Isolation of the Major Phenolic Compounds in the Pits of Brined Green Olive Drupes: Structure Elucidation by Comprehensive 1 H/ 13 C-NMR Spectroscopy

Abstract

Pits of representative green olive drupes from Italy and Greece were evaluated for their content of phenolic compounds. The major purified polyphenols quantitated in mg/kg in the Italian and Greek pit varieties, respectively, were tyrosol (157 and 523), hydroxytyrosol (3 386 and 1723), vanillin (151 and 83), vanillic acid (58 and 101), phloretic acid (85 and 0), (+)-pinoresinol (954 and 877), (+)-1-acetoxypinoresinol (114 and 105), dehydrodiconiferyl alcohol (340 and 386) and the corresponding aldehyde (398 and 553), the erythro and threo isomers of guaiacylglycerol-β-coniferol ether (78 and 147) and (93 and 273) and their corresponding aldehydes (238 and 349) and (232 and 537), and finally nüzhenide (186 and 364). The data indicate that the waste products of olive mills should be a rich source of a variety of polyphenolic compounds. Detailed ¹H- and ¹³C-nuclear magnetic resonance data are presented, representing the most comprehensive and unambiguous results available to date for the compounds studied.

As a possible major contributor to the health-protective effects of the Mediterranean diet, ^{1 -4} the phenolic antioxidant content of olives and olive oil has been under investigation for many years. However, only in the last 2 decades has comprehensive identification ^5,6 and quantitation of these substances been provided. ^2,3,7-9 The study of olive pits or seeds has been more limited. ^{10 -12}

During the production of olive oil, either by traditional or conventional methods, the press cake or pomace is discarded (in some cases after further extraction with organic solvents), or used as an agricultural fertilizer. To date, little attention has been focused on the content of potentially useful phenolic antioxidants in these waste products. Therefore, olive pits, which represent a major portion of the press cake (traditional methods) or the so-called vegetation waste water (conventional methods), of green olives were lyophilized, macerated and extracted with organic solvents. The extracts were analyzed for their content of phenolic compounds.

In this paper we describe the isolation, purification, and structure elucidation of the major phenolic compounds present in the pits of green olives of Italian and Greek varieties. The techniques used included high-performance liquid chromatography (HPLC), nano-electrospray ionization-mass spectrometry (nano-ESI-MS), gas-chromatography mass spectrometry (GC-MS), and ¹H- and ¹³C-nuclear magnetic resonance (NMR) spectroscopy. The NMR data represent the most comprehensive and unambiguous results available to date for the compounds studied.

Profiles of the Phenolic Compounds in Olive Pits

The structures of the major phenolic compounds isolated from the methanol extracts of the olive pits used in this study are shown in Figure S1, (Supplementary data). These include the monophenolics tyrosol (I), hydroxytyrosol (II), vanillin (III), vanillic acid (IV), and phloretic acid (dihydro-p-coumaric acid, V). In addition, several dimeric phenolics were obtained, which are derived from the dehydrogenative dimerization of 2 molecules of coniferyl alcohol ¹³ or from the reaction of the alcohol with its corresponding aldehyde. These coniferyl parent compounds are also shown in Figure S1, (Supplementary data). The isolated dimeric compounds were the so-called β-β dimer (+)-pinoresinol (VI) and its derivative (+)-1-acetoxypinoresinol (VII), the β-5′ dimer dehydrodiconiferyl alcohol (VIII) and the corresponding alcohol-aldehyde dimer (IX), the β-O-4′ dimer guaiacylglycerol-β-coniferyl alcohol ether, which occurs as erythro and threo isomers (X, XI), and the corresponding aldehydes (XII, XIII). Finally, the iridoid glycoside (8E)-nüzhenide (XIV) was also isolated. HPLC chromatograms of the fractions containing the compounds obtained after column chromatography on silicic acid are shown in Figure S2, (Supplementary data).

Spectroscopic Analysis and Structure Elucidation

GC-MS data for derivatized phenolic olive pit components are summarized in Table S1, (Supplementary data). ESI-MS-MS data and UV data for the isolated compounds are presented in Tables S2 and S3, (Supplementary data), respectively (Supplementary data). Detailed ¹H- and ¹³C-NMR data with complete and unambiguous assignments for all compounds discussed in this study are summarized in Tables S4–S11, (Supplementary data).

Monomeric Phenols

For compounds I, II, and V, the UV data, nano-ESI-MS data, as well as the GC-MS results for the TMS derivatives matched the data obtained previously for tyrosol, hydroxytyrosol, ⁸ and phloretic acid. ⁹ The UV absorption data (Supplementary data, Table S3) of compounds III and IV suggested a phenol structure while the GC-MS spectra (Supplementary data, Table S1) of the TMS derivatives were identical to those of authentic vanillin and vanillic acid, respectively. The structures of the underivatized compounds were confirmed by nano-ESI-MS spectra in negative-ion mode (Supplementary data, Table S2), giving molecular weights of 152 (III) and 168 (IV), respectively.

The ¹H- and ¹³C-NMR data for compounds I and II were in excellent agreement with our published results for tyrosol and hydroxytyrosol ⁸ ; therefore, these results will not be repeated here. For vanillin (III), vanillic acid (IV), phloretic acid (V), and the reference compounds coniferyl alcohol and coniferyl aldehyde, the ¹³C and ¹H chemical shift data are summarized in Tables S4 and S5, (Supplementary data), using the numbering scheme of Figure S1, (Supplementary data). Highly resolved ¹H-coupled ¹³C spectra were obtained for compounds III, IV, and V, and the J _CH coupling constants determined by first-order analysis are summarized in Table S6, (Supplementary data). Finally, the ¹H spectra for III, IV, V and coniferyl alcohol were analyzed in detail using spin system simulation/iteration to give the J _HH coupling constants summarized in Table S7, (Supplementary data). For coniferyl aldehyde the sample quality was insufficient for detailed study, and only a first-order analysis was performed. The phloretic acid data agree well with previous results of limited resolution obtained on a sample containing paramagnetic impurities. ⁹ The current data were obtained on a metal-free sample, which provided very high resolution.

Our NMR results for the monophenols in CD₃OD showed good agreement with data obtained with other solvents in the metabolomics database of the Biological Magnetic Resonance Data Bank (BMRB). ¹⁴ The detailed spin system analyses presented here, taken together with the MS data, provide unambiguous confirmation of the structures presented in Figure S1, (Supplementary data) for compounds I-V.

Lignans

Two major lignans were isolated from the pits of green olives. The GC-MS data (Supplementary data, Table S1), nano-ESI-MS data (Supplementary data, Table S2), and detailed NMR data (not shown) for compounds VI (exact mass 358.1) and VII (exact mass 416.1) were entirely consistent with the results described previously ⁸ for (+)-pinoresinol, the β-β dimer of coniferyl alcohol, and its derivative (+)-1-acetoxypinoresinol, respectively. Therefore, the NMR results for VI and VII will not be repeated here.

Neolignans

Several dimeric neolignans derived from coniferyl alcohol were also isolated from olive pits. The ESI-MS data (Supplementary data, Table 2) provided m/z values for the [M–H], [2M–H]^–, and [M + Na]⁺ ions, indicating an estimated exact mass of 358.1 for VIII, consistent with the formula C₂₀H₂₂O₆ (calculated: 358.14) and 356.0 for IX, consistent with the formula C₂₀H₂₀O₆ (calculated: 356.126). The ¹H- and ¹³C-NMR data confirmed very similar C₂₀ skeletons for both compounds, consisting of a trisubstituted phenol attached to a C₃ aliphatic unit and a tetrasubstituted phenol connected to a C₃ trans olefin unit. Two phenolic –OMe groups were also present. The only significant difference between the 2 compounds was the presence of an olefinic aldehyde –CHO group in IX instead of the olefinic alcohol –CH₂OH in VIII, consistent with the major differences in the UV spectra (Supplementary data, Table S3). Thus, the candidate structures were the 8(β)-5′ dimer of coniferyl alcohol, dehydrodiconiferyl alcohol (VIII), and its corresponding aldehyde (IX). In these compounds the A and B ring systems are also connected via a 7-O-4′ linkage. The ¹H- and ¹³C-NMR data obtained for compounds VIII and IX are summarized in Tables S8–S10 (Supplementary data). Each compound was isolated by HPLC as a single NMR-distinguishable isomer (or enantiomeric pair).

Two pairs of isomeric dimers containing an olefinic alcohol or aldehyde group were isolated from the olive pits. The ESI-MS data (Supplementary data, Table S2) provided m/z values for the [M–H]^–, [2M–H]^–, and [M + Na]⁺ ions derived from the HPLC peaks X and XI (Supplementary data, Figure S2), trace cc-18), with estimated exact mass 376.2, consistent with the formula C₂₀H₂₄O₇ (calculated: 376.15). A second pair of isomers, HPLC peaks XII and XIII(Supplementary data, Figure S2, trace cc-14), gave exact mass 374.1 consistent with the formula C₂₀H₂₂O₇ (calculated: 374.14). Appropriate candidate structures for these compounds (Supplementary data, Figure S1) were the 8-O-4′ dimers guaiacylglycerol-β-coniferol ether, which exists in erythro and threo forms, and the corresponding guaiacylglycerol-β-coniferaldehyde ethers. The GC-MS data (Supplementary data, Table S1) and the UV data (Supplementary data, Table S3) also supported the assignment of X and XI as the isomeric alcohol ethers and XII and XIII as the aldehyde ethers. The NMR data for these 4 compounds are summarized in Tables S8–S10, (Supplementary data), and provide unambiguous proof for the structures and stereochemistry as shown in Figure S1, (Supplementary data).

(8E)-Nüzhenide

The final major phenolic component of olive pits, XIV, provided ESI-MS data (Supplementary data, Table S2) indicating an exact mass of 686.3, consistent with the formula C₃₁H₄₂O₁₇ (calculated: 686.242). The ESI-MS-MS spectrum in negative-ion mode indicated the loss of 1 glucose moiety from the molecular ion to give m/z 523.3. The origin of the other major fragments in the spectrum was unclear at first, but given that the UV spectrum (Supplementary data, Table S3) showed similarities to a secoiridoid structure containing elenolic acid, this enabled characterization of the fragmentation pattern. Significant ions at m/z 384.9, 299.0, and 222.7 were due, respectively, to the loss of tyrosol, elenolic acid, and tyrosol glucoside from deglucosylated nüzhenide (m/z, 523.3). GC-MS of the parent molecule gave no signals, consistent with a glycoside. Acid hydrolysis yielded 1 major product with a molecular ion at m/z 299, consistent with tyrosol glucoside. A minor amount of elenolic acid was also observed. Table S11 (Supplementary data) summarizes the completely assigned ¹H- and ¹³C-NMR data for XIV. Four ¹H spin systems were identified by two-dimensional (2D) NMR techniques and analyzed by simulation: elenolic acid, tyrosol, and 2 β-d-glucose units. The connectivity of these fragments and the stereochemistry of the elenolic acid unit shown in Figure S1, (Supplementary data) were confirmed by long-range C-H correlations and nuclear Overhauser effect (NOE) difference spectra.

Quantitation of Phenolic Compounds in Olive Pits

The concentrations of the major phenolic substances in the 2 types of olive pits studied, as determined by HPLC analysis, are summarized in Table S12 (Supplementary data). Pits of the Italian olives contained a total of 6.47 g/kg phenolic compounds, consisting of monophenols (59%), lignans (17%), the dehydrodiconiferyl neolignans (11%), guaiacylglycerol-β-coniferyl compounds (10%), and nüzhenide (3%). Within these classes the phenols were 88% hydroxytyrosol. The lignans comprised 89% (+)-pinoresinol and 11% (+)-1-acetoxypinoresinol; dehydrodiconiferyls were 46% alcohol and 54% aldehyde; guaiacylglycerol-β-coniferyls were present as 27% alcohols and 73% aldehydes with roughly equal amounts of erythro and threo isomers.

The phenolic profile for the Greek olive pits was similar, but phloretic acid was not detected. Pits of the Greek olives contained 6.01 g/kg total phenolics, represented by monophenols (40%), lignans (16%), dehydrodiconiferyls (22%), guaiacylglycerol-β-coniferyl compounds (16%), and nüzhenide (6%). The monophenols were again dominated by hydroxytyrosol (71%) but with a greater contribution from tyrosol (21%). The lignans comprised 89% (+)-pinoresinol and 11% (+)−1-acetoxypinoresinol, as in the Italian pits; dehydrodiconiferyls were 41% alcohol and 59% aldehyde while guaiacylglycerol-β-coniferyl was present as 23% alcohols and 77% aldehydes with a predominance of the threo isomers. The distributions of alcohols vs aldehydes were very similar for Italian and Greek olive pits.

Structure Elucidation by NMR

For the current NMR investigations all isolated or reference compounds were examined in the same deuterated solvent, CD₃OD, which we have used extensively for studying phenolic compounds. This solvent has the advantage of simplifying spectra by removing labile –OH protons and their coupling or line-broadening effects. Very high resolution can often be achieved, depending on sample quality, so that detailed analysis of coupling constants with simulation of second-order spectra is frequently possible. Narrow resonance lines are also advantageous for improving the detectability of 2D correlations from long-range couplings, for example.

A disadvantage of deuteromethanol as solvent is the fact that comparisons with literature data are often difficult since other solvents tend to be used more frequently (CDCl₃, acetone-d₆, DMSO-d₆, pyridine-d₅), as in the BMRB metabolomics or USDA lignin databases. Older literature data often suffer from incomplete analysis and lower information content due to the lower magnetic field strengths used. Thus, in this and other studies our goal has been to assemble the most complete and precise data sets available for phenolic compounds in deuteromethanol, and we hope that these data will be useful for future studies.

Monomeric Phenols

The NMR analyses for the monophenols proved to be relatively straightforward (Supplementary data, Tables 4–7). ¹H spectra were subjected to iterative spin system simulation, which provided accurate values for ortho, meta, and para proton couplings within the phenol ring (not obtainable by first-order multiplet analysis) and, given sufficient sample quality, also long-range couplings between ring and side chain protons. Carbon assignments were readily achieved through one-bond C-H correlations and two- and three-bond C-H couplings obtained from¹H-coupled ¹³C spectra. In general, our results for chemical shifts agree reasonably well with the BMRB database for CDCl₃ or acetone-d₆ as solvent. However, the chemical shifts for C3 and C4 in coniferyl aldehyde are reversed in the BMRB. Some of the interesting features of our data are as follows.

The presence of an –OMe group at position 3 in vanillin, vanillic acid, and the coniferyls leads to a reduction of the ortho and meta couplings in the phenol ring, compared to phloretic acid which lacks the –OMe group. The order of ¹³C shifts for C3 and C4 in the coniferyls is reversed for the aldehyde vs the alcohol. C4 is shifted downfield when the side chain contains a double bond, and this effect also occurs in the vanillins.

In the phloretic acid side chain the methylene protons at positions 7 and 8 do not exhibit nonequivalence in chemical shifts but do exhibit nonequivalence in their vicinal couplings (AA'BB' spin system). This is the result of unequal populations for rotomers about the C7-C8 bond. A preference for the trans orientation of the phenyl ring and carboxyl group results in trans and gauche orientations of H7 protons to H8 protons, and vice versa. Spin system simulation reveals this situation with the determination of trans and gauche couplings as 8.6 and 6.7 Hz, respectively. Very similar results were obtained for dihydrocaffeic acid, ⁹ leading to an estimated population of 70% for the preferred trans rotomer.

5′ Dehydrodiconiferyl Neolignans

The complete assignments for ¹H and ¹³C spectra for the well-known dimeric neolignans VIII and IX as alcohol and aldehyde were carried out using COSY 2D with long-range H-H connectivities and¹³C-detected one-bond and multiple-bond C-H connectivities (Supplementary data, Tables 8–10). The¹H spin systems for fragments A and B were readily recognized and could be analyzed as first-order multiplets. The alcohol and aldehyde exhibited very similar NMR parameters for positions 1-9 (A ring) but significant differences for positions 1′-9′, similar to those observed for the parent coniferyl aldehyde vs alcohol. The high-field 3-OMe group in ring system A was assigned via its long-range proton couplings to H2 and C3. The low-field 3′-OMe group was assigned via its couplings in ring B. C4 and C7 in each ring were assigned via ³J couplings with the protons at positions 2 and 6. The 8-5′ linkage was detected as a strong downfield shift of C5′ compared to coniferyl aldehyde, by the ³J coupling of C8 to H6′ and the ⁴J coupling of H8 to H6′. The 7-O-4′ linkage in the alcohol results in a reversal of the C3′, C4′ shift order in the B ring compared to the A ring or to coniferyl alcohol.

The stereochemistry shown in Figure S1, (Supplementary data) corresponds to a (+)-dehydrodiconiferyl moiety with S,R configuration at C7,8 and with H7,H8 in a quasi trans axial orientation to one another. ¹⁵ This configuration is consistent with our measured vicinal coupling ³ J ₇₈ = 6.3 and 6.4 for the alcohol and aldehyde, respectively. Our complete set of ¹H chemical shifts for the alcohol in CD₃OD is in good agreement with the shift data and the estimated ³ J ₇₈ coupling of 7 Hz for the alcohol in acetone-d₆. ¹⁶ Our complete set of ¹³C chemical shifts for the alcohol in CD₃OD agrees well with the assignments and order of shifts presented for the alcohol in acetone-d₆. ¹³ Using 3D modeling of the alcohol with molecular dynamics (MD) we determined an average dihedral angle for H7,H8 of –127°, which is quite consistent with the observed coupling constant. Finally, the MD results gave average distances of 2.98 and 2.6 Å for H7-H8 and H7-H9ab, respectively. Quantitative NOE measurements gave a ratio of 2.2:1 for the NOE of H7 to the H9 protons vs H8, in agreement with the expected ratio of 2.1:1 based on the modeled distances.

Since we do not have optical rotation data for VIII and IX, we cannot provide information concerning the stereochemical purity of our samples. Thus, the enantiomeric (–) form with (R,S) configuration may also be present (indistinguishable by NMR). In fact, either form or mixtures have been isolated from various natural sources, and a variety of glucoside derivatives of both (+) and (–) enantiomers have been isolated. ^15,17

8-O-4′ Guaiacylglycerol-β-Coniferyl Neolignans

The well-known ether-linked neolignans X-XIII have been isolated from various sources, both as alcohol ^13,18,19 and aldehyde forms. ^17,20,21 Both forms occur as a pair of chromatographically separable and NMR-distinguishable isomers, erythro with (S,R) configuration at C7,8 (or its enantiomer) and threo with (S,S) configuration at C7,8 (or its enantiomer), as shown in Figure S1, (Supplementary data). Under C-18 HPLC conditions similar to ours ¹⁹ or with capillary zone electrophoresis, ¹³ the erythro isomer of the alcohol ether eluted before the threo isomer. For our NMR samples the isomeric purity was 90% or greater, and the order of elution on C-18 was erythro before threo for both the alcohol and aldehyde ethers.

The ¹H spin systems were analyzed in general as first-order multiplets with the following exceptions: the B ring protons in X showed near-equivalence for H5′,6′, requiring simulation with iteration; the aliphatic protons H7,8,9a,9b of XII required simulation with manual optimization. The NMR data (Supplementary data, Tables S8–S10) exhibit a very high similarity for phenol ring A in all 4 compounds. Phenol ring B shows the typical differences in chemical shifts for the aldehyde vs alcohol, e.g. upfield shift of C1′ and downfield shift of C4′ and all B ring protons. Compared to the 8-5′ dimers VIII and IX, the ether-linked dimers show significant differences in chemical shifts for H7 and H8 (circa 0.8 ppm upfield or downfield, respectively), consistent with the 8-O-4′ linkage and chemical shift predictions.

Discrimination between erythro and threo isomers is expected to be possible using the NMR parameters of the propyl side chain C7-C9. In the literature there are examples of the use of ³ J ₇₈ as a diagnostic parameter, whereby a large coupling of circa 7 Hz is typical for the threo isomer while a small coupling of circa 4.5 Hz is typical for erythro. ^16,18,19,22 However, this rule appears to apply only for the neolignans with additional bulky substituents at the C7 or C9 hydroxyl group (acetate, methoxy, ethoxy). In such cases rotation about the C7-C8 bond will be more hindered than in the parent compounds, which is expected to lead to significant differences in rotamer populations and average dihedral angles for the 2 isomers. However, our results for the unsubstituted neolignans X-XIII show that ³ J ₇₈ lies in the range 5.5-6 Hz (Supplementary data, Table 10) and is, therefore, not diagnostic for the isomer form. For the aldehyde ether in pyridine-d_5, ³ J ₇₈ values of 5.6 and 5.2 Hz for the erythro and threo isomers were observed. ²¹ Using 3D modeling one finds that if the C7-C8 rotamer with phenol rings A and B in transaxial position is predominant, then H7, H8 should be transaxial in the threo isomer but gauche in the erythro isomer. This would lead to a larger ³ J ₇₈ coupling in threo compared to erythro, as observed for derivatives with bulky substituents at C7 or C9. The nearly equal couplings observed in our unsubstituted compounds implies that considerable C7-C8 rotamer equilibration occurs.

Close examination of the chemical shifts in Tables S8 and S9, (Supplementary data) indicate that there is a small, systematic downfield shift for C8 and H7 in threo vs erythro for both the alcohol and aldehyde and upfield shifts for H8. The most dramatic effect, however, is for the chemical shifts and vicinal couplings of the nonequivalent H9a,b. In the erythro isomer these shifts differ by <0.1 ppm at circa 3.8 ppm. For the threo isomer the shift difference increases to circa 0.25 ppm with the high-field proton (by convention labeled 9a in each case) appearing at circa 3.5 ppm. Similar behavior for C8 and H9a,b has been documented for the alcohol ether in DMSO-d₆ solution ¹⁸ and for C8 in acetone-d₆. ¹³ In contrast, ¹H-NMR results reported for the aldehyde ether in pyridine-d₅ revealed no significant differences for H9a,b for the 2 isomers, but this may be the result of specific solvent effects of pyridine. In summary, the chemical shift data in Tables S8 and S9, (Supplementary data) show clearly that the compounds isolated as HPLC peaks X and XII are of one isomeric form while XI and XIII are of the other form. Comparison with the literature data for the alcohol ether isomers in DMSO-d₆ or acetone-d₆, as discussed above, leads to the assignment of isomer configuration as shown in the tables and Figure S1, (Supplementary data). Thus, in our study we find that erythro isomers elute before threo isomers, in agreement with other chromatographic results in the literature, as mentioned above.

Examination of the vicinal couplings between H8 and H9a,b in Table 10 (Supplementary data) reveals that the couplings appear to be interchanged for the threo vs erythro isomers, suggesting that the high-field proton labeled 9a with the larger vicinal coupling in the threo isomer becomes the lower-field proton 9b in the erythro isomer, retaining the larger vicinal coupling at the same stereochemical location. The moderate couplings observed of circa 4 and 5.5 Hz indicate that H9a and H9b adopt primarily gauche-gauche positions relative to H8.

In the literature the NMR data for the dimeric neolignans are rather fragmentary, for a variety of solvents and often only partially analyzed. To our knowledge the data presented here represent the most uniform, comprehensive, precise and unambiguous collection for the compounds studied.

(8E)-Nüzhenide

The first report of the isolation and NMR characterization of nüzhenide from olive seeds ¹¹ provided detailed ¹H- and ¹³C-NMR data for an ethanol-d₆ solution confirming the basic structural fragments and linkages as shown in Figure S1, (Supplementary data) but the structure was presented without the stereochemistry of the elenolic acid ring system and with the olefinic methyl group in the (8Z) configuration. Although a 2D NOE experiment was performed, no pertinent information regarding stereochemistry was mentioned.

A later report ²³ on iridoid glycosides from the leaves of Syringa reticulate provided detailed NMR characterization in CD₃OD of both (8Z)- and (8E)-nüzhenide in the (1S,5S) configuration shown in Figure S1, (Supplementary data). Detailed NOE measurements and long-range couplings confirmed the stereochemistry and conformation of the elenolic acid ring. A comparison of the reported ¹³C data with our results in the same solvent showed excellent agreement with the data for (8E)-nüzhenide, with shift differences of <0.07 ppm for all carbons. For the (8Z) isomer, shift differences of >0.3 ppm were found for all carbons 1 to 9. In particular, carbons 4, 5, 6, 8, and 9 provide excellent diagnostics for distinguishing between the 2 isomers. Comparison of the ¹³C shifts reported by Servili et al ¹¹ with those of Machida et al ²³ demonstrates that, after correcting for a general 1.0 ppm difference due to the methods of referencing, the earlier data for nüzhenide from olive seeds are consistent only with the (8E) isomer. As further confirmation, our ¹H-NMR shifts and couplings show excellent agreement with those for the (8E) isomer of Machida et al ²³ as well as with the data of Servili et al. ¹¹

Finally, NMR results (in CD₃OD) presented for nüzhenide isolated from Chionanthus retusus seeds ²⁴ show excellent agreement with our data for the elenolic acid and tyrosol fragments but poor agreement with our ¹³C data for the glucose fragments (¹H assignments were not reported). However, it appears certain that the (8E) isomer was isolated by Chien et al. ²⁴ A more recent study of secoiridoids extracted from Portuguese olive kernels, after removal from the pits, ¹² used HPLC and mass spectrometry alone (no NMR) to detect nüzhenide as the major phenolic along with other more complex oleosides.

We have taken the ¹H-NMR analysis somewhat further than in the previous studies. The spin systems for the elenolic acid, tyrosol, and 2 β-d -glucose units were detected via COSY 2D including long-range couplings. The estimated chemical shifts were compared with the ¹H shifts obtained from the 2D one-bond C-H correlation experiment to assign all protonated carbons. The quaternary carbons were assigned via the multiple-bond C-H correlation, which also gave the linkages between the fragments and discrimination of the glucose residues. A first-order multiplet analysis gave an initial data set for the elenolic acid and tyrosol fragments and partial data for the glucose residues, whereby shifts for strongly overlapping glucose resonances (H2-H5) were taken from the 2D data. For each structural fragment of nüzhenide a spin system simulation was carried out and parameters were adjusted manually to give a best fit to the resolution-enhanced spectrum. Automatic iteration was not possible because of overlap between the individual fragment spectra. The final results are shown in Table S11, (Supplementary data), whereby more precise data and numerous long-range couplings are now available. For example, the couplings of elenolic H1 to H3, H5, H9, and H10 could be resolved, as well as couplings between H3 and H5 and between H5 and H9 and H10. In addition, the nonequivalence of the H7''' and H8''' protons of the tyrosol group could be resolved and the distinct trans and gauche vicinal couplings due to restricted rotamer equilibration were determined. Finally, NOE difference spectra provided evidence for the (8E) configuration of the double bond (strong and nearly equal NOEs from H10 to H5 and H9, no NOE to H1). The (1S,5S) configuration for the elenolic residue was confirmed by the following NOEs: H1 to H6a > H3 >H6 b, no NOE to H9; H3 to H1 and H1′; H5 to H6b > H6 a. These results require a ring conformation with H1 axial, glucose equatorial, H5 equatorial, C6 axial, H5 and H6a transaxial. This is the conformation found previously ²³ for (8E)-nüzhenide, and we have confirmed this low-energy conformation by molecular modeling.

Phenolic Profiles of Green Olive Pits

The profiles of phenolic compounds in methanol extracts of brined green olive pits were quite different from their respective pericarps (Italian pericarp; Owen et al, ⁹ Greek pericarp, this study). While the pericarp of the Italian variety contained ostensibly only hydroxytyrosol ⁹ plus traces of other phenolics, the Greek variety contained a higher percentage of tyrosol (data not shown). Fourteen phenolic compounds (including isomers) were detected in the pits of the Italian olives, and of these only phloretic acid was not detected in the Greek olive pits.

Dehydrodiconiferyl and guaiacyl phenolic derivatives were first described in ²⁵ as constituents of pine tree wood. However, this is the first report whereby HPLC chromatographic separation of the guaiacylglycerol-β-coniferol ether isomers (X and XI) and guaiacylglycerol-β-coniferaldehyde ether isomers (XII and XIII) was achieved for extracts from olive pits, allowing their unequivocal identification by NMR. The synthesis of these compounds from coniferol alcohol, using the enzyme laccase (isolated from mushrooms, Psalliota) or via inorganic catalysis with copper sulfate and hydrogen, has also been described by Freudenbergs group, ²⁵ (see also synthetic work described previously ^13,20 ). All compounds are derived by dehydrogenation and dimerization of coniferol alcohol or its aldehyde. To date, no such compounds have been associated with olives, but the current study shows significant quantities within the Italian (21%) and Greek olive (37%) pits. The aldehyde is favored over the alcohol species with nearly identical ratios of 1.78 and 1.79:1 for the pit groups, respectively.

A recent study of secoiridoids extracted from Portuguese olive pit kernels after removal from the stones ¹² used HPLC and mass spectrometry alone (no NMR) to detect nüzhenide as the major phenolic along with other more complex oleosides. Similarly, Servili et al ¹¹ isolated (8E)-nüzhenide as the predominant phenolic from Italian olive kernels. In these studies of kernels the dimeric neolignans, which we found to be important components of pits, were not detected and are probably restricted to the woody pit structure. Thus, nüzhenide, a minor component of our pit extracts, is likely to be present in the kernels only. The quantity and variability of phenolic compounds in olive pits have not been previously described or reported, and the data indicate that waste products of olive oil mills may be a rich source of natural phenolic antioxidants, which, based on epidemiologic data, should be regarded as potentially healthy additives to various food products.

Experimental

Reagents and Reference Compounds

Acetic acid, acetonitrile, n-hexane, methanol, and pentane were obtained from Merck (Darmstadt, Germany). Coniferyl alcohol, coniferyl aldehyde, tyrosol, vanillin, and vanillic acid were obtained from Aldrich Chemie (Steinheim, Germany). N-methyl-N-(trimethylsilyl)-trifluoroacetamide was obtained from Sigma Chemie (Deisenhofen, Germany). Phloretic acid was obtained from Extrasynthese (Lyon Nord, Genay, France). Hydroxytyrosol was synthesized according to ⁸ while (+)-pinoresinol was synthesized according to the method of Cooper et al. ²⁶ (+)-1-Acetoxypinoresinol was extracted and purified from an olive oil known to contain only this lignan as previously assayed ⁸ All solutions were made up in either doubly distilled water or methanol, unless otherwise stated.

Olive Varieties

Olive varieties were selected, which were known to contain considerable quantities of phenolic substances, in order to facilitate the isolation of sufficient quantities for detailed analytical and spectroscopic analyses. Italian varieties were Il Trullo, Sapori di Puglia, Prodotto Naturale Senza Conservanti, Conservare in Luogo Fresco, Dopo L´Apertura; the Greek variety: Crete.

Extraction of the Olive Pits

The olives were dehulled, and the pits were macerated and lyophilized to constant weight (Christ Gefriertrocknungsanlagen, Osterode, Germany), after which 100 g was extracted twice (3 H) with pentane in a Soxhlet apparatus to remove unwanted lipid components. After drying, the pits were further extracted (3 H, 3 times) with methanol. The extracts were pooled, and the solvent was evaporated in vacuo at 35°C. The residue was suspended in acetonitrile (50 mL), and extracted three times with n-hexane (20 mL) to remove residual lipids. The n-hexane extracts were discarded, and the acetonitrile solution was dried over anhydrous magnesium sulfate. The solvent was removed in vacuo at 35°C prior to fractionation by column chromatography (Supplementary data).

Nuclear Magnetic Resonance Spectroscopy

¹H- and ¹³C-NMR spectra were recorded at 303 K on Bruker spectrometers (Bruker BioSpin, Rheinstetten, Germany), namely AC-250, AM-300, AM-500, and AV-600, operating at ¹H frequencies of 250.13, 300.13, 500.13 and 600.13 MHz, respectively. All compounds were dissolved in CD₃OD (99.8% D), typically 3-10 mg in 0.4 mL. In general, 1D and 2D Fourier transform techniques were employed as necessary to achieve unequivocal signal assignments and structure proof for all compounds independently. In addition to 2D shift-correlation experiments (H-H COSY with long-range connectivities, C-H correlation via ¹ J _CH, C-H correlation via long-range ⁿJ _CH), ¹H-coupled ¹³C spectra and selective ¹H-decoupling were used in some cases to determine long-range J _CH coupling constants and to assign all quaternary carbons unambiguously. Where necessary, stereochemical assignments were made with the aid of 1D NOE difference spectra or 2D ROESY experiments. All data sets were processed using Bruker’s TopSpin 2.1 software package, and resolution-enhanced ¹H spectra were analyzed for shifts and couplings either by first-order multiplet analysis or by spin system simulation and iteration with the WIN-DAISY subroutine (total lineshape fitting for chemical shifts, couplings, and linewidths). ¹H and ¹³C chemical shifts δ are reported in ppm relative to CHD₂OD (δ = 3.31 ppm for ¹H) or CD₃OD (49.053 ppm for ¹³C).

Molecular Modelling and Data base References

Diagrams with stereochemistry and NMR chemical shift predictions for preliminary assignments were produced using the software ChemDraw Ultra 10.0 (CambridgeSoft, Cambridge, MA, USA); 3D modeling was performed with Chem3D Ultra 10.0 and the MM2 force field. For several compounds NMR reference data for comparison with our results were obtained from the USDA NMR Database of Lignin and Cell Wall Model Compounds ²⁷ hosted by the metabolomics section of the BMRB http://www.bmrb.wisc.edu/metabolomics/ metabolomics_standards_jr.shtml]. ¹⁴

Other Methods

Column chromatography, mass spectrometry, and synthetic methods are described in the accompanying Supplementary data.

Supplemental Material

Supplementary data - Supplemental material for Isolation of the Major Phenolic Compounds in the Pits of Brined Green Olive Drupes: Structure Elucidation by Comprehensive ¹H/¹³C-NMR Spectroscopy

Supplemental material, Supplementary data, for Isolation of the Major Phenolic Compounds in the Pits of Brined Green Olive Drupes: Structure Elucidation by Comprehensive ¹H/¹³C-NMR Spectroscopy by Farid Khallouki, William E. Hull, Gerd Würtele, Roswitha Haubner, Gerhard Erben, and Robert W. Owen in Natural Product Communications