1. Introduction
Life on leaf surfaces that are exposed to extreme variations in meteorological and other environmental factors looks like the life of a fugitive. However, existing evidence suggests that microbial communities are well established, that only a subset of the air-introduced taxa can colonize them and that leaves are not passive acceptors of microbes deposited from the air [
1,
2,
3]. There is also growing evidence indicating the involvement of leaf bacteria in important interactions that may affect plant fitness [
4,
5] and quality and the production of crop plants [
6,
7,
8], and even in processes at a global scale such as nitrogen fixation [
9].
Phyllospere microorganisms can arrive as bioaerosols, via rainfall or irrigation water [
10,
11], by animals [
12], particularly herbivorous insects [
10], and also as colonizers of other plant parts at earlier stages of the plant ontogeny [
10,
13]. They may be of plant, animal, water or soil sources, but the relative contribution of each source is still unclear [
14]. Air is an important medium for their transfer and further deposition on leaf surfaces. Models of atmospheric circulation suggest that particles of the size of bacterial cells can move readily between continents within a year [
15], but dispersal limitations may make local sources more important than distant ones for the microbial colonization of the phyllosphere [
10]. In a study using high-throughput sequencing (HTS), it was estimated that up to 50% of airborne bacteria in downwind air samples were of local plant origin [
16].
Culture-independent methods have shown that leaf microbial communities are much more diverse than previously thought and provided insight on the composition, the physiological aspects and the niches of the leaf-associated microorganisms [
17,
18]. New tools such as HTS enabled studies dealing with the structure and variation of the epiphytic bacterial community, and its functions and interactions with biotic and abiotic factors, to rapidly increase [
5,
14,
19]. However, phyllosphere studies concern mostly man-made environments [
20], crops or other plants of economic importance [
18] and their pathogens [
21], and usually one [
22] or a few plant species [
23]. Few have examined epiphytic communities in natural environments, at the ecosystem level, and even fewer examined their features in more than one time slots [
2,
24] and/or in relation to the air inoculum [
25].
Studies at the ecosystem level can provide a better understanding of the factors that determine phyllosphere colonization and expand our knowledge regarding complex dynamics at play. Published studies that focus on the structure of the epiphytic bacterial community concern the tropical [
4,
9], the Atlantic [
26,
27], the temperate [
28] and other forests [
29] or ecosystem types [
3,
30]. To our knowledge, very few studies deal with the epiphytic bacterial communities from areas with typical Mediterranean climate—those that do include the study by Peñuelas et al. [
31] and those exploring the epiphytic microbial community of an ecosystem at Halkidiki, northern Greece [
1,
32,
33,
34]. With culture-dependent methods, it was found that phyllosphere bacterial populations of the latter ecosystem are lognormally distributed, with their size ranging from non-detectable to a maximum of ~10
7 CFU g
−1 [
32] and that phyllosphere colonization is influenced primarily by the leaf water content followed by the phosphorus content and the thickness of the adaxial epidermis [
33]. Using first-generation molecular techniques, denaturing gradient gel electrophoresis (DGGE), in particular, Vokou et al. [
1] evidenced differences between the airborne and the epiphytic communities, prominence of bacteria associated with only one plant species and a frequent occurrence of lactic acid bacteria on leaves.
An important question regarding the microbial community of the phyllosphere, which is exposed to large variations in meteorological parameters, is how it behaves in response to a very seasonal environment such as the Mediterranean. Does it keep its structure with only minor fluctuations, or does it change overall? To answer to this question, we focused on bacteria and studied their phyllosphere community in summer and winter. These are the two most contrasted seasons of the Mediterranean climate, characterized by the stress factors of drought and low temperatures, respectively. By studying the microbial community at these times, we expect to detect the seasonal differences at their maximum. We performed this study in the same Mediterranean ecosystem mentioned above, at Halkidiki, Greece, sampling from the same species and in the same way as before [
1,
32,
33,
34,
35]. Using HTS, we aimed at identifying the microbial taxa present, estimate their contribution to the epiphytic microbial community, detect their dominant life strategies (specialists, generalists), and further examine if seasonal changes are only quantitative (in abundance), or whether they are of more fundamental character associated with differences in the number, identity, occurrence, and dominant functional traits of the taxa present. Considering the air as a most important medium for microbial transfer [
2], we also examined how much the established community in the phyllosphere deviates at the time of sampling from the transient air community. These explorations allow a better understanding of the structure of the microbial community of the Mediterranean phyllosphere, of the dominant players and their strategies, and also of their responses to the marked seasonality of the Mediterranean environment.
3. Results
As determined by quantitative real time PCR, the number of bacterial 16S rRNA gene copies per gram of plant tissue of the Mediterranean phyllosphere studied was of the order of magnitude of 10
8 and did not differ between seasons (
Table 1). The number of bacterial 16S rRNA gene copies per cubic meter of air was of the order of 10
6. Values for microbial abundance in the air and in the phyllosphere cannot be directly compared as the units differ—i.e., in number of copies per unit of weight, in the case of the phyllosphere, and per unit of volume, in the case of the air. Nevertheless, taking into consideration that the density of air at sea level is approximately 1/800th the density of water, what corresponds to 1.25 kg m
−3 [
57], we can estimate the microbial abundance in the air on a per unit of weight basis (
Table 1). Values are of the order of 10
3 per gram of air.
After read processing, the removal of singletons, non-bacterial sequences and normalization, a total of 890 different OTUs were detected. Rarefaction curves that were calculated for all samples (
Supplement Figure S1) indicate that a satisfying part of the diversity was recovered with the sequencing effort applied in most samples. There were 771 OTUs in the overall summer dataset and 430 in the winter one. Of these, 750 and 420 OTUs, respectively, corresponded to the phyllosphere. Air samples had far lower numbers of OTUs: 128 in summer and 86 in winter. The Shannon diversity index of the epiphytic community, estimated after the OTU abundances, was higher in summer (3.63 ± 0.18) than in winter (2.95 ± 0.15). Similarly, the Pielou’s evenness index was higher in summer (0.68 ± 0.02 vs. 0.59 ± 0.03), but the Simpson diversity index did not differ between seasons (
Table 1). On average, there were 233.8 ± 22.4 OTUs per phyllosphere sample in summer, whereas 146.0 ± 10.9 in winter (
Table 1). Summary results regarding the different OTUs in leaf and air samples are given in
Table 2.
Proteobacteria was the prominent phylum in the bacterial community of the Mediterranean phyllosphere examined (
Table 3), followed by Actinobacteria. Bacteroidetes, Firmicutes, Acidobacteria, Deinococcus, Chloroflexi and Planctomycetes had a considerable participation, though two to three orders of magnitude lower than that of the previous two phyla. Other participating phyla were Gemmatimonadetes, Armatimonadetes, Verrucomicrobia and also Latescibacteria, Nitrospirae, Saccharibacteria, Spirochaetae and the WD272 group; there were also a few unidentified bacteria. Within Proteobacteria, Alphaproteobacteria was the dominant class, followed by Betaproteobacteria. The rank at the phylum and class levels remained quite similar if instead of OTU abundance, the number of different OTUs representing each of the above taxa (OTU richness) was taken into consideration (
Figure 1).
At a lower taxonomic level, Rhizobiales was the bacterial order represented in the phyllosphere by the highest number of OTUs (89) (
Table 4). Rhodospirillales, Micrococcales, Sphingomonadales, and Burkholderiales followed, all with similar number of different OTUs (47–57). Rhizobiales was first in rank in the air, too, but represented by a far lower number of OTUs (18); again Burkholderiales, Sphingomonadales and Micrococcales followed, represented by 12–16 OTUs. There were also striking differences. Representatives of Acidimicrobiales and Planktomycetales, with a high participation in the phyllosphere (18 and 16 OTUs, respectively), were absent in the air, whereas other major taxa were at a far lower rank in the air than in the phyllosphere and vice versa (
Table 4).
The most abundant OTU in the phyllosphere was closely related to a Sphingomonas strain. Of the next four most abundant OTUs, one was closely related to another Sphingomonas strain, one to a Methylobacterium strain and the other two to Rhizobiales and Burkholderiales strains.
The vast majority of OTUs (89%) in the whole database were categorized as rare. Only 15 OTUs were categorized as abundant (
Table 2 and
Table 5); these were present in all phyllosphere habitats and in the air and were all locally abundant in one and up to four habitats. There were 82 universal OTUs that were detected in all phyllosphere habitats (
Figure 2). Corresponding to 9% of all recovered OTUs, these universal OTUs belong to Proteobacteria, Actinobacteria, Bacteroidetes, Firmicutes and Acidobacteria. Markedly higher was the number of OTUs (381) that were found in only one phyllosphere habitat (
Figure 2). These narrow-niche OTUs made 44% of the phyllosphere total (
Table 2).
According to the Levins’ index, the vast majority of OTUs (89%) were specialists, with only 1% being generalists. All 10 generalist OTUs belong to Alphaproteobacteria and Actinobacteria; these were also among the 82 universal OTUs and, with the exception of one, they were also present in the air (
Table 5). Of the specialist OTUs, six were abundant. None of the members of Alphaproteobacteria, a major bacterial class in the Mediterranean epiphytic community both in terms of OTU richness and OTU abundance, participated in the group of abundant specialists.
There was a pronounced seasonal difference in the structure of the phyllosphere microbial community. OTU richness was higher in summer for the entire community (
Table 1), for all major bacterial phyla, i.e., Proteobacteria Actinobacteria, Bacteroidetes, Firmicutes, Acidobacteria, and Chloroflexi, for other minor ones, such as Deinococcus and Gemmatimonadetes, and also for the Proteobacteria classes, except for the dominant Alphaproteobacteria that did not differ between seasons (
Figure 1). For the taxa differing between seasons in terms of OTU richness, the summer value was in general one to three times higher than the winter value, but Chloroflexi was essentially a summer taxon; all but one of the Chloroflexi OTUs were detected in summer. Gemmatimonadetes was similarly a summer taxon but it was represented by far fewer OTUs than Chloroflexi. In terms of OTU abundance, there was no difference between seasons of the phyla represented, except for Chloroflexi, Gemmatimonadetes and Firmicutes that were more abundant in summer. Also, the class of Betaproteobacteria was more abundant in summer, whereas that of Alphaproteobacteria in winter (
Table 3). In summer, there was also a higher number of OTUs with exclusive occurrence in only one phyllosphere habitat (
Table 1).
Given that our sampling did not allow direct comparisons between the phyllosphere and the airborne microbial communities, we subjected our samples to Cluster analysis, according to Bray–Curtis dissimilarities. Three major clusters emerged at >20% level of similarity (
Figure 3). Cluster A consisted of the two air samples. Clusters B and C, separated at a 35% level of similarity, consisted of phyllosphere samples, primarily of winter and summer, respectively. The relative participation of the phyla and Proteobacteria classes in each of these three clusters is shown in
Figure 4. Bacteria that do not belong to Proteobacteria, Actinobacteria, Bacteroidetes and Firmicutes had a very low contribution in the air, less than 1%, and whereas Alphaproteobacteria was the dominant Proteobacteria class in the phyllosphere, Betaproteobacteria was the dominant one in the air. Also, Actinobacteria had a far higher relative participation in the summer phyllosphere samples.
The SIMPER analysis, based on within cluster similarities, indicated the assemblages of OTUs that were responsible for the formation of each cluster: 17 OTUs for cluster A, 24 for cluster B, and 44 for cluster C (
Supplement Table S2). Seven OTUs were responsible solely for the formation of cluster A. Of these, three belonged to Alphaproteobacteria, two to Firmicutes and the remaining two to unknown strains that had been reported to be isolated from the soil environment. For the formation of cluster B, three OTUs were solely responsible, of which two belonged to Alphaproteobacteria and one to Acidobacteria. For the formation of cluster C, 19 OTUs were solely responsible, with > 50% of them belonging to Actinobacteria strains (
Supplement Table S2).
The correlations of the OTUs responsible for the formation of each cluster were calculated according to Maximal Information Coefficient (MIC) values. Network analysis was implemented on the OTUs exhibiting strong positive correlations (high MIC values, corresponding to a
p-value < 0.05) in each cluster (
Figure 5). Even though there are no physical interactions among OTUs found on the leaves of different host plants, network analysis provides indications of similar or opposing trends in OTU occurrence patterns. From the topological parameters calculated for each network, it was evident that the microbial community associated with cluster C was the most connected of the three. In particular, the clustering coefficient, the number of shortest paths, and the average number of neighbors took the highest values in cluster C (
Table 6).
Network analysis also provided a visualization of the OTU correlations, indicating that different taxonomic groups had strong correlations both within and among them in each cluster. In cluster B, consisting mainly of winter samples, Alphaproteobacteria showed the highest number of correlations followed by Actinobacteria (
Figure 5). In cluster C, consisting mainly of summer samples, Actinobacteria presented the highest number of correlations, while Betaproteobacteria, with a negligible role in the network structure for the other two clusters, presented high connectivity here. The topological parameters of the networks suggest that cluster C has a denser network structure than the other clusters: the Clustering Coefficient and the Average Number of Neighbors were higher in the network of cluster C (
Table 6). Furthermore, topological parameters, such as the Clustering Coefficient, Centralization, and Heterogeneity, were approximately 2.5, 2 and 1.8 times higher, respectively, for the observed cluster C network relative to a random network of the same size suggesting indeed a denser structure than expected by chance alone [
58].
4. Discussion
The study of the phyllosphere microbial community in a Mediterranean ecosystem showed that Proteobacteria is the dominant phylum in terms of both richness and abundance and that this holds true irrespective of season. This is also true for the airborne microbial community. However, the relative abundance in terms of average number of reads of bacterial classes within Proteobacteria does not remain constant between seasons. The contribution of the dominant Alphaproteobacteria decreased in summer, whereas that of Betaproteobacteria increased.
Results regarding the composition of the microbial community of the Mediterranean phyllosphere present similarities with reports from other ecosystem types of the world with different assemblages of plant species. Proteobacteria is most often the dominant group, with a participation ranging from less than 40% in tropical [
4] to more than 80% in temperate forests [
28,
29], on floating macrophytes [
59] or in the phyllosphere of individual crop species, such as
Spinacea oleracea [
60]. Depending on the conditions prevailing, Actinobacteria, Acidobacteria, Firmicutes, Bacteriodetes are reported as second in rank phyla.
Two
Sphingomonas- and one
Methylobacterium-related OTUs were amongst the five most abundant OTUs in the Mediterranean phyllosphere. Epiphytic microbial populations of soybean, clover and
Arabidopsis [
14] are also reported to be largely composed of
Sphingomonas and
Methylobacterium, and a
Sphingomonas strain was among the five most abundant strains in the tree phyllosphere of the Brazilian Atlantic forest [
27]. The dominance in the phyllosphere of methylotrophic and other taxa consuming one-carbon compounds, or of enzymes involved in the related metabolic reactions is also reported in a number of other cases [
4,
5,
18,
25,
61] and has been associated with degradation processes of compounds that are toxic to plants, to humans or to the environment that are carried out by epiphytic microorganisms [
5]. Dominance of these taxa in different plants, ecosystem types or geographical areas suggests major advantages from this type of symbiosis, which needs to be further explored.
It is reported that the phyllosphere microbial communities are very diverse in terms of richness, but not so in terms of evenness and that, generally, they comprise a few very well represented taxa and a large number of very rare ones [
62]. This is also the case for the Mediterranean phyllosphere that we studied. Only 2% of the phyllosphere OTUs were identified as abundant, whereas 87% were identified as rare. Furthermore, the microbial community in this Mediterranean ecosystem is made up almost exclusively of habitat specialists with only very few habitat generalists.
A comparison of the epiphytic community with the airborne inoculum reveals large differences. Expressed on a per gram basis, the size of the airborne microbial community is approximately five orders of magnitude lower than that of the phyllosphere. This much thinner community bears far fewer taxa than those colonizing the phyllosphere, whereas microbes in abundance in the phyllosphere are not present in the air at the time of sampling and vice versa. A number of reasons can explain the differences between the two communities. For instance, taking air samples for a limited amount of time may not suffice to capture the full diversity of the airborne community; or, the sensitivity of the method does not allow detection at very low abundance. However, air is only a medium of microbial transfer constantly inoculating leaves with a range of taxa from various sources [
63,
64,
65]. Microbes may land on leaves but then they are sorted out [
2,
66]. The epiphytic microbial community should not necessarily mirror the airborne inoculum at the time of sampling. Inocula and selection processes of the past play major roles in determining its structure.
With culture-dependent methods, it was found that the marked seasonality of the Mediterranean climate is not reflected in the size of the epiphytic microbial community [
32]. Similar is the result with the culture-independent method that we used: the level of colonization did not differ between summer and winter. This was also true for Proteobacteria and Actinobacteria, which make the largest part of the community, and for several other phyla, but not for Firmicutes, Gemmatimonadetes and Chroroflexi, and the major Proteobacteria classes. In contrast, there was a seasonal effect on richness. This was higher in summer for the entire bacterial community and for most of the taxonomic groups examined, with none of the other taxa showing higher values in winter. This higher summer richness did not always lead to higher values of the diversity indices. The Pielou’s evenness index was higher in summer suggesting a more homogenous OTU distribution in this season compared to winter. The Shannon diversity index differed also between seasons (higher in summer) but the Simpson index did not. This is explained by the fact that the Simpson index puts very little weight to rare species, which are numerous in the community, particularly in summer.
Experimental evidence suggests that the richness of the epiphytic bacterial community responds to climatic stressors and that it is very much influenced by drought [
31]. The epiphytic microbial community of the Mediterranean phyllosphere that we studied is clearly richer in summer. This combined with the fact that the overall abundance does not change in summer, although the relative abundance of the dominant Alphaproteobacteria falls, suggest that the stressful summer conditions affect primarily the dominant members of the community. The relaxation of communities from extreme dominance of some of its members gives the opportunity to others to establish. This could be the case of Chloroflexi, which seems to be a typically summer phylum in the Mediterranean phyllosphere, although its representatives have been reported from various environments including anaerobic [
67,
68] and high-mountain ones [
69].
The highest number of co-occurring/co-abundant OTUs that were detected through network analysis in winter corresponds mainly to Alphaproteobacteria, whereas in summer to Actinobacteria. Betaproteobacteria, with a negligible role in the winter phyllosphere network, increased considerably their co-occurrence patterns in summer. This change is in accordance with the increase in Betaproteobacteria in both richness and abundance and suggests an important role of this class in community structure in summer. Additionally, the summer network exhibits denser inter-associations among the correlated OTUs, as shown by the topological parameters calculated for this network compared to those for the winter and random networks. It has been suggested that robustness can increase with higher connectivity and denser taxa interactions, independent of taxa abundance [
70], in the same way that increased taxa richness can stabilize community structure against environmental changes [
71].
This is not the first time that season is found to play an important role in determining the composition of the phyllosphere bacterial community. Rastogi et al. [
22] reported Proteobacteria, Firmicutes, Bacteroidetes and Actinobacteria to be the most abundantly represented phyla on lettuce foliage, as we found for the Mediterranean phyllosphere community, and a clear separation between winter and summer samples. In another study of a single
Magnolia grandiflora tree, in which sampling took place at four seasons in one year and repeatedly at one season for three years [
25], great temporal changes were detected, with the summer leaf community being very distinct. These temporal changes were not only seasonal; communities sampled at the same time in different years showed considerable differences, what made authors argue that seasonal patterns may not be predictable from year to year.
Knowledge of the structure of the epiphytic microbial communities in different environments and how these change with time will contribute to answering open questions on the specific functions of the microbiome on plant leaves and of the specific benefits of the partnership [
72]. We found the microbial community of the Mediterranean phyllosphere to differ considerably from the air inoculum at the time of sampling, indicating selection by plants of the microbial community to be established on their leaves. This epiphytic community is dominated by habitat specialists and becomes much richer, more distinct (according to the number of OTUs on a single sample), even (according to the Pielou’s eveness index) and diverse (according to the Shannon index), denser and more connected (according to network analysis) in summer. These summer features of the epiphytic microbial community could be considered as contributing to community stability under the adverse, hot and dry conditions of the Mediterranean summer, whereas the similarity in bacterial abundance in both seasons suggests that resources may not suffice for much higher population sizes.