Pyridine-type alkaloid composition affects bacterial community composition of floral nectar

Pyridine-type alkaloids are most common in Nicotiana species. To study the effect of alkaloid composition on bacterial community composition in floral nectar, we compared the nicotine-rich wild type (WT) N. attenuata, the nicotine biosynthesis-silenced N. attenuata that was rich in anatabine and the anabasine-rich WT N. glauca plants. We found that the composition of these secondary metabolites in the floral nectar drastically affected the bacterial community richness, diversity and composition. Significant differences were found between the bacterial community compositions in the nectar of the three plants with a much greater species richness and diversity in the nectar from the transgenic plant. The highest community composition similarity index was detected between the two wild type plants. The different microbiome composition and diversity, caused by the different pyridine-type alkaloid composition, could modify the nutritional content of the nectar and consequently, may contribute to the change in the nectar consumption and visitation. These may indirectly have an effect on plant fitness.


Methods
Study system -nectar chemical composition. irPMT N. attenuata plants are unable to synthesize nicotine, due to the silencing of the key nicotine biosynthetic enzyme, putrescine-N-methyl transferase, in an inverted repeat orientation. Full characterization of the irPMT plants can be found in Steppuhn et al. 27 . irPMT plants have drastically reduced transcripts of both PMT genes, and they do not produce any detectable quantities of nicotine. All compounds found in the floral scent of WT plants, with the expected exception of nicotine (t-test, t 14 = − 4.86, P < 0.001), can be detected in statistically equivalent amounts in the floral scent of the transformed plants. Sugar concentrations in nectar did not differ among WT and irPMT plants (t-test, t 24 = 0.32, P = 0.75) 21 . As for nectar metabolites, irPMT plants accumulate the alkaloid anatabine up to the concentration of 41.35 ± 1.75 ppm (Ian T. Baldwin, unpublished data). Anatabine is not detectable in the WT plants. Elevated anatabine levels did not affect transcript levels of other gene-encoding enzymes involved in alkaloid metabolism 28 . irPMT plants did not differ from WT plants in any other measured secondary metabolite or growth parameter.
Nicotiana plants and nectar collection. No specific permissions were required for plant samplings, since the field studies did not involve endangered or protected species. Floral nectar was collected from two Nicotiana species: N. attenuata and N. glauca. The N. attenuata WT and irPMT plants were sampled at the glasshouse of Max Planck Institute for Chemical Ecology, Jena, Germany, in October -November 2011 and January -February 2014, for culture-independent and culture-dependent techniques, respectively. N. glauca was sampled in open areas near Be' er Sheba, Israel, in November 2013. N. attenuata seeds (irPMT and WT plants) were germinated as previously described by Kruegel et al. 29 and grown in individual pots, side by side in the same environmental conditions in a glasshouse. These plants were not exposed to any floral visitors. Pollinator visitation is known to differ between the WT and irPMT 22 and because pollinators can act as vectors, this would have been a confounding variable. Using greenhouse-grown plants allowed a clearer test of the effects of nicotine alone.
All plants were sampled between 07:00 am-08:00 am. For each plant sample, nectar was collected from 20-30 flowers (at least 100 μ l), using sterile tips (under sterile conditions to avoid contamination).
For the culture-dependent approach, the nectar was collected from five N. attenuata (WT) and five irPMT plants. For the culture-independent approach, nectar was collected from six N. attenuata WT, five N. attenuata irPMT, and five N. glauca WT (ISWT) plants.
Culture-dependent approach. One nectar sample from each plant got split onto four different isolation media: (i) tap water-yeast extract agar [TWYE; 0.25 g of yeast extract (Roth), 0.5 g of K 2 HPO 4 , and 16 g of agar (Roth) per L of tap water]; (ii) Streptomycetes isolation media (SIM; 0.4 g casein, 1.0 g starch, 0.5 g KNO3, 0.2 g K 2 HPO4, 0.1 g MgSO4, 0.1 g CaCO 3 and 16 g agar in 1L distilled water); (iii) R2A agar (Fluka Analytical, Germany) and (iv) Nutrient agar (NA, Fluka Analytical, Germany). No colonies were observed in SIM and TWYE media. After incubation, colonies were picked from R2A and NA plates, sub-cultured and purified. For long-term use, each culture was suspended in 50% glycerol solution and stored at − 80 °C.
Species identification procedure. A total of 83 16S rRNA amplifications were performed for the bacterial isolates, using Ready Mix Taq PCR reaction mix (Sigma-Aldrich), template DNA and 50 μ M primers [27F (5'-AGAGTTTGATCCTGGCTCAG-3') and 1492R (5'-GGTTACCTTGTTACGACTT-3')] 29 . The reaction conditions were as follows: 95 °C for 1 min, followed by 30 cycles of denaturation at 95 °C for 30 s, annealing at 53 °C for 30 s and primer extension at 72 °C for 30 s, followed by a final extension at 72 °C for 5 min. Direct sequencing using the primer 783R (5'-CTACCAGGGTAT C TAATCCTG-3') 30 was conducted in Big Dye Mix (Applied Biosystems, Foster City, CA, USA), and purification of the sequencing reactions was performed using the Nucleo-SEQ Kit (Macherey-Nagel, Düren, Germany). Analysis of all sequences was carried out in EzTaxon server (http://eztaxon-e.ezbiocloud.net/) 31 . Based on the genera level, a phylogenetic tree was generated for all the 16S rRNA sequences, using neighbor-joining algorithms drawn from the MEGA 5 package 32 . An evolutionary distance matrix for the neighbor-joining analysis was prepared using the Jukes and Cantor model 33 . The robustness of the inferred tree topologies was evaluated after 1000 bootstrap replicates of the neighbor-joining data using MEGA 5 software. Nucleotide sequence accession numbers of the partial 16S rRNA sequences determined in this study were deposited in the GenBank (LK020716-LK020798).
Statistical analyses of the cultured communities. To determine whether the cultured bacterial communities in WT and irPMT nectar were significantly different, we used phylogenetic indices such as UniFrac 34 and P-test significance test 35 . Culture-independent approach. DNA was extracted from the nectar samples using a DNA isolation kit (DNeasy Blood and Tissue, Qiagene, Germany) according to the manufacturer's instructions.
PCR amplifications were performed in 25 μ l reactions in PCR tubes. A mastermix was made using the EmeraldAmp MAX HS PCR Master Mix (Takara bio Inc). Final concentration of CS1_515F and CS2_806R primers was 0.5 ng/μ l. Genomic DNA in the amount of 10-100 ng was added to each PCR reaction. Cycling conditions were as follows: 95 °C for 5 minutes, followed by 28 cycles at 95 °C for 30", 55 °C for 45" and 68 °C for 30". A final 7 minute elongation step was performed at 68 °C. Each sample was analyzed using agarose gel electrophoresis. Reactions were verified to contain visible amplification, in addition to no visible amplification in the no-template control, prior to the second stage of PCR amplification.
Illumina MiSeq sequencing. MiSeq sequencing was performed at DNA Services (DNAS) Facility -University of Illinois at Chicago (UIC). Before sequencing, a second PCR amplification was performed in a 10μ l reaction in 96-well plates. A mastermix for the entire plate was made using the 2X AccuPrime SuperMix II. Each well received a separate primer pair, obtained from the Access Array Barcode Library for Illumina Sequences. The final concentration of each primer was 400 nM, and each well received a separate primer set with a unique 10-base barcode (Fluidigm, South San Francisco, CA; Item# 100-4876). Separate reactions with unique barcodes were included for positive control, no-template control (reaction 1), and a second no-template control reaction containing only Access Array Barcode library primers. Cycling conditions were as follows: 95 °C for 5 minutes, followed by 28 cycles at 95 °C for 30", 60 °C for 30" and 68 °C for 30". A final, 7 minute elongation step was performed at 68 °C. PCR yields of positive and negative controls and select samples were validated using Qubit fluorometric quantitation with the Qubit 2.0 fluorometer (Life Technologies) and with size and quantification employing an Agilent TapeStation2200 device with D1000 ScreenTape (Agilent Technologies, Santa Clara, California). After assessing no amplification in the negative controls, samples were pooled in equal volume and purified using solid phase reversible immobilization (SPRI) cleanup, implemented with AMPure XP beads at a ratio of 0.6X (v:v) SPRI solution to sample. This ratio removes DNA fragments shorter than 300 bp from the pooled libraries. The final quality control was performed using TapeStation2200 and Qubit analysis, prior to dilution to 6 pM for emulsion PCR.
Pooled, diluted libraries were sequenced on an Illumina MiSeq instrument, using a MiSeq 600-cycle sequencing kit version 3, and analyzed with Casava1.8 (pipeline 1.8). Reads were 200 nucleotides in length (paired end, 2 × 200). PhiX DNA was used as a spike-in control. Barcode sequences from Fluidigm were provided to the MiSeq server, and sequences were automatically binned according to 10-base multiplex identifier (MID) sequences. Raw reads were recovered as FASTQ files.
Sequence analysis. Bioinformatics were performed using MOTHUR v.1.33.3 38 . The operational taxonomic unit (OTU)-based approach of the MiSeq Standard Operating Procedure (SOP) was followed (Kozich et al. 2013 39 ). Briefly, any sequences with ambiguities or homopolymers longer than 8 bases were removed from the data set. Sequences were aligned using the SILVA-compatible alignment database available within MOTHUR. Sequences were trimmed to a uniform length of 295 base pairs, and chimeric sequences were removed using Uchime (Edgar et al. 2011 40 ). We classified sequences using the MOTHUR-formatted version of the RDP training set (v.9), and any unknown, chloroplast, mitochondrial, archaeal, or eukaryotic sequences were removed. Sequences were clustered into OTUs, based on 97% sequence identity. To avoid biases associated with uneven numbers of sequences across samples, the entire dataset was randomly subsampled to the minimum number of sequences per sample (lowest number of sequences that were obtained in a sample): 2146 sequences per sample.
All of the sequence data analyses reported in this paper can be downloaded from the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) with accession number SRP049451.
Microbial richness and similarity estimations. Alpha Diversity. We calculated microbial richness in the floral nectar of each of the three plants (WT, irPMT, ISWT) on the subsampled OTUs and genera tables. The expected taxon (either OTU or genera) richness for the complete collection of each of the three plants was calculated by the abundance-based Chao 1. The Chao 1, a nonparametric, abundance-based richness estimator that predicts the true number of taxa based on the proportion of rare taxa in a sample 41 , was found to be suitable for microbial diversity analysis 42 . To remove the effect of sample order, the sample order was randomized 100 times and the mean richness estimate was computed for each sample accumulation level. Beta-Diversity. To evaluate the similarities of the bacterial community composition between each pair of the three plant types, we calculated the observed shared subsampled OTUs (Sobs) and the expected number of shared OTUs, using Chao's coverage-based estimator 43 . In addition, we calculated Chao's Estimated Jaccard and Sørensen incidence-based similarity indices, based on the OTU analysis. The two latter indices are incidence-based estimates of the probability that, if one individual is chosen from each of two plant types, both belong to OTUs that are shared between the plants 44 . This model also accounts for the probability that some OTUs were not sampled when actually present. This method allows for a bootstrapping approach (n = 200) to generate SD for the estimated similarities. All above calculations were performed using Estimate S9.1.0 To visualize beta-diversity among the three plant types (WT, irPMT, ISWT), we processed the data using the MOTHUR program (version 1.33.3; Schloss et al. 38 ), to generate the following: (a) A Venn diagram was generated to show the number of unique and overlapping taxon memberships between the three plant types at a distance of 0.03. (b) A dendrogram describing the similarity of the samples to each other was generated using the jclass calculator within the tree.shared command, which returns the traditional Jaccard index measuring the dissimilarity between two communities. Parsimony analysis and the UniFrac weighted algorithm were used to determine whether the clustering within the resulting dendrogram was statistically significant. (c) Ordination plots based on the jclass distance matrix between samples were plotted using the dist.shared command with Principal Coordinates Analysis (PCoA) in 2 dimensions. Two methods were used to statistically analyze the plots: (1) analysis of molecular variance (AMOVA), which tests whether the centers of the clouds of samples representing each plant type are more separated than the variation among the samples of the same plant species, and (2) Bartlett's test for homogeneity of variance, which determines whether the variation in each of the plant type samples is different.

Results
Culture-dependent approach. Isolates belonging to a total of 33 different species and three classes (Table 1,  isolates) were recovered only from the irPMT nectar. Five novel bacterial species (with 97-98% similarities in the 16S rRNA gene sequences to known species) were isolated (Table 1).
Bacterial communities in the WT nectar were evenly dominated by three different bacterial classes (Actinobacteria, Bacilli and Alphaproteobacteria, 32-35%). In contrast, the irPMT nectar was highly dominated by Actinobacteria (70%), followed by Alphaproteobacteria (20%), and Bacilli (10%). At the genus level, the relative position and diversity found among nectar bacterial communities of the WT and the irPMT plants was also reflected in the 16S rRNA phylogenetic tree (Fig. S1).
A phylogenetic tree was constructed based on species level and subjected to UniFrac and P-test significance test (http://unifrac.colorado.edu/). The bacterial communities from the WT and the irPMT nectar samples were significantly different [UniFrac (P = 0.01) and P-test (P = 0.03)], demonstrating that nicotine has a direct effect on nectar's bacterial community composition.  S2a) and subsampled data (Fig. S2b)]. These indicate that the surveying effort covered almost the full extent of taxonomic diversity at this genetic distance (Fig. S2).
Composition and richness of bacterial communities in the nectar. Bacterial isolates found in N. attenuata WT nectar were predominantly of the class Bacilli (38.5%), followed by Actinobacteria (22.8%), and Gammaproteobacteria (20.3%). In the irPMT nectar samples, bacterial isolates were predominantly of the Bacilli class (36.4%), followed by Actinobacteria (20.5%), and Gammaproteobacteria (11.8%). Sequence analysis showed significant differences between the bacterial communities from the nectar of the wild type (N. atttenuata) and the transgenic N. attenuata (AMOVA, F 1,9 = 2.34; P = 0.01), as was also demonstrated by the PCoA ordination that was based on OTU relative abundances (Fig. 1).
We further compared the bacterial communities from nectar of N. attenuata WT and transgenic plants with nectar that was obtained from another Nicotiana plant species (N. glauca -ISWT) as an outlier. The majority of the bacterial sequences from nectar of ISWT plants were classified as Gammaproteobacteria (61.8%), followed by Bacilli (14.9%) and Actinobacteria (9.8%). The nectar of the three plant types differed in their bacterial taxonomic richness, both at the OTU and at the genus levels. The nectar of the transgenic N. attenuata (irPMT) had 4-5 times and 7-9 times higher observed and expected bacterial  N. attenuata WT, transgenic N. attenuata irPMT and N. glauca ISWT samples. The first two principal coordinates (PC1 and PC2) from the principal coordinate analysis of unweighted UniFrac are plotted for each sample. The variance explained by the PCs is indicated on the axes.   OTU richness compared to the N. attenuata WT and N. glauca nectar, respectively (Sobs Mean and Chao1 estimators, Table 2). ISWT had the lowest number of taxa, whereas the N. attenuata WT was intermediate. Thus, the transgenic N. attenuata (irPMT) nectar harbored a much richer bacterial community than did the two wild-type plants.
The two incidence-based similarity indices, Chao-Sørensen and Chao-Jaccard, which were used to compare bacterial community composition among the three plant types, demonstrated that the bacterial community composition in the nectar of the irPMT plants was about as different from the composition in the WT nectar as it was from that in the ISWT nectar ( Table 3). The highest similarity values were found between the nectar samples of the two wild type plants (WT and ISWT, Table 3).
The principal coordinates analysis (PCoA) of the sequenced data showed a clear clustering of the various samples from each plant type in a distinct region of the ordination (with the exception of one sample from the irPMT plant that was clustered with the WT samples, Fig. 1). The AMOVA that was used to assess the statistical significance of the spatial separation observed among the three plant types in the PCoA ordination plots showed statistically significant dissimilarities across the three groups (F 2,6 = 2.98, P < 0.001). Bartlett's test for homogeneity of variances demonstrated that there was a significant difference in the variation of the samples obtained from the three plant types (P = 0.005).
The dendrogram of the community cluster analysis under the Jaccard coefficient (jclass) structural diversity measure showed separated positions of the bacterial community in the floral nectar of each of the three plant types (Fig. 2). N. attenuata WT samples were closer to N. glauca ISWT samples than to the samples obtained from N. attenuata irPMT (Fig. 2). Parsimony analysis of this UniFrac dendrogram showed a significant difference for the data corresponding to the clustering of the samples between N. glauca ISWT and N. attenuata WT (P < 0.05), between irPMT and WT (P < 0.05), and marginally significant between irPMT and ISWT plants (P = 0.08). UniFrac weighted analysis, which determines whether two or more communities have statistically significant structural dissimilarities, indicated statistically significant dissimilarities among the samples of each pair from the three studied plant types (P < 0.001).
The Venn diagram (Fig. 3) showed that N. attenuata irPMT had the highest number of unique OTUs (417/578 = 72.1%). The two wild types had only two (0.3%) unique OTUs. It was also demonstrated that  Table 3. Pairwise similarity indices of bacterial community composition in the floral nectar of the three plant types, based on subsample of the OTU table. the irPMT plants shared a low fraction of OTUs with the WT plants (7.6%) and with the ISWT plants (3.4%). Only 93 (16.1%) OTUs overlapped among the three plant types (Fig. 3). Species belonging to the genera Gardnerella and Lactobacillus were detected only in irPMT plants, Enterococcus was found only in the WT plants, and Enhydrobacter and Vibrio were detected only in the ISWT plants (Table S2). Full taxonomic classifications in the genera level are shown in Table S2.

Discussion
Here we have demonstrated that the elimination of a single secondary metabolite -nicotine, which caused an elevation of another secondary metabolite -anatabine, drastically affected the composition and diversity of bacterial communities in floral nectar. This was shown both by culturable and as well as by unculturable methods. The culture-dependent method demonstrated that only five out of 33 different bacterial species were isolated from both N. attenuata WT and the transgenic irPMT plants (Table 1). By using the MiSeq illumina technique, we found that only 23.7% of the total observed OTUs (137/578) were identified in both the N. attenuata WT and the transgenic irPMT plant. The bacterial community composition in the nectar of the outlier plant species within the genus (N. glauca) was different from both N. attenuata WT and the transgenic irPMT plants. However, the similarity indices between the two wild type plants were much higher than between each one of them and the irPMT plants.
A possible explanation for this observation is a differential response of the bacterial community to the different pyridine-type composition of the nectar of these three plants. Studies showed that nicotine interferes with the growth of microorganisms 45 . Nicotine is undetectable in irPMT plants while both wild type plants contain nicotine, although about 55 times lower in N. glauca compared to its concentration in N. attenuata WT plants. Yet, the effect of nicotine on the bacterial community might be crucial even in relatively low concentrations and thus leads to a relatively similar community composition of the two wild type plants. Furthermore, anabasine which is the most dominant alkaloid in N. glauca is known to possess pronounced antibacterial properties 46,47 . Thus, anabasine may have a similar prominent impact on the bacterial community as nicotine.
Furthermore, the taxonomic richness of the bacterial communities in the floral nectar of the irPMT plants was 574 and 1134 for Sobs and Chao1 expected OTUs, respectively ( Table 2). These values were much higher than in the two wild type plants. Such high richness and diversity are probably due to the elimination of nicotine, which is considered an extremely toxic alkaloid 25,48 , from the irPMT plants and/or due to the presence of high anatabine concentrations. This suggests that anatabine, which is characterized only in N. attenuata irPMT plants, enabled the inhabitation of many more bacteria than nicotine and anabasine. Indeed, Steppuhn et al. (2004) 27 demonstrated strong evidence that nicotine but not anatabine efficiently protects plants against herbivorous insects. There are no parallel studies that compare the antibacterial effects of nicotine and anatabin. However, we suggest that while the presence of nicotine and anabasine inhibited some bacterial species proliferation in the nectar, the presence of anatabine supported it.
Studies showed that some bacteria have developed nicotine resistance 18 . For example, a study on bacterial communities inhabiting Nicotiana leaves demonstrated that some bacterial species were adapted to nicotine as a growth substrate, and developed biochemical strategies to decompose this organic heterocyclic compound (reviewed by Brandsch 49 ). In the current study, we observed a relatively low bacterial OTU numbers in the nectar of the WT and the ISWT plants. Some of these OTUs are probably adapted to nicotine as a growth substrate. However, as far as we know, the influence of anatabine on bacterial growth was never studied before.
The nectar from N. attenuata WT and from the transgenic irPMT plants that were grown in a closed greenhouse had significantly different bacterial community composition. Secondary metabolism is known to be sensitive to environmental parameters that differ between glasshouse and field conditions (e.g., UV-B influence 50 ). However, in a previous study 27 , no significant difference was found between plants that were grown in the field and in the glasshouse in nicotine and anatabine levels (either in the WT or in the irPMT plants). Recently, we showed that nectar consumers may act in a natural environment as bacterial vectors 7,9 and affect the transfer of bacteria into the nectar. Although WT and irPMT plants in the current study were not exposed to visitors, as they both were grown in a greenhouse with no access to nectar consumers or pests, they nevertheless possessed diverse bacterial communities. This indicates that bacteria are probably transferred to the nectar also by other means, such as soil, irrigation, pesticides, and air and not only through insects or other nectar consumers.
Nectar of N. attenuata with all its constituents has the potential to filter flower visitors, favoring some and deterring others. Here we demonstrated how the pyridine-alkaloids (nicotine, anatabine and anabasine) influence the bacterial community composition in the floral nectar of N. attenuata. The differences in the nectar bacterial communities between the three plants may also have an effect on nectar consumption by floral visitors. This is because different bacterial species may degrade different nutritional substrates into different products; thus, these secondary metabolites, through their effect on the bacterial community composition, may have also an indirect effect on the nectar's nutritional composition and consequently, on the floral visitors and pollination.
Little is known about how specific microbial communities are biostimulated by an individual plant's secondary metabolites. Uhlik et al. 51 showed that plant secondary metabolites have a strong effect on the bacterial communities' activity and degradative ability. Yuan et al. 48 demonstrated that nectar odors directly impact plant reproductive fitness through pollinator attraction or deterrence of nectar robbers and florivores. Kessler and Baldwin 21 tried to understand the purpose of nicotine presence in nectar. The results of the current study indicate that secondary metabolites play a role in shaping the composition of the bacterial communities in nectar. These may indirectly have an effect on attracting or deterring plant visitors and thus, may explain the evolutionary role of nicotine (or other secondary metabolites) in the nectar. Furthermore, bacterial nectar could provide direct benefit to insect pollinators, for example, by assisting in the insects' digestion, especially because insects lose their gut microbes each time they molt 52 .

Conclusion
The aim of this study was to examine whether the bacterial community composition in nectar varied with an artificially-induced change in a single metabolite (nicotine). This change led to a different pyridine-alkaloid composition among the studied Nicotiana species. By using both a culture-dependent approach and MiSeq illumine sequencing, we showed that bacterial community composition in floral nectar varies between different Nicotiana plant types and especially between N. attenuata WT and the genetically transformed irPMT plants, which are unable to produce nicotine. Yeast metabolism has been shown to significantly contribute to floral scent in other plant genera 53 ; therefore, bacteria in nectar might contribute to variation in plant signaling and eventually, even to pollination success and plant fitness. This study is the first step toward examining our hypothesis that different bacterial communities in nectar that are formed by a change in a single metabolite (which increased another metabolite) and by the pyridine-alkaloid composition, may affect nectar consumption, plant pollination and plant fitness.