Evaluation of the botanical origin of commercial dry bee pollen load batches using pollen analysis: a proposal for technical standardization

High quality of bee pollen for commercial purpose is required. In order to attend the consumer with the best identification of the botanical and floral origin of the product, 25 bee pollen batches were investigated using two techniques of pollen grain preparation. The first started to identify pollen loads of different colors in two grams of each well mixed batch, and the second to identify pollen grains in a pool made of all the pollen loads comprised in two grams. The best result was obtained by this last technique, when a pollen grain suspension was dropped on a microscope slide and circa 500 pollen grains were counted per sample. This analysis resulted in the recognition of monofloral and bifloral pollen batches, while the use of the first technique resulted in all samples receiving a heterofloral diagnosis.


INTRODUCTION
Bee pollen production increased during the last years as a response to commercial demand.Apis mellifera L. is the best pollen supplier mainly in tropical countries where entomophilous plant species are dominant.Beekeepers use pollen traps of different types in order to obtain the pollen loads from bees when coming home to hives.This pollen is humid and has to be dried before commercialization.Then, it may be distributed in vials receiving an identification that comprises, in addition, its botanical origin.This procedure is important to avoid missing the accurate scientific definition of the botanical name, since beekeepers report a lot of common names of plants that were visited by the bees.
Pollen grains present a huge variation of morphological features that are established by its genetic hered-ity and are not influenced by environmental events or changes.Pollen analysis is the tool to recognize from where the pollen grains are coming.Several laboratorial techniques are used to prepare pollen grains for microscope observation and identification.The difficulty remains in the interpretation of the obtained data and the evaluation of numerical and personal informations.Several aspects are to be considered, the first one regarding the bee behavior upon flower visitation for pollen extraction.In this case, single bees are captured and both of their pollen loads were analyzed (Carvalho and Marchini 1999, Marques-Souza et al. 2002, Noor et al. 2009).Secondly, when pollen traps are used, several pollen loads perform one sample that, for analysis, has to be well mixed; a group of pollen loads, in general selected by its weight or color (Almeida-Muradian 2005, Barth et al. 2009, Modro et al. 2009a), is used for the recognition of the botanical species composing this pollen sample.
The main issue to address is the significance or not of the choice by color.An accurate description of the pollen load sample processing and analysis is frequently missed during experiment descriptions (Modro et al. 2007(Modro et al. , 2009a)).Commonly, a fragment of a pollen load of each color is analyzed; no sample centrifugation is used.In another way (denoted technique 1), ten pollen loads (Barth et al. 2009) or 25 ones of each color batch (Barth, personal information and in the present paper) are mixed using ethanol, and a little amount is analyzed after centrifugation.As pollen grains have different weight, a gradient is obtained, from which pollen grains are removed using a piece of glycerin jelly.The significance of the obtained results is not clear.Another technically very complex preparation of pollen batches identification was recently used comprising several steps of pollen drying, hydrating and acetolysis (Novais et al. 2009) and, thus, missing important information besides pollen morphology.
The proposed technique by this study (technique 2) may help to get quick and accurate information about the botanical and regional origin of commercial bee pollen load products.It neither demands long time to be performed nor is expensive.

MATERIALS AND METHODS
Twenty five samples of commercial batches of bee pollen loads were obtained in apiaries of several states in Brazil and analyzed (Table I).Each sample was analyzed using two techniques.The first technique considers sub samples according to the pollen load colors; each sample was made of well-mixed two grams and each sub sample was made of 25 units or pollen loads of each color.When two or more types of color were mentioned in a same sub sample (Table I, column 2), it indicates that no sufficient pollen loads of a unique color were available to perform n=25, and several poorly represented color batches were mixed.Color was determined according to the Red Green Blue (RGB) color classification system.Pollen grain identification used the available literature, meanly Barth (1989), Roubik and Moreno (1991), and the Palynological Slide Collection of the Laboratory of Palynology.Pollen frequency standard classes (Louveaux et al. 1978) were used, making an additional subdivision of the dominant pollen class (Table I).

Preparation of pollen load samples:
The pollen loads of two grams of each sample were distributed in batches or sub samples according to its color.

Preparation of microscope slides:
Twenty five pollen loads of each sub sample were macerated using 10 mL of 70% ethanol distributed into two 15 mL centrifuge tubes and centrifuged during 3 minutes at 1500 rpm.The sediment was resuspended with 10 mL of 70% ethanol and centrifuged again.In sequence, 5 mL of a water/glycerin 1:1 mixture were added to each tube, stirred, and left for 30 minutes.After centrifugation, the tubes remained with their aperture down on absorbent paper during a few minutes.Then, the pollen sediment was mixed and pollen slides were prepared using a little piece of glycerin jelly to capture the pollen grains, and slides were sealed with paraffin.Counting and identification comprised at least 500 pollen grains.

Preparation of pollen load samples:
Dry pollen loads of two grams of a sample (this is a pool of circa 300 pollen loads) were weighted into a 15 mL Falcon centrifuge tube, mixed using 70% ethanol just to complete 13 mL, and left for 30 minutes.Treatment with ultrasound (if disposable) during 5 minutes can be suitable for particle dispersal.The sediment obtained after centrifugation may be extracted with ethanol, and submitted to ultrasonic treatment again.A solution of distilled water/glycerin 1:1 was added to the sediment just to complete 13 mL, and left for circa 30 minutes.

Preparation of microscope slides:
One drop of the well-mixed pollen grain suspension was applied on a microscope slide, covered with a 22×22 mm cover slide, and sealed with enamel.The slide may be An Acad Bras Cienc (2010) 82 (4) STANDARDIZATION OF DRY BEE POLLEN ORIGIN EVALUATION 895 Fig. 1 -A pool of pollen loads prepared using technique 2; pollen grains are well dispersed all over the slide.Fig. 2 -A pool of pollen loads prepared using technique 2 three weeks later; the slide started to dry, and the pollen grains overlapped and were pressed against the air bubbles.maintained in a horizontal position for circa one month or less, when drying starts, and the pollen grains were selectively pressed against the air bubbles.Then discard the slide (Figs. 1 and 2).

Preparation of stock samples:
One mL of the well mixed pollen grain suspension was put into a 1.5 mL Eppendorf centrifuge tube and centrifuged.A 15 mL Falcon tube can be used as a support for the Eppendorf tube.After discarding the supernatant, one mL of glycerin was added, mixed, and then the tube was closed and identified for additional preparations.It may be kept at room temperature.

Preparation of additional pollen slides:
The content of the Eppendorf tube has to be homogenized.One drop of this pollen grain suspension was mixed with one drop of distilled water in a clean Eppendorf tube.A microscope slide, as formerly described may be prepared after 30 min waiting for water impregnation.

Counting and identification of pollen grains and structured elements:
At least 500 pollen grains have to be counted and identified using a 400× magnification, preferentially within 10 days, before drying starts.

RESULTS AND DISCUSSION
The obtained data are shown in Table I.Each of the 25 samples analyzed presented two or more colors of pollen loads, establishing several sub samples.No sample was single colored.When no sufficient pollen loads of a unique color were available, two, rarely three different colored pollen loads were joined together in order to make up a group of 25 unities (column three); these never constituted a dominant sub group in a sample or batch.
Fifty two pollen types presenting a frequency equal or higher than 3% of the pollen sum were recognized.Technique 1 presented 46 pollen types, 22 being exclusives.Technique 2 presented 29 pollen types, six being exclusives.
No monofloral pollen load batch was recognized by technique 1.Among the 25 analyzed samples, three were bifloral presenting two dominant pollen types; 22 samples were heterofloral comprising three or more pollen types.Considering technique 2, eight pollen load batches were monofloral, six bifloral and 11 heterofloral.
Nevertheless, technique 1 detected more pollen types (46) than technique 2 (29), this last procedure showed a better evaluation of the whole pollen batch.Thus, commercial qualification has to follow this way.Consequently, it presented more monofloral batches.
Technique 1 was useful in order to teach about pollen loads botanical origin.As explained below, the color of pollen of a plant species may vary (Barth et al. 2009 and in the present paper).Therefore, it is not a good characteristic to inform about the botanical origin of a pollen load batch.While it is not possible to analyze the pollen grain composition of each pollen load, and several plants may contribute to form a unique load, technique 2 is a quicker and more accurate procedure to make a diagnosis of a large commercial product.

TABLE I
Comparison of 25 pollen load samples (2g of each batch) from Apis mellifera evaluated by its color (sub samples in columns two, three and four) and evaluated by a pool of 2g of each batch (columns five and six).Each sub sample comprised 25 units of pollen loads.Only the pollen types with frequency >3% were considered: very frequent (++++, circa >85%), frequent (+++, circa 45 to 85%), few frequent (++, circa 15 to 45%), rare (+, circa 3 to 15%).Preparing slides in five folds, the same pollen spectrum was obtained.
Detailed discussion of the data obtained by the use of the two techniques of pollen load batches analysis (Table I).
Pollen grain color of a plant taxon (third column in Table I) can change from bright to dark as it may be observed in Cocos (samples 55, 74, 77, 115, 145, 146), Vernonia (samples 58, 111), Eucalyptus (samples 86, 87) and Baccharis (sample 88) pollen grains, and in the Mimosa scabrella pollen type (sample 115).Time to air exposition of pollen grains resulting in exine and cytoplasm oxidation could be responsible for different colored pollen loads of a plant species.
The monofloral samples presented five dominant pollen types: Cocos nucifera and Mimosa scabrella (= M. sensitiva) pollen types from the state of Bahia, Cecropia and Eucalyptus from São Paulo State, Antigonon leptopus (a garden species) and Eucalyptus from the state of Minas Gerais.These commercial batches have obtained the best evaluation.
The bifloral samples presented six important pollen types: Cocos nucifera and Cyperus from the state of Piaui, Mimosa scabrella and Cocos nucifera in three samples from the state of Sergipe, Mimosa caesalpiniaefolia and Cocos nucifera also from the state of Sergipe, and Cecropia and Myrcia from the state of Minas Gerais.These commercial batches may receive a good evaluation.
The heterofloral pollen load batches analyzed by techniques 1 and 2 do not have palynological definition of any dominance.Their botanical origin is variable depending upon several factors and their reproduction must not be effective.
A curious composition of some pollen loads was made of uredospores of fungi detected by technique 1 (in samples 59,72,74,100).Uredospores never appeared using technique 2. As so, their contribution to any of the pollen batches was not significant.Cladosporium sp.spores were also collected by bees in the state of Minas Gerais during an alimentary scarcity (Modro et al. 2009b).
In conclusion, using technique 1 based upon color analysis, more pollen types were identified, but no dominance of pollen type or plant species in a commercial pollen batch was reported.When using the technique 2, besides the monofloral batches, more bifloral and less heterofloral batches were recognized.This result shows that a better characterization of a large pollen load batch composition of commercial interest was obtained when using the last technique.