Bacterial microbiota associated with flower pollen is
influenced by pollination type, and shows a high
degree of diversity and species-specificity
Binoy Ambika Manirajan,
Institute of Applied Microbiology, Research Center for
BioSystems, Land Use, and Nutrition (IFZ), Justus-
Liebig-University, Giessen, Germany.
ur Integrative Biologie, Stiftung Old Herborn
University, Herborn, Germany.
MVZ Institut f€
okologie GmbH, Herborn,
Diverse microorganisms colonise the different plant-
microhabitats, such as rhizosphere and phyllo-
sphere, and play key roles for the host. However,
bacteria associated with pollen are poorly investigat-
ed, despite its ecological, commercial and medical
relevance. Due to structure and nutritive composi-
tion, pollen provides a unique microhabitat. Here the
bacterial abundance, community structure, diversity
and colonization pattern of birch, rye, rapes and
autumn crocus pollens were examined, by using
cultivation, high-throughput sequencing and micros-
copy. Cultivated bacteria belonged to Proteobacteria,
Actinobacteria and Firmicutes, with remarkable dif-
ferences at species level between pollen species.
High-throughput sequencing of 16S rRNA gene
amplicon libraries showed Proteobacteria as the
dominant phylum in all pollen species, followed by
Actinobacteria,Acidobacteria and Firmicutes. Both
plant species and pollination type significant influ-
enced structure and diversity of the pollen
microbiota. The insect-pollinated species possessed
a more similar microbiota in comparison to the wind-
pollinated ones, suggesting a levelling effect by
insect vectors. Scanning electron microscopy as well
as fluorescent in situ hybridisation coupled with
confocal laser scanning microscopy (FISH-CLSM)
indicated the tectum surface as the preferred niche of
bacterial colonisation. This work is the most compre-
hensive study of pollen microbiology, and strongly
increases our knowledge on one of the less investi-
Plant-associated microbiome studies have increased enor-
mously because of their recognised importance. The
majority of such studies focused on the plant microhabitats
corresponding to rhizosphere, phyllosphere and endo-
sphere (Ryan et al., 2008; Turner et al., 2013; Aleklett
et al., 2014), while other niches, such as anthosphere, car-
posphere and spermosphere remained less investigated
so far (Vandenkoornhuyse et al., 2015). It has been shown
that plants usually exhibit a high species-specificity level of
associated bacteria (Redford et al., 2010; Berg et al.,
2014). A large microbiome spreads on the surface of
plants and most abundant members of this are bacteria
(Meyer and Leveau, 2012). Plant organs carry number of
microorganisms, especially bacterial communities, and
they can be found in and on various organs, including pol-
len (Junker and Keller 2015). Certain bacterial species
could be isolated and cultivated from pollen extracts by
Colldahl and Carlsson (1968). This observation was later
confirmed with scanning electron microscopy by Colldahl
and Nilsson (1973). The sources of bacterial colonisation
of pollen were reported to be honey bees, weather, plant
materials, other insects, animals and human activities
(Hani et al., 2012). Although some research indicated the
presence of various bacterial species on pollen, little is
known on this microhabitat (Gilliam, 1979; Hani et al.,
Many plants are emitting large quantity of pollen during
spring to autumn and several types of plant pollen may
cause serious pollen-related diseases (Carinanos and
Casares-Porcel, 2011; Oldenburg et al., 2011). Therefore,
pollen-associated bacteria may have a potential ecological
and medicinal impact. In addition, they may also enter the
Received 1 July, 2016; accepted 4 September, 2016. *For corre-
spondence. E-mail Sylvia.Schnell@umwelt.uni-giessen.de; Tel.
149(0)6419937351; Fax +49(0)6419937359.
C2016 Society for Applied Microbiology and John Wiley & Sons Ltd
Environmental Microbiology (2016) 18(12), 5161–5174 doi:10.1111/1462-2920.13524
plant reproduction processes and be directly transmitted to
the next generation as seed endophytes.
Two recent studies indicated that flower organs harbour
a unique microbiota different from that of the leaves of the
same plant species and independent from environmental
conditions (Junker and Keller, 2015). Recently, it could be
shown for birch and timothy grass that the microbial com-
munity associated with pollen is plant species-specific
(Obersteiner et al., 2016). Furthermore, pollen are a prom-
ising source of unknown bacteria, as a new genus was
isolated from pollen (Jojima et al., 2004). Since the bacteri-
al community of pollen is mainly unknown, well-structured
hypothesis-driven studies on pollen microbial communities
are necessary. This will allow to disclose the structure,
composition, dynamics as well as specificity and functions
with respect to various types of plants and pollens.
Given the most recent evidence, we hypothesised (i)
that the influence of the specific characteristics of flowers
is stronger than that of the geographic location and (ii) that
pollination will influence the pollen microbiota. Hence,
insect-pollinated plants will differ from wind-pollinated
ones, due to the spreading of insect-carried microbes
across plant species boundaries.
The aim of this study was to compare the bacterial
microbiotas associated with the pollen of four plant species
(two wind-pollinated and two insect-pollinated respectively;
Supporting Information Fig. S1) from a restricted geo-
graphical area by cultivation-dependent and -independent
methods. And thus to allow to (i) compare bacterial abun-
dance, structure and diversity between the four species,
(ii) identify the “core” pollen microbiotas, (iii) assess the
contribution of the pollination type to the variability of the
pollen microbiotas and (iv) to estimate the extent of the cul-
turable bacterial fraction by using commercial AC medium
(“all culture” medium) and a pollen-enriched mineral medi-
um. In addition, we studied the niches of colonisation of
pollen bacteria by scanning electron microscopy as well as
fluorescent in-situ hybridisation and confocal microscopy.
This work is the first comprehensive report on bacterial
microbiota associated with pollen, which integrates the
results of cultivation-dependent, cultivation-independent
and microscopy analysis of different pollen species
Cultivation-dependent analysis of bacterial microbiota
Isolation of bacteria from pollen. Total of 61 (18 from rye,
16 from birch, 15 from rape and 12 from autumn crocus)
morphologically different bacterial colonies were isolated
from the pollen species on the two different agar media.
Colony morphologies and colony numbers were similar
between the AC medium and pollen medium. Total num-
bers of CFUs of the different pollen species (birch, rye,
rape and autumn crocus) were significantly different from
each other on AC medium (Kruskal–Wallis test p50.024;
Fig. 1A). The highest number was found in autumn crocus
(7.5 67.4 310
), while the lowest in birch (4.163.1 3
). Pollen medium gave similar results (data not shown).
Identification of the bacterial isolates by 16S rRNA gene
sequencing. All 16S rRNA gene sequences of the bacterial
isolates were identified by the EzTaxon database (Kim et al.
2012). A total of 44 bacterial species were identified from
the morphologically different types of bacterial colony iso-
lates (15 species from rye, 15 from birch, 10 from rape and
9 from autumn crocus respectively) (Supporting Information
Table S1). Four major phyla were identified from the pollen
isolates: Proteobacteria (45.5%), Actinobacteria (31.8%),
Fig. 1. (A) Numbers of cultivable bacteria on flower pollen (CFUs per gram of dry pollen weight). The values are means of three samples per
pollen species (n 536SE). Kruskal–Wallis test, p50.024. (B) Venn diagram shows the isolated bacterial species shared by the analysed
5162 B. Ambika Manirajan et al.
C2016 Society for Applied Microbiology and John Wiley & Sons Ltd, Environmental Microbiology,18, 5161–5174
Firmicutes (18.2%) and Bacteroidetes (4.5%) (Fig. 2). The
bacterial richness at the species level of birch and rye pollen
was significantly higher than in rape and autumn crocus
(Kruskal–Wallis test p50.026). The isolates belonging to
Proteobacteria (especially Enterobacteriaceae) were more
abundant in birch, rape and autumn crocus, while Actino-
bacteria (especially Microbacteriaceae and Nocardiaceae)
were found to be more abundant in rye (Fig. 2). At the bac-
terial species level, each pollen hosted a very unique
community, while only three species were shared by rape
and rye. Rosenbergiella nectarea was found on all pollens
except rye (Fig. 1B and Supporting Information Table S1).
Cultivation-independent analysis of the bacterial
Ion torrent sequencing. In total 1,140,212 raw sequences
were obtained from two independent runs. After removal of
technical sequences (primer/barcode/linker), length trim-
ming (range: 350–450 nucleotides) and quality filtering
(threshold520, calculated on 50-nucleotides sliding
windows, option -w and -g in QIIME: as soon as a window
is below the quality threshold, the whole sequence is elimi-
nated), 361,518 remaining sequences were grouped into
OTUs at a 97% similarity (hereafter OTU
). After removal
of 111,692 plastidic OTUs, 147,282 mitochondrial OTUs,
181 chimeric OTUs and 1076 singletons, 92,880 sequen-
ces remained (14,420 from rye, 28,604 from birch, 38,188
from rape and 11,668 from autumn crocus, with an aver-
age of 8444 65344 reads per pollen sample), grouped
into 1056 bacterial OTUs (532 in rye, 441 in birch, 377 in
rape and 133 in autumn crocus). In order to keep a higher
number of sequences, one sample of autumn crocus was
eliminated because it had a notably lower number of
sequences (611). This would have sensibly reduced the
coverage of the sequencing. However, an additional analy-
sis of the data including this sample showed no variation in
the output of any of the alpha- and beta-diversity metrics
(data not shown).
Analysis of ion torrent sequences. Proteobacteria was
Fig. 2. Multilevel doughnut chart shows the relative abundance of cultivated bacterial taxa from birch, rape, rye and autumn crocus pollen.
The bacterial microbiota of flower pollen 5163
C2016 Society for Applied Microbiology and John Wiley & Sons Ltd, Environmental Microbiology,18, 5161–5174
pollen sample of rape. Actinobacteria and Acidobacteria
were the next dominant phyla in both birch and rye pollen
samples, while Actinobacteria and Firmicutes dominated
in autumn crocus and Firmicutes in rapes. At the family
taceae,Burkholderiaceae and Acidobacteriaceae were the
most abundant in birch pollen; Enterobacteriaceae,
Streptococcaceae,Pseudomonadaceae and Enterococca-
ceae in rape; Oxalobacteraceae,Xanthomonadaceae,
Micrococcaceae and Acetobacteraceae in rye; Enterobac-
teriaceae and Pseudomonadaceae in autumn crocus
The rarefaction curves of observed species and Chao1
richness estimator showed clear differences between pol-
len species (Fig. 3B and C).
For alpha- and beta-diversity analysis, the dataset was
rarefied to 3468 sequences in order to eliminate the biases
due to the different sequence depths across samples.
Eight different alpha diversity measures were calculated
on the rarefied dataset and all of them were found to be
significantly different between pollen species (ANOVA,
p<0.01). Birch and rye pollen showed higher values than
rape and autumn crocus (Table 1).
The ten most abundant OTUs (relative abundance >1%
of the total rarified dataset) were significantly different
between pollen species (ANOVA, FDR-corrected p<0.05)
and included 60.63% of the total reads (rarified dataset).
Among those, the OTU identified as Enterobacteriaceae was
more abundant in autumn crocus and to a minor extent in
rape, Xanthomonadaceae was more abundant in rye, three
Fig. 3. (A) Occurrence of bacterial families in the pollen samples, retrieved by 16S rRNA gene amplicon library sequencing (Ion Torrent); for
better visibility, only families with a relative abundance of 0.5% in any sample were included in the bar graph legend. (B) Chao1-rarefraction
curves of individual samples based on OTUs
. (C) Rarefaction curve describing the observed bacterial richness (number of OTUs
pollen species. Means of three replicates (n 536SD).
5164 B. Ambika Manirajan et al.
C2016 Society for Applied Microbiology and John Wiley & Sons Ltd, Environmental Microbiology,18, 5161–5174
Oxalobacteraceae OTUs more abundant in rye and to a
minor extent in birch, Buchnera was only found abundant in
birch, Enterococcaceae was abundant in rape and two Aceto-
bacteraceae OTUs were more abundant in birch (Fig. 4).
The BLAST similarity search of representative sequen-
ces from these ten most abundant OTUs gave more
precise hints about their taxonomic affiliation (Supporting
Information Table S2): the Enterobacteriaceae OTU was
most closely related to Rosenbergiella spp. (100%); the
Xanthomonadaceae OTU Stenotrophomonas rhizophila
(100%); the three Oxalobacteriaceae OTUs to Massilia sp.
(99.4%), Noviherbaspirillum suwonense (100%) and
Duganella sp. (100%) respectively; the Enterococcaceae
OTU to Enterococcus spp. (100%) and the Buchnera OTU
as Buchnera aphidicola (98.2%). The identification of the
three Acetobacteraceae OTUs did not improve, since sev-
eral genera, such as Acetobacter,Gluconacetobacter,
Asaia,Neoasaia and Kozakia, gave the same similarity
score (94.8%–96.4%). BLAST search of the representative
sequence of one of these Acetobacteraceae OTUs
(denovo2199, Supporting Information Table S2), against
sequences of type strains only, resulted in a best match
with Gluconacetobacter tumulisoli (NR114383) with a simi-
larity of just 94.8%. This suggests the occurrence of
potentially new bacterial species or even new genera asso-
ciated with pollen. Interestingly, when aligned with
sequences of uncultured material, it showed a best match
with a clone obtained from indoor dust (AM697168, Rintala
et al., 2008).
Beta diversity analysis by PCoA-plot of jackknifed-
weighted Unifrac distances at OTU
level showed that
pollen species samples were grouped into very significant
discrete clusters (Adonis R
50.764, p<0.001). Rye and
birch being the most clearly differentiated as well as rape
and autumn crocus closely segregated (Fig. 5A). Interest-
ingly, the pollination type was highly significant (Adonis
50.418, p<0.001) (Fig. 5A). Sampling site did not
affect the structure of the microbiota (p50.096 and 0.133
for adonis and anosim tests respectively) (Fig. 5A).
Profile clustering network analysis indicated the shared
as well as the exclusive OTUs
between pollen species
(Fig. 5B). Birch and rye shared more OTUs
Oxalobacteraceae (Massilia sp.), Methylocystaceae and
Acetobacteraceae being the most abundant. Xanthomona-
daceae (Stenotrophomonas rhizophila) was found
abundant in birch rye and autumn crocus, but was absent
in rape, while Enterococcus, found abundant in rye, rape
and autumn crocus, was absent in birch (Fig. 5B). Rape-
birch and rye-autumn crocus, respectively, shared very few
, which is coherent with their relative position in the
beta-diversity plot (Fig. 5A). However, rape and autumn
crocus shared abundant OTUs
(Rosenbergiella sp.) and Pseudomonas], which explains
the reduced Unifrac distances (Fig. 5A).
The core microbiome, defined as all OTUs
all analysed species, comprised only 15 OTUs
to the Alpha-, Beta- and Gammaproteobacteria, Actino-
bacteria and Firmicutes. Moreover, only four core-OTUs
were found in all eleven pollen samples analysed (Fig.
5C). Most of these core OTUs were differentially distribut-
ed across pollen species, showing characteristic patterns
(Fig. 5C). One of the core OTUs was identified by QIIME
as Sinobacteriaceae (Fig. 5C); further BLAST and
EzTaxon alignment of the representative sequence
Table 1. Alpha diversity indices based on OTUs
(mean 6standard deviation) and statistical comparison between different pollen species.
equitability Dominance Simpson
reciprocal PD whole tree Chao1
Birch 4.35 60.06 a 0.58 60.01 a 0.11 60.02 b 0.89 60.02 a 9.32 61.32 a 13.9 60.98 a 251 644.1 a 185 621.6 a
Rye 5.37 60.37 a 0.65 60.04 a 0.08 60.02 b 0.92 60.02 a 12.6 63.90 a 24.4 63.57 b 428 631.5 b 308 626.2 b
Rape 2.51 60.70 b 0.37 60.08 b 0.37 60.12 a 0.63 60.12 b 2.88 61.03 b 6.91 60.25 c 183 683.0 b 112 645.5 bc
2.25 60.13 b 0.35 60.00 b 0.42 60.01 a 0.58 60.01 b 2.36 60.06 b 5.70 62.47 c 121 620.6 b 86 622.6 c
Different letters indicate statistically significant differences between means across pollen species and within diversity indices (ANOVA p<0.05,
Fig. 4. Abundance of the 10 biggest OTUs was significantly
different between pollen of the four plant species (ANOVA, FDR-
The bacterial microbiota of flower pollen 5165
(against sequences of type strains only) resulted in a best
match with the species Hydrocarboniphaga daqingensis
with a similarity of 91%, indicating that potential new bacte-
rial genera or even new families occur ubiquitously in
pollens. BLAST alignment against uncultured material
gave a match of 100% with a clone from freshwater habi-
tats of the Amazon River (JX673475, unpublished).
Comparison of cultivation-dependent and -independent
results. We detected 18 families by cultivation, while by
IonTorrent 104 families were annotated, according to the
QIIME classification. So, the cultured fraction accounted for
about 17% of the total bacterial diversity on the analysed
pollen, at family level. One family (Flavobacteriaceae)was
detected by isolation (isolate Bd-AC-1 from birch pollen,
Supporting Information Table S1) but not by IonTorrent
sequencing (Supporting Information Table S3).
Scanning electron microscopy (SEM)
Microscopic observations revealed the surface colonisation
on the tectum (outer surface) of the pollen. The bacterial
cells were found adhering to the outer surface of the pollen
grains as single cells or clusters or as a thin biofilm. Differ-
ent bacterial morphotypes were found in the four pollen
species (Fig. 6A–D).
Fluorescent in-situ hybridisation and confocal laser
scanning microscopy (FISH-CLSM)
Fluorescent signals were detected with the universal bac-
terial probe and with the FISH probes specific to
Actinobacteria (Fig. 6E–F). Bacterial cells were mainly
detected as single cells on the tectum. No signal was
obtained in the negative controls stained with the non-
sense FISH probes (Fig. 6G).
This study compared the structure, diversity and colonisa-
tion pattern of the bacterial microbiotas associated with
pollen of four different plant species from the same geo-
graphic region, but with different pollination strategies.
Flower pollen is one of the less known plant micro-
Fig. 5. (A) Principal Components Analysis (PCoA) of weighted jackknifed UniFrac distances, comparing all pollen samples at OTU
Samples (spheres) are coloured by species and grouped by pollination type. Adonis significance test: R
50.764, p<0.001 for the factor
50.418, p<0.001 for the factor “pollination type.” (B) Profile clustering network showing occurrence and relative abundance of
across the analysed pollen species. Abundant OTUs are labelled with the original QIIME identification of the OTUs. Edge size
proportional to the mean OTUs abundance per pollen species. Node sizes defines total OTU abundance. Colour of each cluster represents the
different pollen species. (C) Structure and abundance of the pollen core microbiome, defined as the OTUs
detected in all pollen species.
Pies are coloured by pollen species and show the distribution of the respective core OTU. Pie size indicates the absolute abundance (number
of reads) of the respective OTU, while numbers at the edges indicate the relative abundance (percentage of total reads). Black border around
the pie indicates that the OTU was found in all pollen samples. Labels indicate the original QIIME identification of the OTU and, in brackets,
the next relative (unambiguous) genus as retrieved by BLAST alignment (in some cases, BLAST alignment did not result in closer
unambiguous relatives than the QIIME identification).
5166 B. Ambika Manirajan et al.
Fig. 6. Scanning electron microscopy images showing the bacterial colonization of autumn crocus (A), birch (B), rape (C) and rye (D) pollen.
(E) Volume rendering of a confocal laser scanning microscopy image-series, showing FISH-stained bacteria on autumn crocus (Colchicum
autumnale L.) pollen grains. Yellow: Actinobacteria (double stained by both probe HGC236 and universal probe EUB338-MIX); red: other
bacteria (stained by the universal-probe only). Cyan: Pollen autofluorescence. (F) Three-dimensional model of panel E: bacteria were localized
on the tectum, the outer surface of pollen grains. (G) Negative control stained with a mix of non-sense FISH-probes labelled with all the
fluorochromes used in E. Scale bars in all panels represent 10 mm.
The bacterial microbiota of flower pollen 5167
habitats. While pollen-associated fungi were extensively
investigated, only a few studies focused on bacteria. How-
ever, these studies relied only on the cultivated fraction
(Colldahl and Nilsson, 1973), amplicon libraries sequenc-
ing (Junker and Keller, 2015) or molecular fingerprinting
(Obersteiner et al., 2016). Here, for the first time,
cultivation-dependent, cultivation-independent and micros-
copy analysis of four different pollen species were
performed, thus achieving the most comprehensive bacte-
rial microbiota analysis of pollen.
Cultivation-dependent analysis of bacterial microbiota
Comparison of CFU numbers showed that the population
size of cultivable bacteria was significantly different
between pollen species. Autumn crocus and rape had a
higher bacterial load, followed by rye and birch (Fig. 1A). It
is known, that various chemical and physical factors limit
bacterial growth and survival in and around the plant sur-
face (Lindow and Brandl, 2003). The ultrastructure
examination of 34 species of plant pollen showed that the
structure and pattern of the pollen outer layer (exine) dif-
fered between plant species and genera (Kosenko, 1999).
The pollen showed a species-specific structure (Ariizumi
and Toriyama, 2011), which influences the occurrence and
abundance of microbes (Blackmore et al., 2007). Further-
more, the bacterial population size on the plant surface
depends on the level and availability of nutrition (Remus-
Emsermann et al., 2012). Nutritional composition studies
of pollen revealed that hand-collected, bee-collected and
stored pollen contained different amounts and types of pro-
teins, lipids, ash and carbohydrates (Clark and Lintas,
1992; Human and Nicolson, 2006; Nicolson and Human,
2013). Proteobacteria (especially Gammaproteobacteria,
family Enterobacteriaceae) were predominant among the
isolates, followed by Actinobacteria, Firmicutes and Bac-
teroidetes in all the pollen species except for rye, were
Actinobacteria dominated, followed by Proteobacteria (Fig.
2). Floral nectar-related major bacterial phyla from different
plant species were Proteobacteria,Actinobacteria and Fir-
erez et al., 2012; Mortazavi et al.,
2015). The extreme low overlap of bacterial species
between the investigated pollens (Fig. 1B) demonstrated,
that, the culturable fraction, of the pollen microbiota had a
surprisingly high level of species-specificity. Only Rose-
nbergiella nectarea (Halpern et al., 2013) was isolated
from three of the four investigated pollen species, thus
confirming that flower organs are the preferred habitat of
this genus (Lenaerts et al., 2014).
Cultivation-independent analysis of bacterial microbiota
Our culture independent approach showed that a highly
diverse bacterial community is inhabiting the pollen. The
taxa inventory obtained by high-throughput sequencing
corroborated the results of cultivation-dependent analysis:
Proteobacteria was the most abundant phylum and Gam-
maproteobacteria was one of the most abundant classes.
At the family level, Enterobacteriaceae was one of the
most abundant family in all pollen except rye (Fig. 3A). The
bacterial species richness in wind-pollinating pollen spe-
cies was significantly higher than in insect-pollinating
pollen species (Fig. 3C). According to the alpha diversity
indices, bacterial diversity significantly differed across the
pollen species (Fig. 3B; Table 1). Beta diversity clearly indi-
cated that bacterial communities were significantly
different between pollen species, with the additional factor
“pollination type” being also highly significant (Figs. 4 and
5A). These differences might be related to the different pol-
len coat structure between wind-pollinated and insect-
pollinated species. It could be shown that insect-pollinated
pollen has a more lipid-rich coat than wind-pollinated pol-
len (Edlund et al., 2004). Bacterial and fungal patterns on
different pollen species (birch and timothy grass) were
found to be species-specific, and the birch pollen showed
higher microbial diversity than timothy grass (Obersteiner
et al., 2016). However, these two pollen species have been
collected from different regions, while in our work we ana-
lysed samples originating from four sampling sites within a
restricted geographic area, with different pollen species
sampled from the same sites. Beta-diversity analysis dem-
onstrated that the site of origin is not a significant factor
shaping the composition of pollen microbiota. The statisti-
cal significance of “pollination type” we found, based on
two species per category, should be taken as good evi-
dence rather than as prove, and the analysis of more
pollen species would be suitable for further confirmation.
The network analysis indicated that there were more
species-specific bacterial taxa than shared ones (Fig. 5B),
which is in agreement with bacterial diversity analyses of
floral nectar (
erez et al., 2012). It also revealed
that there were more common bacterial taxa between
wind-pollinating species than between insect-pollinated
ones. This suggests an impact of the pollen characteris-
tics, similarly to the microorganisms in the floral nectar,
which were influenced by chemistry of the nectar (Fridman
et al., 2012), and impacted by plant pollinator interactions
(Ushio et al., 2015; Aizenberg-Gershtein et al., 2013) and
wind pollination (Glushakova and Chernov, 2007). The
identified core-microbiota (Fig. 5C) included only a few
OTUs, both abundant (including Rosenbergiella,Pseudo-
monas and Lactococcus) and rare ones. Rosenbergiella
sp. was isolated from floral nectar (Halpern et al., 2013),
while previous studies reported Lactococcus sp. detected
in bee-collected pollen and Pseudomonas sp.onpollen
samples (Vanneste et al., 2011; Basim et al., 2006). It can
be speculated that by increasing the sequencing effort
additional minor members shared by all pollen species
5168 B. Ambika Manirajan et al.
would be revealed, thus increasing the diversity. However,
this would probably not significantly affect the size of the
core bacterial microbiota of pollen. Considering the high
level of specificity encountered, we suppose that adding
other plant species would further reduce the size of the
core-microbiome, although this remains to be verified by
Hitherto, there are only a few examples of habitat-
related core-microbiomes. Neotropical plant phyllospheres
consisted of a core bacterial biome of 32 bacterial species
(Kembel et al., 2014). A study of three different wild plants
showed more than ten common root-associated bacterial
families (Aleklett et al., 2015). Thus, the bacterial core-
microbiome of pollen appears to be smaller than that of
other plant parts. More research is needed to better
assess the habitat-related core-microbiomes next to the
single-species core-microbiomes, since same plant organs
play the same roles for the plants and might then be colo-
nized by functionally similar microbes.
Comparison of cultivation-dependent and -independent
The comparison of the bacterial families detected by isola-
tion and by high-throughput sequencing showed the typical
bias related to the culturability of most environmental
microbes: about 17% of all the families detected by the
cultivation-independent approach were also isolated. Only
the family Flavobacteriaceae was isolated (one isolate of
birch pollen) but not present in the total microbiota. The
reason might be that the shorter sequences delivered by
the IonTorrent method did not allow the annotation of the
corresponding OTU at family level, although even the
higher taxonomic ranges including this missing family
(order Flavobacteriales, class Flavobacteriia) were not
identified by the QIIME pipeline. Overall, we found a good
correspondence between isolated families and total fami-
lies for the analysed pollen species: as an example, the
family Enterobacteriaceae was one of the most abundant
by cultivation-independent approach for all the pollens
except rye; interestingly we got several isolates of Entero-
bacteriaceae, but only one isolate from rye (Supporting
Information Table S3). Further examples are the families
Bacillaceae and “Exiguobacteriaceae,” which were only
detected in two pollen species by cultivation-independent
approach and, accordingly, were only isolated from the
same pollen species (Supporting Information Table S3).
Both Burkholderiaceae and Enterococcaceae families
were abundant only in one pollen species (birch and rape
respectively), and the two only isolates found belonging to
these families came from the same pollen species (Sup-
porting Information Table S3). As a final example, the
family Microbacteriaceae occurred in all pollen species by
cultivation-independent analysis and was isolated from all
pollen species, too.
SEM and FISH-CLSM microscopy were used to investi-
gate the colonisation pattern of bacteria on pollen, in order to
gain insight into the pollen–bacteria interactions (Gamalero
et al., 2003; Moter and G€
obel, 2000). With both microscopic
methods morphologically and taxonomically different bacte-
ria were detected colonising the pollen surface (tectum).
Furthermore, the difference of the tectum structure between
pollen species was visualised (Fig. 6). Microscopic analysis
thus supported the results of both culture-dependent and -
independent methods. We hypothesise that the adhered
bacteria colonise, multiply and eventually form biofilms on
the sticky suitable habitat of pollen tectum. Bacterial adhe-
sion is the preliminary stage of colony/biofilm formation, and
SEM or confocal microscopy are key tools to evaluate this (El
Abed et al., 2012; Cardinale, 2014). Colldahl and Nilsson
(1973) observed with SEM that the surface of birch and timo-
thy pollens carried microorganisms-like particles. FISH-
CLSM analysis of the pumpkin anthosphere showed the
presence of Alphaproteobacteria,Gammaproteobacteria
and Firmicutes on pollen grains, densely colonised by micro-
urnkranz et al., 2012). Our microscopy observa-
tions clearly indicated that bacteria colonise ubiquitously the
pollen micro-habitat as single cells, colonies as well as
biofilm-like structures. The multilayered pollen walls might be
impermeable to FISH probes, which would hinder the detec-
tion of bacterial cells inside them; however we did not find
any internal bacterial cell even in the broken pollen grains
that we often found in our confocal microscopy observations.
The comparison with the electron microscopy images let us
argue that there might be the possibility that some bacterial
cells get detached and lost during the washing steps of
FISH: in fact, no washing step is performed for the SEM anal-
ysis, and more bacterial cells were eventually detected there
on the pollen grains.
Our results highlighted the complexity of the pollen micro-
biota and provided for the first time its comprehensive
characterization, including clear identification of cultivated
and molecularly detected bacteria and identification of drivers
of community structure and diversity. More research is need-
ed to understand the implications of our discoveries on plant
ecology and reproduction, which should focus on the func-
tion(s) of pollen-inhabiting microbes and on their horizontal
as well as vertical transmission. Plant microbiome is not a
single biological unit but a puzzle of complementary parts:
our study on pollen microbiota significantly contributed to
reveal one additional piece of this mosaic of microhabitats.
Birch (Betula pendula Roth., “B” when abbreviated), rape
(Brassica napus L., “R” when abbreviated), rye (Secale cereal
The bacterial microbiota of flower pollen 5169
L., “Ry” when abbreviated) and autumn crocus (Colchicum
autumnale L., “Au” when abbreviated) were the selected plant
species to collect the pollen samples. Birch and rye are wind-
pollinated species, while rape and autumn crocus are insect-
pollinated. Flowers of each plant species were collected in
sterile separate DNA-free sterile plastic tubes from four loca-
tions in a restricted geographical region (within Giessen and in
the 50 km surrounding, Supporting Information Table S4) dur-
ing the flowering season (B1 – 2015-04-28, B2 and B3 –
2015-05-06 R1 and R2 – 2015-05-27, R3 – 2015-01-06; Ry1,
Ry2 and Ry3 – 2015-06-10; Au1 – 2015-09-07, Au2 – 2015-
09-11 and Au3 – 2015-09-14). For each species, three biologi-
cal samples (for birch, rape and rye about 200–1000 flowers
from the same plant or from closely growing plants, for autumn
crocus about 30 flowers each) were collected. Anthers were
immediately dissected sterile from the flowers into separate
plastic tubes, kept in silica desiccator for one day to dry out.
Then pollen grains were collected and extracted by shaking
the tubes gently with a vortex mixer for 20–30 sec. A subsam-
ple was immediately frozen at 2208C for DNA extraction and
other two subsamples were fixed for fluorescence in-situ
hybridisation (FISH; see below for details). The remaining
sample was used for bacterial isolation on agar medium
Cultivation-dependent analysis of bacterial biome
Isolation of bacteria from pollen. The pollen samples
(Birch: 23, 32 and 36 mg; Rape: 28, 33 and 89 mg; Rye: 20,
39 and 74 mg; Autumn crocus: 100, 110 and 250 mg) were
mixed each with 5 ml of shaking solution (0.05% Tween 80
and 0.18% Na
; Musovic et al., 2006) and agitated for 30
min. This was followed by serial dilutions with 0.02% Tween
80 10.085% NaCl to a dilution of 10
. About 100 mlofeach
dilution was plated in triplicate onto 1:10 diluted AC agar medi-
um (Sigma-Aldrich Chemie GmbH, Steinheim, Germany) and
pollen medium; the latter consisted of minimal salt medium
(Widdel and Bak, 1992) amended with commercial flower pol-
uten pollen, Bergland Pharma GmbH, Heimertingen,
Germany) and 10 ml of pollen extract obtained from the same
pollen sample used for isolation (10 g of autoclaved flowers in
500 ml sterile water for 30 min and filter-sterilized) (Supporting
Information Table S5). Cycloheximide (200 mg l
) was added
to the media to avoid growth of fungi. The plates were incubat-
ed aerobically for five days at 258C. Colony counts were
determined for both AC agar and pollen medium. The number
of colonies of the three replicates of each dilution were aver-
aged, and the total colony forming units (CFUs) per gram of
sample were calculated and compared between pollen spe-
cies by the Kruskal–Wallis test (Kruskal and Wallis, 1952).
Morphologically different colonies were selected and sub-
cultured from single colonies for pure culture isolation. Each
culture was grown in liquid AC medium to confirm purity by
microscopic observation of the cellular morphology. For con-
servation of the pure culture it was grown in liquid AC (1:10)
medium until an optical density of 0.8 to 1.2 at 600 nm was
reached, and a 20% glycerol stab was stored at 2808C.
Identification of the bacterial isolate by 16S rRNA gene
sequencing. Genomic DNA was isolated from the pure cul-
tures with the NucleoSpin DNA isolation kit (MACHEREY-
NAGEL GmbH & Co. KG, D€
uren, Germany). The isolated
genomic DNA was used as template for the polymerase chain
reaction (PCR) using a MyCycler
Germany) to amplify the 16S rRNA gene with primers EUB9F
(50-GAGTTTGATCMTGGCTCAG-30) and EUB1492R (50-
ACGGYTACCTTGTTACGACTT-30) (Lane, 1991). The PCR
products were cleaned using the QIAquick PCR purification kit
(QIAGEN GmbH, Hilden, Germany) and sequenced by LGC
genomics (Berlin, Germany). The high-quality region of the
sequences was manually selected by visualising the electro-
pherograms with the software MEGA version 6.0 (Tamura
et al., 2011) and then compared with the reference sequences
by BLAST (Zhang et al., 2000) and Ez Taxon (Kim et al.,
2012) alignment. Final figures were assembled with Photo-
shop CS6 (Adobe Systems Inc, San Jose). Venn diagram was
drawn by Venny 2.1 (http://bioinfogp.cnb.csic.es/tools/venny/).
The 16S rRNA gene sequences obtained were submitted to
NCBI gene library with the following accession numbers:
Cultivation-independent analysis of bacterial microbiota
Isolation of total DNA from pollen. Metagenomic DNA
from different pollen species was obtained by a modified DNA
extraction procedure (Burgmann et al., 2001). For each pollen
sample, 1 ml of extraction buffer (2.5 g l
SDS, 0.2 M sodium
phosphate buffer (pH 8), 50 mM EDTA and 0.1 M NaCl, pH 8)
was added to the reaction tube containing 40–500 mg of pol-
len grains and 100 mg sterile zirconium beads were added
into the same tube. Cells were then disrupted for 2 min at
maximum speed in a cell mill MM200 (Retsch, Haan, Germa-
ny). Then the samples were centrifuged at 48Cand10,0003
gfor 6 min in a microcentrifuge (Heraeus Fresco, Thermo
Fisher Scientific Inc., Waltham). The supernatant was trans-
ferred into a new 2 ml microcentrifuge tube (Laborhaus
Scheller GmbH & Co KG, Euerbach, Germany).
The supernatant was incubated with 10 mlRNaseA(10mg
)at378C for 30 min, then 850 ml of phenol/chloroform/iso-
amyl alcohol (25:24:1) were added and mixed with gentle
swirling. The tubes were then centrifuged again at 10,000 3g
for 5 min at 48C, the aqueous phase was collected in a new
sterile tube and chloroform/isoamyl alcohol (24:1) was added,
mixed by gentle swirling and centrifuged 10,000 3gfor 5 min
at 48C. Aqueous phase were transferred into new tube, DNA
was precipitated with 1 ml of precipitation buffer [20% Poly
(ethylene glycol) and 2.5 M NaCl] at 48C for 1 h and centri-
fuged again at 10,000 3gfor 5 min at 48C. The DNA was
washed with ice cold 75% ethanol, dried out, dissolved in
nuclease free water. An aliquot of the DNA was visualized in
1% agarose gel stained with gel red.
Ion torrent sequencing. The partial sequence of the hyper-
variable regions (V4&V5) of the 16S rRNA gene of each DNA
samples was PCR amplified using the primer 520 F (50-
AYTGGGYDTAAAGNG-30) (Claesson et al., 2009) and 907 R
comp (50-CCGTCAATTCMTTTRAGTTT-30) (Engelbrektson
et al., 2010). PCR was carried out at a final volume of 15 ml
contain 10 ng of genomic DNA, 3 mlof53KAPAHiFi (KAPA
Biosystems, Wodurn) buffer, KAPA dNTP mix 200 mMeach,
primer 5 pmol each and KAPAHiFi polymerase 0.3 units. The
PCR was performed with MyCycler
(Bio-Rad) for 25 cycles
5170 B. Ambika Manirajan et al.
with the initial denaturation for 3 min at 958C, cyclic denatur-
ation for 20 sec at 988C, annealing for 30 sec at 558Cand
extension for 30 sec at 728C with final extension for 5 min. The
amplified PCR product was confirmed with agarose gel elec-
trophoresis and this DNA fragments used as the template of
second PCR reaction. The second PCR was done with primer
520 F and 907 R comp, adapter (50-CCATCTCATCCCT
GCGTGTCTCCGACTCAG-30) and barcodes (Supporting
Information Table S6) with KAPA2G Robust HotStart Ready-
Mix (KAPA Biosystems). The final volume of 50 mlcontain2ml
of template, 10 mlof53KAPAHiFi buffer, KAPA dNTP mix 600
mM, primer 5 pmol and KAPAHiFi polymerase 1 unit. The PCR
was performed with MyCycler
(Bio-Rad) for 8 cycles with
the initial denaturation for 3 min at 958C, cyclic denaturation for
20 sec at 988C, annealing for 30 sec at 558C and extension for
30 sec at 728C with final extension for 7 min. The PCR prod-
ucts were purified using QIAquick PCR purification kit
(QIAGEN GmbH) and further purified with DNA purification
NGS clean-up kit (MACHEREY-NAGEL
GmbH & Co. KG) to remove primer-dimers. Concentration of
the amplified DNA was estimated with Qubit dsDNA HS assay
kit by Qubit
3.0 fluorometer (Life Technologies, Carlsbad).
Concentration of each sample was adjusted to 1 mM, the sam-
ples were pooled to the final concentration of 26 pM and used
for emulsion PCR with Ion PGM Template OT2 400 kit (Life
Technologies) by using Ion OneTouch 2 (Life Technologies).
Further the enrichment was done with Ion PGM Hi-Q OT2
(Life Technologies) kit by using Ion OneTouch ES (Life Tech-
nologies). After the enrichment, samples were loaded on a
314 chip for posterior sequencing (Ion PGM Hi-Q Sequencing
kit & Ion 314v2 chip) (Life Technologies) by using Ion PGM
Sequencer (Life Technologies).
PCR and sequencing was repeated two times (with the
same metagenomic DNAs) and the sequences were merged
in a unique dataset after checking that both delivered very
similar outputs in terms of taxonomical composition of the
Sequences were submitted to EMBL under the project num-
bers PRJEB14477 and PRJEB14487.
Analysis of ion torrent sequences. The fastq file from Ion
Torrent was analysed using QIIME, version 1.9 (Caporaso
et al., 2010a, b). The sequences were quality (threshold: 20)
and length (350–450 bp) filtered, maximum homopolymer
run <6. The sequences were dereplicated and assigned to
specific samples by corresponding barcodes. Operational tax-
onomic units were generated with a sequence similarity
97% using the uclust method (Edgar, 2010). Chimera slayer
(Haas et al., 2011) was used to remove chimeric OTUs previ-
ously aligned to the Greengenes core reference alignment
(DeSantis et al., 2006) by pyNAST method (Caporaso et al.,
2010a, b). Taxonomy was assigned on the Greengenes data-
base (McDonald et al., 2012). Profile clustering networks was
constructed using Cytoscape, version 3.3.0 (Shannon et al.,
2003) based on node and edge table created from QIIME with
the script make_otu_network.py.Alphadiversityindices
(Shannon-Wiener, Shannon-equitability, Dominance, Simp-
son, Simpson-reciprocal, PD-whole tree, Chao1 and observed
species) were statistically compared between pollen species
by ANOVA and Tukey HSD post-hoc test, using the software
Statistica (Statsoft Inc., Tulsa). For Beta diversity analysis,
jackknifed-weighted Unifrac distances (Lozupone and Knight,
2005) were used to assess the similarity of the community
structure between pollen samples, and were visualized by
Principal Component Analysis (PCoA)-plot (Jost, 2007; Milani
et al., 2013; Seedorf et al., 2014; Schneider et al., 2015) using
Emperor (Vazquez-Baeza et al., 2013). Statistical significance
of the factor “species,” “pollination type” and “sampling site”
was tested by permutational multivariate analysis of variance,
as implemented in QIIME (script compare.categories.py,
method “Adonis,” 999 permutations). Final figures were
assembled with Photoshop CS6 (Adobe Systems Inc, San
Scanning electron microscopy (SEM)
Freshly collected pollen samples (10–100 mg) were dried out
in silica desiccator for 24 hours and stored at 2808Cuntil
scanning electron microscopy examination. The tip of a small
laboratory spoon of each pollen sample was hydrated 1 h at
room temperature, then mounted on metal stubs and then
coated with a thin layer of gold. They were observed with
scanning electron microscope XL30 (Philips, Amsterdam, The
Fluorescent in-situ hybridisation and confocal laser
scanning microscopy (FISH-CLSM)
For FISH staining of Gram-negative bacteria, one pooled sam-
ple per pollen species consisting of 10–100 mg of pollen
grains was fixed with a 3:1 mixture of 4% paraformaldehyde
and 13PBS, by incubation at 48C for 6 hours. The samples
were then washed three times with 13PBS for increasing
times, and then stored in 250–750 ml of 99.8% ethanol:1X
PBS (1:1) at 2208C. Additionally, one pooled sample per pol-
len species was also fixed directly in 99.8% ethanol: 13PBS
(1:1) and stored at 2208C, for FISH staining of Gram-positive
bacteria. For FISH staining, 25 ml of each sample were dried
onto a poly-L-lysine-coated slide and treated with 1 mg ml
lysozyme for 10 min at room temperature, in order to increase
the permeability of the cell walls to the FISH probes. Then, the
slides were dipped stepwise into 50%, 80% and 96% ethanol,
respectively, for 3 min each. About 15 ml of hybridisation buffer
(180 ml5MNaCl,20ml 1 M Tris HCl, 5 ml2%SDS,300ml
formamide and 495 mlH
O) containing 2.5 ng ml
Cy3-labeled EUB338MIX FISH probe and one of the Cy5-
labeled specific probes (either LGC 354-MIX for Firmicutes or
HGC236 for Actinobacteria; Table 2) were added to each sam-
ple and incubated at 458C for 2 hours in the dark in a humid
chamber saturated with hybridisation buffer. After removing
the hybridisation buffer and washing with pre-warmed (468C)
washing buffer (1.02 ml5MNaCl,1ml 1 M Tris HCl, 0.5 ml
0.5 M EDTA and 47.48 mlddH
O), the slides were dipped for 5
sec into ice-cold water, dried out with soft compressed air,
immediately mounted with antifade reagent and finally covered
with a coverslip and sealed with transparent nail polish. The
occurrence of false positive signals derived from a specific
adhesion of FISH probes or fluorochromes to pollen structures
was checked by staining a sample of each pollen species with
ATTO488-, Cy3- and Cy5-labeled NONEUB probes (non-
The bacterial microbiota of flower pollen 5171
sense probes with sequence complementary to that of
FISH-stained pollen samples were observed with a confocal
laser scanning system Leica SP8 (Leica Microsystems
GmbH, Mannheim, Germany), equipped with a microscope
DM 6000CS and solid state Ar and HeNe lasers. Cy3 signal
emitted by the universal bacterial probe EUB338MIX was
excited with the 561 nm laser light and detected in the range
563–610, while Cy5 signal emitted by the phylum-specific
probes was excited with the 633 nm laser light and detected in
the range 645–690. Confocal stacks were acquired with a Z-
step of 0.5–0.8 mm using the objective Leica HC PL APO CS2
633/1.20 W. Volume-rendering and three-dimensional models
of the confocal stacks were created with the software Imaris
8.2 (Bitplane AG, Z€
urich, Switzerland). Final figures were
assembled with Photoshop CS6.
We are very grateful to Bellinda Schneider for her excellent
technical support with the Ion Torrent sequencing. We appre-
ciate and acknowledge Karl-Heinz Kogel, Institute of Phytopa-
thology, JLU-Giessen, for allowing M.C. to use the confocal
microscope. We are thankful to Ulrich G€
artner Department of
Anatomy and Cell Biology, JLU-Giessen, for allowing us to use
the scanning electron microscope, and Anika Seipp for her
technical support with scanning electron microscopy.
Aizenberg-Gershtein, Y., Izhaki, I., and Halpern, M. (2013) Do
honeybees shape the bacterial community composition in
floral nectar? Plos One 8: e67556.
Aleklett, K., Hart, M., and Shade, A. (2014) The microbial
ecology of flowers: An emerging frontier in phyllosphere
research. Botany 92: 253–266.
Aleklett, K., Leff, J.W., Fierer, N., and Hart, M. (2015) Wild
plant species growing closely connected in a sualpine
meadow host distinct root-associated bacterial communi-
ties. PeerJ 3: e804.
erez, S., Herrera, C.M., and de Vega, C. (2012)
Zooming-in on floral nectar: A first exploration of nectar-
associated bacteria in wild plant communities. FEMS Micro-
biol Ecol 80: 591–602.
Amann, R.I., Binder, B.J., Olson, R.J., Chisholm, S.W.,
Devereux, R., and Stahl, D.A. (1990) Combination of 16S
rRNA-targeted oligonucleotide probes with flow cytometry
for analyzing mixed microbial populations. Appl Environ
Microbiol 56: 1919–1925.
Ariizumi, T., and Toriyama, K. (2011) Genetic regulation of
sporopollenin synthesis and pollen exine development.
Annu Rev Plant Biol 62: 437–460.
Basim, E., Basim, H., and €
Ozcan, M. (2006) Antibacterial
activities of Turkish pollen and propolis extracts against
plant bacterial pathogens. J Food Process Eng 77:
Berg, G., Grube, M., Schloter, M., and Smalla, K. (2014)
Unraveling the plant microbiome: Looking back and future
perspectives. Front M icrobiol 7: 148.
Blackmore, S., Wortly, A.H., Skvarla, J.J., and Rowley, J.R.
(2007) Pollen wall development in flowering plants. New
Phytol 2007: 483–498.
Burgmann, H., Pesaro, M., Widmer, F., and Zeyer, J. (2001) A
strategy for optimizing quality and quantity of DNA extracted
from soil. J Microbiol Methods 45: 7–20.
Caporaso, J.G., Bittinger, K., Bushman, F.D., DeSantis, T.Z.,
Andersen, G.L., and Knight, R. (2010a) PyNAST: A flexible
tool for aligning sequences to a template alignment. Bioin-
formatics 26: 266–267.
Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K.,
Bushman, F.D., Costello, E.K., et al. (2010b) QIIME allows
analysis of high-throughput community sequencing data.
Nat Methods 7: 335–336.
Cardinale, M. (2014) Scanning a microhabitat: Plant-microbe
interactions revealed by confocal laser microscopy. Front
Microbiol 5: 94.
Carinanos, P., and Casares-Porcel, M. (2011) Urban green
zones and related pollen allergy: A review. Some guidelines
for designing spaces with low allergy impact. Landscape
Urban Plan 101: 205–214.
Claesson, M.J., O’sullivan, O., Wang, Q., Nikkil€
Marchesi, J.R., Smidt, H., et al. (2009) Comparative analy-
sis of pyrosequencing and a phylogenetic microarray for
exploring microbial community structures in the human dis-
tal intestine. Plos One 4: e6669.
Clark, C.J., and Lintas, C. (1992) Chemical composition of
pollen from kiwifruit vines. NZJCropHort20: 337–344.
Colldahl, H., and Carlsson, G. (1968) Allergens in pollen. Acta
Allergol 28: 387–397.
Colldahl, H., and Nilsson, L. (1973) Possible relationship
between some allergens (pollens, mites) and certain micro-
organisms (bacteria and fungi). Acta Allergol 28: 283–295.
Daims, H., Br€
uhl, A., Amann, R., Schleifer, K.H., and Wagner,
M. (1999) The domain-specific probe EUB338 is insufficient
for the detection of all Bacteria: Development and
Table 2. FISH probes used in this work.
Probe name Sequence (5’–3’) Fluorochrome Target
(at 428C) Reference
EUB338* GCTGCCTCCCGTAGGAGT Cy3 Most bacteria 15 Amann et al. (1990)
EUB338II* GCAGCCACCCGTAGGTGT Cy3 Planctomycetales 15 Daims et al. (1999)
EUB338III* GCTGCCACCCGTAGGTGT Cy3 Verrucomicrobiales 15 Daims et al. (1999)
NONEUB** ACTCCTACGGGAGGCAGC Cy3 / 15 Wallner et al. (1993)
NONEUB** ACTCCTACGGGAGGCAGC Cy5 / 15 Wallner et al. (1993)
HGC236 AACAAGCTGATAGGCCGC Cy5 Actinobacteria 15 Erhart et al. (1997)
* and ** used mixed in equimolar concentration.
5172 B. Ambika Manirajan et al.
evaluation of a more comprehensive probe set. Syst Appl
Microbiol 22: 434–2444.
DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie,
E.L., Keller, K., et al. (2006) Greengenes, a chimera-
checked 16S rRNA gene database and workbench compat-
ible with ARB. Appl Environ Microb 72: 5069–5072.
Edgar, R.C. (2010) Search and clustering orders of magnitude
faster than BLAST. Bioinformatics 26: 2460–2461.
Edlund, A.F., Swanson, R., and Preuss, D. (2004) Pollen and
stigma structure and function: The role of diversity in polli-
nation. Plant Cell 16: S84–S97.
El Abed, S., Hamadi, F., Latrache, H., and Ibnsouda, S.K.
(2012) Scanning electron microscopy (SEM) and environ-
mental SEM: Suitable tools for study of adhesion stage and
biofilm formation. In: Scanning Electron Microscopy.
Kazmiruk, V. (ed). New York: INTECH Open Access Publisher.
Engelbrektson, A., Kunin, V., Wrighton, K.C., Zvenigorodsky,
N., Chen, F., Ochman, H., et al. (2010) Experimental factors
affecting PCR-based estimates of microbial species rich-
ness and evenness. ISME J 4: 642–647.
Erhart, R., Bradford, D., Seviour, R.J., Amann, R., and
Blackall, L.L. (1997) Development and use of fluorescent in
situ hybridization probes for the detection and identification
of Microthrix parvicella in activated sludge. Syst Appl Micro-
biol 20: 310–318.
Fridman, S., Izhaki, I., Gerchman, Y., and Halpern, M. (2012)
Bacterial communities in floral nectar. Environ Microbiol
Rep 4: 97–104.
urnkranz, M., Lukesch, B., M€
uller, H., Huss, H., Grube, M.,
and Berg, G. (2012) Microbial diversity inside Pumpkin:
Microhabitat-specific communities display a high antagonistic
potential against phytopathogens. Microb Ecol 63: 418–428.
Gamalero, E., Lingua, G., Berta, G., and Lemanceau, P.
(2003) Methods for studying root colonization by introduced
beneficial bacteria. Agronomie 23: 407–418.
Gilliam, M. (1979) Microbiology of pollen and bee bread: The
yeast. Apidologie 10: 43–53.
Glushakova, A.M., and Chernov, I.Y. (2007) Seasonal dynam-
ic of the numbers of epiphytic yeasts. Microbiologiya 76:
Haas, B.J., Gevers, D., Earl, A.M., Feldgarden, M., Ward,
D.V., Giannoukos, G., et al. (2011) Chimeric 16S rRNA
sequence formation and detection in Sanger and 454-
pyrosequenced PCR amplicons. Genome Res 21: 494–
Halpern, M., Fridman, S., Atamna-Ismaeel, N., and Izhaki, I.
(2013) Rosenbergiella nectarea gen. nov., sp. nov., in the
family Enterobacteriaceae, isolated from floral nectar. Int J
Syst Evol Microbiol 63: 4259–4265.
Hani, B., Dalila, B., Saliha, D., Daoud, H., Mouloud, G., and
Seddik, K. (2012) Microbiological sanitary aspects of pollen.
Adv Environ Biol 6: 1415–1420.
Human, H., and Nicolson, S.W. (2006) Nutritional content of
fresh, bee-collected and stored pollen of Aloe greatheadii
var. davyana (Asphodelaceae). Phytochemistry 67: 1486–
Jojima, Y., Mihara, Y., Suzuki, S., Yokozeki, K.,
Yamanaka, S., and Fudou, R. (2004) Saccharibacter flo-
ricola gen. nov., sp. nov., a novel osmophilic acetic acid
bacterium isolated from pollen. Int J Syst Evol Microbiol
Jost, L. (2007) Partitioning diversity into independent alpha
and beta components. Ecology 88: 2427–2439.
Junker, R.R., and Keller, A. (2015) Microhabitat heterogeneity
across leaves and flower organs promotes bacterial diversi-
ty. FEMS Microbiol Ecol 91: fiv097.
Kembel, S.W., O’conner, T.K., Arnold, H.K., Hubbell, S.P.,
Wright, S.J., and Green, J.L. (2014) Relationship between
phyllosphere bacterial communities and plant functional
traits in a neotropical forest. pnas 111: 13715–13720.
Kim, O.S., Cho, Y.J., Lee, K., Yoon, S.H., Kim, M., Na, H.,
et al. J., (2012) Introducing EzTaxon-e: A prokaryotic 16S
rRNA gene sequence database with phylotypes that repre-
sent uncultured species. Int J Syst Evol Microbiol 62:
Kosenko, V.N. (1999) Contributions to the pollen morphology
and taxonomy of the Liliaceae.Grana 38: 20–30.
Kruskal, W.H., and Wallis, W.A. (1952) Use of ranks in one-
criterion variance analysis. JAmStatAssoc47: 583–621.
Lane, D.J. (1991) 16S/23S rRNA sequencing. In Nucleic Acid
Techniques in Bacterial Systematics. Stackebrandt, E., and
Goodfellow, M. (eds). New York: John Wiley and Sons Inc,
erez, S., De Vega, S., Van Assche, A.,
Johnson, S.D., Willems, K.A., et al.(2014)Rosenbergiella
australoborealis sp. nov., Rosenbergiella collisarenosi sp.
nov. and Rosenbergiella epipactidis sp. nov., three novel
bacterial species isolated from floral nectar. Syst Appl
Microbiol 37: 402–411.
Lindow, S.E., and Brandl, M.T. (2003) Microbiology of the
phyllosphere. Appl Environ Microb 69: 1875–1883.
Lozupone, C., and Knight, R. (2005) UniFrac: A new phyloge-
netic method for comparing microbial communities. Appl
Environ Microbiol 71: 8228–8235.
McDonald, D., Price, M.N., Goodrich, J., Nawrocki, E.P.,
DeSantis, T.Z., Probst, A.P., et al. (2012) An improved Green-
genes taxonomy with explicit ranks for ecological and evolu-
tionary analyses of bacteria and archaea. Isme J 6: 610–618.
Meyer, K.M., and Leveau, J.H.J. (2012) Microbiology of the
phyllosphere: A playground for testing ecological concepts.
Oecologia 168: 621–629.
Milani, C., Hevia, A., Foroni, E., Duranti, S., Turroni, F., Lugli,
P. , et al. (2013) Assessing the fecal microbiota: An opti-
mized ion torrent 16S rRNA gene-based analysis protocol.
Plos One 8: e68739.
Mortazavi, R., Attiya, S., and Ariya, P.A. (2015) Arctic microbial
and next-generation sequencing approach for bacteria in
snow and frost flowers: Selected identification, abundance
and freezing nucleation. Atmos Chem Phys 15: 6183–6204.
Moter, A., and G€
obel, U.B. (2000) Fluorescence in situ hybridi-
zation (FISH) for direct visualization of microorganisms.
J Microbiol Meth 41: 85–112.
Musovic, S., Oregaard, G., Kroer, N., and Sorensen, S.J.
(2006) Cultivation-independent examination of horizontal
transfer and host range of an IncP-1 plasmid among gram-
positive and gram-negative bacteria indigenous to the bar-
ley rhizosphere. Appl Environ Microb 72: 6687–6692.
Nicolson, S.W., and Human, H. (2013) Chemical composition
of the ‘low quality’ pollen of sunflower (Helianthus annuus,
Asteraceae). Apidologie 44: 144–152.
Obersteiner, A., Gilles, S., Frank, U., Beck, I., H€
aring, F., Erns,
D., et al. (2016) Pollen-associated microbiome correlates
The bacterial microbiota of flower pollen 5173
with pollution parameters and the allergenicity of pollen.
Plos One 11: e0149545.
Oldenburg, M., Petersen, A., and Baur, X. (2011) Maize pollen
is an important allergen in occupationally exposed workers.
J Occup Med Toxicol 6: 32.
Redford, A.J., Bowers, R.M., Knight, R., Linhart, Y., and
Fierer, N. (2010) The ecology of the phyllosphere: Geo-
graphic and phylogenetic variability in the distribution of
bacteria on tree leaves. Environ Microbiol 12: 2885–2893.
Remus-Emsermann, M.N., Tecon, R., Kowalchuk, R., and
Leveau, J.H. (2012) Variation in local carrying capacity and
the individual fate of bacterial colonizers in the phyllo-
sphere. ISME J 6: 756–765.
Rintala, H., Pitk€
aranta, M., Toivola, M., Paulin, L., and
Nevalainen, A. (2008) Diversity and seasonal dynamics of bac-
terial community in indoor environment. BMC Microbiol 8: 56.
Ryan, R.P., Germaine, K., Franks, A., Ryan, D.J., and
Dowling, D.N. (2008) Bacterial endophytes: Recent develop-
ments and applications. FEMS Microbiol Lett 278: 1–9.
Schneider, D., Engelhaupt, M., Allen, K., Kurniawan, S.,
Krashevska, V., Heinemann, M., et al. (2015) Impact of low-
land rainforest transformation on diversity and composition
of soil prokaryotic communities in Sumatra (Indonesia).
Front Microbiol 6: 1339.
Seedorf, H., Kittelmann, S., Henderson, G., and Janssen, P.H.
(2014) RIM-DB: A taxonomic framework for community
structure analysis of methanogenic archaea from the rumen
and other intestinal environments. Peer J 2: e494.
Shannon, P., Markie, A., Ozier, O., Baliqa, N.S., Wang, J.T.,
Ramage, D., et al. (2003) Cytoscape: A software environ-
ment for integrated models of biomolecular interaction net-
works. Genome Res 13: 2498–2504.
Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M.,
and Kumar, S. (2011) MEGA5: Molecular evolutionary
genetics analysis using maximum likelihood, evolutionary
distance, and maximum parsimony methods. Mol Biol Evol
Turner, T.R., James, E.K., and Poole, P.S. (2013) The plant
microbiome. Genome Biol 14: 10–1186.
Ushio, M., Yamasaki, E., Takasu, H., Nagano, A.J., Fujinaga,
S., Honjo, M.N., et al. (2015) Microbial communities on
flower surfaces act as signatures of pollinator visitation. Sci
Rep-UK 5: 8695.
Vandenkoornhuyse, P., Quaiser, A., Duhamel, M., Le Van, A.,
and Dufresne, A. (2015) The importance of the microbiome
of the plant holobiont. New Phytol 206: 1196–1206.
Vanneste, J.L., Giovanardi, D., Yu, J., Cornish, A.D., Kay, C.,
Spinelli, F., et al.(2011)DetectionofPseudomonas syrin-
gae pv. Actinidiae in kiwifruit pollen samples. N Z Plant Prot
Vazquez-Baeza, Y., Pirrung, M., Gonzalez, A., and Knight, R.
(2013) Emperor: A tool for visualizing high-throughput
microbial community data. Gigascience 2: 16.
Wallner, G., Amann, R., and Beisker, W. (1993) Optimizing
fluorescent in situ hybridization with rRNA-targeted oligonu-
cleotide probes for flow cytometric identification of microor-
ganisms. Cytometry 14: 136–143.
Widdel, F., and Bak, F. (1992) Gram-negative mesophile
sulfate-reducing bacteria. In The Prokaryotes. Balows, A.,
uper, H.G., Dworkin, M., Harder, W., and Schleifer, K.-H.
(eds). New York: Springer, pp. 3352–3378.
Zhang, Z., Schwartz, S., Wagner, L., and Miller, W. (2000) A
greedy algorithm for aligning DNA sequences. JComput
Biol 7: 203–214.
Additional supporting information may be found in the online
version of this article at the publisher’s web-site
Additional Supporting Information may be found in the
online version of this article at the publisher’s web-site:
Table S1. Cultured bacteria associated with four different
flower pollen species, identified using the EzTaxon server
on basis of 16S rRNA gene sequence data.
Table S2. BLAST similarity search details of representative
sequence of biggest OTUs significantly different between
Table S3. Comparison of bacterial families detected by iso-
lation (identification by sequencing of the 16S rRNA gene
and EzTaxon database comparison) and by IonTorrent high-
throughput sequencing (identification by Qiime pipeline
Table S4. Details of the sampling sites were the pollen
samples were collected.
Table S5. Composition of pollen medium.
Tab le S6. Details of barcode sequences, linker primer sequen-
ces and reverse primer used for Ion Torrent sequencing.
Fig. S1. Flower samples used in this work for the pollen
collection: rape (A), rye (B), birch (C) and autumn
5174 B. Ambika Manirajan et al.