16S rRNA gene-based microbiome analysis identifies candidate bacterial strains that increase the storage time of potato tubers

Sprouting behaviour of potato tubers

In this study, the stored tubers exhibited clear variety-specific differences in sprouting behaviour, with the Agata variety being the earliest and the Hermes variety the latest to sprout. However, within each variety, there were differences in sprouting time depending on the soil in which the tubers had grown (Suppl. Table 1). These soil-dependent differences in sprouting were greatest in the Fabiola variety (sprouting time ranged from 135 to 169 days after harvest) and least significant in the Hermes variety (158 to 168 days after harvest). By examining the storage stability of tubers broken down by soil type, we found that tubers grown in soil from the field site in Karnabrunn sprouted on average up to 9 days later than the tubers from other soils. The chemical profile of the soil collected in Karnabrunn was similar to that of the other farmland soils, especially those from Kettlasbrunn. The soil collected in Tulln showed reduced K and P contents compared to those of the other farmland soils. As expected, the potting soil differed most significantly from the other soils, i.e., there was no clay, and the exchange capacity was substantially higher than that of the farmland soils (Suppl. Table S2). Statistical analysis revealed that the chemical parameters of the soils did not correlate with the sprouting behaviour of tubers (data not shown).

Sequencing results

The sequencing of 16S rRNA gene amplicons of tubers, sprouts and soil samples yielded 9,158,550 high-quality merged reads, corresponding to an average of 33,920 reads per sample. All sequences were cut to a read length of 360 bp. To ensure sufficient diversity by maintaining an adequate sequencing depth, samples with a low read number were excluded from the analysis. Therefore, the data of four samples were excluded from further analysis (LC_T3_PS; LC_T7_K; H_T6_K: LC_T7_T; H_T6_K; F_T7_K). A detailed explanation for the sample naming can be found in Table 1. Sequencing reads from all 270 samples were clustered into operational taxonomic units (OTUs), which resulted in 11,485 OTUs with an average of 42,537 OTUs per sample.

The bacterial community of potato tubers during storage

To test whether the bacterial community in potato tubers changes during storage, the 16S rRNA gene amplicon data of the potato tubers of the Agata, Lady Claire and Hermes varieties, cultivated in five different soil types (T2), stored until dormancy break (T6) and until sprouting (T7) were analysed as well as samples of the sprouts (T7_Sprouts) (Fig. 1). Cultivation of the cultivar Fabiola in the soil Kettlasbrunn B did not yield sufficient tubers for analysis throughout the whole storage period, i.e., no data were available for T7 and the sprouts. Therefore, community data for the Fabiola variety were not included in the following statistical analysis.

For the statistical analysis, we selected only OTUs that showed at least 0.01% relative abundance (1425 OTUs) and that were present in at least two of three replicates (1395 rOTUs). We considered them “reproducibly occurring OTUs” (rOTUs)².

For calculation of alpha diversity values, read numbers were rarefied to 6785 reads in each sample. The permutation ANOVA of rOTU richness (9999 permutations) revealed significant differences in the bacterial community of tubers and sprouts between cultivars, time points and soil types (observed for cultivar (F value = 3.963, P value = 0.0211*), for time point (F value = 4.762, P value = 0.0021*) and for soil type (F value = 3.779, P value = 0.0057**). The same results were obtained with Simpson’s index for the factors cultivar (F value = 4.594, P value = 0.01*), time point (F value = 6.741, P value = 0.0004***) and soil type (F value = 5.123, P value = 0.0008***) (Suppl. Table S3). Overall, the statistical analysis showed that the soil type had the strongest impact on the microbial community in potato tubers, followed by time point and cultivar.

Similarly, the bacterial community composition (beta diversity) in potato tubers was significantly affected by the storage time, soil type and plant variety. The CAP ordination plot indicated a shift in the bacterial community of tubers from harvest to dormancy break and sprouting, which could be proven by PERMANOVA on the Bray–Curtis dissimilarity distance matrix (R2 = 0.113, P value = 0.0004***) (Fig. 2A). Additionally, the bacterial communities in tubers grown in different soil types (R2 = 0.255, P value = 0.0004***) and of different varieties were significantly different from one other (R2 = 0.027, P value = 0.0021**) (Suppl. Table S4). The results for the factor time point (P value = 0.01**), genotype (P value = 0.01**) and soil type (P value = 0.01**) could be confirmed by the multivariate generalized linear model for multivariate abundance data on the Bray–Curtis dissimilarity distance matrix (999 permutations) (Suppl. Table S4).

Constrained analysis of principal coordinates (CAP) of Bray–Curtis dissimilarities. CAP based on the V5–V7 regions of the 16S rRNA gene investigated for (A) tubers of the varieties (Agata, Lady Claire and Hermes) sampled at T2, T6 and T7 and (B) tubers of the varieties (Agata, Lady Claire and Hermes) sampled at T2-7 as well as sprout samples at T7. An overview of potato cultivars, soil types and sampling time points is shown in Fig. 1.

The dynamics of bacterial communities in potato tubers during storage

To obtain more in-depth insight into the changes in community composition during storage, we compared the bacterial communities in tubers of the Agata, Lady Claire and Hermes varieties grown in potting soil at six different timepoints: T2 (harvest), T3 (2 weeks after harvesting), T4 (5 weeks after harvesting), T5 (10 weeks after harvesting), T6 (dormancy break), T7 (sprouting) and in the corresponding sprouts (T7_Sprouts) (Fig. 1). We focused here on tubers grown in potting soil, since the cultivation of potatoes in farmland soil did not yield sufficient tubers to be analysed throughout the entire storage period.

Again, we filtered data for OTUs with at least 0.01% relative abundance and “reproducibly occurring OTUs”. After both filtering steps, 1108 rOTUs remained. Before calculating alpha values, read numbers were rarefied to 6761 reads in each sample. The permutation ANOVA of richness and evenness (9999 permutations) did not reveal differences in the alpha diversity of the tuber microbiome between cultivars but between time points (observed species F value = 6.159, P value = 0.0004*** and Simpson’s index F value = 2.216, P value = 0.0263*) (Suppl. Table S3). The bacterial richness of each cultivar declined significantly during the period between harvest (T2) and the dormancy break (T6) (Suppl. Figure S1B). The CAP scaling plot of six different time points during storage and the corresponding sprouts indicated a shift from harvest to sprouting (Fig. 2B). The PERMANOVA on the Bray–Curtis dissimilarity distance (9999 permutations) revealed that the microbiomes of tuber samples differed significantly between the cultivars (R2 = 0.069, P value = 0.0004***) and time points (R2 = 0.221, P value = 0.0001***) (Suppl. Table S4). The test results for cultivar (P value = 0.01**) and time point (P value = 0.01**) were confirmed by the multivariate generalized linear model for multivariate abundance data on the Bray–Curtis dissimilarity distance matrix (999 permutations) (Suppl. Table S4). To identify the bacterial taxa that changed over time during storage, differentially abundant rOTUs were calculated with the random forest function and visualized in Fig. 3 at the genus level. The relative abundance of Staphylococcus sp. increased during the period from harvesting (T2) to dormancy breaking (T6) from approximately 0% to 11% relative abundance. A similar result was visible for Propionibacterium sp. and Acinetobacter sp. The relative abundance of both was approximately 2% in tubers after harvesting (T2) and increased during storage to 8% and 9% in dormancy broken tubers (T6), respectively, whereas the relative abundance of the taxa Iamia sp. and Nocardioides sp. decreased from approximately 10% after harvesting to 4% and 6% in already sprouted tubers (T7), respectively (Suppl. Tables S5 and S6).

Differentially abundant taxa at the genus level. Visualization of differentially abundant taxonomic groups at the genus level of the bacterial community in tubers of three different potato cultivars (Agata, Hermes and Lady Claire) at all sampling time points T2-7 as well as in sprout samples at T7. Differentially abundant taxa were calculated with the function varSelRF⁴⁸ and visualized in barplot with the function group.abundant.taxa of the R package RAM⁴⁷.

Taxa associated with short, medium or long storage stability

To identify the bacterial taxa in the potato tuber microbiome associated with early or late potato tuber sprouting, sample data were grouped into short, medium and long storage abilities based on the individual storage time (in days from harvest until sprouting) of the samples. The data of all samples, including those of cultivar Fabiola, were considered for the following analysis (Fig. 1). After filtering for the OTUs with at least 0.01% relative abundance and presence in at least two of three replicates, 813 rOTUs were obtained. To identify rOTUs associated with short, medium or long storage stability at the beginning and end of the storage period of the potato tubers, two different analysis steps were performed. In the first step, differentially abundant rOTUs depending on storage stability (short, medium or long) and storage time points (T2, T6 and T7) were calculated with the random forest function (Suppl. Table S7). In a second step, a correlation matrix based on Spearman’s rank correlation was calculated based on the previously described factor storage stability and storage time points to identify positive or negative interactions between rOTUs (Suppl. Table S8). The same rOTUs obtained with random forest analysis, as well as with the correlation matrix, based on Spearman analysis, were filtered and named key OTUs (Suppl. Table 9). Even if both analysis steps do not provide the same evidence, they provide a meaningful indication of which OTUs are related to the factors, storage stability and storage time point. In total, we identified 24 key OTUs. The relative abundance of nine key OTUs was associated with long storage stability, meaning that the OTUs were significantly increased in samples where dormancy break (T6) and sprouting (T7) started late. These key OTUs are members of the orders Flavobacteriales, Cytophagales, Sphingobacteriales, Gaiellales, Corynebacteriales, Caulobacterales, Methylophilales and Solirubrobacterales. Additionally, the relative abundance of eight rOTUs was associated with short storage stability. These rOTUs are members of the orders Enterobacteriales, Pseudomonadales, Myxococcales, Rhizobiales, Bacillales and Burkholderiales. The relative abundance of seven key OTUs was associated with medium storage stability.

Testing the effect of selected bacterial taxa on potato tuber sprouting

In the next step, we tested whether the bacterial taxa that correlated with longer storability can directly affect the sprouting of potato tubers. Therefore, we screened a collection of bacteria isolated from seed potatoes^18,20 for isolates that are homologous in the 16S rRNA gene to the key OTUs identified in the statistical analysis. We identified two isolates that were homologous to OTU_14 (Flavobacterium sp.), which correlated with late sprouting. The selected isolates were tested in an in vitro sprouting assay adapted from Hartmann et al.²¹. Tuber discs treated with cultures of Flavobacterium sp. isolates (AIT1165 and AIT1181) resulted in sprout growth inhibition compared to control discs treated with sterile tryptic soy broth (Fig. 4). Furthermore, we observed significant individual differences in the sprouting behaviour of the tested buds. Differences in sprouting behaviour are only partly genotype-dependent, but these differences might also be due to the natural developmental variability between buds. The apical eye on a tuber usually begins to sprout first, marking the start of the apical dominance stage²². For the sprouting assay performed in this study, we took several buds from one tuber independently of their position.

In vitro potato tuber sprouting assay results. Tuber buds of three different varieties (Lady Claire, Ditta and Agata) were treated with two different isolates of Flavobacterium sp. (AIT1165 and AIT1181). For evaluation, bud growth was assessed according to the first principal growth stages of the BBCH scale. Hereby, the first stage 00 is considered innate or enforced dormancy with no sprouting at all, followed by stages 01 and 02, which represent the beginning of sprouting when sprouts are visible with sizes up to < 1 mm and 2 mm, respectively. According to the BBCH scale, dormancy is broken at stage 03. When tuber buds treated with maleic hydrazine reached stage 01, the assay was ended, and the sum of all growth stages of each replicate was calculated. Afterwards, the average of two repetitions was visualized in a boxplot diagram. The boxplot diagram shows that treatments with both isolate cultures led to sprouting inhibition in comparison to the assay control. The assay control gibberellic acid represents an effective sprouting promoter, whereas maleic hydrazine was used as the negative control. Ten percent tryptic soy broth (used as the medium for culture of the isolates) and sterile water were used as neutral controls.

Screening bacterial isolates for modulation of potato tuber sprouting

The next step was to examine the common sprouting inhibition activity among potato endophytes to eliminate the possibility of a random association. Therefore, 218 randomly selected bacterial isolates from seed potatoes of the Agata, Ditta, Hermes and Lady Claire varieties, isolated elsewhere^18,20, were tested for their effect on the potato tuber buds of the Lady Claire variety in the in vitro sprouting assay. The isolates AIT 1165 and AIT 1181 were also included in this assay and served as a benchmark. In total, 16 different bacterial isolates suppressed potato bud outgrowth (Suppl. Table 10), and the treatment of 10 of these 16 isolates was significantly different from the treatment with sterile tryptic soy broth (P value < 0.04) (Suppl. Table 11). These strains were selected for further sprouting assays with the Hermes, Tosca, Fontane and Agria varieties. Two of the 10 strains were tested, Bacillus sp. (AIT 1184) and Arthrobacter sp. (AIT 1143), resulting in sprout growth inhibition in all tested potato varieties compared to the control discs treated with sterile tryptic soy broth (Suppl. Figure S3).