Journal of Chemical Ecology, Vol. 30, No.

9, September 2004 ( C 2004)

SEASONAL VARIATION IN THE CONTENT OF HYDROLYZABLE TANNINS, FLAVONOID GLYCOSIDES, AND PROANTHOCYANIDINS IN OAK LEAVES

JUHA-PEKKA SALMINEN,1 TOMAS ROSLIN,2, MAARIT KARONEN,1 JARI SINKKONEN,3 KALEVI PIHLAJA,1,3 and PERTTI PULKKINEN4
1 Laboratory of Environmental Chemistry Department of Chemistry, University of Turku FI-20014 Turku, Finland 2 Metapopulation Research Group Department of Biological and Environmental Sciences P.O. Box 65 (Viikinkaari 1), University of Helsinki FI-00014 Helsinki, Finland 3 Structural

Chemistry Group, Department of Chemistry University of Turku, FI-20014 Turku, Finland 4 Finnish Forest Research Institute Haapastensyrj Breeding Station a Karkkilantie 247, FI-12600 L yli inen, Finland a a

(Received August 26, 2003; accepted May 23, 2004)

AbstractOaks have been one of the classic model systems in elucidating the role of polyphenols in plantherbivore interactions. This study provides a comprehensive description of seasonal variation in the phenolic content of the English oak (Quercus robur). Seven different trees were followed over the full course of the growing season, and their foliage repeatedly sampled for gallic acid, 9 individual hydrolyzable tannins, and 14 avonoid glycosides, as well as for total phenolics, total proanthocyanidins, carbon, and nitrogen. A rare dimeric ellagitannin, cocciferin D2 , was detected for the rst time in leaves of Q. robur, and relationships between the chemical structures of individual tannins were used to propose a biosynthetic pathway for its formation. Overall, hydrolyzable tannins were the dominant phenolic group in leaves of all ages. Nevertheless, young oak leaves were much richer in hydrolyzable tannins and avonoid glycosides than old leaves, whereas the opposite pattern was observed for proanthocyanidins. However, when quantied as individual compounds, hydrolyzable tannins and avonoid glycosides showed highly variable seasonal patterns. This large variation in temporal trends among compounds, and a generally weak correlation between the concentration of any individual compound and the total

To whom correspondence should be addressed. E-mail: tomas.roslin@helsinki.

1693
0098-0331/04/0900-1693/0
C

2004 Springer Science+Business Media, Inc.

1694

SALMINEN ET AL.
concentration of phenolics, as quantied by the FolinCiocalteau method, leads us to caution against the uncritical use of summary quantications of composite phenolic fractions in ecological studies. Key WordsQuercus robur, hydrolyzable tannins, ellagitannins, avonoid glycosides, proanthocyanidins, HPLC, quantication of phenolics, biosynthetic pathways, compound-specic patterns.

INTRODUCTION

Oaks in the genus Quercus have long been a popular target for ecologists studying the chemical interplay between plants and herbivorous insects (e.g., Feeny, 1970; Faeth, 1986; Rossiter et al., 1988; Tikkanen and Julkunen-Tiitto, 2003). One of the earliest and still most inuential ecological explorations of oak leaf chemistry was conducted by Feeny (1970). In his seminal paper, he related the phenology of an oak-feeding moth (Operopthera brumata) to seasonal variation in the chemical contents of oak leaves. By quantifying relatively crude fractions of phenolic compounds, he inferred that the nutritional value of oak leaves declines throughout the summer, and that this may be the ultimate factor causing spring-feeding in O. brumata. Phenolics may well play a central part in the oaks defense against its herbivores (e.g., Schultz and Baldwin, 1982; Rossiter et al., 1988). However, past phytochemical and experimental studies leave no doubt that individual phenolic compounds vary substantially with respect to biological activity (Zucker, 1983; Ozawa et al., 1987; Clausen et al., 1990; Ayres et al., 1997; Feldman et al., 1999; Kilkowski and Gross, 1999; Kraus et al., 2003) and that even chemically closely related compounds may encounter a different fate in the digestive tract of an insect (e.g., Salminen and Lempa, 2002). Hence, to understand the interplay between the oak and its herbivores, it seems preferable to measure the concentrations of individual compounds. The quantication of composite phenolic fractions will only make sense if (a) concentrations of individual phenolics vary in concert and (b) the quantied compounds have matching biological activitiespresumably a rare combination in real organisms. So far, a relatively large number of compound-specic studies have investigated the hydrolyzable tannin composition of oak wood (e.g., Masson et al., 1994; Viriot et al., 1994; Conde et al., 1998; Mosedale et al., 1998; Fern ndez de a Sim n et al., 1999; M mmel et al., 2000; Cadaha et al., 2001). This interest has o a a primarily been spurred by the needs of the wine industry, as most wine barrels are made of English oak (Quercus robur) or sessile oak (Q. petraea). Of the named hydrolyzable tannins, the heartwood of Q. robur has been found to contain castalagin, vescalagin, grandinin, and roburins AE (Herv du Penhoat et al., 1991a,b; e Masson et al., 1994; Vivas et al., 1995).

SEASONAL VARIATION IN OAK LEAF CONTENTS

1695

Substantially less attention has been aimed at the phenolic composition of oak leaves. A handful of earlier studies have shown the foliage of Q. robur to contain pedunculagin, castalagin, vescalagin, casuarictin, and avonol glycosides (Scalbert and Haslam, 1987; Scalbert et al., 1988; Grundh fer et al., 2001). Howo ever, this information has only rarely been utilized in ecological studies that try to relate the polyphenolic composition of oak leaves to the performance of folivorous insects. Most often, only total phenolics or the total concentrations of summary tannin groups have been quantied (e.g., Feeny, 1970; Faeth, 1986; Rossiter et al., 1988; Lill and Marquis, 2001; Abrahamson et al., 2003; Forkner et al., 2004; but see Tikkanen and Julkunen-Tiitto, 2003), and how well such composite measures will reect variation at the level of individual compounds has remained an open question. In this paper, we provide the rst comprehensive description of seasonal variation in the phenolic contents of oak leaves as quantied at the level of both individual compounds and two summary groupstotal phenolics and total proanthocyanidins. For the rst time, we also report the presence of a rare dimeric ellagitannin, cocciferin D2 , in the foliage of Q. robur and propose a biosynthetic pathway for its formation. On the basis of seasonal patterns observed in individual compounds, we ask: (1) To what extent can seasonal variation in the concentrations of individual compounds be described by a single summary measure such as the total concentration of phenolic compounds? (2) How well is variation among compounds explained by biochemical connectionswill the knowledge of central metabolic pathways help us understand temporal changes in concentrations of individual compounds? (3) What does variation in phenolic content and variation in the total content of carbon and nitrogen reveal about changes in the nutritional value of oak leaves over the season?
METHODS AND MATERIALS

Study Object. The deciduous English oak is native to Europe and Western Asia (Jalas and Suominen, 1976). In Finland, its natural distribution is limited to the southernmost part of the country, where new oak leaves are produced in late May and early June (Hoffman and Lyr, 1973; Niemel and Haukioja, a 1982). Although the period of active leaf production is relatively short and wellsynchronized in oak as compared to other tree species (Niemel and Haukioja, a 1982), its onset may vary by several weeks between both years and trees (cf. Crawley and Akhteruzzaman, 1988). Within trees, leaf production is apparently more synchronized on short shoots than on actively growing long shoots (Erkki Haukioja, personal communication, 2003). As a result, leaves on short shoots will be of similar age throughout the summer; only in some oak individuals is there a second growth period in August, resulting in so-called lammas shoots with new leaves. All leaves typically senesce in October, but variation among trees and

1696

SALMINEN ET AL.

years is again substantial. While most leaves are shed in the autumn, some trees retain a high proportion of dead leaves until the following spring. Study Site. In the spring of 2001, seven oaks were selected within a small (about 0.1 ha) stand of 21 oaks growing at the Haapastensyrj Tree Breeding a Center (60 36 N, 24 26 E), run by the Finnish Forest Research Institute. These oaks had been planted in 1978 and were now approximately 78 m in height. Each tree was originally created by grafting a small oak twig to a sapling reared from a randomly collected acorn. Any twigs produced by the original sapling were later pruned, and hence the current canopy of the oak consists of the grafted genotype only. The grafted twigs had been collected in oak stands across the full Finnish range of the English oak. Leaf Sampling. Leaf samples were collected from each of the seven oaks on 11 dates throughout the summer of 2001. The sampling dates ranged from May 29, which was the 1st d when each tree had leaves larger than 1 cm, to September 26, when the leaves were already senescing. Individual sampling dates were May 29, June 7, June 15, June 24, July 7, July 16, July 27, August 16, August 31, September 13, and 26. All samples were collected between 8 and 12 A.M. A similar number of leaves (730, depending on the date and leaf size) were randomly collected from the lower branches of each tree. To reduce variation among leaves within a sample, we specically avoided the hard and waxy sun leaves of the outer canopy (cf. Feeny, 1970) as well as leaves on actively growing long shoots. Leaves were sealed in polyethylene bags in the eld and placed into a cooler with ice. Upon return to the laboratory, they were air-dried for 3 d at room temperature in a ventilated fume cupboard, and samples were weighed and subsequently stored at 18 C until ground into a ne powder before extraction. Although air-drying may not be the optimal drying method, it is known to alter the levels of hydrolyzable tannins and avonoid glycosides only minimally in birch leaves (Salminen, 2003, unpublished data) or oak leaves (Salminen, unpublished pilot study) as compared to levels observed in freeze-dried samples. For each sample, the average biomass per leaf was calculated by dividing the total weight of the sample by the number of leaves it contained. Extraction. Dried and ground oak leaves (200 mg per sample) were extracted four times (4 1 hr) with 70% aqueous acetone (4 8 ml) on a planary shaker. The freeze-dried aqueous phase of the extract was dissolved in water (3 2 ml); the supernatant of the centrifuged (10 min at 2000 g) sample was ltered through a 0.45-m PTFE lter and kept frozen at 20 C until analyzed with HPLC-DAD or HPLCESI-MS. Analysis of Phenolics with HPLC-DAD and HPLCESI-MS. HPLC-DAD analysis of oak leaf extracts was performed at 280 and 349 nm with MerckHitachis LaChrom HPLC system (Merck-Hitachi, Tokyo, Japan). Column and chromatographic conditions were as described earlier (Salminen et al., 1999),

SEASONAL VARIATION IN OAK LEAF CONTENTS

1697

except that 0.1 M H3 PO4 was replaced with 0.05 M H3 PO4 . Phenolic compounds were quantied using pedunculagin, 1-O-galloylglucose, gallic acid, quercetin, and kaempferol as external standards. HPLCESI-MS analysis was performed using a Perkin-Elmer Sciex API 365 triple quadrupole mass spectrometer (Sciex, Toronto, Canada) equipped with an ion-spray interface. The HPLC system consisted of two Perkin-Elmer Series 200 micro pumps (Perkin-Elmer, Norwalk, CT, USA) connected to a Series 200 autosampler. The column used and chromatographic and ESI-MS conditions were as described previously (Salminen et al., 1999). Isolation and Identication of Cocciferin D2 . Cocciferin D2 (120 mg) was isolated from oak leaves with a combination of Sephadex LH-20 (40 2.5 cm i.d.) and Merck LiChroprep RP-18 (44 3.7 cm i.d., 4063 m) columns following the methods outlined by Salminen et al. (1999, 2001). Part of the pure isolate was partially hydrolyzed in mild conditions (40 C water, 1 hr) and the reaction products analyzed by HPLC-DAD and HPLCESI-MS. The NMR spectra of Cocciferin D2 were acquired using a JEOL JNM-A-500 spectrometer operating at 500.16 MHz for 1 H and 125.78 MHz for 13 C. Spectra were recorded at 25 C using acetone-d6 as a solvent. In addition to standard proton and carbon spectra, DEPT, DQF-COSY, HMQC, and HMBC spectra were measured. Analyses of Total Phenolics, Proanthocyanidins, Carbon, and Nitrogen. The total phenolic content of the extracts was determined by a modication of the FolinCiocalteau method (Nurmi et al., 1996), using a Perkin-Elmer Lambda 12 UVVIS spectrometer (Norwalk, CT, USA). Three replicates of each sample were analyzed and their average was used as the nal reading. A standard curve was prepared on the basis of known concentrations of gallic acid. The total content of proanthocyanidins was measured with the butanol-HCl assay as in Ossipova et al. (2001). Again, measurements were based on the average reading of three replicate samples, and a standard curve prepared on the basis of known concentrations of puried birch leaf proanthocyanidins. The total concentration of carbon and nitrogen in the leaves was performed with a Perkin-Elmer Series II CHNS/O Analyzer 2400 (Norwalk, CT, USA). A subset of six samples was included in this analysis, corresponding to sampling dates May 29, June 7, June 15, July 7, August 16, and September 26. Data Analysis. Seasonal changes in leaf chemistry were described by visual plots of compound-specic patterns. To further evaluate the extent to which the concentrations of different compounds vary in unison among sampling dates, we calculated simple Spearman rank correlation coefcients (rS ). All trees were rst ranked within compounds and sampling dates, from the tree richest in this particular compound to the tree with the lowest concentration. Then, two types of comparisons were made on the basis of samples taken on the date with the highest phenolic readings (May 29). First, to illustrate the concordance between

1698

SALMINEN ET AL.

the concentration of an individual compound and the pooled concentration of larger phenolic groups, correlation coefcients were calculated for the rank of the compound in question and the rank of the pooled concentrations of all individual compounds as quantied by HPLC. Second, to depict the consistency between the concentration of an individual compound and the total concentration of phenolics as quantied by a summary method, correlation coefcients were calculated for a trees rank for the compound in question and its rank for total phenolics as quantied by the FolinCiocalteau method. From an applied perspective, what we ask is specically: if we know the precise concentrations of individual compounds, will patterns at the level of total phenolics summarize patterns at the level of individual compounds?

RESULTS

Characterization of Phenolic Compounds. Twenty-four phenolic compounds were detected in the HPLC-DAD analyses. From their UV spectra, two were preliminarily identied as gallic acid or its derivatives (compounds 12; Salminen et al., 1999), and eight as ellagitannins (310; Salminen et al., 1999). The rest of the compounds were classied on the basis of both UV and mass spectral characteristics as quercetin (1119; Ossipov et al., 1995, 1996) or kaempferol glycosides (2024; Ossipov et al., 1995, 1996). Since tannins are generally supposed to play a more important role in plant herbivore interactions than avonoid glycosides, we focused more on the identication of compounds 110 than on 1124. On the basis of their retention times (Rt s; compared to those given in the literature) and molecular masses recorded from a negative ion HPLCESI-MS run, the structures of 15 were identied as gallic acid (1, 6.7 min, 170 g/mol), 1-O-galloylglucose (2, 6.4 min, 332 g/mol), tellimagrandin II (3, 19.2 min, 938 g/mol), casuarictin (4, 18.0 min, 936 g/mol), and pedunculagin (5, 9.4 and 11.7 min, anomeric mixture, 784 g/mol; Salminen et al., 2001). Ellagitannin 6 (14.3 min, 936 g/mol) could be either stachyurin or casuarinin; only one of these isomers was detected in the Q. robur extract, thus making it impossible to utilize the Rt data of Okuda et al. (1982) for differentiating between the two. In contrast, the other two isomeric ellagitannins, i.e., 7 (8.4 min, 934 g/mol) and 8 (10.2 min, 934 g/mol), were clearly detected and identied as vescalagin and castalagin, respectively (Fern ndez de Sim n a o et al., 1999; Zhentian et al., 1999). At this point, the structures of two ellagitannins, i.e., 9 (7.6 min, 1102 g/mol) and 10 (15.2 min, 1868 g/mol), remained unresolved. The mild hydrolysis of 10 yielded 5 and an unidentied ellagitannin with a molecular mass of 1084 g/mol; thus, 10 was shown to be a dimeric ellagitannin

SEASONAL VARIATION IN OAK LEAF CONTENTS

1699

with 5 as one of the monomeric constituents. The whole chemical structure of 10 was unravelled by a combination of NMR experiments. All proton and carbon chemical shifts, as well as the protonproton coupling constants, matched perfectly with those reported for cocciferin D2 , an ellagitannin recently found in the leaves of Q. coccifera and Q. suber (Ito et al., 2002). Therefore, 10 was identied as cocciferin D2 , now detected for the rst time in leaves of Q. robur. This dimeric ellagitannin consists of two monomeric units, i.e., 5 and castavaloninic acid. The latter of these compounds has a molecular mass of 1102 g/mol, corresponding to that of monomeric ellagitannin 9. Interestingly, according to Yoshida et al. (1992), the valoneoyl group of castavaloninic acid may undergo lactonization (loss of water) upon hydrolysis resulting in a depsidone molecule having a molecular mass of 1084 g/mol (see also hydrolysis of 10 above). This is exactly the same as that of a signicant fragment of 9 produced in negative ion HPLCESI-MS, which is known to fragment hydrolyzable tannins in a manner similar to chemical hydrolysis (cf. Salminen et al., 1999, 2001; Salminen, 2002). For these reasons, 9 was identied as castavaloninic acid. A Biosynthetic Pathway for the Hydrolyzable Tannins of Oak Leaves. By examining the relationships among the chemical structures of individual tannins identied in this study, we propose a biosynthetic pathway for their formation. It is well accepted (as reviewed by Gross, 1999) that the rst compound in the general hydrolyzable tannin pathway, 2, is formed from 1 and UDP-glucose, and that the galloylations then continue consecutively and position-specically to nally yield 1,2,3,4,6-penta-O-galloylglucose. On the other hand, the formation of the rst ellagitannin of the pathway, 3, directly from pentagalloylglucose was proven only recently (Niemetz et al., 2001). The subsequent biosynthetic steps from 3 onwards have not been experimentally proven, but it is generally assumed that e.g., 4 is a product of further oxidative coupling of two spatially adjacent galloyl groups of 3 (see, e.g., Helm et al., 1999). Furthermore, it has been suggested that the C-glycosidic ellagitannins 7 and 8 are formed from 5, and that this step contains at least stachyurin and casuarinin (6) as intermediates (see also Okuda et al., 1982; Hatano et al., 1986; Haslam, 1992; Vivas et al., 1995; Helm et al., 1999). We are not aware of any biosynthetic speculations about the origin of 9 and 10, but on the basis of their structures it seems safe to assume that 9 is formed by galloylation at the hexahydroxydiphenoyl (HHDP) group of 8, and that 10 is a product of dimerization of 5 and 9. On the basis of these considerations, we arrive at the biosynthetic pathway depicted in Figure 1. Seasonal Changes in Phenolic Contents. Throughout the summer, the phenolic contents of oak leaves were dominated by hydrolyzable tannins (Figure 2A and B). Although the pooled concentration of all hydrolyzable tannins declined by 54% between late May and September, their total concentration still overshadowed that of proanthocyanidins by a factor of 7.7 on September 26. The concentrations

1700

SALMINEN ET AL.

FIG. 1. The proposed biosynthetic pathway for the formation of ellagitannins in leaves of Quercus robur L. The biosynthetic steps involve (a) galloylation at C-1 of glucose; (b) four consecutive galloylation steps; (c) oxidative coupling between galloyl groups at C-4 and C-6; (d) oxidative coupling between galloyl groups at C-2 and C-3; (e) cleavage of galloyl group at C-1; (f) glucopyranose ring opening at C-1, followed by galloylation at C-5; (g) oxidative coupling between benzene rings at C-4 and C-5; (h) further galloylation thus forming a valoneoyl group at C-4/C-5; and (i) dimerization of pedunculagin and castavaloninic acid to form cocciferin D2 .

of proanthocyanidins showed an opposite seasonal pattern, steadily increasing over the summer from undetectable levels in the young leaves to an average of 10.8 mg/g in old leaves. The most rapid changes in leaf chemistry occurred during the early part of the summer, as the oak leaves grew and matured (Figure 2). In the majority

SEASONAL VARIATION IN OAK LEAF CONTENTS

1701

FIG. 2. Seasonal variation in leaf traits. Each data point represents the mean of seven trees, and vertical bars show standard errors. (A) Concentrations of total hydrolyzable tannins (sum of compounds 210) and proanthocyanidins (quantied by the butanol-HCl assay). (B) Concentrations of total phenolics (quantied by the FolinCiocalteau assay) and avonoid glycosides (sum of compounds 1124). (C) Total concentrations of carbon and nitrogen. (D) Biomass per dried oak leaf. (E and F) Concentrations of individual hydrolysable tannins and gallic acid. (G) Concentrations of individual kaempferol glycosides. (H and I) Concentrations of individual quercetin glycosides. Compounds numbered as in Figure 1.

1702

SALMINEN ET AL.

of individual compounds, there was a distinct peak in concentration around the time of maximum leaf expansion (Figure 2), followed by an extended period of relatively stable values. After July 7, when the leaves had reached their nal size, their phenolic composition remained more or less unchanged until leaf senescence (Figure 2). However, when inspected in more detail, different seasonal patterns emerge among individual compounds, with a few compounds peaking later in the season than the rest (notably 2, 9, and 10 among the tannins; Figure 2EI). Interestingly, the maximal concentrations of many phenolic compounds in late May and early June coincide with a minimum in the total amount of carbon, followed by a continuous buildup of carbon over the course of the summer (Figure 2C). For nitrogen, the pattern was exactly the opposite, with the level steadily decreasing over the summer (Figure 2C). Consistency Among Individual Compounds and Larger Phenolic Groups. Variation at the level of total phenolic content did not capture variation at the level of individual compounds: a sample rich in total phenolics, or in the pooled concentration of all individual compounds, was not necessarily rich in any individual compound (Figure 3). Some of the correlations between the concentration of an individual compound and the total concentration of a larger phenolic group are trivial because we are comparing an element to a sum of which it is a part. In the absence of any correction for multiple tests, chance alone is also expected to render one in every 20 results signicant. Still, among correlations between individual compounds and total phenolic content, only one out of 48 rS values was signicant, and most values were relatively low or even negative (Figure 3). At the level of the pooled concentrations of chemically similar compounds, only the concentration of hydrolyzable tannins was signicantly correlated with the pooled concentration of all individual compounds (and perfectly so; rS = 1.0). Finally, our two different measures of total phenolic contentstotal phenolics as quantied by the FolinCiocalteau method and pooled phenolics as calculated by pooling the HPLC readings of individual compoundswere not signicantly correlated with each other (rS = 0.71, N = 7, P = 0.07). Hence, we conclude that the rough quantication of a composite phenolic fraction tells us virtually nothing about how different trees rank compared to each other in terms of individual compounds.

DISCUSSION

Oaks have formed one of the classic model systems in elucidating the role of polyphenols in plantherbivore interactions. This study provides, to our knowledge, the most elaborate description of temporal variation in the phenolic contents of oak leaves conducted so far. As such, it adds substantial detail to the already classical image of oak leaf chemistry drawn by Feeny in 1970. However, several

SEASONAL VARIATION IN OAK LEAF CONTENTS

1703

FIG. 3. Spearman rank correlations (rS ) among different phenolic fractions in samples from May 29. PPh (i.e., Pooled Phenolics) shows the consistency in rank (rS ) between the concentration of individual compounds and the pooled concentration of all individually quantied phenolic compounds; FC (i.e., FolinCiocalteau) shows rS values for the compound in question compared to FolinCiocalteau readings of total phenolics. For each data point, the number shows the identity of the compound (numbers as in Figures 1 and 2). Data points labelled with letters refer to pooled concentrations of different phenolic subgroups; PFG = Pooled Flavonoid Glycosides (1124), PHT = Pooled Hydrolyzable Tannins (210), PQG = Pooled Quercetin Glycosides (1119), and PKG = Pooled Kaempferol Glycosides (2024). For all values of rS , N = 7 trees. The horizontal bars at the top and bottom of each panel shows the critical rS for P < 0.05 before (solid line) and after (stippled line) Bonferroni correction for 29 individual tests.

discrepancies with Feeny show the need for some substantial specications to current descriptions of oak leaf chemistry. In his seminal account of seasonal variation in oak leaf quality, Feeny (1970, p. 574) reported a general increase in the tannin contents of oak leaves over the course of the summer. On the basis of two-dimensional paper chromatography, he attributed this pattern to a progressing dominance of proanthocyanidins over hydrolyzable tannins, the levels of which appeared to remain approximately constant over time. Our results strongly contrast with those ndings. First, we

1704

SALMINEN ET AL.

found hydrolyzable tannins, not proanthocyanidins, to be the dominant group of phenolic compounds over the full course of the growing season. The peak levels of hydrolyzable tannins were strikingly highwhile heartwood of Q. robur can contain up to 10% of ellagitannins by weight (Scalbert et al., 1988), young oak foliage contained levels as high as ca 18% (see Figure 2A). Second, the concentration of hydrolyzable tannins did not remain stable over time, but dropped to less than half between late May and September. Early-summer maxima in the levels of hydrolyzable tannins have been reported from Q. robur (e.g., Tikkanen and Julkunen-Tiitto, 2003), from other oak species (e.g., Faeth, 1986; Rossiter et al., 1988; Maufette and Oechel, 1989), and from other trees (e.g., Riipi et al., 2002). In all these cases, the early peak in hydrolyzable tannins is associated with a late-summer peak in condensed tannins, suggesting a common pattern across species and upsetting Feenys (1970) initial notion. Yet, there is one important difference between our material and Feenys: where Feeny focused on the upper sun leaves of the canopy, we collected our leaf samples on the lower branches of the trees. As light levels have been found to affect foliar phenolics (e.g., Dudt and Shure, 1994), part of the observed difference may be due to differences in leaf exposure. This calls for further analyses of seasonal changes in different parts of the canopy. Feeny (1970) focused on seasonal patterns at the level of summary phenolic groups. When we examined patterns at the level of individual compounds, temporal changes at the level of pooled phenolic contents were found to mask variation in the concentration of individual hydrolyzable tannins and avonoid glycosides over time (Figures 2 and 3). The ultimate, evolutionary reasons for the observed variation in compound-specic patterns are largely unknown, as consensus has yet to be reached even regarding the exact biological roles of hydrolyzable tannins and avonoid glycosides (cf. Appel, 1993; Close and McArthur, 2002). Proximate reasons are better understood, as seasonal changes in the concentration of individual hydrolyzable tannins can sometimes be mapped onto proposed or established biogenetic pathways (cf. Hatano et al., 1986; Salminen et al., 2001). In Q. robur, differences in the way individual compounds changed over time appeared closely connected to their biosynthetic relationships. First, compound 2 was the only galloylglucose present in oak leaves in levels detectable with HPLC-DAD. This was surprising, since the biosynthetic pathway of hydrolyzable tannins also contains di-, tri-, tetra-, and pentagalloylglucoses before the rst ellagitannin, i.e., 3 (Figure 1; cf. Gross, 1999; Niemetz et al., 2001; Salminen et al., 2001). Evidently the synthesis of ellagitannins in oak is effective enough not to let these galloylglucose intermediates accumulate in detectable amounts. Consistent with this view of active ellagitannin synthesis, the concentration of 2 (i.e., the compound from which a galloyl group is utilized in the formation of other galloylglucoses and ellagitannins therefrom) decreased dramatically after June 15 (see Figure 2F).

SEASONAL VARIATION IN OAK LEAF CONTENTS

1705

Second, temporal variation in the concentrations of compounds 7, 8, 9, and 10 appears intimately interrelated (Figure 2). Although all the other ellagitannins showed almost linear reductions in their concentrations already from May 29 onwards, ellagitannin 9, i.e., the monomeric building block of 10, showed an increase until June 7, and its biosynthetic successor 10 until June 15. Simultaneously, the concentration of 8, the building block of 9, showed a more steep decrease than that of 7. This was presumably because 8 was utilized further in the biosynthesis of oak leaf ellagitannins (for formation of 9 and 10) unlike its isomer 7 (see Figure 1). For the same reason, the concentration of 7 was approximately four times as high as that of 8 throughout the season (e.g., May 29; 57.6 mg/g vs. 14.0 mg/g). However, by summing the concentration of 8 with that of its biosynthetic successors 9 and 10, almost equal values were obtained as with 7 only (e.g., May 29; 55.8 mg/g vs. 57.6 mg/g, respectively). This implies that oak leaves invest a closely similar amount of resources in the production of 7 and 8, although this cannot be seen in their foliar levels as such. Importantly, the potential for such invisible investment in a given compound suggests that rm conclusions on the specic pattern of resource allocation in oak trees cannot be reached until compounds are quantied individually and biosynthetic pathways unravelled. Considering the proposed biogenetic pathway for Q. robur leaves (Figure 1), it is intriguing to note that it does not contain grandinin or roburins AE compounds that occur as biosynthetic successors of 7 and 8 in the heartwood of this species (Herv du Penhoat et al., 1991a,b; Masson et al., 1994; Vivas et al., e 1995). At the same time, ellagitannins 9 and 10 have not been encountered in the heartwood, but have now been found for the rst time in leaves of Q. robur. Therefore, it is evident that enzymatic systems controlling the directions of the ellagitannin pathways onwards from e.g., 7 and 8 differ even between plant parts, not just among plant species as stated by Hatano et al. (1992). Interestingly, Scalbert et al. (1988) noted that the proportions of 7 and 8 vary between samples from different parts of Q. robur, the former being predominant in the leaves and the latter in the wood. However, they were unable to identify one major peak from the HPLC chromatogram of the leaf extract, thus possibly neglecting the presence of the dimeric ellagitannin 10. If the unidentied peak was indeed due to 10, that might explain the lower level of 8 in the foliage. Nevertheless, the observed differences in ellagitannin synthesis pathways caution against uncritical generalizations among different plant tissues when interpreting, e.g., patterns of seasonal change in phenolic composition. Variation in oak leaf chemistry will affect a broad range of oak-associated taxa, including herbivorous mammals and arthropods, pathogenic and endosymbiotic fungi, and other microorganisms. Among Lepidoptera alone, more than 200 species feed on oak leavesa gure higher than for any other European tree species (Feeny, 1970). The majority of these species attack young oak leaves in the spring (Feeny, 1970; Niemel and Haukioja, 1982), and the ratio between oak a

1706

SALMINEN ET AL.

specialists and generalists similarly peaks early in the season (Niemel , 1983). a Feeny (1970) attributed this pattern to a clearcut decline in the nutritional quality of oak leaves over the course of the summer, largely because of an accumulation of phenolics in general and proanthocyanidins in particular, is presumed to precipitate proteins in the digestive system of the larva (Feeny, 1970; Niemel , 1983; a cf. Herms and Mattson, 1992). This view is partly upset by the current results, as the total concentrations of phenolics clearly decreased over time, and the concentration of hydrolyzable tannins was found to dominate over proanthocyanidins. Hence, if there is a general decline in oak leaf quality over time, it can hardly be linked to crude changes in phenolic contentsif it were, we would instead expect an increase (cf. Figure 2A and B). Given different seasonal patterns both among proanthocyanidins and hydrolyzable tannins as groups, and among individual compounds within these groups, the total effect of phenolics on seasonal variation in oak leaf quality will also depend on the relative biological activities of each individual compound. Nevertheless, under no circumstances will herbivore performance be determined by phenolic contents aloneas emphasized by Haukioja (2003), the impacts of phenolics should be interpreted against the background of seasonal changes in crucial nutrients. Our elemental analysis showed that the nitrogen concentration of leaves declined by more than 50% between late May and September, suggesting a rapid decline in the availability of proteins and free amino acids. At the same time, the carbon contents of the leaves increased, indicating a buildup of lignin causing increased toughness of the leaves. Changes in nitrogen content and leaf toughness were also observed by Feeny (1970) and may interact with other attributes of the leaves such as water content (Mattson and Scriber, 1987; Haukioja et al., 2002; Henriksson et al., 2003) in determining their nutritional quality. In the end, a wealth of physical and biochemical factors may inuence the quality of growing leaf tissue from an herbivores perspective (Kause et al., 1999), and much work remains to be done before the chemical contents of an oak leaf can be linked to its perceived nutritional value. From a methodological perspective, the observed idiosyncrasies among individual compounds cast some doubt on the common use of so-called total methodse.g., FolinCiocalteau for total phenolics, the sodium nitrite method for total ellagitannins (Wilson and Hagerman, 1990), and the rhodanine method for total gallotannins (Inoue and Hagerman, 1988). Clearly, when concentrations of individual compounds vary more or less independently of each other, any method quantifying their pooled concentration will be a poor descriptor of patterns at the level of single compounds. This fact was graphically demonstrated by a general lack of correlation between the contents of individual compounds and total phenolics as quantied from the same samples (Figure 3). Furthermore, total methods do not provide an unbiased measure even of the sum of individual compounds, as shown by a discrepancy between the

SEASONAL VARIATION IN OAK LEAF CONTENTS

1707

pooled amount of individual hydrolyzable tannins, avonoid glycosides, and condensed tannins on the one hand, and total phenolics as quantied by the FolinCiocalteau method on the other (Figure 3). The problems are compounded by the type of reactions used in the quantication processes. To illustrate this point, let us consider the determination of total ellagitannins with the sodium nitrite method. This method relies on the hydrolysis of ellagitannins and on the reaction of ellagic acid (the most common hydrolysis product of ellagitannins) with sodium nitrite to yield a nitrosylated chromophore that is subsequently quantied by a spectrophotometer. The production of ellagic acid requires the presence of hexahydroxydiphenoyl (HHDP) group(s) in the hydrolyzed ellagitannins. Unfortunately, HHDP groups are not found in the structures of all ellagitannins, and, even when they are, they may be just a small part of a larger ellagitannin molecule. This may lead us to underestimate the ellagitannin content of plant samples, orin extreme caseseven to overlook the mere presence of ellagitannins. In Q. robur, we found the leaves to contain substantial amounts of 5, 7, 8, 9, and 10the structures of which contain highly variable proportions of HHDP groups (77.6, 32.5, 32.5, 0, and 16.3%, respectively). Compound 9 is the extreme example as it contains only biosynthetically modied HHDP groups, i.e., a nonahydroxytriphenoyl and a valoneoyl group. Hence, for samples of oak leaves, the sodium nitrite method is likely to yield a highly biased estimate of total ellagitannin content. Earlier, the same pattern of underestimation was shown to be true with quantication of total galloylglucoses by the rhodanine method (Salminen, 2003) and with quantication of total phenolics by the Folin (FolinCiocalteau or FolinDenis) assay (Appel et al., 2001). Moreover, the Folin assay also underestimated the total phenolic content of our oak leaves (compare Figure 2A and B). Despite these shortcomings, total methods are still commonly used in ecological studies (e.g., McKinnon et al., 1999; Fisher et al., 2000; Inbar et al., 2001; Lill and Marquis, 2001; Abrahamson et al., 2003; Forkner et al., 2004). We realize that such methods may be useful under certain circumstances (Appel et al., 2001; Salminen, 2003), but stress their limitations in studies aiming to pinpoint the role of, e.g., hydrolyzable tannins as determinants of herbivore performance. Simply speaking, if we do not know the structures of the hydrolyzable tannins of our target species, we know neither what we are quantifying nor our measurement error. Hence, we recommend that chemically minded ecologists and ecologically minded chemists should focus their future analyses on individual tannins and establish the specic biological activities of these polyphenolic compounds through rigorous bioassays (cf. Salminen and Lempa, 2002).
AcknowledgmentsWe thank Soa Gripenberg, Aulis Lepp nen, and Markku Salo for help a with the eld work, P ivi Fr nti for measuring total phenolics and proanthocyanidins, and Tiina Buss a a for conducting the elemental analyses. Comments by Ann Hagerman, Erkki Haukioja, Ky sti Lempa, o

1708

SALMINEN ET AL.

Vladimir Ossipov, Janne Suomela, and an anonymous referee helped us improve earlier drafts of this article; comments by Gavin Hinten helped us improve the language. This project was funded by the Academy of Finland (project numbers 45748, 51789, and 204209). REFERENCES ABRAHAMSON, W. G., HUNTER, M. D., MELIKA, G., and PRICE, P. W. 2003. Cynipid gall-wasp communities correlate with oak chemistry. J. Chem. Ecol. 29:209223. APPEL, H. M. 1993. Phenolics in ecological interactions: The importance of oxidation. J. Chem. Ecol. 19:15211552. APPEL, H. M., GOVENOR, H. L., DASCENZO, M., SISKA, E., and SCHULTZ, J. C. 2001. Limitations of folin assays of foliar phenolics in ecological studies. J. Chem. Ecol. 27:761778. AYRES, M. P., CLAUSEN, T. P., MCLEAN, S. F., REDMAN, A. M., and REICHARDT, P. B. 1997. Diversity of structure and antiherbivore activity in condensed tannins. Ecology 78:16961712. CADAHA, E., VAREA, S., MUNOZ, L., FERNANDEZ DE SIMON, B., and GARCA-VALLEJO, M. C. 2001. I I Evolution of ellagitannins in Spanish, French, and American oak woods during natural seasoning and toasting. J. Agric. Food Chem. 49:36773684. CLAUSEN, T. P., PROVENZA, F. D., and BURRITT, E. A. 1990. Ecological implications of condensed tannin structure: A case study. J. Chem. Ecol. 16:23812392. CLOSE, D. C. and MCARTHUR, C. 2002. Rethinking the role of many plant phenolicsProtection from photodamage not herbivores. Oikos 99:166172. CONDE, E., CADAHA, E., GARCA-VALLEJO, M. C., and FERNANDEZ DE SIMON, B. 1998. Polyphenolic I I composition of Quercus suber cork from different Spanish provenances. J. Agric. Food Chem. 46:31663171. CRAWLEY, M. and AKHTERUZZAMAN, M. 1988. Individual variation in the phenology of oak trees and its consequences for herbivorous insects. Func. Ecol. 2:409415. DUDT, J. F. and SHURE, D. J. 1994. The inuence of light and nutrients on foliar phenolics and insect herbivory. Ecology 75:8698. FAETH, S. 1986. Indirect interactions between temporally separated herbivores mediated by the host plant. Ecology 67:479494. FEENY, P. P. 1970. Seasonal changes in oak leaf tannins and nutrients as a cause of spring feeding by winter moth caterpillars. Ecology 51:565581. FELDMAN, K. S., SAMBANDAM, A., LEMON, S. T., NICEWONGER, R. B., LONG, G. S., BATTAGLIA, D. F., ENSEL, S. M., and LACI, M. A. 1999. Binding afnities of gallotannin analogs with bovine serum albumin, ramications for polyphenol-protein molecular recognition. Phytochemistry 51:867 872. FERNANDEZ DE SIMON, B., CADAHA, E., CONDE, E., and GARCA-VALLEJO, M. C. 1999. Evolution of I I phenolic compounds of Spanish oak wood during natural seasoning. First results. J. Agric. Food Chem. 47:16871694. FISHER, A. E. I., Hartley, S. E., and Young, M. 2000. Direct and indirect competitive effects of foliage feeding guilds on the performance of the birch leaf-miner Eriocrania. J. Anim. Ecol. 69:165 176. FORKNER, R. E., MARQUIS, R. J., and LILL, J. T. 2004. Feeny revisited: Condensed tannins as antiherbivore defences in leaf-chewing herbivore communities of Quercus. Ecol. Entomol. 29:174 187. GROSS, G. G. 1999. Biosynthesis of hydrolyzable tannins, pp. 799826, in B. M. Pinto (ed.). Comprehensive Natural Products Chemistry, Vol. 3. Elsevier, Amsterdam. GRUNDHOFER, P., NIEMETZ, R., SCHILLING, G., and GROSS, G. G. 2001. Biosynthesis and subcellular distribution of hydrolyzable tannins. Phytochemistry 57:915927.

SEASONAL VARIATION IN OAK LEAF CONTENTS

1709

HASLAM, E. 1992. Gallic acid and its metabolites, pp. 169194, in R. W. Hemingway and P. E. Laks (eds.). Plant Polyphenols. Plenum Press, New York. HATANO, T., KIRA, R., YOSHIZAKI, M., and OKUDA, T. 1986. Seasonal changes in the tannins of Liquidambar formosana reecting their biogenesis. Phytochemistry 25:27872789. HATANO, T., OKONOGI, A., and OKUDA, T. 1992. Oligomeric hydrolyzable tannins from Liquidambar formosana and spectral analysis of the orientation of valoneoyl groups in their molecules, pp. 195 207, in R. W. Hemingway and P. E. Laks (eds.). Plant Polyphenols. Plenum Press, New York. HAUKIOJA, E. 2003. Putting the insect into the birchinsect interaction. Oecologia 136:161 168. HAUKIOJA, E., OSSIPOV, V., and LEMPA, K. 2002. Interactive effects of leaf maturation and phenolics on consumption and growth of a geometrid moth. Entomol. Exp. Appl. 104:125136. HELM, R. F., ZHENTIAN, L., RANATUNGA, T., JERVIS, J., and ELDER, T. 1999. Towards understanding monomeric ellagitannin biosynthesis, pp. 8399, in G. G. Gross, R. W. Hemingway, and T. Yoshida (eds.). Plant Polyphenols, 2: Chemistry, Biology, Pharmacology, Ecology. Kluwer Academic/Plenum, New York. HENRIKSSON, J., HAUKIOJA, E., OSSIPOV, V., OSSIPOVA, S., SILLANPAA, S., KAPARI, L., and PIHLAJA, K. 2003. Effects of host shading on consumption and growth of the geometrid Epirrita autumnata: Interactive roles of water, primary and secondary compounds. Oikos 103:316. HERMS, D. A. and MATTSON, W. J. 1992. The dilemma of plants: To grow or defend. Q. Rev. Biol. 67:283335. HERVE DU PENHOAT, C. L. M., MICHON, V. M. F., OHASSAN, A., PENG, S., SCALBERT, A., and GAGE, D. 1991a. Roburin A, a dimeric ellagitannin from heartwood of Quercus robur. Phytochemistry 30:329332. HERVE DU PENHOAT, C. L. M., MICHON, V. M. F., PENG, S., VIRIOT, C., SCALBERT, A., and GAGE, D. 1991b. Structural elucidation of new dimeric ellagitannins from Quercus robur L. J. Chem. Soc., Perkin Trans. 1, 16531660. HOFFMAN, G. and LYR, H. 1973. Charakerisierung des Wachstumsverhaltens von Pantzen durch Wachstumsschemata. Flora 162:8198. INBAR, M., DOOSTDAR, H., and MAYE, R. 2001. Suitability of stressed and vigorous plants to various insect herbivores. Oikos 94:228235. INOUE, K. and HAGERMAN, A. 1988. Determination of gallotannin with rhodanine. Anal. Biochem. 169:363369. ITO, H., YAMAGUCHI, K., KIM, T.-H., KHENNOUF, S., GHARZOULI, K., and YOSHIDA, T. 2002. Dimeric and trimeric hydrolyzable tannins from Quercus coccifera and Quercus suber. J. Nat. Prod. 65:339345. JALAS, J. and SUOMINEN, J. 1976. Atlas Florae Europaea. Distribution of vascular plants in Europe. Quercus robur, Map 301. Committee for Mapping the Flora of Europe and Societatis Biologica Fennica, Helsinki. KAUSE, A., OSSIPOV, V., HAUKIOJA, E., LEMPA, K., HANHIMAKI, S., and OSSIPOVA, S. 1999. Multiplicity of biochemical factors determining quality of growing birch leaves. Oecologia 120:102112. KILKOWSKI, W. J. and GROSS, G. G. 1999. Color reaction of hydrolyzable tannins with Bradford reagent, Coomassie brilliant blue. Phytochemistry 51:363366. KRAUS, T. E. C., YU, Z., PRESTON, C. M., DAHLGREN, R. A., and ZASOSKI, R. J. 2003. Linking chemical reactivity and protein precipitation to structural characteristics of foliar tannins. J. Chem. Ecol. 29:703730. LILL, J. T. and MARQUIS, R. J. 2001. The effects of leaf quality on herbivore performance and attack from natural enemies. Oecologia 126:418428. MAMMELA, P., SAVOLAINEN, H., LINDROOS, L., KANGAS, J., and VARTIAINEN, T. 2000. Analysis of oak tannins by liquid chromatographyelectrospray ionisation mass spectrometry. J. Chromatogr. A 891:7583.

1710

SALMINEN ET AL.

MASSON, G., PUECH, J.-L., and MOUTOUNET, M. 1994. Localization of the ellagitannins in the tissues of Quercus robur and Quercus petraea woods. Phytochemistry 37:12451249. MATTSON, W. J. and SCRIBER, J. M. 1987. Nutritional ecology of insect folivores of woody plants: Nitrogen, water, ber, and mineral considerations, pp. 105146, in F. Slansky Jr. and J. G. Rodriquez (eds.), Nutritional Ecology of Insects, Mites, Spiders and Related Invertebrates. Wiley, New York. MAUFETTE, Y. and OECHEL, W. C. 1989. Seasonal variation in leaf chemistry of the coast live oak Quercus agrifolia and implications for the California oak moth Phryganidia californica. Oecologia 79:439445. MCKINNON, M. L., QUIRING, D. T., and BAUCE, E. 1999. Inuence of tree growth rate, shoot size and foliar chemistry on the abundance and performance of a galling adelgid. Func. Ecol. 13:859 867. MOSEDALE, J. R., FEUILLAT, F., BAUMES, R., DUPOUEY, J.-L., and PUECH, J.-L. 1998. Variability of wood extractives among Quercus robur and Quercus petraea trees from mixed stands and their relation to wood anatomy and leaf morphology. Can. J. Forest Res. 28:9941006. NIEMELA, P. 1983. Seasonal patterns in the incidence of specialism: Macrolepidopteran larvae on Finnish deciduous trees. Ann. Zool. Fenn. 20:199202. NIEMELA, P. and HAUKIOJA, E. 1982. Seasonal patterns in species richness of herbivores: Macrolepidopteran larvae on Finnish deciduous trees. Ecol. Entomol. 7:169175. NIEMETZ, R., SCHILLING, G., and GROSS, G. G. 2001. Ellagitannin biosynthesis: Oxidation of pentagalloylglucose to tellimagrandin II by an enzyme from Tellima grandiora leaves. Chem. Commun. 2001:3536. NURMI, K., OSSIPOV, V., HAUKIOJA, E., and PIHLAJA, K. 1996. Variation of total phenolic content and individual low-molecular-weight phenolics in foliage of mountain birch trees (Betula pubescens ssp. tortuosa). J. Chem. Ecol. 22:20232040. OKUDA, T., YOSHIDA, T., HATANO, T., YAZAKI, K., and ASHIDA, M. 1982. Ellagitannins of the casuarinaceae, stachyuraceae and myrtaceae. Phytochemistry 21:28712874. OSSIPOV, V., NURMI, K., LOPONEN, J., HAUKIOJA, E., and PIHLAJA, K. 1996. HPLC separation and identication of phenolic compounds from leaves of Betula pubescens and Betula pendula. J. Chromatogr. A 721:5968. OSSIPOV, V., NURMI, K., LOPONEN, J., PROKOPIEV, N., HAUKIOJA, E., and PIHLAJA, K. 1995. HPLC isolation and identication of avonoids from white birch Betula pubescens leaves. Biochem. Syst. Ecol. 23:213222. OSSIPOVA, S., OSSIPOV, V., HAUKIOJA, E., LOPONEN, J., and PIHLAJA, K. 2001. Proanthocyanidins of mountain birch leaves: Quantication and properties. Phytochem. Anal. 12:128 133. OZAWA, T., LILLEY, T. H., and HASLAM, E. 1987. Polyphenol interactions: Astringency and the loss of astringency in ripening fruit. Phytochemistry 26:29372942. RIIPI, M., OSSIPOV, V., LEMPA, K., HAUKIOJA, E., KORICHEVA, J., OSSIPOVA, S., and PIHLAJA, K. 2002. Seasonal changes in birch leaf chemistry: Are there trade-offs between leaf growth and accumulation of phenolics? Oecologia 130:380390. ROSSITER, M. C., SCHULTZ, J. C., and BALDWIN, I. T. 1988. Relationships among defoliation, red oak phenolics and gypsy moth growth and reproduction. Ecology 69:267277. SALMINEN, J.-P. 2002. Birch leaf hydrolysable tannins: Chemical, biochemical and ecological aspects. PhD Dissertation, University of Turku, Finland. SALMINEN, J.-P. 2003. Effects of sample drying and storage, and choice of extraction solvent and analysis method on the yield of birch leaf hydrolysable tannins. J. Chem. Ecol. 29:1289 1305. SALMINEN, J.-P. and LEMPA, K. 2002. Effects of hydrolysable tannins on an herbivorous insect: Fate of individual tannins in insect digestive tract. Chemoecology 12:203211.

SEASONAL VARIATION IN OAK LEAF CONTENTS

1711

SALMINEN, J.-P., OSSIPOV, V., HAUKIOJA, E., and PIHLAJA, K. 2001. Seasonal variation in the content of hydrolysable tannins in leaves of Betula pubescens. Phytochemistry 57:1522. SALMINEN, J.-P., OSSIPOV, V., LOPONEN, J., HAUKIOJA, E., and PIHLAJA, K. 1999. Characterisation of hydrolysable tannins from leaves of Betula pubescens by high-performance liquid chromatographymass spectrometry. J. Chromatogr. A 864:283291. SCALBERT, A. and HASLAM, E. 1987. Plant polyphenols and chemical defense. Part 2: Polyphenols and chemical defense of the leaves of Quercus robur. Phytochemistry 26:3191 3195. SCALBERT, A., MONTIES, B., and FAVRE, J.-M. 1988. Polyphenols of Quercus robur: Adult tree and in vitro grown calli and shoots. Phytochemistry 27:34833488. SCHULTZ, J. C. and BALDWIN, I. T. 1982. Oak leaf quality declines in response to defoliation by gypsy moth larvae. Science 217:149151. TIKKANEN, O.-P. and JULKUNEN-TIITTO, R. 2003. Phenological variation as protection against defoliating insects: The case of Quercus robur and Operophtera brumata. Oecologia 136:244251. VIRIOT, C., SCALBERT, A., HERVE DU PENHOAT, C. L. M., ROLANDO, C., and MOUTOUNET, M. 1994. Methylation, acetylation and gel permeation of hydrolysable tannins. J. Chromatogr. A 662:77 85. VIVAS, N., LAGUERRE, M., GLORIES, Y., BOURGEOIS, G., and VITRY, C. 1995. Structure simulation of two ellagitannins from Quercus robur L. Phytochemistry 39:11931199. WILSON, T. C. and HAGERMAN, A. E. 1990. Quantitative determination of ellagic acid. J. Agric. Food Chem. 38:16781683. YOSHIDA, T., HATANO, T., KUWAJIMA, T., and OKUDA, T. 1992. Oligomeric hydrolyzable tannins Their 1 H NMR spectra and partial degradation. Heterocycles 33:463482. ZHENTIAN, L., JERVIS, J., and HELM, R. F. 1999. C-glycosidic ellagitannins from white oak heartwood and callus tissues. Phytochemistry 51:751756. ZUCKER, W. V. 1983. Tannins: Does structure determine function? An ecological perspective. Am. Nat. 121:335365.