Introduction

Proanthocyanidins (PAC) constitute a class of oligomeric compounds present in grape skin and seeds composed of flavan-3-ol ((epi)catechin) units [1,2,3,4]. In recent pioneering reports, the presence of tetrameric and pentameric novel cyclic PAC (therein referred as crown proanthocyanidins) in red wines from the Bordeaux region was proposed for the first time [5, 6]. MS and NMR characterizations of the isolated molecules from purified extracts were provided. Despite the thorough characterization performed, this observation (the cyclic ‘crown’ structure of these PAC) was affected by an unforeseen ambiguity: the proposed novel cyclic B-type PAC share the same elemental composition with an already known class of (non-cyclic) PAC (seldom observed in wine) containing the additional A-type O-C linkage, usually present between the n and n-1 monomers in a n-mer [7, 8]. An example of a compound of that non-cyclic PAC class in red wine is procyanidin A2 [9]. In addition, this A-type PAC class up to the tetramer was reported to be abundant in cranberries [10]. The proposed novel cyclic B-type PAC and the non-cyclic A-type PAC possess identical elemental compositions (see the Results section for details). Therefore, the application of mass spectrometry for directly clarifying the actual structure of the new cyclic oligomers in wine may not be straightforward. MSn methods may also yield insufficient results since the A-type linkage can be broken during MS/MS fragmentation leaving only the B-type linkage [7, 8]. In this respect, a direct approach supporting the structural NMR characterization would be helpful [5, 11]. This approach may be achieved by devising a chemical modification able to differentiate between the two different chemical bond arrangements of the two PAC classes (cyclic B-type and non-cyclic A-type).

Among the most exploited derivatization approaches [12,13,14,15], hydrogen/deuterium (H/D) exchange offers a valuable option for structural resolution whenever exchangeable protons are present. Moreover, the H/D exchange rates influence conformational mobility, hydrogen bonding strength, and solvent accessibility of the molecule studied [16,17,18,19,20]. Nonetheless, this modification is completely reversible, and care must be applied to obtain a complete deuteration. Accordingly, HPLC methods have been modified for fully profiting from this transformation, for example by employing deuterium oxide as replacement of the water phase in HPLC-MS gradient methods [19]. A critical step is the equilibration time for a complete H/D exchange, in particular for proteins, which is influenced by the substrate, the solution pH, and the temperature [20]. H/D exchange was infrequently but successfully applied to wine extracts to characterize anthocyanins and pyranoanthocyanins, demonstrating how the H/D exchange rate followed an apparent first order kinetic independent from the solution pH [21].

In this report, H/D isotopic exchange was applied to confirm the B-type cyclic structure for a tetramer PAC recently proposed [5, 6] possessing the same theoretical mass and elemental composition of the non-cyclic A-type tetrameric proanthocyanidin. The different number of exchangeable protons between the novel cyclic oligomers and their non-cyclic A-type analogs implies that a complete H/D exchange should remove the isobaric property and reveal the actual structure. For comparison, the same procedure was applied on a concentrated cranberry extract, which was shown to possess more of the A-type PAC [10]. Therefore, a direct comparison of the HPLC-HRMS/MS spectra could add valuable information. In addition, the presence of a novel cyclic B-type hexameric PAC is proposed in red wine.

Experimental

LC-MS grade solvents and additives were purchased from Sigma Aldrich Srl (Milano, Italy). D2O used was 99.9% of D atoms. Wine samples (Lagrein, 2016 vintage) were donated by a local winery (Kellerei Bozen, Bolzano, Italy). Dried cranberries were purchased from a local market.

Sample Preparation

No preliminary extraction was applied to the wine samples. Dried cranberries were initially extracted according to a published protocol [10]. Briefly, 15 g of dried and refrigerated cranberries were extracted with 135 mL of acetone/water/acetic acid (70:29.5:0.5 v/v). The mixture was sonicated (99 W for 5 min at 50 oC), kept at 50 oC for 30 min, vortexed, and centrifuged at 3300 rpm. All samples (wine and cranberries) were concentrated at low pressure (9–10 mbar) followed by 30 min of gentle N2 flux; then they were reconstituted to a final concentration 10 to 30 times higher than the starting sample. When H/D exchange was performed, the samples were first dried and then reconstituted 3–4 times in pure deuterium oxide before fluxing N2 and reconstituting in the deuterated mobile phase A. The samples were always filtered (0.2 μm, regenerate cellulose) before injection.

HPLC-HRMS/MS Analysis

The HPLC-HRMS/MS method applied was adapted from a published report [22]. The HPLC-HRMS system used consisted of a QExactive HRMS instrument (Thermo Fisher Scientific, Rodano, Milano, Italy) coupled to an Agilent 1260 HPLC (Agilent Technologies Italia S.p.A., Cernusco sul Naviglio, Milano, Italy) with a 16-channel DAD detector. The separation was performed at a flow rate of 1 mL·min-1 with a ODS Hypersyl C18 LC column (125 mm × 4.6 mm i.d., 5 μm, Thermo Fisher Scientific) protected with a HPLC pre-column filter (ODS Hypersil, 5 μm pore size, 10 × 4 mm drop-in guards, Thermo Fisher Scientific). A contact closure electronic board provided synchronization of the HPLC with the mass spectrometer detector. The mobile phase consisted of a combination of solvent A (0.1% v/v formic acid in 0.02 mol·L-1 ammonium formate in water or 0.1% v/v deuterated formic acid in 0.02 mol·L-1 fully deuterated ammonium formate in deuterium oxide) and solvent B (0.1% v/v formic acid in saturated ammonium formate acetonitrile or 0.1% v/v deuterated formic acid in saturated fully deuterated ammonium formate acetonitrile, LC-MS grade). The gradient was set as follows: from 5% B at 0 min to 25% B at 21 min, then to 95% B at 22 min to 27 min, to 5% at 28 min, followed by a re-equilibration step (5% B) at 32 to 35 min. The DAD spectra were recorded from 210 to 600 nm and provided real-time monitoring at 280, 320, 365, 420, and 520 nm (+/– 4 nm). A post-column flow splitter was used to feed both analyzers in parallel (DAD and HRMS) at a fixed ratio. For full MS analysis, the mass spectrometer was operated in positive ionization mode using the following conditions for the HESI ionization source: sheath gas at 20 (arbitrary units), aux gas at 5 (arbitrary units), aux temperature 250 °C, spray voltage at +3.00 kV, capillary temperature at 320 °C, and rf S-lens at 70. The mass range selected was from 500 to 2000 m/z with a full MS set resolution of 70,000 (@200 m/z), AGC target at 3e6, max. injection time of 300 ms. To maximize the sensitivity, data-dependent LC-MS/MS experiments were run on the N2 concentrated samples and on a selection of 35 ions only (manually specified in the inclusion list): full-MS parameters were kept as shown, MS/MS AGC target 106, max. injection time 300, FT-MS set resolution 35,000, loop count 5, isolation window 2 m/z, normalized collision energy 15 eV. For data-dependent settings: minimum AGC target 3.103, apex trigger 2 to 8 s, charge exclusion 3–8 and higher, dynamic exclusion 3 s, “if idle” tool set to “do not pick others.” LC-MS/MS experiments were also tested in negative ionization mode (spray voltage –3.50 kV) but with lower response factor; therefore, they are not discussed any further. Lock masses were consistently employed to correct mass deviations across the Full MS acquisition range throughout the experiments. When deuterium oxide was employed, the lock masses were modified accordingly into the main instrument method. The HPLC-DAD data were collected and analyzed by OpenLab software whereas the MS data and results were collected and analyzed by Xcalibur 3.1 software. MS/MS fragmentation interpretation was performed on the basis of comparison with the cited literature, and it was aided by Mass Frontier 7.0 software.

Results and Discussion

Red wine samples were directly applied for a preliminary HPLC-HRMS characterization of the PAC species. In Figure 1, the extracted ion chromatograms (EIC) for the novel B-type cyclic tetramer (CTP) and pentamer (CPP) and their relative MS/MS spectra (Figure 2) are shown, in agreement with recent reports for red wines [5, 6]. The species compatible with the cyclic hexamer analog (CHP) is also proposed. The assignment of CHP in Figures 1e and 2e (C90H72O36, CHP) was given according to the theoretical structure of the hexamer (adding an additional monomer unit in the cyclic structure of the known pentamer). The associated [M+H]+ (m/z 1729.3876) EIC showed a retention time lower than its relative M+2 non-cyclic analog such as previously observed for the tetramer and pentamers [6]. Moreover, the retention times’ distribution was very different between the cyclic and the linear analogs, indicating a much more limited variability of structures for the cyclic one; this is presumably compatible with species with fewer possible conformational arrangements, regio- and stereochemistries. MS/MS fragmentations are also consistent with previous reports on the fragmentation pathways for the linear B-type PAC [8]. The interpretation was given for CHP, but the same considerations held true for the other two cyclic oligomers (CTP and CPP). The fragmentations of the linear B-types are well represented in the literature [8], and they were exploited here only as a mean of comparison with the cyclic ones. PAC MS/MS fragmentations were shown to follow three main fragmentation pathways, namely a retro Diels-Alder ring opening (RDA), a quinone methide monomer loss (QM), and a heterocyclic ring fission (HRF) [8]. For these proanthocyanidins (containing only (epi)catechin units), the RDA pathway yields losses of 152.047 Da (one 3,4-dihydroxyphenylacetaldehyde) and a further 18.010 Da (water); the QM mechanism removes one monomer although in an oxidized (–2[H]) form (288.063 Da); HFR yields a loss of 126.03 Da (one phloroglucinol unit). The losses are indicated in Figure 2e (CHP) by their respective abbreviations along with the neutral loss between the closest relative (parent) and its derived fragments. QM in particular is responsible for the loss of an entire monomer unit, albeit oxidized (–2[H]). This specific fragmentation pathway was particularly useful since losses of entire monomer units produce fragments easily interpretable as the smaller oligomers. In our case, this was even more interesting considering the different fragmentation preferences imposed on a closed cyclic structure. Analogous considerations could be extended to the MS/MS spectra of the pentamer CPP and tetramer CTP in Figure 2. For all the cyclic analogs, the main fragmentation pathways observed were again RDA followed by water loss and QM. In particular, from the pseudo-molecular ion it was possible to appreciate the progressive losses of (epi)catechin moieties down to the monomer unit (e.g., for CHP, m/z 1729.361, 1153.262, 865.197, 577.132, and 289.069). Moreover, the other main fragmentation pathway was RDA (with and without a further loss of water, e.g., m/z 1577.334, 1289.269) from the QM generated (just mentioned) oligomeric fragments. HFR (which yields a phloroglucinol unit from one (epi)catechin moiety) contributed less than the other two mechanisms and was observed only at the dimeric stage (m/z 451.100). These considerations may not hold true for higher collision energies. Notably, different fragmentation preferences (at this same collision energy) were observed for the linear B-type forms (compared in Figure 2), concerning in particular the number of monomers lost: whereas the linear forms showed fragments corresponding to all the progressive monomer (–288.063 Da) losses (from hexamer to monomer), the 5- and 6-members cyclic analogs showed fewer fragments (e.g., tetramer and trimer for CHP). Additionally, the cyclic species were less activated (fragmented) than their linear analogs. The major difference was the trimeric fragment: whereas in the linear forms it gave m/z 867.211 and m/z 865.197, for the cyclic forms no m/z 867.211 was observed at this fragmentation energy. However, the fragmentation pathways for B-type linear PAC are well known and discussed in the literature [8].

Figure 1
figure 1

FullMS (ESI+, extracted ion chromatograms – EIC – shown with 4 ppm filter applied) analysis of a 2016 Lagrein sample (see HPLC and MS settings in Material and Methods section). Extracted ion chromatogram of (a) 1153.2608 m/z (CTP), (b) 1155.2765 m/z (LTP), (c) m/z 1441.3242 (CPP), (d) m/z 1443.3399 (LPP), (e) m/z 1729.3876 (CHP), and (f) m/z 1731.4032 (LHP)

Figure 2
figure 2

MS/MS (ESI+, MS/MS) analysis of a 2016 Lagrein sample (see HPLC and MS settings in Material and Methods section). MS/MS spectra (NCE = 15 eV) of (a) CTP (3.6–4.0 min), (b) LTP (8.8–13.9 min), (c) CPP (4.0-4.2 min), (d) LPP (9.8-13.7 min), (e) CHP (10 min), and (f) LHP (11.8–13.3 min). Spectra were averaged over their correspondent TIC in the indicated r.t. ranges. The main MS/MS fragmentation pathways are proposed in the picture for CHP. The same consideration can be done for the corresponding spectra of CPP and CTP

With respect to the cyclic analogs, the linear ones showed a higher level of fragmentation, in particular through the QM mechanism. In all spectra, m/z 577.131 was the most intense peak, whereas in the cyclic forms the pseudo-molecular ions (m/z 1729.381, 1441.317, and 1153.259 for CHP, CPP, and CTP, respectively) were the highest. We could consider this as an interesting marker of distinction between the cyclic and the linear forms. However, also RDA (with or without further water loss) contributed in the MS/MS of the linear forms (e.g., m/z 1291.281, 1273.279, 1003.225, 715.161, and the three main m/z 695.136, 427.101, 425.085, and 409.090). HFR contributed as well (e.g., m/z 1317.816 and 451.100). However, the most remarkable difference between cyclic and linear species is the preference towards the trimeric and dimeric fragments. For the cyclic PAC, m/z 865.197 and 577.131 are the main (and only) trimeric and dimeric peaks, whereas for the linear PAC, m/z 867.209 and 579.147 also contribute. Considering that these fragments are produced via the reported QM mechanism (in which the monomer released from the main chain is virtually oxidized), we could infer that this difference is directly related to the cyclic nature of CHP, CPP, and CTP. In fact, the presence of the additional head-tail C–C bond forced the main chain also to be in a –2[H] state, and this may generate the –2 Da difference in certain peaks for the cyclic forms. The other main differences, as already stated, are the fragmentation preferences: (1) trimeric and tetrameric forms are almost negligible in the cyclic pentamer and hexamer spectra; (2) the much higher fragmentation level of the linear PAC with respect to the cyclic species.

Despite all these findings, it was realized that a fundamental point had to be clarified before proceeding further: the elemental compositions of the proposed novel cyclic oligomers are identical to those of analog compounds belonging to another already known PAC class, namely the A-type proanthocyanidins. This similarity is exposed plainly in Table 1 for the tetramer.

Table 1 Theoretical masses (pseudo-molecular ion, [M H +H]+ and [M D +D]+) and elemental compositions for native and deuterated tetramers

In this report, only the tetrameric species could be investigated any further; low concentrations and a decreased sensitivity with deuterium oxide prevented the analysis of the less abundant higher oligomers.

H/D exchange applied to the tetrameric m/z 1153.2608 species (ESI+) should be able to distinguish between the isobaric cyclic B-type and linear A-type, due to their different number of exchangeable protons (20 and 19, respectively). In Figure 3, the molecular structures of the tetramers considered herein are presented along with their deuterated analogs in the applied conditions. Figure 3 shows the theoretical molecular structures for the tetrameric species under investigation. All the stereogenic centers’ configurations are not resolved and the inter-flavanol bonding preferences (C4-C6 or C4-C8) are arbitrarily shown. Both must not be considered defined within this work since their resolution was beyond the purpose of this study.

Figure 3
figure 3

(a) A-type non-cyclic tetrameric PAC (A-LTP). (b) B-type cyclic tetrameric PAC (B-CTP). (c) B-type non-cyclic tetrameric PAC (B-LTP). (d) Deuterated A-type non-cyclic tetrameric PAC (A-LTP-d 19 ). (e) Deuterated B-type cyclic tetrameric PAC (B-CTP-d 20 ). (f) Deuterated B-type non-cyclic tetrameric PAC (B-LTP-d 20 ). The A-linkage (in Figure 3a) is indicated by an arrow. Likewise, the head-tail B-type C–C bond that characterizes the proposed novel cyclic B-type PAC (Figure 3b) is shown by an arrow. Inter-(epi)catechin bonds are shown as C4-C8 for simplicity but also C4-C6 cannot be excluded

The A-linkage (in Figure 3a) is indicated with an arrow. Likewise, the B-linkage that differentiates the proposed novel cyclic B-type PAC (Figure 3b) from the non-cyclic B-type PAC (Figure 3c) is indicated with an arrow.

The species in Figure 3a and b are isobaric. However, as aforementioned, all the O–H protons should undergo H/D isotopic exchange, whereas all the C–H protons should not. Therefore, with an A-linkage (A-type LTP, Figure 3a) one exchangeable proton (O–H) is lost with respect to the B-type LTP species (Figure 3c) since the A-type C–O bond is in its place. Conversely, no exchangeable proton is lost with a new C–C B-linkage (B-type CTP, Figure 3b) in place of two C–H bonds. As a result, the two deuterated species (in Figure 3d and e, respectively) should not be isobaric anymore, as shown in Table 1.

The A-type PAC (e.g., 3-A) were recently characterized for the first time in grape seeds by a MSn approach [23]. Notably, in that paper the same difference of 2 Da from the linear B-type PAC (e.g., 3-C) was found; however, this was directly associated only with the linear A-type. A-type PAC are reported to be produced from B-type PAC (3-C) upon oxidation. Their abundance in some fruits suggested further study comparing cranberry extracts with wine [10]. To facilitate interpretation, B- and A-type oligomers are given with their exchangeable protons’ positions highlighted (Figure 3). According to these similarities in the elemental composition, three possibilities were considered: (1) the tetramer m/z 1153.2608 does not correspond to a cyclic B-type tetrameric PAC (3-B, B-type CTP) proposed by [5, 6], but it does correspond to the already known non-cyclic A-type tetrameric PAC (3-A, A-type LTP), which was shown to be scarce but not completely unknown in red wines [7, 9]; (2) this tetramer (m/z 1153.2608) really belongs to a novel class of cyclic PAC (B-type CTP) in both wine and cranberries, with further deep consequences discussed later on; or (3) the tetramers are isobaric but structurally different in wine (3-B) and cranberries (3-A). The H/D exchange was applied to distinguish between the possible structure of 3-A and 3-B by comparison of retention times, mass spectra, and MS/MS fragmentations. The isobaric structures of 3-A and 3-B showed one major difference that could be revealed by H/D exchange HPLC-HRMS/MS technique: both tetramers (B-type CTP and A-type LTP) have two fewer hydrogen atoms than 3-C (B-type LTP), but in 3-A only one of these two hydrogens is an exchangeable phenolic proton, whereas in 3-B they both are non-exchangeable protons (one benzylic and the other arylic). Benzylic and arylic protons are in fact not expected to exchange with deuterium under the mild conditions used. The task was pursued: (a) by means of a HPLC-HRMS analysis of the samples before and after a complete H/D exchange and applying pure deuterium oxide in the HPLC mobile phase, so that the m/z 1153.2608 ([C60H49O24]+) pseudo-molecular ion of 3-A should add 20.1256 Da (3-D, [C60H29D20O24]+) whereas the pseudo-molecular ion of 3-B should add 21.1318 Da (3-E, [C60H28D21O24]+); (b) by comparing the obtained MS/MS patterns of wine and cranberries in water and deuterium oxide. For wine, the results are reported in Figure 4.

Figure 4
figure 4

H/D exchange HPLC-HRMS/MS analysis of a Lagrein sample. Extracted ion chromatograms (EIC) in FullMS are reported at a 2 ppm applied filter set. Mass shifts from the corresponding peak in H2O spectrum are reported in italic

The retention times (of the cyclic and the linear oligomers alike) in deuterium oxide were delayed for over 60 s with respect to water (in Figure 1), which was reasonable since deuterium oxide exerts a weaker hydrogen bonding than does water. The retention time differences upon H/D exchange were already reported as dependent upon the specific compound eluted [19].

A loss in relative abundance was observed with deuterium. Nonetheless, these preliminary results suggest that the presence of the proposed cyclic compounds only (and not the non-cyclic A-type LTP) would properly explain the results for the tetramer. Traces of a co-eluting species compatible with the A-type LTP were indeed observed (Figure 4). However, this was due to a contribution of an overlapping m+2 isotopic peak of a yet unidentified m/z 1171.536 (z = +2) eluting at 5.4 min (EIC not shown). Nonetheless, the relative abundance of the peak at m/z 1173 is negligible (less than 2% of B-type CTP). The MS/MS fragmentation pattern was also recorded. Notably, all peaks were shifted with respect to the corresponding spectrum in water (Figure 2) after H/D exchange. These values are indicated in italics in the spectrum. MS/MS spectrum analysis has been performed in the spectrum in H2O and reported in the discussion (and in Figure 2). The fragmentation pattern for deuterated CTP reported in Figure 4 is completely analogous to the one reported in Figure 2-A. The pseudo-molecular ion peak at m/z 1174.387 showed a similar level of fragmentation with respect to the native (hydrogenated) species. The fragmentation mechanisms were altogether the same. Just as for the cyclic tetramer of Figure 2, the trimer, dimer, and monomer fragments were also produced here via the QM mechanism (m/z 881.294, 588.201, and 295.107, respectively). Moreover, RDA mechanism accounted for the losses of 3,4-dihydroxyphenylacetaldehyde-d3 with or without the further loss of deuterated water (m/z 1019.321, 706.204, and 415.127). Notably, looking at the progressive loss of monomer units of Figure 4, it is possible to notice a difference of exactly one monomer unit containing five exchangeable protons. Taking into account the difference given by the deuterium needed for providing the ionization, exactly 20 u are added to the tetramer pseudo-molecular ion, 15 u to the trimeric form, 10 u to the dimeric form, and 5 u to the monomeric unit.

The results for the cranberry extract are shown in Figure 5. As seen for wine, the best explanation for the observed results is the presence of B-type CTP species and not the A-type LTP, in both cranberries and wine. However, the prevalence of A-type PAC in cranberries has been proposed in previous works [10]. On the contrary, the retention times of the shown m/z 1174.391 EIC in the two samples (wine and cranberries) were exactly the same (+/– 0.1 min). Again, a small contribution from the nearby m/z 1171.536 (z = +2) m+2 isotopic peak was observed (here the contribution was slightly greater than in wine, close to 3%). The MS/MS fragmentation pattern was also recorded and the considerations made for the MS/MS spectra of the wine hold here also.

Figure 5
figure 5

H/D exchange HPLC-HRMS/MS analysis of a cranberry extract. Extracted ion chromatograms (EIC) in FullMS are reported at a 2 ppm applied filter set

The related experiments in water for wine and cranberries are shown together in Figure 6. Figure 6 outlines the likelihood for the m/z 1153.2608 tetramer (r.t. 3.8 ± 0.1 min) to be exactly the same in red wine and in the cranberry extract. The fragment at m/z 1153.2608 was shown to convert into m/z 1174.3926 upon H/D exchange in both wine and the cranberry extract (Figure 4 and Figure 5). Therefore, the m/z 1153.2608 tetramer (r.t. 3.8 ± 0.1 min) can be associated only with the cyclic B-type PAC (CTP) as proposed in recent reports [5, 6]. Instead, in previous literature this compound had been associated only with the A-type LTP [7]. The MS/MS experiment (Figure 6, lowest layer) supported the hypothesis that the main tetrameric species (r.t. 3.8 ± 0.1 min) was the B-type CTP. In fact, the MS/MS spectra for wine and cranberries in water were completely indistinguishable. The obtained fragments in wine and cranberries differed from those previously reported for A-type PAC [7]; in particular, the fragment at m/z 863 found in that report and associated with an A-type trimer fragment with two A-linkages was not observed here. The fragmentation pattern may be more complex than assumed, since the m/z 865 fragment associated to a trimer (shown in Figure 6) was assigned in that report to an A-type trimeric fragment with one A-linkage only. Nonetheless, in that report the fragmentation energy used was 30 eV, and here it was only 15 eV. Similarly, another fragment reported for an A-type dimer was m/z 575 [24], but that fragment was not found here. Instead, m/z 577 was observed in all the samples analyzed and for all the oligomers analyzed in wine (Figure 2), which again may be a confirmation of the hypothesis that the compounds observed are of B-type only.

Figure 6
figure 6

HPLC-HRMS/MS analysis of a wine sample and a cranberry extract in H2O. Extracted ion chromatograms (EIC) in FullMS are reported at a 2 ppm applied filter set

The EIC profiles of m/z 1153.2608 for the cranberry extract (Figure 6) showed isobaric peaks eluting at later retention times (>7 min). A similar profile was shown with deuterium oxide (Figure 5). This profile suggests that some linear A-type tetrameric PAC (A-type LTP) may be present in cranberries, even if the major component referring to m/z 1153.2608 (r.t. 3.8 ± 0.1 min) has to be the B-type CTP, as it is in wine. In order to investigate whether the m/z 1153.2608 peaks at r.t. >7 min were in fact due to A-type LTP, the MS/MS spectra of m/z 1153.26 at 3.9 min, 11–14 min, and the MS/MS spectrum of m/z 1155.27 at 9.2–9.4 min were compared in Figure 7. As expected, the MS/MS of peaks at r.t >7 min for m/z 1153.26 (7-B) is compatible with the proposed A-type non-cyclic tetramer (A-type LTP) [24]. 7-B displays fragments with –2 Da values with respect to the spectrum at 3.8 ± 0.1 min (7-A, e.g., the m/z 863 and m/z 575 discussed previously). They still derive from the QM main fragmentation mechanisms; however, the presence of an A-type linkage (thus, more oxidized) in the pseudo-molecular ion might be kept in the fragments (e.g., see also the dimeric m/z 575.117). Moreover, the fragment m/z 533.105 is present only for 7-B species (likely to be produced via the HFR mechanism, and further, a double loss of water from m/z 695.136, which is in fact present for all of the species in all of the spectra).

Figure 7
figure 7

MS/MS spectra of: (a) m/z 1153.26 at 3.8–3.9 min (B-type tetrameric CTP); (b) m/z 1153.26 at 11–14 min (A-type tetrameric LTP); (c) m/z 1155.27 at 9.2–9.4 min (B-type tetrameric LTP)

The MS/MS at 3.8 ± 0.1 min (7-A) matches instead most of the ion fragments of m/z 1155.27 MS/MS (7-C, B-type LTP), suggesting an M-2 species (only for the pseudo-molecular ion) compatible with the proposed B-type cyclic tetramer (B-type CTP). Additional evidence supporting our hypothesis is that 7-A and 7-B (despite being isobaric species) showed different relative abundances of fragments with respect to their pseudo-molecular ion: for 7-A the second most intense fragment (m/z 289.07) is about 30% of the pseudo-molecular ion relative abundance, whereas for 7-B (m/z 533.10) the most intense fragment is higher than 75%. A similar difference is noted between all the cyclic and linear oligomers in Figures 2 and 7. Hence, it is probable that the linear analogs (A- and B-type alike) can be fragmented more easily than the cyclic analogs, at least under these conditions.

Conclusions

Hydrogen/deuterium exchange provided a direct method to support the structural resolution of unconventional PAC in red wine and cranberry extract. With this technique, it was demonstrated that the main, most abundant PAC tetramer (referring to m/z 1153.2608, eluting at 3.9 min) in wine belonged to the proposed novel class of cyclic oligomeric B-type PAC. Moreover, this compound was identical (same retention times, exact masses, and MS/MS fragmentations) to the major tetrameric PAC found in cranberries. Traces of A-type LTP were, however, observed in cranberries and (much less) in wine at later retention times, but they were negligible with respect to the B-type CTP in both samples. In addition, a novel B-type cyclic hexameric proanthocyanidin (CHP) was also tentatively identified for the first time in wine and was assigned by analogy with the previously discovered cyclic oligomers. In fact, the identification of CHP was proposed: (1) on the basis of its exact mass; (2) on the basis of its polarity (hence its retention time) with respect to the linear analogs; (3) on the basis of the presence of only one isomer (similarly to what was observed with CTP and CPP) versus the many isomers of the linear form; (4) on the basis of its MS/MS fragmentation pattern, similar to that observed with CTP and CPP and showing fragments typical for B-type PAC.