Characterization of Pollen Carotenoids with in situand High-Performance Thin-Layer ChromatographySupported Resonant Raman Spectroscopy

Franziska Schulte,†,‡ Jens Mader,§ Lothar W. Kroh,§ Ulrich Panne,†,‡ and Janina Kneipp*,†,‡

Chemistry Department, Humboldt Universitat zu Berlin, Brook-Taylor-Strasse 2, D-12489 Berlin, Germany, FederalInstitute for Materials Research and Testing, I.42, Richard-Willstatter-Strasse 11, D-12489 Berlin, Germany, andDepartment of Food Analysis, Institute of Food Technology and Food Chemistry, Technical University Berlin,Gustav-Meyer-Allee 25, D-13355 Berlin, Germany

Raman signatures of the carotenoid component are stud-ied in individual pollen grains from different species oftrees. The information is obtained as differences in thestrong pre-resonant Raman spectra measured before andafter photodepletion of the carotenoid molecules. Theresults provide the first in situ evidence of interspeciesdifferences in pollen carotenoid content, structure, and/or assembly between plant species without prior purifica-tion. The analysis of carotenoids in situ is confirmed byhigh-performance thin-layer chromatography (HPTLC)-supported resonance Raman data measured directly onthe HPTLC plates after separation of carotenoids in pollenextracts. Utilization of the in situ, extraction-free proce-dure in carotenoid analysis will improve sensitivity andstructural selectivity and provides insight into carotenoidstructure and composition in single pollen grains.

Raman spectroscopy has become an important tool in theanalysis of complex and heterogeneous biomaterials. As evidencedby results from our and other groups, it is possible to detect,chemically characterize, and image pollen by Raman spec-troscopy.1-6 Raman spectroscopic analysis is based on thefingerprint-like information that results from the spectral contribu-tions of all molecules present in individual pollen grains, namely,proteins, lipids, nucleic acids, carbohydrates, pigments, and sofar unknown organic structures such as the constituents of thesporopollenin outer pollen coat. We have recently found thattaxonomic relationship between plant species is reflected by thesimilarity of the Raman spectra from their pollen and have shown

the capability of a fast Raman-based identification of single pollengrains.6 In such an identification approach, the multivariateinformation from pollen Raman spectra is analyzed using auto-mated data evaluation. In ref 6 we had based such an analysis onall types of biological molecules contained in the pollen. In someof the experiments, very intense pre-resonant Raman scatteringfrom carotenoids superimposed spectral contributions from thesecompounds. The spectral signatures of the carotenoid pigmentscan be the main reason for misclassification of biological speciesin multivariate analysis, and strategies to overcome this problemhave been suggested.6,7 In our recent study on classification ofpollen grains, we applied photodestruction of the carotenoidmolecules, using 633 nm laser light prior to the acquisition ofRaman spectra that were excited with light from a 785 nm laser.6

In principle, a second method for eliminating the intensespectral contribution from carotenoid molecules is the correctionof whole pollen spectra after data acquisition. This is difficult toachieve using spectra from model substances, since the carotenoidsignature in situ is a result of many different carotenoid speciesand includes effects that are due to their association with thebiological matrix.

In this paper, instead of removing the carotenoids, we aim atselectively extracting Raman spectral information also from thesepigment molecules by the following approach: Due to photo-bleaching at the applied 633 nm pre-resonant excitation, thecontribution from the carotenoid molecules in the spectra de-creases over time. To extract the spectral signature of thecarotenoids, we use the difference between the Raman spectrabefore and after their depletion. In this way, by generating thesedepletion difference spectra, pure carotenoid spectral features canbe analyzed. The in situ carotenoid spectrum would enable areliable correction of pollen spectra. In addition, another motiva-tion for the analysis presented here is to gather insight into insitu carotenoid structure and composition in single pollen grainsof different plant species. Using the Raman spectrum this can beachieved with improved sensitivity and structural selectivity.

It has long been known that carotenoids are present in pollen.Initially, it was suggested that the outer layer of the pollenstheexineswas formed by an oxidative polymerization of carotenoids

* To whom correspondence should be addressed. E-mail: janina.kneipp@chemie.hu-berlin.de. Phone: +49-30-2093-7171. Fax: +49-30-2093-7175.

† Humboldt Universitat zu Berlin.‡ Federal Institute for Materials Research and Testing.§ Technical University Berlin.

(1) Manoharan, R.; Ghiamati, E.; Britton, K. A.; Nelson, W. H.; Sperry, J. F.Appl. Spectrosc. 1991, 45, 307–311.

(2) Laucks, M. L.; Roll, G.; Schweiger, G.; Davis, E. J. J. Aerosol Sci. 2000,31, 307–319.

(3) Ivleva, N. P.; Niessner, R.; Panne, U. Anal. Bioanal. Chem. 2005, 381,261–267.

(4) Sengupta, A.; Laucks, M. L.; Davis, E. J. Appl. Spectrosc. 2005, 59, 1016–1023.

(5) Kano, H.; Hamaguchi, H. O. Chem. Lett. 2006, 35, 1124–1125.(6) Schulte, F.; Lingott, J.; Panne, U.; Kneipp, J. Anal. Chem. 2008, 80, 9551–

9556.(7) Scholtes-Timmerman, M.; Willemse-Erix, H.; Schut, T. B.; van Belkum, A.;

Puppels, G.; Maquelin, K. Analyst 2009, 134, 387–393.

Anal. Chem. 2009, 81, 8426–8433

10.1021/ac901389p CCC: $40.75 2009 American Chemical Society8426 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009Published on Web 09/24/2009

and carotenoid esters,8 which was contradicted by tracer experi-ments later.9 Up to now, the importance of carotenoids in pollenhas not been clarified, but it seems possible that carotenoids arenecessary to attract insects and/or to protect plants againstvermin.10 So far the analysis of carotenoids from pollen has beenrestricted to very few species that were studied mostly bychromatographic methods. In particular, the following carotenoidswere identified in pollen obtained from the species Cucurbita pepo,Calendula officinalis, and Lilium tigrinium: �-carotene, cryptox-anthin, �-carotene-5,6,5,6-diepoxide, zeaxanthin, antheraxanthin,violaxanthin, neoxanthin, flavoxanthin, lutein, 9/9-(Z)-lutein, �-car-otene, and luteoxanthin.9,11-13 Thin-layer chromatography (TLC)and high-performance thin-layer chromatography (HPTLC) havealso been applied to analyze carotenoids in many other biologicalsamples, among them leaves,14 wine,15 and snails.16 Because ofresonant enhancement of their Raman scattering at visibleexcitation,17-20 carotenoids were successfully studied withoutinterference from other molecules in several different complexmatrixes, such as fungi,21 lichens,22 molluscs,23 birds’ feathers,24

and whole fruit,25 by resonant Raman spectroscopy.Here, we discuss in situ Raman spectra of carotenoids collected

from different pollen species. As we demonstrate, the depletiondifference spectra can be used to obtain information on carotenoidcomposition directly in situ in single pollen grains withoutpurification procedures. We compare the situ Raman results frompollen grains with the information obtained by HPTLC and applya combination of HPTLC and resonant Raman scattering (RRS)on pollen extracts as well. By using these two procedures, weanalyze the carotenoid composition of tree pollen of differentspecies (horse chestnut, large-leaved linden, European ash, sallow,mahaleb cherry, tree of heaven).

METHODS SECTIONChemicals and Samples. Acetone, n-hexane, tetrahydrofuran,

and methylene chloride were purchased from VWR International(Darmstadt, Germany), and pure carotenoids were from Rot-ichrom (Carl Roth, Karlsruhe, Germany). Lutein was obtained

from a softgel product (nutrition supplement) from a drugstorein the United States. Pollen of six different species were collectedduring their flowering season: white horse chestnut (Aesculushippocastanum), sallow (Salix caprea), tree of heaven (Ailanthusaltissima), European ash (Fraxinus fragilis), mahaleb cherry(Prunus mahaleb), and large-leaved linden (Tilia platyphyllus). Partof the samples was snap-frozen in liquid nitrogen and stored at-20 °C. The pollen grains were snap-frozen within very shorttimes upon removal from their flowers.

Sample Preparation for HPTLC. For HPTLC analysis, thecarotenoids were extracted directly after harvesting with acetoneuntil the supernatant solution remained colorless. In particular,the amounts of pollen of the different species used for extractionwere for horse chestnut 6.2 mg (680 µL acetone), sallow 2.7 mg(300 µL acetone), tree of heaven 6.2 mg (400 µL acetone),European ash 10 mg (300 µL acetone), mahaleb cherry 62 mg(900 µL acetone), and for large-leaved linden 10.7 mg (600 µLacetone). The extracted and pure carotenoids were kept in thedark and cold (-17 °C). All measurements were performed indimmed laboratories to avoid rearrangements or decompositionof carotenoid molecules.

Preparation of Standards for HPTLC. As external standards,stock solutions of zeaxanthin, cryptoxanthin, �-carotene, and luteinin acetone were prepared and diluted with n-hexane. For calibra-tion, the stock solutions were mixed 1:1:1:1 (v/v) and subsequentlydiluted with n-hexane. For qualification, single standard solutionsand the standard mix were applied on one plate. Volumes of0.5-15 µL were applied onto the plates.

Chromatography. Chromatographic plates (20 cm × 10 cm,HPTLC plates silica gel 60, without F, Merck, Darmstadt,Germany) were used without prewashing. Samples wereapplied bandwise onto the plate using an automated TLC-Sampler IV (CAMAG, Muttenz, Switzerland). Sample applica-tion took place by spraying and using the following parameters:band length, 5.3 mm; distance between tracks, 9.4 mm;application volume, 0.5-15.0 µL; dosage speed, 160 nL/s.Directly after air drying, the plates were developed in anautomated multiple development II device (CAMAG, Muttenz,Switzerland) using an elution system of tetrahydrofuran,methylene chloride, and n-hexane with the parameters shownin Table 1. Development was carried out without preconditioningwith a drying time of 4 min between each developing step.

Densitometry was performed using a TLC-Scanner III (CA-MAG, Muttenz, Switzerland). Densitometric scanning had to beaccomplished within 5 min after development due to the fastbleaching of the carotenoid color. UV-vis absorption spectra ofthe standard and sample zones were recorded in the range of250-650 nm in order to identify the optimum detection wave-length for quantification. Quantitative densitometry took place byscanning at 425 nm with a slit dimension of 4.0 mm × 0.3 mmand a scanning speed of 20 mm/s.

Depending on the desired range, calibration was done by linearor polynomial regression. Graph height was used for quantitativeevaluation. All data are based on at least two repeated experiments.

Raman Spectroscopy. All Raman experiments were per-formed using a LabRam HR800 Raman setup (Jobin Yvon,Bensheim, Germany), coupled with a BX41 microscope (Olympus,Hamburg, Germany). To reject the Rayleigh-scattered light, notch

(8) Brooks, J.; Shaw, G. Nature 1968, 219, 532–533.(9) Prahl, A. K.; Springstubbe, H.; Grumbach, K.; Wiermann, R. Z. Naturforsch.,

C: Biosci. 1985, 40, 621–626.(10) Masson, G.; Baumes, R.; Puech, J. L.; Razungles, A. J. Agric. Food Chem.

1997, 45, 1649–1652.(11) Bako, E.; Deli, J.; Toth, G. J. Biochem. Biophys. Methods 2002, 53, 241–

250.(12) Karrer, P.; Oswald, A. Helv. Chim. Acta 1935, 18, 1303–1305.(13) Schulz, H.; Baranska, M.; Baranski, R. Biopolymers 2005, 77, 212–221.(14) Dauradelevagueresse, M. H.; Bounias, M. Chromatographia 1991, 31, 5–

10.(15) de Pinho, P. G.; Ferreira, A. C. S.; Pinto, M. M.; Benitez, J. G.; Hogg, T. A.

J. Agric. Food Chem. 2001, 49, 5484–5488.(16) Evans, R. T.; Fried, B.; Sherma, J. Comp. Biochem. Physiol., Part B: Biochem.

Mol. Biol. 2004, 137, 179–186.(17) Sufra, S.; Dellepiane, G.; Masetti, G.; Zerbi, G. J. Raman Spectrosc. 1977,

6, 267–272.(18) Salares, V. R.; Mendelsohn, R.; Carey, P. R.; Bernstein, H. J. J. Phys. Chem.

1976, 80, 1137–1141.(19) Okamoto, H.; Sekimoto, Y.; Tasumi, M. Spectrochim. Acta, Part A 1994,

50, 1467–1473.(20) Saito, S.; Tasumi, M.; Eugster, C. H. J. Raman Spectrosc. 1983, 14, 299–

309.(21) Arcangeli, C.; Cannistraro, S. Biopolymers 2000, 57, 179–186.(22) Edwards, H. G. M. Spectrochim. Acta, Part A 2007, 68, 1126–1132.(23) Barnard, W.; de Waal, D. J. Raman Spectrosc. 2006, 37, 342–352.(24) Veronelli, M.; Zerbi, M.; Stradi, R. J. Raman Spectrosc. 1995, 26, 683–692.(25) Baranska, M.; Schutz, W.; Schulz, H. Anal. Chem. 2006, 78, 8456–8461.

8427Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

filters were used. The spectrometer was equipped with a nitrogen-cooled CCD detector (1024 × 256 pixels). The excitation wave-lengths 488, 633, and 785 nm were supplied by a He-Ne laser,an argon ion laser (Melles Griot, Aalsbergen, Netherlands), anda diode laser (Toptica, Graefelfing, Germany), respectively.

Forty-five minutes prior to the Raman experiments, pollensamples from the species horse chestnut, sallow, and large-leavedlinden were thawed in a desiccator on CaF2 slides to avoidaccumulation of condensing water and possible hydration.Raman spectra of individual pollen grains were measured withan excitation wavelength of 633 nm (∼106 W/cm2) in acollection time of 10 s. Irradiation with the 633 nm laser alsocaused decomposition of �-carotene, and depletion in caro-tenoid content was verified by the decrease of the carotenoidRaman signals, such as the ν1 carotenoid signal around 1520cm-1. As described in the introduction, to obtain spectra fromcarotenoid pigments, the difference spectrum of the spectra ofthe unbleached and bleached sample was generated. Typically,the time spans for bleaching varied between 30 and 60 min forthe different species.

Separated carotenoids were measured directly on the HPTLCplate, with 488 nm resonant excitation wavelength. No differenceswere observed between spectra measured directly from powderedreferences that were applied to HPTLC plates as substrate andfrom solutions run in an HPTLC experiment. To avoid degradationof the samples, the laser intensity was adjusted to an intensity of∼102 W/cm2. Raman spectra from HPTLC spots were acquiredover a range from 350 to 4000 cm-1 using integration timesfrom 1 to 10 s.

RESULTS AND DISCUSSIONIn Situ Raman Spectra. The Raman spectra of pollen grains

from many plant species are dominated by typical carotenoidfeatures.6 This was found even in the case of 785 nm excitation(Figure 1 and ref 6), which is ∼330 nm away from typicalabsorption maxima of carotenoids. With this near-infrared (NIR)excitation, the spectra of the pollen remained unaltered at theapplied excitation intensity of 1.8 × 106 W/cm2, independent ofexposure time. The bands ascribed to carotenoid moleculesare not stable and are found to diminish steadily due to

photodestruction when the pollen grains are exposed to laserlight in the visible wavelength range.6 Therefore, in our recentRaman spectroscopic classification and identification studieswe eliminated carotenoid signatures by photodestruction with633 nm laser light and used the remaining nonbleachableconstituents of pollen for classification based on Raman spectracollected at 785 nm excitation. In contrast, in the in situexperiments reported here, we use the difference spectra ofthe unbleached and bleached pollen samples. This procedurecancels out the Raman signals of the nonbleached componentsin the pollen grain and can be used to analyze the spectralfeatures of the carotenoid pigment in different pollen species.The difference spectra, displayed for three tree species inFigure 2, reveal typical carotenoid signatures known from numer-

Table 1. AMD Parameters for HPTLC Separation of theCarotenoids Zeaxanthin, Cryptoxanthin, �-Carotene,and Lutein

runno.

tetrahydrofuranvol [%]

methylene chloridevol [%]

n-hexanevol [%]

solvent frontposition [mm]

1 10 90 0 182 9 91 0 223 8 92 0 264 7 93 0 305 6 94 0 346 5 95 0 387 4 96 0 428 3 97 0 469 0 80 20 5010 0 60 40 5411 0 40 60 5812 0 30 70 6213 0 20 80 6614 0 10 90 7015 0 0 100 74

Figure 1. Raman spectra of individual untreated pollen grains fromhorse chestnut, large-leaved linden, and sallow, excited with 785 nm(106 W/cm2, acquisition time 1 s).

Figure 2. In situ spectra of the carotenoid molecules contained inpollen grains from horse chestnut, large-leaved linden, and sallow.The spectra are differences of Raman spectra measured at thebeginning and end of an irradiation period with 633 nm laser light(excitation intensity 106 W/cm2).

8428 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

ous studies of purified carotenoid molecules.17,18,26,27 Interspeciesvariations of the in situ carotenoid spectra are observed for boththe position and the width of the bands: some modes vary inposition by more than 10 cm-1, such as the ν1 band in horsechestnut (1524 cm-1) and large-leaved linden (1538 cm-1). Thewidth of the ν1 band is much greater in the spectra fromhorse chestnut pollen compared to the two other species. Thefrequency of the ν2 mode, which is lower in the horse chestnutpollen as well compared to the other two species, is a thirddifference apparent in the spectra of Figure 2.

The variation in the in situ Raman spectra suggests differencesbetween different tree species in the carotenoid composition butalso in the structure and packaging of the carotenoid moleculeswith respect to other highly structured components of the pollencoats. Examples of the structures of some typical carotenoidsreported to occur in pollen are given in Figure 3. As was revealedin systematic studies, the Raman spectrum of a carotenoiddepends on the number of conjugated double bonds in its polyenicchain and on the substituents on the terminal ring structures.24,27

Example spectra of four reference carotenoids used in this studyare given in Figure 1 of the Supporting Information. In particular,the position of the ν1 frequency around 1520 cm-1 of the CdCstretching vibration can be used to monitor the degree ofconjugation through the π electron system.28 In addition, theposition of ν2 around 1160 cm-1 has been used to deduceinformation about the presence of CH3 groups, which may exertinfluence on the ν2 C-C vibration.28 In comparative studies of�-carotene, high frequencies of ν1 and ν2 were also associatedwith various cis isomers of the molecule.29 In the case of thepollen grains, conformational distortions or specific orientationsin the pollen coat matrix of a similar set of carotenoids couldoccur, dictated by the species-specific molecular and ultra-structure of the pollen coat. The integration of the carotenoidmolecules into the sporopollenin biopolymer matrix could leadto an altered excitation profile for all or some of the carotenoid

species. This would then result in changes of the in situcarotenoid spectrum. Veronelli et al. have discussed the spectraof carotenoids in parrots’ feathers in terms of an effectiveconjugation of the pigments, based on their in vivo surround-ings.24 Binding of carotenoids to proteins can also be identifiedfrom their RR spectrum.30 However, as there is no informationso far on the type of association of the carotenoids with thesporopollenin polymer of the outer pollen wall, the extent ofits influence on the pigment spectra has to remain speculativeat this point.

Independent of the association or conformation of particularcarotenoids with their biological matrix, it is very likely thatdifferences in the spectra arise from the variations betweendifferent carotenoid species that are present at the same time inthe pollen grains of different plant species. On the basis of effectiveconjugation coordinate (ECC) theory, which correlates the posi-tion of the ν1 CdC stretching and the number of double bondspresent in the molecule,31 it can be concluded that all caro-tenoid molecules must contain a similar number of doublebonds and ∼40 carbon atoms. Hence, variation of the ν1

frequencies (Figure 2) is likely to result from varying substituentson the ring structures such as presence or absence of alcoholicor carboxylic groups in the vicinity of the conjugated bonds. Inhorse chestnut pollen (Figure 2), such molecules could bezeaxanthin and lutein, both known to occur in pollen of otherspecies9 and to display down-shifted ν1 and ν2 frequenciescompared to those of �-carotene.24,28 In our comparisons withthe spectra from reference molecules (see Supporting InformationFigure 1), ν3 has a lower frequency in other carotenoids, e.g.,cryptoxanthin, and both ν1 and ν2 are lowered in zeaxanthincompared to the position in �-carotene.

Carotenoid Composition of Pollen as Obtained by HPTLC.Molecules such as �-carotene, lutein, violaxanthin, zeaxanthin, andantheraxanthin have been reported to occursin varying pro-portionssin several pollen species.9 However, analyses on thecarotenoid composition of pollen based on molecules isolateddirectly from the pollen grains9 rather than from other planttissues9,11-13 are relatively rare. As discussed above, the in situobtained Raman spectra can be assumed to represent all caro-

(26) Connors, R. E.; Burns, D. S.; Farhoosh, R.; Frank, H. A. J. Phys. Chem.1993, 97, 9351–9355.

(27) Weesie, R. J.; Merlin, J. C.; Lugtenburg, J.; Britton, G.; Jansen, F.; Cornard,J. P. Biospectroscopy 1999, 5, 19–33.

(28) Merlin, J. C. Pure Appl. Chem. 1985, 57, 785–792.(29) Koyama, Y.; Takii, T.; Saiki, K.; Tsukida, K. Photobiochem. Photobiophys.

1983, 5, 139–150.(30) Merlin, J. C. J. Raman Spectrosc. 1987, 18, 519–523.(31) Rimai, L.; Heyde, M. E.; Gill, D. J. Am. Chem. Soc. 1973, 95, 4493–4501.

Figure 3. Structure of (a) R-carotene, (b) �-carotene, (c) lutein, (d) zeaxanthine, (e) cryptoxanthine (all trans isomers).

8429Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

tenoid molecules that were present in the pollen grains at thetime of measurement. In order to interpret the interspeciesvariation of the in situ carotenoid signatures, we studied thecarotenoid composition of the concerned fresh pollen by acombination of HPTLC and resonant Raman spectroscopy.

Although a single TLC run led to insufficient separation,utilization of intensive automated multiple development (AMD)optimization experiments, which resulted in a final gradient withdecreasing polarity (details in Table 1), turned out to be successfulin separation of the carotenoids (Figure 4). Figure 4 illustratesthe resulting resolution for a mixture of reference carotenoids.The separation achieved using HPTLC/AMD was sufficient todistinguish between zeaxanthin and lutein, which differ only inthe steric configuration of one oxygen atom and one double bond(compare also the molecular structures displayed in Figure 3).The separation by TLC of compounds with a similar structure asis the case for these carotenoids has been known to be difficult,32

and also with high-performance liquid chromatography (HPLC)the separation of lutein and zeaxanthin was rather insufficient.15

Figure 5 displays the densitograms of the different carotenoidspecies after separation by HPTLC. In addition to the plant speciesthat were studied by in situ Raman experiments, three morespecies were analyzed by HPTLC/AMD. A schematic for qualita-tive comparison is given in Figure 2 of the Supporting Information.Table 2 contains the results and assignments of the observedretention factors to four reference molecules. As can be seen fromthese data, various different carotenoids were found in extractsfrom the pollen samples of horse chestnut, sallow, large-leavedlinden, and mahaleb cherry. There exist overlaps in the carotenoidpalettes of the extracts from the different species. In particular,the carotenoid compositions of pollen from sallow and mahalebcherry resulted in very similar chromatograms (see Figure 5 andTable 2). The HPTLC/AMD data also indicated that molecularvariety of the carotenoids was greatest in the horse chestnut pollenextract. This finding supports the assumption that the broadspectral features in the in situ Raman spectra of horse chestnutpollen are indeed caused by a great variety of different molecules.On the other hand, the HPTLC data also show that it is unlikelythat the presence of zeaxanthin, lutein, and cryptoxanthin causesthe great difference in the in situ Raman spectra of the threespecies shown in Figure 2, as at least two of these carotenoidsare contained in each of the trees (Table 2).

Quantitative analysis of the carotenoids on the HPTLC platewas performed based on UV-vis absorption spectra of referencecarotenoids. Upon extraction from the same amount of pollenmaterial, carotenoid concentration varied greatly in the differentextracts. No carotenoids were detected in the sample obtainedfrom European ash. �-Carotene and lutein were found in allsamples. The amounts varied from 4.7 to 17.5 ppm for �-caroteneand from 1.9 to 4.3 ppm for lutein. Cryptoxanthin was found insallow, mahaleb cherry, and large-leaved linden with concentra-tions ranging from 4.3 to 9.5 ppm. Zeaxanthin was not detectedin large-leaved linden, but in the other pollen species it could befound with an average amount of 5.7 ppm.

Apart from their use for quantification on the HPTLC plate,the UV-vis absorption spectra also contain information about thecarotenoids’ structure. In addition to the characteristic transitionaround 450 nm, the spectra of some of the molecules, e.g., thoseof �-carotene (Rf ∼ 0.94), showed an absorption maximumaround 350 nm (examples shown in Supporting InformationFigure 3). The signal around 350 nm is helpful to decide whichstereoisomer of the molecule is present, as it is characteristic ofcarotenoid cis isomers.33 The indication of both cis- and trans-specific UV-vis spectra in different spots of the same Rf valueimplies that the stereoisomers of �-carotene are present invarying amounts in the different pollen species that wereinvestigated. However, the UV-vis data are hampered byextremely low signals.

Raman Spectroscopy on HPTLC-Separated CarotenoidSpecies. For comparison with the Raman spectra collected fromintact pollen grains, Raman spectral data were also obtained fromextracted carotenoid after chromatographic separation. In general,the combination of TLC with Raman spectroscopy has beenproposed before to analyze amino acids and other organicmolecules. In most of these cases, excitation with NIR to avoidfluorescence from the TLC plate was identified as optimumcondition,34,35 and it was also proposed to enhance the weakRaman signal.36,37 Here, we apply 488 nm excitation. Due to theintense resonant Raman signal it was possible to conduct Ramanexperiments with the molecules separated on the plate after theHPTLC run. As fading of the spots occurred extremely fast,selection of the positions on the HPTLC plate had to rely on visualinspection. Figure 4 in the Supporting Information illustrates someexamples of spots where spectra could be collected.

Figure 6 displays spectra of carotenoids with different retentionfactors for the species horse chestnut and large-leaved linden. Inaddition to the typical carotenoid fingerprint between 1000 and1600 cm-1, the high-frequency region of the resonant Ramanspectra reveals several bands. They can be assigned toovertones of ν2 at ∼2320 cm-1 and of ν1 at ∼3045 cm-1, as wellas to the combinations ν1 + ν3 (∼2530 cm-1) and ν1 + ν2

(∼2690 cm-1).20

The positions of the ν1 and ν2 bands in the spectra of theseparated carotenoids vary greatly for the different retention

(32) Schoefs, B. Trends Food Sci. Technol. 2002, 13, 361–371.

(33) Zechmeister, L.; Polgar, A. J. Am. Chem. Soc. 1943, 65, 1522–1528.(34) Everall, N. J.; Chalmers, J. M.; Newton, I. D. Appl. Spectrosc. 1992, 46,

597–601.(35) Rau, A. J. Raman Spectrosc. 1993, 24, 251–254.(36) Istvan, K.; Keresztury, G.; Szep, A. Spectrochim. Acta, Part A 2003, 59,

1709–1723.

Figure 4. HPTLC densitogram obtained from a mixture of carotenoidstandards by automated multiple detection (AMD). The detectionwavelength was 425 nm.

8430 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

factors. In Figure 6A, this is illustrated for the example of thehorse chestnut pollen extract. The position of the ν1 band,adopting positions between 1528 and 1536 cm-1 in the spectrafrom the purified molecules, supports the observation of thevery wide ν1 band in the in situ spectrum (compare also withFigure 2). However, no spectrum was obtained from fractionswith ν1 frequencies as low as the frequency observed in the insitu measurements. We conclude that, in the extracts preparedfor HPTLC, these molecules were not contained in sufficientconcentration and/or had decomposed on the plate before theRaman measurement. From overall carotenoid spectra obtainedfrom the extracts, we can see that the difference between thein situ information and the spectra of the HPTLC-separatedspecies must have occurred already during the extraction

procedure (Supporting Information Figure 6), as is revealed by,e.g., a difference in ν1 frequency between the in situ and theextract spectra. For the pollen of sallow (not shown) and large-leaved linden (Figure 6B) mainly carotenoid molecules withhigher ν1 frequency were observed by Raman spectroscopy onthe HPTLC plates. On the basis of the absence of carotenoidswith lower ν1 frequencies there, we assume that carotenoidswith the ν1 signal at smaller wavenumbers contribute more tothe in situ spectrum of horse chestnut pollen than to the insitu spectra of these other species. The ν1 position in the insitu spectra agrees well with the positions obtained for theextracted molecules (compare bottom spectrum in Figure 2with Figure 6B).

(37) Sequaris, J. M. L.; Koglin, E. Anal. Chem. 1987, 59, 525–527.

Figure 5. HPTLC densitograms of the extracted carotenoids from (A) large-leaved linden, (B) sallow, (C) horse chestnut, (D) mahaleb cherry,and (E) tree of heaven. The detection wavelength was 425 nm.

8431Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

The identity of molecules that show the same retention factorsis also confirmed by their identical Raman spectra. Figure 7

displays the spectra of carotenoids from four different plantspecies, with Rf ) 0.94. These molecules were identified as�-carotene based on the retention factor of the reference. Thepositions of the C-C (ν2) at 1164 cm-1 and the CdC (ν1)stretch at 1533 cm-1 are identical for these samples. Thespectra suggest that �-carotene in the investigated spots ispresent as trans isomer.20 This is in accord with the observationthat trans isomers of carotenoids are more stable and thereforeoccur more often in nature.13 However, as mentioned above,directly after the HPTLC run, the presence of cis isomers hadbeen observed by UV-vis spectroscopy (Supporting Informa-tion Figure 3). We conclude that the same HPTLC spot Rf ) 0.94contained both cis- and trans-�-carotene and that the cis fractionhad decomposed in the HPTLC spots very fast before the timeof the Raman measurement.

Table 2. Averaged Content and Standard Deviations of�-Carotene, Lutein, Zeaxanthin, and Cryptoxanthin inExtracts of Six Pollen Species Determined byDensitometry

pollen speciesretention

value assignmentaverage[ppm]

no. ofrepetitions

horse chestnut 0.090.170.21 zeaxanthin 3.9 ± 0.9 20.27 lutein 2.5 ± 1.2 40.520.600.640.750.810.94 �-carotene 4.7 ± 0.8 4

sallow 0.21 zeaxanthin 7.0 ± 0.2 20.27 lutein 3.5 ± 0.7 20.56 cryptoxanthin 9.6 ± 2.5 20.640.700.760.830.94 �-carotene 6.8 ± 1.4 4

mahaleb cherry 0.060.22 zeaxanthin 6.4 ± 0.8 20.28 lutein 4.3 ± 0.4 20.56 cryptoxanthin 4.3 ± 1.3 40.640.690.750.820.94 �-carotene 17.5 ± 0.8 4

large-leaved linden 0.32 lutein 1.9 ± 0.14 20.53 cryptoxanthin 7.9 ± 1.2 20.720.760.820.910.95 �-carotene 6.1 ± 0.71.02

tree of heaven 0.91 �-carotene n.d.aEuropean ash b

a Amount of �-carotene was too low for quantification. b No signalscould be detected by densitometry.

Figure 6. Raman spectra of carotenoids extracted and separated by HPTLC from (A) horse chestnut pollen and (B) large-leaved linden pollen.The respective retention factors are indicated. Excitation wavelength was 488 nm (∼102 W/cm2); accumulation time was 1 s. Assignments arebased on ref 20. The signal marked with the asterisk in panel A originates from the HPTLC plate.

Figure 7. Raman spectra of carotenoids extracted from horsechestnut, mahaleb cherry, large-leaved linden, and sallow pollen witha retention factor of 0.91 (excitation wavelength 488 nm, accumulationtime 1 s, ∼102 W/cm2).

8432 Analytical Chemistry, Vol. 81, No. 20, October 15, 2009

As can be concluded from Figure 6, no trend or correlation isfound for the change in position of either the ν1 or ν2 signalswith Rf of the respective molecule. As illustrated by the RRspectra of horse chestnut pollen extract (Figure 6A), twocarotenoids can have very different retention factors yet verysimilar Raman spectra; compare, e.g., the spectra obtained formolecules with Rf ) 0.09 and Rf ) 0.81 (Figure 6A). The findingagrees with the different molecular qualities that are revealed byboth methods. Subtle changes in side chains and molecular groupsaffect HPTLC separation but do not necessarily influence effectiveconjugation and therefore have much less impact on the RRspectrum. On the other hand, the advantage of the Raman analysisis certainly that structural information can be obtained regardlessof the availability of reference molecules and a priori assumptions.

In order to further utilize the combination of HPTLC and RRspectroscopy in the study of carotenoids, faster methods such asautomated scanning of HPTLC plates and online detection willbe employed.

Considering the disadvantages of absorption spectroscopy, RRspectroscopy could serve for sensitive detection and a simulta-neous structural characterization and may help to characterizecarotenoid from pollen and many other types of complex samples.

CONCLUSIONSAs our data indicate, the in situ Raman spectra of the

carotenoid molecules measured in single intact pollen grainsprovide in situ evidence of interspecies variations in pollencarotenoid content, structure, and/or assembly without priorpurification. The results from HPTLC confirmed that carotenoidcomposition varied greatly between species and that the differentin situ Raman spectral signatures reflect indeed differences incarotenoid composition.

Moreover, comparison of the HPTLC data with resonantRaman spectra measured in situ illustrated that RR spectroscopyis very sensitive compared to approaches that need to rely onextraction procedures: For example, although the presence oftypical carotenoid bands in the Raman spectra from single intactpollen grains from European ash had verified the presence of themolecules (see Figure 5 of the Supporting Information), nocarotenoids could be detected in the extract of this pollen thatwas generated from several milligrams of material (Table 2).

Although the extraction process and application to the platesis time-consuming and prone to loss of material, no molecule islost unanalyzed in the in situ Raman experiment. Utilization of anextraction-free procedure in carotenoid analysis such as in situ

Raman microspectroscopy also circumvents the problem ofcontrolling the concentrations of carotenoids in such extracts. Asecond problem with purification is the presence of irrelevantmaterial such as parts of anthers that can contaminate preparationsif large amounts of pollen material need to be generated forextraction. It should be noted that only very few carotenoid studieson pure pollen material have been conducted due to complicatedpreparation and extraction.

Raman data obtained from carotenoid species separated byHPTLC provide evidence that the in situ Raman difference spectrareally represent an average of the overall carotenoid constitutionin pollen. The chromatograms and the HPTLC-RR spectraindicate that pollen species with very broad in situ carotenoidsignals contain a great variety of carotenoids with varyingcharacteristic frequencies. A full interpretation of average pollencarotenoid signals should in principle be achieved by an RRexploration of the plethora of pollen carotenoids, e.g., upon HPTLCseparation. However, the influence of the sporopollenin matrixon an in situ spectral signature is not considered in such apurification-based approach.

Therefore, comparing the results of the different analyticalapproaches and taking into account the great instability ofcarotenoid molecules, the Raman spectral information obtainedin situ during the photodepletion process is clearly of advantage.In particular, the in situ difference spectra already include spectralfeatures that may arise due to association of the carotenoidmolecules with the biological matrix. A database of the species-specific in situ carotenoid spectra could in the future be used forautomated spectrum correction and identification. This wouldfulfill an important prerequisite for fast analysis of pollen, and alsoof other plant tissues, e.g., in quality control of produce.

ACKNOWLEDGMENTFunding of the project by Deutsche Forschungsgemeinschaft

(GrantDFGKN557/9-1andPA716/9-1) isgratefullyacknowledged.

SUPPORTING INFORMATION AVAILABLERaman spectra of carotenoid extract before HPTLC separation,

untreated pollen grains, photographs of the HPTLC plate, sche-matic for qualitative comparison of HPTLC densitograms, andUV-vis spectra. This material is available free of charge via theInternet at http://pubs.acs.org.

Received for review June 25, 2009. Accepted August 26,2009.

AC901389P

8433Analytical Chemistry, Vol. 81, No. 20, October 15, 2009