micro347 remained stable at 10 6 bacteria per g of leaf. In contrast, the bacterial densities for 257 Willi354 and Pst slightly rose to 10 7 bacteria per g of leaf within 4 dpi (Fig. 2a). Interestingly, 258 the bacterial density of Micro347 significantly (p < 0.001, Tukey's HSD test) dropped to ~ 10 4 259 bacteria per g of leaf 7 dpi. This two-magnitude drop in bacterial density cannot be 260 explained by the increase in plant weight (Fig. S1) and, therefore, suggests that the bacteria 261 were dying.  Two out of the six bacterial leaf colonisers, Acido84 and Pedo194, failed to consistently 277 establish densities above the threshold of detection. Acido84 was recovered from some, 278 but not all plants. Whenever Acido84 was recovered it reached densities of ~ 10 5 -10 6 CFU 279 g -1 . This heterogeneity in colonisation success was unlikely to have been caused by non-280 homogeneous spray inoculation, as other inoculants exhibited considerably lower plant to 281 plant variation (Fig. 2a). Further, all plants sampled 168 hpi harboured ~ 10 4 -10 6 CFU g -1 . and Willi354, were investigated further by qPCR (Fig. 2b). Early temporal changes in the 285 plant immune response were tracked using previously reported marker genes that follow 286 the levels of the three major phytohormones in plant immunity: ET, JA and SA (Kim et al., weak. The strongest changes did not exceed 5-fold (maximum mean: 2.5-fold; maximum 293 individual replicate: 4.6-fold) in gene expression, relative to the mock-treated control (Fig.  294   2b). The strongest changes were observed in the expression of the ET marker, ARL2. Early 295 after inoculation, its expression dropped significantly, at 1 and 3 hpi for Pst and 3 hpi for 296 Willi354. After recovering to the expression levels found in mock-treated plants the 297 relative expression of the ET marker dropped again at 12 hpi, which was significant in the 298 case of Willi354, with Pst seemingly following the same trend. Expression levels then rose 299 above mock-treated control by ~ 2.5-fold at 96 hpi, with changes being statistically 300 significant for 48 and 96 hpi for Pst and 96 hpi for Willi354 (Fig. 2b). In addition, Micro347 301 showed a significant ~ 1.8-fold rise in ET marker gene expression at 96 hpi (Fig. 2b). 302

303
Regarding the JA marker, statistically significant changes were only observed in response 304 to Sphingo34. A drop in expression was observed between 12 and 48 hpi with the latter 305 being statistically significant, but rather weak at ~ 1.5-fold (Fig. 2b). Expression of the SA 306 marker, PR1, fluctuated strongly, but not significantly between up-and downregulation 307 upon Willi354 treatment. Strong fluctuations between no expression change and a strong 308 downregulation were observed following Micro347 treatment with a significant ~ 3-fold 309 decrease in expression at 24 hpi. Sphingo34 and Pst elicited no changes in PR1 expression 310 (Fig. 2b). 311

312
The increase in ET marker expression within the last three sampling times (24, 48, 96 hpi) 313 seemed to follow the increase in bacterial density of the inoculants (Fig. 2a, b). 314 Interestingly, 57% (adj. R 2 = 0.57, p = 0.0027) of the change in gene expression relative to the 315 mock control can be explained by the bacterial density, irrespective of the inoculant 316 (Fig. 3).  those quantified by RT-qPCR confirming that RNA sequencing was performed correctly 337 (Fig. S2). As expected, the transcriptomes of mock-treated plants were distinct from those 338 of inoculated plants, as shown by multi-dimensional scaling (MDS) and k-means 339 clustering (Fig. 4a, Fig. S3c). The transcriptomes separate by treatment along the first 340 dimension of the MDS plot (Fig. 4a) within Willi354 samples, corresponding to a leading FC of ~ 3-fold (Fig. 4a). Willi354 are also differentially expressed in Pst (Fig. 4b). This shows that plant responses 356 to these three leaf-colonising strains are largely similar but differ in their strength 357 depending on the leaf coloniser. A closer look at individual gene expression changes 358 further highlights the similarity in the responses. Most changes follow a sequence with either increasing or decreasing expression from mock-treated control over Micro347 and 360 Willi354 to Pst-treated plants (Fig. 4c, Fig. S3c)

380
To gain a better resolution of gene expression changes, the 770 genes that were 381 differentially expressed in any of the treatments were further separated by k-means 382 clustering. K-means clustering was performed based on moderated log2(cpm). Ten k-383 means were chosen, based on the 'elbow' of the total sum of squares as a function of the 384 number of k-means (Fig. 4c, Fig. S3b,c). The 433 upregulated genes are in clusters 1-7, and 385 the 337 downregulated genes are in clusters 8-10 (Fig. 4c, Fig. S3c). In addition to more 386 genes being significantly upregulated than downregulated, FCs were greater in 387 upregulated genes. Clusters 1 and 2 contain genes with the strongest upregulation and 388 cluster 10 genes with the strongest downregulation at FCs in moderated log2(cpm) of ~ 16-389 fold and ~ 6-fold, respectively (Fig. 4c). 390 391 Transcriptional responses depend on bacterial load 392 As seen above, responses to bacterial colonisation seem to be largely similar (Fig. 4b,c) in 393 response to the tested strains. This was especially surprising as Pst is an arabidopsis 394 pathogen, whereas Micro347 and Willi354 were isolated from leaves of asymptomatic 395 plants (Cuppels, 1986;Bai et al., 2015). In addition, bacterial density, irrespective of the 396 bacterial coloniser, had a highly significant effect on ethylene responses (Fig. 3). Taken together, this suggests that pathogenicity is to some extent dependent on bacterial density, 398 following Paracelcus' theory "the dose makes the poison" (Paracelsus, 1538). This raises the 399 question whether non-pathogenicity of bacteria is merely a case of the plant balancing 400 their proliferation, or the bacteria doing so in order to avoid being penalised by the plant. 401 To gain a better understanding of plant responses to non-pathogenic leaf-colonising 402 bacteria and to determine whether the bacterial load changes the nature of the response, 403 plants were inoculated with different concentrations of Willi354 followed by RNA 404 sequencing. Willi354 was chosen for this experiment as it exhibited stronger responses 405 than Micro347 in the previous experiment (Fig. 4b,c). 406 407 Six-weeks-old axenically-grown arabidopsis plants were spray inoculated with Willi354 408 with inoculation densities ranging from 10 5 CFU ml -1 to 10 8 CFU ml -1 . Four days after 409 inoculation, the bacterial densities on the plants ranged from 6.57 × 10 6 CFU g -1 to 3.22 × 10 8 410 CFU g -1 and strongly correlated with the inoculation density (adj.R 2 = 0.9265, p = 3.594 × 10 -411 14 ) (Fig. 5a). The transcriptomic response of the plants changed gradually with increasing 412 inoculation density, as seen in the MDS plot and k-means clustering (Fig. 5b, Fig. S4). The 413 transcriptomes separate along the first dimension of the MDS plot, which explains 55% of 414 the variation between the different samples, with those of mock-treated plants and those 415 of plants inoculated with Willi354 at 10 8 CFU ml -1 being most dissimilar at a leading FC of ~ 416 16-fold (Fig. 5b).

456
To further explore the nature of plant responses to inoculation with Willi354, a GO term 457 enrichment analysis was performed on two different sets of genes. The first set comprised genes that were exclusively differentially expressed in plants treated with Willi354 at 459 10 8 CFU ml -1 , whereas the second set comprised genes that were also differentially 460 expressed at lower densities of Willi354. The functional profiles of both sets of genes were 461 markedly different (Fig. 5d). Genes differentially expressed exclusively at the highest 462 density of Willi354, were greatly enriched for genes related to plant immunity, including 463 perception of the biotic environment, such as 'response to molecule of bacterial origin' and 464 'response to insect', metabolism of secondary metabolites such as 'indole glucosinolate 465 metabolic process', local defence responses such as 'cell wall thickening' and 'defence 466 response by callose deposition in cell wall', and systemic defence responses such as 467 'regulation of salicylic acid mediated signalling pathway' and 'systemic acquired 468 resistance' (Fig. 5d). By contrast, genes that were also differentially expressed at lower 469 densities of Willi354, were greatly enriched for genes related to plant nitrogen 470 homeostasis, such as 'response to nitrogen compound' and 'response to organonitrogen 471 compound', plant oxygen levels, such as 'response to oxygen levels' and 'response to 472 hypoxia' and the plant's 'response to light intensity' (Fig. 5d).  (Fig. 6a). Both sets exhibited an enrichment of states 1 and 2, 487 which are both characterised by the presence of the histone variant H2A.Z accompanied either by activating marks such as H3K4me3, H3K36me3 or by a combination of activating 489 (H3K4me3) and repressive (H3K27me3) marks, respectively. They also displayed an 490 underrepresentation of states 8 and 9, which contain heterochromatic marks such as 491 H3K9me2 and H3K27me1. In addition, genes which are upregulated by Willi354 492 irrespective of inoculation density showed an underrepresentation of states 3, 5, 6 and 7. 493 By contrast, genes solely induced by the highest density of Willi354 either showed an 494 enrichment (state 6) or no difference to the reference. These results reveal that, similarly to 495 the distinct functional enrichments, genes induced specifically by higher bacterial 496 densities exhibit a partially different chromatin signature than genes that are also induced 497 by lower densities. 498

499
As the genes induced solely by the higher densities of Willi354 displayed an enrichment 500 for plant immunity and defence-related terms (Fig. 5d), we wondered whether those genes 501 would have a similar chromatin profile as genes induced by Pst. The genes were divided 502 into two sets, depending on whether they were induced specifically by Pst or whether they 503 were also induced by Willi354 or Micro347 (Fig. 6b). Similarly, to the previous analysis, 504 both sets presented an enrichment for states 1 and 2 and an underrepresentation for states 505 8 and 9. They also showed an underrepresentation of state 5, characterised by the presence 506 of the repressive mark H3K27me3. Additionally, genes induced by both Pst or 507 Willi354/Micro347 displayed an underrepresentation of states 6 and 7, which was not 508 observed or of reduced magnitude for the genes induced specifically by Pst. Despite limited 509 overlap between the two transcriptomic experiments (Fig. S6), the chromatin profiles of 510 genes induced specifically either by high Willi354 densities or Pst were comparable and 511 distinct from the profiles of genes induced also by non-pathogenic bacteria or by lower 512 bacterial densities. This observation supports the idea that inoculation with higher 513 densities of non-pathogenic bacteria leads to a transcriptomic response similar to that 514 triggered by pathogenic bacteria. 515  inoculated with Willi354 at 10 8 CFU ml -1 , but not at lower densities, exhibited necrotic 529 lesions on a few leaves 14 and 21 dpi (Fig .7a-f, Fig. S7). In addition, plant weight was 530 negatively correlated with inoculation density of Willi354 (adj.R 2 = 0.11, p = 6.48 × 10 -5 ) 531 (Fig. 7h). consistently reach bacterial densities above the threshold of detection, which was on 553 average ~ 2500 CFU g -1 of leaf fresh weight (Fig. 2a). This was rather surprising as both 554 genera were previously found to make up more than 1% of the total bacterial population on 555 arabidopsis (Vorholt, 2012). Further, both strains were recently shown to successfully 556 colonise arabidopsis . However, in the study by Vogel and colleagues, 557 colonisation density was measured nine days after drop inoculation on seedlings in an 558 agar-based system. In this study the 'Litterbox' system was employed, which reliably 559 mimics environmental population densities, as opposed to agar-based systems that exhibit 560 unnaturally high population densities (Miebach et al., 2020). In addition, Acido84 reached 561 population densities of 10 4 -10 6 CFU g -1 in all sampled plants at 168 hpi (i.e. 7 dpi). Since 562 Acido84 was also successfully recovered straight after inoculation this indicates (1) that 563 Acido84 was not harmed during the spraying procedure and (2) that it was able to thrive 564 on leaves at later time points. Overall, this suggests that Acido84 had to acclimatise to its 565 new environment after growth on R2A media, even though R2A is, like the phyllosphere, 566 oligotrophic. Pedo194, in contrast, was only successfully recovered from one plant 567 immediately after spray-inoculation at a density ~ 2 magnitudes lower than the inoculum, 568 suggesting that the inoculation procedure might have been detrimental to it. 569 570 Pst and Will354 rose in population size to ~ 10 7 CFU g -1 at 96 hpi. Population size then 571 declined to 10 6 CFU g -1 at 168 hpi (Fig. 2a). Whether this was due to exhaustion of resources 572 or plant immune responses remains to be determined. Interestingly though, the expression 573 of the ET marker ARL2 could be explained in large parts by the bacterial density of the 574 coloniser, irrespective of the inoculant. This suggests that ARL2 expression must be either triggered by a common MAMP, shared between the isolates, or was triggered by many 576 MAMPs which the plant did not distinguish between. Further, it suggests that ARL2 577 expression is proportional to the MAMP titer. This agrees with previous findings, which 578 described stronger transcriptional responses to both higher pathogen and higher MAMP 579 titers (Thilmony et al., 2006;Denoux et al., 2008).

581
The observed expression changes upon bacterial treatment were overall rather weak (Fig.  582   2b). The strongest changes in gene expression were observed in the ET marker, ARL2, and  Melotto et al., 2006). In addition, the flg22 receptor, FLS2 is highly expressed in leaves 594 near bacterial entry sites, such as stomata, which are predominantly found on the abaxial 595 (lower) leaf surface and hydathodes, as well as in leaf veins (Beck et al., 2014). Vacuum 596 infiltration of bacterial suspensions would render leaf veins more exposed to MAMPs and 597 a change in liquid media would render stomata and hydathodes more exposed to MAMPs, 598 than in the more 'natural' scenario of topical application. coloniser. Remarkably, the responses observed were largely similar, but weaker in 606 response to colonisation by the non-pathogenic bacteria. Most genes significantly 607 expressed in response to one strain were also significantly expressed in response to 608 strains that elicited stronger responses and thus, higher numbers of DEGs. None of the 770 609 genes with a significant FC in any of the bacterial treatments was upregulated by one 610 strain and downregulated by another. Changes in gene expression of genes belonging to 611 clusters 2, 4, 5, 6, 7, 8, 9 and 10 were either progressively increasing or decreasing when 612 treatments were sorted by the number of DEGs that they elicited (Fig. 4c). This indicates 613 that those genes were similarly regulated in response to bacterial colonisation irrespective 614 of the symbiotic relationship of the inoculant with the plant, although less severely in 615 response to non-pathogenic bacteria. Such similarity in the plant response to various leaf 616 colonisers was also described recently by Maier and colleagues, although without the 617 context of a pathogenic bacterium (Maier et al., 2021). Interestingly, the response strength hpi irrespective of the bacterial coloniser (Fig. 3). In addition, genome-wide transcriptional 628 responses to bacterial colonisation were largely similar but differed in number of DEGs ( Fig. 4b), as well as in the expression strength of individual genes (Fig. 4c). This was