Seasonal microbial dynamics on grapevine leaves under biocontrol and copper fungicide treatments

Winemakers have long used copper as a fungicide on grapevine. However, the potential of copper to accumulate on soil and affect the biota poses a challenge to achieving sustainable agriculture. One recently developed option is the use of biocontrol agents to replace or complement traditional methods. In the present study, a field experiment was conducted in South Africa in which the leaves in two blocks of a vineyard were periodically treated with either copper sulphate or sprayed with Lactobacillus plantarum MW-1 as a biocontrol agent. This study evaluated the impact of the two treatments on the bacterial and fungal communities as they changed during the growing season. To do this, NGS was combined with quantitative strain-specific and community qPCRs. The results revealed the progression of the microbial communities throughout the season and how the different treatments affected the microbiota. Bacteria appeared to be relatively stable at the different time points, with the only taxa that systematically changed between treatments being Lactobacillaceae, which included reads from the biocontrol agent. Cells of Lactobacillus plantarum MW-1 were only detected on treated leaves using strain-specific qPCR, with its amount spanning from 103to 105cells/leaves. Conversely the fungal community was largely shaped by a succession of different dominant taxa over the months. Between treatments, only a few fungal taxa appeared to change significantly and the number of ITS copies was also comparable. In this regards, the two treatments seemed to affect the microbial community similarly, revealing the potential of this biocontrol strain as a promising alternative among sustainable fungicide treatments, although further investigation is required.


31
Copper (Cu) compounds have traditionally been the means to combat phytopathogenic microbes, 32 especially fungi, in crop plants (Banik and Pérez-de-luque 2017). In organic agriculture including organic 33 vineyards, the use of Cu-based pesticides is currently the only chemical treatment allowed, although it is 34 limited to a maximum of 6 kg Cu ha -1 per year in the EU (Commission Regulation, 2002). Such a strict 35 regulation is explained by the adverse effects of Cu on soil organisms and its long-term persistence on 118 This section focuses on two different parameters that describe the microbial community on leaves: 119 phylogenetic diversity (PD) and evenness. The alpha diversity results are summarised in Figure 1. 120 Looking at the bacterial community on leaves, it was noted that PD was significantly higher in the Cu-

Beta diversity on leaves
This section shows the differences between the samples using Bray-Curtis dissimilarity, looking at the 139 distribution and relative abundances of the single feature between samples. The beta diversities of 16S 140 and ITS sequence datasets, coloured by treatment or collection time, were visualised by PCoA plots and 141 are shown in Figure 2. With regard to the bacterial community in Figure 2a, it can clearly be seen that 142 the 16S distribution on leaves strongly clustered by treatment (p-value=0.001), while no significant 143 pattern (p-value>0.05) was found in relation to the sampling collection time (Fig. 2b). The opposite trend 144 was apparent for the fungal community. In fact, although ITS distribution does not appear to be affected 145 by the treatment (p-value >0.05) as displayed in Figure 2b, it does strongly depend on the period of 146 collection during the season, not only at month level (p-value= 0.001) as shown in Figure 2d, but 147 sometimes even at day level (p-value= 0.001), as presented in Figure S2.

Taxonomical composition on leaves
The results below show the microbial composition in terms of taxa assigned to the different features 156 found, highlighting the microbial representation on grapevine leaves. All the information regarding 157 bacterial and fungal community are summarised in Figure 3. Figure 3a shows  Lactobacillaceae resembled the levels found on copper-treated leaves.

172
The fungal communities, displayed in Figure 3b, were relatively similar for the biocontrol and Cu-treated 173 samples when looking at the entire growing season. The dominant taxa belonged to Pleosporaceae, 174 Cladosporium, Alternaria, Capnodiales, Sporobolomyces and Aureobasidium pullulans. Although no 175 differences appeared between the two treatments, another pattern was seen with respect to the collection 176 time. In fact, during the growing season, there was a significant change in the taxonomical composition 177 of the fungal communities. Several fungal outbreaks led to a variation in relative abundances during the 178 period studied. For instance, Pleosporacea members tended to increase in relative abundance from 179 September to December, reaching 55 % of the total community, before suddenly decreasing to 9 % in     198 To test which taxa change in a statistically relevant way, ANCOM was run using treatment and period 199 as discriminants for these samples. The result was that for 16S between different treatments, the only 200 taxa that changed in a statistically significant way was Lactobacillaceae, the family to which this study's 201 biocontrol agent belongs, with the main sequence variant assigned to Lactobacillus plantarum (Table   202   S1). Only two statistically relevant differences were found when looking at the period regarding 203 mitochondria and Ralstonia, but none were associated with grapevine and furthermore they appeared in 204 low abundance (Table S2). 205 In the fungal community there were only two taxa that changed significantly between treatments. These 206 taxa belong to the species Kondoa aeria and to the order of Filobasidiales (Table S3). They were in very 207 low abundances and did not seem to be related to any known disease or have an impact on the plant itself.

208
Instead, for this period a large amount of taxa were observed that changed during the season, 42 of which 209 are classified at least at class level (Table S4). These findings provide information that can be used to 210 understand the evolution of the fungal community as the season progresses. These 42 taxa included 211 potential pathogens such as Alternaria, Cladosporium or members of Botryosphaeriaceae, with the main 212 sequence assigned to Diplodia seriata using BLAST. As the period shaped the fungal community, an 213 investigation was carried out on which taxa changed between different treatments during specific periods 214 in the season. This ANCOM test returned 50 taxa, of which 46 were classified at order level (Table S5). 215 Among the taxa that appeared to be differentially distributed between treatments, in a short period of 216 time (such as September alone), there were Botryosphaeriaceae and Leptosphaeriaceae, which were 217 previously highlighted when looking at the taxonomical composition in Figure 3b.

218
Gneiss was then run on the bacterial community dataset to evaluate the impact of MW-1 on other bacteria.

219
Interestingly, Lactobacillacea appeared to be sensitively different (p-value <0.05) from the rest of the 220 microbial community, based on the position occupied in the tree (Fig. S3), but the fact that this taxon 221 branches out from the rest means that it does not interact with other species in the community within the 222 same microbial niche. This is a further evidence of that the introduction of this biocontrol agent does not 223 alter the bacterial community that would normally be found on leaves sprayed with copper. September, was the first time that MW-1 was not detectable, with no spraying having previously been 235 done, while the second time was on 9 March, two months after harvest.
The total fungal abundance could be estimated at cell level number without introducing significant biases 237 due to an uneven ploidy variation between the different taxa detected. However, since the community 238 composition between the treatments appeared to be very similar, it was assumed to have a negligible 239 difference in terms of genome distribution. For this reason, a relative quantification was performed 240 between the two fungal communities grouped by treatments. With this in mind, the results obtained 241 showed that the fungal communities between the two treatments followed a similar trend, with no 242 profound variations between the different treatments. It was decided to estimate the number of fungal 243 genomes, assuming a fungal-genome average size of 27.5 Mb. Accordingly, the number of genomes 244 ranging from 10^2 to 10^4 genomes/leaves during this period was calculated. After normalising the raw-245 qPCR signals using the gDNA amount, an ANOVA was run which gave a p-value of 0.56. This 246 confirmed that the variation between the two fungal communities coming from the two treatments was  To further demonstrate the link between initial gDNA amount and qPCR signal, an additional correlation 252 analysis was performed. First two index ratios (IR) were calculated: IR1 between input gDNA in Cu 253 versus biocontrol-treated leaves, and IR2 by dividing qPCR genome-copies detected on Cu versus 254 biocontrol-treated leaves. Finally, the correlation between IR1 and IR2 during the period was determined.

255
The correlation coefficient was 0.81, confirming that the two IRs were highly positively correlated, as 256 shown in Figure 4a. A t-test on the two IR series (the assumption of equal variance was tested with a F-257 test that resulted in a p-value of 0.165) confirmed that the difference was not significant (p-value= 0.699).

258
This means that the tendency of the two IRs did not differ significantly between the treatments. In 259 conclusion, this indicated that not only did the two fungal communities not differ significantly in 260 composition between treatments, but they were also quantitatively comparable.    Table 1.

410
The tubes containing the leaves were thawed at room temperature and 20 ml of a washing solution was  The fungal community was sequenced using the same double-step PCR approach for library preparation, In order to find a region that was unique for the biocontrol agent, part of the genome was uploaded and  Statistical evaluation of these results was performed separately for qPCR and the sequencing dataset. All

498
To relate the differences in qPCR signal on the fungal community to the presence/absence of the 499 biocontrol treatment, the qPCR signal had to be correlated with the initial DNA amount before checking 500 the differences and their statistical relevance. To do this two different index ratios (IR) were created: 501 IR1 dividing the raw qPCR signal from Cu samples from that from biocontrol samples, and IR2 dividing 502 the initial amount of DNA of Cu-treated with that of biocontrol-treated leaves. Two series of IRs were 503 then obtained to test for equal variance with a F-test. This was intended to establish whether the two IRs, 504 deriving from two different datasets, had a comparable variance within their samples. The F-test for 505 variances, with a p-value= 0.16, proved that the two variances were equal and then a t-test was run on 506 two series assuming equal variances between the IRs on the two treatments, Cu and biocontrol. A 507 correlation analysis was also performed between the initial DNA amount and the resulting qPCR signal.

508
Finally correlation analyses and a t-test were repeated on the samples treated with the bacterial agent as 509 well to compare the variation between the total fungal community on biocontrol-treated leaves and the 510 number of cells detected of the biocontrol bacteria. Sequencing data after QIIME 2 pipeline processing 511 were statistically evaluated using the Kruskal-Wallis test for alpha and beta diversity. The resulting p-