Quantification of bacterial population density in tomato plants
To determine the bacterial abundance in the phyllosphere microhabitats targeted real-time quantitative PCR (qPCR) was applied. The qPCR results showed that bacterial density varied between 5.24 x 108 and 6.94 x 109 16S rRNA gene copy numbers per gram of sample. However, statistical significance tested using the Kruskal-Wallis test indicated that there was no effect of sample type (P=0.356) and tomato genotype (P=0.158) on bacterial density (Additional File 2).
In situ visualization of bacteria on tomato trichomes
Confocal laser scanning microscopy (CLSM) was implemented in combination with differential DEAD/LIVE staining to visualize microbes colonizing different trichome types (Fig. 1A - C). Bacteria were found on type VII (Fig. 1A) as well as type IV (Fig. 1B & C) trichomes. The colonization density of the detected bacteria was similar among samples irrespective of the trichome type. Moreover, the micrographs display a diverse bacterial community as indicated by cell morphology. The visualized bacteria were located in close approximation to the trichome head from both genotypes (Fig. 1A - C).
General assessment of amplicon libraries
Composition of bacterial communities were obtained using amplicon sequencing. After quality filtering of the raw reads and exclusion of non-target sequences (mitochondria and chloroplasts), a total of 102,722 high quality reads from bacterial libraries were retained with 442 to 15,788 reads per sample which were assigned to 1421 bacterial amplicon sequences variants (ASVs). Even though a high proportion of reads were assigned to non-target taxa that were detected due to the nature of the samples (high occurrence of chloroplasts and mitochondria), the average ASV number found in the sample was 69.2 (Additional file 3). Interestingly, trichome samples had a relatively lower proportion of non-target reads in comparison to other samples.
Observed alpha and beta diversity of bacterial communities
With respect to sample type, the Kruskal-Wallis test showed that bacterial richness is significantly higher in trichomes compared to other sample types (P< 0.001). A higher bacterial richness indicated by the Shannon index (H’) was observed in trichomes samples (H’= 4.4) when compared to leaves (H’= 2.5) and leaves without trichomes (H’= 2.8). No significant differences were observed between leaves and leaves without trichomes. With respect to tomato genotype, samples from LA1777 had a relatively higher alpha bacterial diversity (H’ = 3.4) compared to LA4024 (H’= 3.0). However, Kruskal-Wallis test showed that the tomato genotype did not influence bacterial richness according to number of ASV and Shannon diversity index (P= 0.157 and P= 0.248, respectively) (Additional file 3, Fig. 2E).
Observed beta diversity visualized using non-metric multidimensional scaling (NMDS) showed discrete clusters according to sample type as well as tomato genotype. These factors influenced the bacterial community structures significantly (P<0.05). Sample type was shown to be the dominant factor that influenced the bacterial variation (23.9%) whereas tomato genotype only explained 8.4% of the variation. Trichome samples tend to cluster closer together whereas leaves with and without trichomes formed another cluster. Despite only a small variation that was statistically explained by tomato genotype, a clear separation of the two genotypes was observed in all sample types (Fig. 2F).
Identification of microbial taxa in different microhabitats
Differences in the taxonomic composition were visualized by assessing the 100 most abundant bacterial ASVs (Fig. 3). The bacterial communities present in all samples showed a high proportion of Proteobacteria (44.4%) and Firmicutes (30.8%). Comparing individual sample types, no substantial differences were found on phylum level. The analysis on class level, however, showed that the bacterial classes Bacilli (39.1%) and Gammaproteobacteria (29.5%) were the most abundant bacterial class in leaves with and without trichomes, whereas in trichome samples, a higher relative abundance of Alphaproteobacteria (23.6%) was observed for both genotypes (Fig. 3A). At family level, the most abundant families were Bacillaceae and Burkholderiaceae that accounted for 38.3% relative abundance regardless of the sample type. In trichome samples, a relatively higher number of bacterial families were generally observed compared to the others. This result supported the higher bacterial diversity generally found in trichomes in comparison to other samples types. Especially the relative abundance of Moraxellacea (6.4%) and Sphingomonadacea (8.4%) were higher in trichomes compared to the other sample types. The family of Bacillaceae (8.6%), which had a relatively high abundance in leaves, showed the opposite pattern (Fig. 3B). Moreover, we observed low abundant bacterial families (relative abundance >0.1%) i.e. Hymenobacteraceae and Alicyclobacillaceae that were found exclusively in trichomes.
Trichome samples that were shown to harbor a higher bacterial diversity compared to the other two sample types, overall also harbored the highest number of ASVs. In total, 25 ASVs were shared between all samples. Trichome samples in contrast harbored 14 ASVs which were exclusively found in those samples (Fig. 4A). DESeq2 analysis was performed to investigate which taxa were specifically increased in trichomes in comparison to leaves without trichomes. A total of 20 ASVs which were dominated by Burkholderiaceae (n=4) and Sphingomonadaceae (n=7) families respectively were significantly increased (Fig. 4B). Six ASVs from the Actinobacteria also showed the same pattern. The analysis of core microbiomes found in the different sample types showed a highly more diverse core microbiome in trichome samples. Especially the genus Sphingomonas, a member of the family Alphaproteobacteria, was identified as a central element in the core microbiome of trichome samples but was not found in the core of leaves without trichomes (Fig. 4C & 4D). A core microbiome was identified in each genotype; however, higher abundance of Actinobacteria and Proteobacteria ASVs were found to be characteristic for trichome samples.
Genotype specific enrichment of trichome-colonizing microorganisms
Overall, trichome samples were found to harbor increased microbial diversities, however, genotype specific differences were also observed. Compared to S. lycopersicum LA4024, trichomes of S. habrochaites LA1777 carried a higher abundance of Bacilli, however a lower abundance of Gammaproteobacteria. At family level, increased levels of Bacillaceae, Moraxellaceae and Sphingomonadaceae, as well as decreased abundances of Burkholderiaceae, Pseudomonadaceae and Xanthomonadaceae were observed (Fig. 3). Significance of taxonomic differences was further accessed using DESeq2 analysis. A total of 26 ASVs were found to be significantly changed between the trichome samples of both genotypes. A total of 18 ASVs, including ASVs from the genus Bacillus, Deinococcus, Acinetobacter, Paracoccus, and Sphingomonas, were significantly increased in trichomes of genotype LA1777 compared to genotype LA4024. In contrast, 8 ASVs from the genus Bacillus, Massilia, Caulobacter, Capnocytophaga, Pseudomonas, Pedobacter, and Luteimonas were significantly increased in genotype LA4024 compared to genotype LA1777 (Fig. 5).