Annals of Botany 113: 191– 197, 2014
doi:10.1093/aob/mct257, available online at www.aob.oxfordjournals.org
TECHNICAL ARTICLE
Flow cytometric analysis of pollen grains collected from individual bees provides
information about pollen load composition and foraging behaviour
Paul Kron*, Allison Kwok and Brian C. Husband
Department of Integrative Biology, University of Guelph, Guelph, Ontario, N1G 2W1, Canada
* For correspondence. E-mail [email protected]
Received: 4 June 2013 Returned for revision: 9 August 2013 Accepted: 13 September 2013 Published electronically: 14 November 2013
† Background and Aims Understanding the species composition of pollen on pollinators has applications in agriculture, conservation and evolutionary biology. Current identification methods, including morphological analysis,
cannot always discriminate taxa at the species level. Recent advances in flow cytometry techniques for pollen
grains allow rapid testing of large numbers of pollen grains for DNA content, potentially providing improved
species resolution.
† Methods A test was made as to whether pollen loads from single bees (honey-bees and bumble-bees) could be classified into types based on DNA content, and whether good estimates of proportions of different types could be made.
An examination was also made of how readily DNA content can be used to identify specific pollen species.
† Key Results The method allowed DNA contents to be quickly found for between 250 and 9391 pollen grains
(750–28 173 nuclei) from individual honey-bees and between 81 and 11 512 pollen grains (243 –34 537 nuclei)
for bumble-bees. It was possible to identify a minimum number of pollen species on each bee and to assign
proportions of each pollen type (based on DNA content) present.
† Conclusions The information provided by this technique is promising but is affected by the complexity of the
pollination environment (i.e. number of flowering species present and extent of overlap in DNA content).
Nevertheless, it provides a new tool for examining pollinator behaviour and between-species or cytotype pollen
transfer, particularly when used in combination with other morphological, chemical or genetic techniques.
Key words: Flow cytometry, pollen, Apis mellifera, Bombus, pollen load composition, DNA content, bumble-bee,
honey-bee, foraging behaviour, pollination.
IN T RO DU C T IO N
The foraging patterns of pollinators are of considerable practical
and scientific interest, in no small part because of their critical
role in flowering plant reproduction. An understanding of pollinator behaviour has informed agricultural practices (e.g.
Delaplane and Mayer, 2000; Ramsay, 2005; Pasquet et al.,
2008) and conservation biology (e.g. Ellstrand, 1992;
Lopezaraiza-Mikel et al., 2007). Pollen movement is also a key
consideration in the study of plant evolutionary and population
processes, such as gene flow and genetic structuring (e.g.
Levin and Kerster, 1974; Slatkin, 1985), reproductive isolation
and speciation (e.g. Coyne and Orr, 2004; Kennedy et al.,
2006; Marques et al., 2012), and the coevolution of plants and
pollinators (e.g. Barrett and Harder, 1996; Morales and
Traveset, 2008).
Several methods have been used to study the complex processes of plant – pollinator interactions, from observation of pollinator foraging behaviour (e.g. Pasquet et al., 2008; Marques
et al., 2012) to genetic analyses of the resulting plant progeny
(e.g. Schaal, 1980). The direct analysis of pollen carried by pollinators ( pollen loads) is a technically challenging but important
step in relating pollinator behaviour to pollination outcomes,
because observed pollinator movements may not be strongly correlated with the composition of pollen loads (Alarcón, 2010), and
because the full complement of pollen on the pollinator may not
correspond to what is deposited on stigmas (Harder and Barrett,
1996; Morales and Traveset, 2008; Alarcón, 2010). The primary
method for studying pollen loads on bees has been through morphological identification of pollen grains collected from bees,
hives or honey (Wodehouse, 1959; Sharma, 1970; Von Der
Ohe et al., 2004; Barth et al., 2011). Less commonly, methods
such as chemical analysis have been used (Ruoff, 2006), as
well as genetic marker analysis of pollen collections (Zhou
et al., 2007; Shoemaker, 2010). Recent improvements to the
study of pollen DNA content using flow cytometry (Kron and
Husband, 2012) raise the possibility of complementing such
methods with the use of DNA content as a species marker.
While DNA content can be a crude marker of species identity,
the level of resolution may be similar to that of morphology
and genetic markers, and, because flow cytometry can be used
to test large numbers of pollen grains quickly, it may serve as
an efficient tool for estimating the proportions of different
pollen types.
In this study, we ask two questions. First, can flow cytometry
be used to identify and count discrete classes of nuclei from
pollen carried on individual bees with sufficient resolution and
numbers to estimate the proportions of each type present?
Secondly, with what level of success can DNA content be used
to characterize the composition of pollen from a single bee: specifically, the number of pollen species present, and the identity of
these species. We asked these questions using both honey-bees
(Apis mellifera) and bumble-bees (Bombus species) in a relatively
complex pollination environment (rich in flowering species with
# The Author 2013. Published by Oxford University Press on behalf of the Annals of Botany Company. All rights reserved.
For Permissions, please email: [email protected]
192
Kron et al. — Flow cytometric analysis of individual bee pollen loads
overlapping genome sizes), and using honey-bees in a much
simpler pollination environment (low number of flowering
species, with ploidy-level differences in DNA content).
MAT E RI ALS A ND METH O DS
Sampling
We collected bees foraging on three plant genera (Monarda,
Centaurea and Solidago) at two locations. Monarda fistulosa
and a mixed patch of C. jacea and C. nigra were located in a
small (150 × 50 m) prairie restoration site located near Arkell,
Ontario, Canada (Arkell site, 43 832′ 44′′ N, 80 809′ 22′′ W) that
contained a diverse array of flowering plants (Supplementary
Data Table S1). A mixed population of two heteroploid
Solidago species was located in a 100 × 60 m old field on the
University of Guelph campus, Guelph, Ontario (Guelph site,
43 831′ 35′′ N, 80 813′ 57′′ W). We initially identified the
Solidago species based on morphology (Semple et al., 1999)
as members of the S. canadensis (diploid, 2x), S. gigantea (tetraploid, 4x) and S. altissima (hexaploid, 6x) complex, and subsequently confirmed them as S. canadensis and S. altissima
based on the relative DNA contents of their 2C leaf nuclei. At
the time of sampling (late August), this population contained
only two other flowering species in low abundance, Euthamia
graminifolia and Erigeron strigosus.
At the Arkell site, we collected the dominant pollinator on
each plant species: bumble-bees on M. fistulosa and honey-bees
on Centaurea. All bumble-bees collected on M. fistulosa were
keyed out to the genus Bombus, although there may have been
some species variation. We collected the first ten bees of each
type that we could capture, with the only criterion being that
they were perched on the flowers. We did not distinguish
between Centaurea species while collecting bees. At the
Guelph site, we haphazardly collected 12 honey-bees without
distinguishing between the Solidago species on which they
were foraging, and sampled a leaf from each bee-capture plant
for later cytotype testing. At both sites, bees were captured in
plastic bottles, immediately placed in a cooler with ice packs,
transferred within 1 h to a – 20 8C freezer and kept at – 20 8C
until testing (20 min to 2 h). We collected additional honey-bees
from Solidago plants at a later date and tested 13 randomly
selected bees after 7 months storage at – 20 8C to assess the
effects of storage. At the time of bee collection, we also collected
leaves and flowers from all species flowering in the two sites for
DNA content estimation.
Flow cytometry
Pollen grains from bees were suspended in LB01 buffer
(Doležel et al., 1989) by adding 1.4 mL of buffer to the capture
bottle, vortexing at high speed for 10– 15 s, collecting the
buffer and repeating once. The pollen nuclei were then extracted
and stained using the filter-bursting method described in Kron
and Husband (2012). Briefly, the buffer was passed through
two filters: a large pore pre-filter to remove contaminants and a
small pore filter to collect the pollen grains. We used 50 mm
and 10 mm filters for the Solidago bees and 100 mm and
20 mm filters for the Monarda and Centaurea bees. Pollen
grains were rubbed against the collection filter with a plastic
rod to burst them, and the released nuclei were washed through
the filter into a clean tube with 2 × 0.25 mL of LB01 buffer
with 100 mg mL21 propidium iodide and 50 mg mL21 RNase.
Samples from Centaurea and Monarda bees were stained
for 20– 27 min and then run at medium speed on a BD
FACSCalibur flow cytometer (BD Biosciences, San Jose, CA,
USA) for 7 min (near exhaustion of sample). Solidago samples
were run at low speed until the combined peaks of nuclei totalled
15 000– 16 000 events or until the sample was exhausted
(5 – 26 min). Particle fluorescence was measured with the 585/
42 nm (FL2) detector on a linear scale, and the relative fluorescence of peaks was determined using the fluorescence area.
Settings were chosen to detect nuclei in the 0.35– 35 pg DNA
range, but we briefly ran samples at a higher voltage to confirm
the absence of peaks below this range and also measured FL3
(670 nm) fluorescence on a log scale to detect higher fluorescence
peaks. We used the same method to test samples of pollen collected directly from flowers of Monarda and Centaurea, using
the same instrument settings as the bee-collected pollen, to establish external reference positions (fluorescence means) for the
nuclei of these species. For Solidago, we used leaf tissue as external standards, assuming that 1C pollen nuclei would have half the
fluorescence of 2C leaf nuclei, an assumption we verified with a
small number of pollen samples collected directly from flowers.
Quality assessment
The quality of fluorescence histograms was assessed using the
following measures: total number of nuclei (and corresponding
number of pollen), coefficient of variation (CV) of the peaks of
nuclei, and background aggregates and debris (BAD, the percentage of debris and aggregate events in the fluorescence
range of peaks of nuclei; Shankey et al., 1993). For each peak
of nuclei, we calculated its proportion ( p) of the total
p nuclei
(n) in a sample, as well as the standard error (s.e. ¼ [ p(1 –
p)/n] and relative standard error (RSE ¼ s.e./p × 100 %) for
each p. Numbers of nuclei (total and for each peak), CVs,
BADs and peak fluorescence means were measured using
ungated histograms and ModFit LT software (Verity Software
House, Inc., 2000) for the relatively simple Centaurea and
Solidago bee histograms. ModFit LT was used to find the BAD
of ungated histograms for the Bombus (Monarda) samples,
but, because of high debris levels, CVs were measured using
histograms gated to remove debris on a fluorescence area by sidescatter plot. CellQuest Pro 4.0 software (1996, BD Biosciences)
was used to measure the means and numbers of nuclei for the less
well-defined peaks observed in these Monarda samples. The
number of pollen grains corresponding to the total nuclei in a
sample depends on the proportions of these nuclei coming
from binucleate and trinucleate pollen species, and was calculated after attributing the individual peaks of nuclei to particular
plant species (see ‘Pollen identification’).
For comparison with an established pollen analysis method,
we determined how many of our samples met the pollen
number criteria applied in melissopalynology, where pollen
morphology is used to identify and quantify pollen types in
honey. Standard practice is to test a minimum of 300 pollen
grains when assigning pollen types to frequency classes
(e.g. ‘frequent’ or ‘rare’), and 500– 1000 when greater accuracy
is desired or when assigning specific percentages to pollen types
Kron et al. — Flow cytometric analysis of individual bee pollen loads
(Behm et al., 1996; Von Der Ohe et al., 2004). We therefore
determined the percentages of our samples that exceeded 300
and 1000 pollen grains, and confirmed that samples with 300+
pollen grains also had RSEs ,30 % for the proportions of different nuclei types (a commonly applied cut-off for data quality; e.g.
Klein et al., 2002). For comparison with a flow cytometry application that requires estimates of the proportions of rare events, we
determined how many of the histograms met the standards for
clinical cell cycle analysis (i.e. CV ,8 %, nuclei ≥10 000 and
BAD ,20 %; Shankey et al., 1993; Ormerod et al., 1998). We
also used t-tests to compare the peak CVs and histogram BADs
of the Solidago honey-bee samples stored for 7 months with
the original samples to test for a decline in sample quality.
Pollen identification
To identify the pollen source of the detected nuclei, we used
flow cytometry to create a library of relative DNA contents for
all species flowering in the populations at the time of bee collection, using M. fistulosa (Arkell) and S. canadensis (Guelph) as
internal reference standards. We co-chopped leaf tissue from
each flowering species with one of these internal standards,
using a razor blade, LB01 buffer and a 30 mm filter. We
assumed that the fluorescence ratio of somatic nuclei (test
plant 2C/standard 2C) would equal the ratio of their pollen 1C
values, but, for a sub-set of species, we also co-prepared pollen
nuclei of the two species and used these 1C/1C ratios preferentially for identification purposes. We used leaf tissue for most
species because pollen can be difficult to collect directly from
some plants in the field, for example when few flowering individuals are present, or when pollinators quickly remove available
pollen. We therefore tested the method using what would typically be the more readily available tissue (leaves), assuming that
this approach may often be a practical necessity for field
researchers. The two Centaurea species at Arkell had very
similar DNA contents relative to M. fistulosa, and the expected
ratios of all species to the Centaurea species were derived
from the M. fistulosa ratios using the following calculation:
(test species/Monarda) × (Monarda/Centaurea) ¼ test species/
Centaurea (Supplementary Data Table S1).
We identified a fluorescence peak on each bee’s histogram that
corresponded to the plant on which it was collected (the capture
plant) by comparing it with an external reference sample for the
capture species. For bees collected on Monarda and Centaurea,
the expected fluorescence mean of the capture plant nuclei was
the average of the peak means of nuclei of three pollen samples
collected directly from flowers. For Solidago, we tested a leaf
from each plant immediately before testing the pollen from the
bee collected thereon, to establish both the species and cytotype
of the plant, and the expected fluorescence of the capture plant’s
1C pollen nuclei (half that of the somatic tissue). We then calculated the ratio of the mean fluorescence of each peak of nuclei
with that of the capture plant peak, and compared these ratios
with the reference library of relative DNA contents for that
population. Each peak of nuclei therefore could be described as
belonging to a pollen ‘type’, defined by its relative DNA content
and identifiable as an individual species or set of species in the
library.
We applied the following rules in assigning candidate species
to peaks of nuclei. The relative DNA content of the candidate
193
species had to be within 10 % of that of the observed peak. We
had to consider whether the pollen of each species was trinucleate or binucleate, because trinucleate pollen grains contain three
nuclei with a single DNA content (1C), while binucleate pollen
grains typically contain a 1C vegetative nucleus and a 2C generative nucleus (Brewbaker, 1967; Bino et al., 1990; Suda et al.,
2007). If the candidate species was binucleate, there had to be
corresponding vegetative and generative peaks with neither containing ,30 % of the nuclei for the species. We also assumed that
each peak consisted of only one species, with the following exception for M. fistulosa bee histograms. Pollen from any of
three binucleate species had either vegetative or generative
peaks of nuclei potentially overlapping those of M. fistulosa
nuclei (Linaria vulgaris, Desmodium canadensis and Trifolium
pratense). It was assumed that some portion of the presumed
M. fistulosa peak could consist of nuclei from one of these
species as well, if a matching vegetative or generative peak of
nuclei was observed.
Large peaks of pollen nuclei are often accompanied by a small
peak at twice their fluorescence, potentially comprising unreduced gamete nuclei and/or doublets (two adhering nuclei).
Such peaks typically account for a small proportion of the
nuclei present, but require relatively complex analysis (doublet
discrimination; e.g. Wersto et al., 2001) to determine which
events should be counted once (unreduced gamete nuclei) vs.
twice (doublets). We therefore made two simplifying assumptions. First, we treated all nuclei in such peaks as unreduced
gamete nuclei rather than doublets, an approach that is conservative in terms of counts of nuclei. Secondly, when such peaks constituted ,2 % of all nuclei present, they were automatically
treated as peaks of unreduced gamete nuclei, their count of
nuclei was added to the corresponding reduced gamete peak
and they were not included in further analysis. Similar peaks
exceeding 2 % of all nuclei were treated as any other peak,
with species assignment including the unreduced gamete
option as well as other species. This approach simplifies the analysis at the expense of potentially overlooking some species
matching the unreduced gametes in DNA content but present
at low frequencies (,2 %). Unreduced gamete nuclei were not
considered as an option if such an interpretation would require
an unreduced gamete production rate exceeding 20 %.
If multiple peaks and/or candidate species per peak were
present, we considered all possible combinations of species, and
determined a range for the number of species present and proportion of pollen attributable to each species. The following summary
data were found for each bee: proportion of pollen matching
the DNA content of the capture plant species (minimum and
maximum across different scenarios), number of non-capture
species present (range across different scenarios) and the
average number of possible pollen species per peak of nuclei.
R E S ULT S
Quality of fluorescence histograms
Pollen nuclei (i.e. fluorescence peaks) were detected for all
samples (Fig. 1). Most honey-bee samples met the numerical criterion applied in morphological studies, with 91 % of samples
having nuclei from at least 300 pollen grains, and 86 % having
nuclei from at least 1000 (Supplementary Data Table S2).
194
Kron et al. — Flow cytometric analysis of individual bee pollen loads
Counts
A 1150
B 500
Sc-1x
C 900
Sa-3x
920
400
720
690
300
540
460
200
360
230
Sc-2x
100
Sa-3x Sa-6x
0
D 650
Sc-1x
180
Sa-6x
0
0
Cen
E
F 100
Mf
200
520
80
Counts
160
390
120
260
60
p
40
200 400 600 800 1000
Relative fluorescence
0
0
Mf
40
80
130
0
0
Cen
p
p
pp
p
200 400 600 800 1000
Relative fluorescence
20
0
0
p p
p
200 400 600 800 1000
Relative fluorescence
F I G . 1. Fluorescence histograms of pollen nuclei collected from single honey-bees foraging on (A) 2x Solidago canadensis, (B) 6x Solidago altissima, (C, D) a mixed
patch of two Centaurea species and (E, F) single bumble-bees on Monarda fistulosa. Vertical dashed lines indicate the expected position of 1n pollen nuclei based on
external standards of a leaf collected from the same plant (A, B) or pollen collected from flowers of the same species (C–F). In (A) and (B), Sc-1x and Sa-3x nuclei
correspond to S. canadensis and S. altissima 1n nuclei, respectively, while Sc-2x and Sa-6x correspond to unreduced (2n) nuclei or doublets from the same species; some
nuclei matching S. canadensis may also be Euthamia graminifolia or Erigeron strigosus. In (C) and (D), Cen indicates pollen nuclei matching Centaurea. In (E) and (F),
Mf indicates nuclei matching 1n M. fistulosa, and ‘p’ indicates other pollen nuclei. Histograms (E) and (F) are gated on a side-scatter vs. fluorescence plot to remove
debris.
For those honey-bee samples with two types of pollen nuclei, the
RSE for the proportion of nuclei belonging to the most common
type never exceeded 2 %, and the RSE for the proportion of the
less common pollen exceeded 10 % in only one sample (19 %;
Supplementary Data Table S2), despite the fact that RSEs increase with decreasing proportion ( p) values within a sample.
In contrast, 70 % of the bumble-bee samples had nuclei from at
least 300 pollen grains, but only 30 % had nuclei from at least
1000. Although the RSE for the proportion of the most
common pollen type (usually M. fistulosa) was always ,3 %,
most bumble-bees carried at least one pollen type whose proportion had an RSE .10 %. Nevertheless, only the sample with the
fewest pollen grains (n ¼ 81) had pollen type proportions with
RSEs exceeding 30 % (Supplementary Data Table S2).
When the higher quality standards of cell cycle analysis were
applied to the histograms, the success rates differed among the
particular bee – forage plant combination. All samples from
honey-bees foraging on Solidago had CVs ,8 % (mean ¼
3.05 %, maximum ,5 %), 83 % had in excess of 10 000 nuclei
and 58 % had BAD ,20 %, so that 58 % met all criteria for
cell cycle analysis (Supplementary Data Table S2). In contrast,
while all honey-bees foraging on Centaurea had CVs ,8 %
(mean ¼ 3.86 %, maximum ,5 %), only 40 % had 10 000
nuclei and only 20 % had BAD ≤20 %, for a total of 10 %
meeting all criteria. For Monarda bumble-bee samples,
ungated histograms could not be fit well using ModFit software
because of high noise/debris levels and BAD exceeded 20 %
for all samples, so no samples met the quality criteria. After
gating out debris on an FL2 area vs. side-scatter plot, all peaks
of Monarda nuclei had CVs ,5 %, but smaller peaks still
frequently could not be fit well by the ModFit software, and
only one sample had at least 10 000 nuclei.
Samples from Solidago honey-bees stored for 7 months at
– 20 8C had higher CVs (mean ¼ 4.0 %) than samples run after
2 h storage (3.5 %, t-test, P ¼ 0.034), but BADs did not differ
(7 months mean ¼ 19 %, 2 h mean ¼ 20 %, P ¼ 0.80).
Identification of pollen sources
Nine out of 10 honey-bees collected from Centaurea had a
single fluorescence peak that corresponded to the expected position for the two Centaurea species (Fig. 1A), which were effectively identical in DNA content within the level of resolution of
the current study (Supplementary Data Table S1). Two other
species present in the population (Arctium lappa and Erigeron
strigosus) were close enough in DNA content to overlap the
Centaurea peak, so this single peak could potentially be attributed to four species (Table 1). In the one sample with a second type
of nuclei (5.2 % of the nuclei and pollen present), the only matching pollen source was Solidago canadensis.
DNA content testing of Solidago leaf tissue at the Guelph site
showed that the 12 plants from which bees were collected
included three S. canadensis (2x) and nine S. altissima (6x),
and pollen collected from bees typically contained a mix of the
two cytotypes (Table 1). Peaks matching 1x S. canadensis and
3x S. altissima nuclei were detected on all bees, regardless of
whether they were collected from 2x plants or 6x plants,
with the exception of one collected from a 2x S. canadensis
(S. canadensis-type pollen only). Most bees carried at least 80
% pollen nuclei matching the cytotype of the plant on which
Kron et al. — Flow cytometric analysis of individual bee pollen loads
TA B L E 1. Percentage of pollen on each bee corresponding in
DNA content to the plant on which it was captured, number of
other pollen species present not matching the capture plant in
DNA content and mean number of possible donor species per
peak across all peaks of nuclei present on a given bee
Bee
% pollen with capture
plant DNA content
Other
species
Mean possible
species/peak
Apis.Cen.01
Apis.Cen.02
Apis.Cen.03
Apis.Cen.04
Apis.Cen.05
Apis.Cen.06
Apis.Cen.07
Apis.Cen.08
Apis.Cen.09
Apis.Cen.10
Apis.Sol.01.2x
Apis.Sol.07.2x
Apis.Sol.12.2x
Apis.Sol.02.6x
Apis.Sol.03.6x
Apis.Sol.04.6x
Apis.Sol.05.6x
Apis.Sol.06.6x
Apis.Sol.08.6x
Apis.Sol.09.6x
Apis.Sol.10.6x
Apis.Sol.11.6x
Bombus.Mf.01
Bombus.Mf.02
Bombus.Mf.03
Bombus.Mf.04
Bombus.Mf.05
Bombus.Mf.06
Bombus.Mf.07
Bombus.Mf.08
Bombus.Mf.09
Bombus.Mf.10
100
100
100
100
100
100
100
94.8
100
100
94.4
100.0
86.2
93.5
90.3
46.2
81.3
96.3
97.7
69.0
49.1
91.2
65.5
47.1– 55.4
26.2– 72.7
49.3– 82.3
85.4
15.2– 15.8
73.4
76.3– 93.1
81.0– 86.0
12.5– 58.8
0
0
0
0
0
0
0
1
0
0
1–2
0
1
1–2
1–2
1–2
1–2
1–2
1–2
1–2
1–2
1–2
2
3–5
5
2–4
1
4
3
1–2
2–3
1
4
4
4
4
4
4
4
2.5
4
4
2
2
1.5
1.33
1.33
1.33
1.33
1.33
1.33
1.33
1.33
1.33
2.33
1.75
3.5
4.67
1
1.67
1.67
2
2
4
Values for ‘% pollen with capture plant DNA content’ and ‘Other species’
can vary because the number of pollen grains and species present depends on
the particular combination of binucleate and trinucleate species to which the
nuclei are assigned, and different combinations of candidate species are
possible for each sample.
Calculation of ‘Mean possible species per peak’ and ‘Other species’ does
not include peaks containing ,2 % of all nuclei and interpreted as unreduced
gamete nuclei according to our criteria.
they were collected, although three had values between 46 and 69
% (Table 1).
Pollen nuclei (1x) from S. canadensis could not be distinguished
from 1x nuclei of E. graminifolia, because these two species overlapped in 2C leaf DNA content; while E. graminifolia was much
less common, some 1x pollen potentially originated from this
species. Also, nuclei from the uncommon Erigeron strigosus
were a close match for 2x S. canadensis nuclei, although only
one sample had sufficient events at this fluorescence to warrant
consideration as either E. strigosus or unreduced nuclei/doublets
from S. canadensis (S01, Table 1). All samples collected from
S. altissima-foraging bees had a 1x peak with two possible
origins (S. canadensis or E. graminifolia) and two peaks attributed
to S. altissima (a 3x peak and a 6x peakexceeding 2 % of nuclei), for
an average of 1.33 candidate species per peak (Table 1). The three
195
S. canadensis samples were more variable: all had the 1x peak
with two candidate species, one had no other peaks (S07), one
had an additional 3x S. altissima peak (S12) and one had a 1x, 2x
and 3x peak (S01), with the 2x attributable to three species
(S. canadensis, E. graminifolia or E. strigosus) (mean candidate
species/peak ¼ 2, 1.5 and 2, respectively, Table 1).
For bees collected from M. fistulosa, a fluorescence peak
matching M. fistulosa nuclei was detected on each histogram,
along with nuclei from at least one other species. The Monarda
peak typically contained the most nuclei. In one exceptional
case, a bee with an unusually high pollen number (.10 000)
had only 15.8 % Monarda nuclei (Table 1). In 60 % of
samples, we could unambiguously determine the number of
non-Monarda pollen species present on the bee (assuming one
species per fluorescence peak), but the identity of these species
was generally ambiguous (Table 1). Only one Bombus had a
set of pollen peaks that could be unambiguously attributed to
specific pollen sources; for the other nine bumble-bees, each
peak of nuclei had an average of 1.67– 4.67 possible candidate
species (averages across peaks on individual bees; Table 1).
Five bumble-bees had nuclei from one or two species that
could not be matched to any of the plants collected in the
population.
DISCUSSION
Flow cytometry can be used to classify and quantify pollen
‘types’ (defined by DNA content) within single bee pollen
loads. We were able to measure the relative DNA content of
nuclei from between 81 and 11 512 pollen grains from individual
bees, using a method that allowed the testing of 2 – 10 bees h21.
Eighty-five per cent of all samples provided data from .300
pollen grains, and 70 % from .1000, meeting or exceeding
requirements used in similar morphological analyses (Behm
et al., 1996; Von Der Ohe et al., 2004; Barth et al., 2011). The
sufficiency of these sample sizes is reflected in the RSEs for
the estimates of proportions of pollen types, all but one of
which were ,25 %, with much lower numbers (,2.5 %)
typical for the numerically dominant types. These numerical
requirements were met despite being limited to pollen collected
from a single bee, as opposed to from honey or pollen traps in
hives. Honey-bees with prominent pollen loads ( pollen packs
visible in corbiculae) typically provided data for .1000 pollen
grains, while most bumble-bees and honey-bees without obvious
pollen loads still met the 300 pollen grain criterion. Furthermore,
more than half of all honey-bees foraging on Solidago met the
more stringent criteria for clinical cell cycle analysis, demonstrating that very high quality output can be obtained under
some conditions (i.e. large pollen loads and pollen species
with low debris). For these samples, meeting the 10 000 nuclei
criterion meant that at least 3333 pollen grains were assayed.
High debris levels resulted in BAD .20 % for most of the bees
on Monarda and Centaurea, but while low debris is always preferable, there are two reasons why the observed levels are not necessarily problematic for finding pollen type frequencies. First,
the BAD ,20 % criterion for cell cycle analysis is largely
driven by the need to estimate S phase (Shankey et al. 1993),
which is not relevant when counting the 1C and 2C nuclei of
mature pollen grains. Secondly, in pollen flow cytometry
samples, a large proportion of debris consists of particles with
196
Kron et al. — Flow cytometric analysis of individual bee pollen loads
distinctive properties (e.g. exine fragments with high side
scatter) that allow them to be readily gated out without affecting
counts of nuclei.
The non-specific classification of pollen into DNA content
‘types’ is comparable with identifying ‘types’ or ‘forms’ of
pollen on the basis of morphology, where it may be possible to
distinguish families, tribes or genera, but rarely species
(Wodehouse, 1959; Louveaux et al., 1978; Van Der Ohe et al.,
2004; Cane and Sipes, 2006). Because of the limits of DNA
content as a species marker, species resolution is much better
in simpler contexts (e.g. the Guelph Solidago site) than in floristically diverse ones (e.g. the Arkell Centaurea and Monarda
site). At the Guelph site, with only four flowering species,
nuclei not identifiable as S. altissima were either the abundant
S. canadensis or the relatively uncommon E. graminifolia or
E. strigosus. At the Arkell site, with 20 flowering species and
many overlapping DNA contents, the species resolution was
much lower. The inherent limits on DNA content as a species
marker were probably exaggerated by the simplified methodology we adopted in this study. Specifically, replication within
species would provide more precise measures for the relative
DNA contents of the reference set, as well as determining
whether our 10 % difference criterion was too broad. Also
there is evidence that fluorescence properties may differ slightly
between pollen and leaf nuclei within a species (Kron and
Husband, 2012), so the use of pollen from all species, rather
than leaves, would almost certainly provide the most accurate
reference values,
A comparison of the species resolution provided by DNA
content with that of morphological analysis demonstrates the
way in which these methods may complement one another.
Wodehouse (1959) described Solidago species (including
Euthamia at that time) as essentially indistinguishable from
one another when using morphological traits. In our study,
DNA content could not distinguish S. canadensis and
E. graminifolia, but S. canadensis and S. altissima were readily
separable. On the other hand, some plants in our study with nearidentical DNA contents would be readily distinguished by
morphology (e.g. T. pratense and D. carota; Crompton and
Wojtas, 1993). Combining the two methods would provide
improved resolution, as would the additional use of other
pollen identification techniques. Genetic marker approaches
have been used to discriminate pollen in specific groups (e.g.
Zhou et al., 2007), but development of a method applicable
across diverse groups (DNA barcoding) is still incomplete for
pollen, and the limited results to date suggest that species-level
identification may be difficult in heterospecific pollen loads
(Shoemaker, 2010). Other techniques familiar to pollination
biologists would complement pollen load analyses by narrowing
the list of possible pollen sources. Observations of bee movements, for example, might answer the question of how likely
bees are to move between certain species pairs (e.g. S. altissima
and the morphologically similar S. canadensis, vs. between
Solidago and Euthamia). We treated all flowering species in a
population as equally likely contributors to pollen loads, but information about the floral preferences of different pollinators, as well
as the abundance and spatial distribution of the different flowering
species, could also narrow down identification possibilities.
Species resolution might also be improved by considering other
flow cytometric measures of the nuclei, such as fluorescence
width, side scatter and forward scatter, all of which can differ
between species because of variation in the size and shapes of
pollen nuclei.
Our analyses of pollen loads demonstrate the potential of this
method to reveal useful information about pollinator biology.
The contrast between the number of pollen types on honey-bees
and bumble-bees foraging in the same population was striking
(Table 1) and consistent with results from barcoding
(Shoemaker, 2010) although not with all studies of Bombus
(Morales and Traveset, 2008). The clear evidence that honeybees were moving regularly between two species and cytotypes
of Solidago is an important result, relevant to the study of
polyploid plant evolution (Kennedy et al., 2006) and, more
generally, to studies of the effects of interspecific pollen transfer
(Morales and Traveset, 2008). Understanding the complex nature
of plant – pollinator interactions, from pollinator behaviour
through to plant reproduction, requires the integration of many
kinds of observations, and flow cytometric analysis of pollen
loads can be a valuable new tool in the pollination biologist’s
toolkit.
S U P P L E M E N TARY D ATA
Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Table S1: plant
species flowering at the Arkell and Guelph sites, with 1C DNA
content relative to M. fistulosa and S. canadensis; Table S2:
number of nuclei, pollen number and quality measures for flow
cytometry samples.
ACK N OW L E DG E M E N T S
We thank Cara Dawson for assistance in identifying bees,
Mélissa Girard for comments on melissopalynology, and
Lindsey Vyvey for field assistance. This work was supported
by the Natural Sciences and Engineering Research Council of
Canada and Canada Research Chair Program to B.C.H. and by
equipment purchased through the Canada Foundation for
Innovation.
LIT E RAT URE CITED
Alarcón R. 2010. Congruence between visitation and pollen-transport networks
in a California plant– pollinator community. Oikos 119: 35– 44.
Barth OM, Da Silva De Freitas A, De Oliveira Sales É, Vit P. 2011.
Palynological evaluation of bee pollen load batches from the Venezuelan
Andes of Misintá. Interciencia 36: 296–299.
Barrett SCH, Harder LD. 1996. Ecology and evolution of plant mating. Trends
in Ecology and Evolution 11: 73–79.
Behm F, Von Der Ohe K, Henrich W. 1996. Zuverlässigkeit der Pollenanalyse
von Honig: Bestimmung der Pollenhäufigkeit. Deutsche LebensmittelRundschau 92: 183– 188.
Bino RJ, Van Tuyl JM, De Vries JN. 1990. Flow cytometric determination of
relative nuclear DNA contents in bicellulate and tricellulate pollen.
Annals of Botany 65: 3 –8.
Brewbaker JL. 1967. The distribution and phylogenetic significance of binucleate and trinucleate pollen grains in the angiosperms. American Journal of
Botany 54: 1069–1083.
Cane JH, Sipes S. 2006. Characterizing floral specialization by bees: analytical
methods and a revised lexicon of oligolecty. In: Waser NM, Olerton J, eds.
Plant–pollinator interactions: from specialization to generalization.
Chicago, IL: University of Chicago Press, 99–122.
Coyne JA, Orr HA. 2004. Speciation. Sunderland, MA: Sinauer Associates, Inc.
Kron et al. — Flow cytometric analysis of individual bee pollen loads
Crompton CW, Wojtas WA. 1993. Pollen grains of Canadian honey plants.
Agriculture Canada, Ottawa, Publication 1892/E
Delaplane KS, Mayer DF. 2000. Crop pollination by bees. Wallingford, UK:
CABI Publishing.
Doležel J, Binarová P, Lucretti S. 1989. Analysis of nuclear DNA content in
plant cells by flow cytometry. Biologia Plantarum 31: 113–120.
Ellstrand NC. 1992. Gene flow by pollen: implications for plant conservation
genetics. Oikos 63: 77– 86
Harder LD, Barrett SCH. 1996. Pollen dispersal and mating patterns in animalpollinated plants. In: Lloyd DG, Barrett SCH, eds. Floral biology: studies in
floral evolution in animal-pollinated plants. New York: Chapman and Hall,
140– 190.
Kennedy BF, Sabara HA, Haydon D, Husband BC. 2006. Pollinator-mediated
assortative mating in mixed ploidy populations of Chamerion angustifolium
(Onagraceae). Oecologia 150: 398–408.
Klein RJ, Proctor SE, Boudreault MA, Turczyn KM. 2002. Healthy People
2010 criteria for data suppression. Healthy People 2010 Statistical Notes
24. DHHS Publication No. (PHS) 2002-1237
Kron P, Husband BC. 2012. Using flow cytometry to estimate pollen DNA
content: improved methodology and applications. Annals of Botany 110:
1067–1078.
Levin DA, Kerster HW. 1974. Gene flow in seed plants. In: Dobzhansky T,
Hecht MK, Steere WC, eds. Evolutionary biology, vol. 7. New York:
Plenum Press, 139– 220.
Lopezaraiza-Mikel ME, Hayes RB, Whalley MR, Memmott J. 2007. The
impact of an alien plant on a native plant–pollinator network: an experimental approach. Ecology Letters 10: 539–550.
Louveaux J, Maurizio A, Vorwohl G. 1978. Methods of melissopalynology.
Bee World 59: 139– 153.
Marques I, Aguilar JF, Martins-Loução MA, Feliner GN. 2012. Spatial– temporal patterns of flowering asynchrony and pollinator fidelity in hybridizing
species of Narcissus. Evolutionary Ecology 26: 1433– 1450.
Morales CL, Traveset A. 2008. Interspecific pollen transfer: magnitude, prevalence and consequences for plant fitness. Critical Reviews in Plant Sciences
27: 221–238.
Ormerod MG, Tribukait B, Giaretti W. 1998. Consensus report of the task
force on standardisation of DNA flow cytometry in clinical pathology.
Analytical Cellular Pathology 17: 103–110.
197
Pasquet RS, Peltier A, Hufford MB, et al. 2008. Long-distance pollen flow assessment through evaluation of pollinator foraging range suggests transgene
escape distances. Proceedings of the National Academy of Sciences, USA
105: 13456–13461.
Ramsay G. 2005. Pollen dispersal vectored by wind or insects. In: Poppy GM,
Wilkinson MJ, eds. Gene flow from GM plants. Oxford: Blackwell
Publishing, 43–77.
Ruoff K. 2006. Authentication of the botanical origin of honey. MSc Thesis, ETH
Zurich, Switzerland.
Schaal BA. 1980. Measurement of gene flow in Lupinus texensis. Nature 284:
450–451.
Semple JC, Ringius GS, Zhang JJ. 1999. The goldenrods of Ontario: Solidago
L. and Euthamia Nutt, 3rd edn. U.W. Biology Series. Department of
Biology, University of Waterloo, Waterloo, Canada.
Shankey TV, Rabinovitch PS, Bagwell B, et al. 1993. Guidelines for implementation of clinical DNA cytometry. Cytometry 14: 472–477.
Sharma M. 1970. An analysis of pollen loads of honey bees from Kangra, India.
Grana 10: 35– 42.
Shoemaker IR. 2010. Tracking floral visitation using DNA barcodes. MSc
Thesis, Columbus State University, USA.
Slatkin M. 1985. Gene flow in natural populations. Annual Review of Ecology
and Systematics 16: 393–430.
Suda J, Kron P, Husband BC, Trávnı́cek P. 2007. Flow cytometry and ploidy:
applications in plant systematics, ecology and evolutionary biology. In:
Doležel J, Greilhuber J, Suda J, eds. Flow cytometry with plant cells: analysis of genes, chromosomes and genomes. Weinheim: Wiley-VCH
Verlag, 103 –130.
Von Der Ohe W, Persano Oddo L, Piana ML, Morlot M, Martin P. 2004.
Harmonized methods of melissopalynology. Apidologie 35: S18–S25.
Wersto RP, Chrest FJ, Leary JF, Morris C, Stetler-Stevenson MA,
Gabrielson E. 2001. Doublet discrimination in DNA cell-cycle analysis.
Cytometry 46: 296–306.
Wodehouse RP. 1959. Pollen grains: their structure, identification and significance in science and medicine. New York: Hafner Publishing Co.
Zhou L-J, Pei K-Q, Zhou B, Ma K-P. 2007. A molecular approach to species
identification of Chenopodiaceae pollen grains in surface soil. American
Journal of Botany 94: 477– 481.