Identification of A-Type Proanthocyanidins in Cranberry-Based Foods and Dietary Supplements by Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry, First Action Method: 2019.05

Abstract

Background

Cranberry proanthocyanidins (c-PAC) are oligomeric structures of flavan-3-ol units, which possess A-type interflavan bonds. c-PAC differs from other botanical sources because other PAC mostly have B-type interflavan bonds. Cranberry products used to alleviate and prevent urinary tract infections may suffer from adulteration, where c-PAC are replaced with less expensive botanical sources of PAC that contain B-type interflavan bonds.

Objective

Identifying the presence of A-type interflavan bonds in cranberry fruit and dietary supplements.

Methods

Thirty-five samples reported to contain A-type PAC (cranberry fruit and cranberry products) and 36 samples reported to contain B-type PAC (other botanical sources) were identified and differentiated using MALDI-TOF MS, deconvolution of overlapping isotope patterns, and principal component analysis (PCA).

Results

Our results show that both MALDI-TOF MS and deconvolution of overlapping isotope patterns were able to identify the presence of A-type interflavan bonds with a probability greater than 90% and a confidence of 95%. Deconvolution of MALDI-TOF MS spectra also determined the ratio of A-type to B-type interflavan bonds at each degree of polymerization in cranberry fruit and cranberry products, which is a distinguishing feature of c-PAC in comparison to other botanical sources of PAC. PCA shows clear differences based on the nature of the interflavan bonds.

Conclusions

MALDI-TOF MS, deconvolution of overlapping isotope patterns of MALDI-TOF MS spectra, and PCA allow the identification, estimation, and differentiation of A-type interflavan bonds in cranberry-based foods and dietary supplements among other botanical sources containing mostly B-type interflavan bonds.

Cranberries (Vaccinium macrocarpon Aiton) grown in North America are mostly processed into products such as juice, sauce, sweetened dried cranberries, and dietary supplements. In the United States, cranberry production for 2018 was estimated to be 9.72 million barrels (one barrel is equivalent to 45.4 kg of cranberries) (1). Cranberry fruit and cranberry products are sources of proanthocyanidins (PAC), which are widely researched because of their putative health benefits (2, 3). The consumption of cranberry PAC (c-PAC) has been associated with the prevention of urinary tract infections. Recent studies suggest that c-PAC agglutinate extra-intestinal pathogenic Escherichia coli (ExPEC) and inhibit ExPEC adhesion and invasion of epithelial cells (2, 4, 5). The bioactivity of c-PAC against E. coli is greater than other PAC because c-PAC contains one or more A-type interflavan bonds (2, 4–6).

PAC are oligomeric structures composed of repeating flavan-3-ol units (7). PAC are structurally differentiated according to the stereochemistry, the pattern of hydroxylation of flavan units (B-ring), the degree of polymerization (DP), and the nature of the interflavan bonds (A- versus B-type) (7–9). In B-type interflavan bonds, PAC form bonds between C4-C8; in A-type interflavan bonds, PAC have an additional bond between C2-O-C7 (7, 9, 10).

Cranberry products used to alleviate and prevent urinary tract infections may suffer from adulteration, where cranberries are replaced with less expensive botanical sources of PAC that contain B-type interflavan bonds (11, 12). Manufacturers are able to adulterate cranberry products while still claiming high PAC content because the majority of claims are based on results from a colorimetric assay, such as the 4-(dimethylamino)cinnamaldehyde (DMAC) reaction to produce a chromophore that is detected at 640 nm (11–13). While useful for quantifying PAC, the DMAC assay is unable to differentiate and authenticate the source of PAC (14, 15).

Current methodologies for identification, classification, and authentication of PAC are inadequate and time consuming due to the structural heterogeneity and complexity of PAC, posing a challenge for qualitative and quantitative analysis (2). The goals of this study were (a) to develop a methodology to identify the presence of A-type interflavan bonds in cranberry fruit and cranberry products (following the AOAC standard method performance requirements (SMPR) 2017.004) (16) and (b) to differentiate amongst other botanical sources containing mostly B-type interflavan bonds. The proposed methodology could be used to support manufacturers’ claims of the presence of A-type interflavan bonds in cranberry products.

AOAC Official Method 2019.05

A-type Proanthocyanidins in Cranberry-Based Foods and Dietary Supplements

Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry

First Action 2019

(Applicable for identifying the presence of A-type proanthocyanidins, interflavan bonds, in cranberry fruit and cranberry products.)

Caution: Refer to Material Safety Data Sheets (MSDS) for all chemicals prior to use. Use all appropriate personal protective equipment and follow good laboratory practices.

A. Principle of Method

Proanthocyanidins (PAC) in samples are extracted with 70% v/v acetone, purified by chromatography on LH-20 column, and analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). The deconvolution of overlapping isotopic patterns of MALDI-TOF MS spectra are used for the identification of A-type interflavan bonds in PAC.

B. Sample Materials

• Cranberry fruit (Vaccinium macrocarpon Aiton).—Varieties Scarlet Knight, BenLear A2, G1 A57, G2A20, and Stevens A10 (Habelman Bros. Company, Tomah, WI, USA).

• Cranberry products.—capsules

• Apple fruit (Malus pumila).—Varieties Cameo, Honeycrisp, Jazz, Ambrosia, SnapDragon, RubyFrost, SweeTango, Kanzi, Eve, Kiku, Red Delicious, McIntosh, Golden Delicious, Braeburn, Cripps Pink, Pink Lady, Fuji, Gala, Granny Smith, and Paula Red.

• Cocoa (Threobroma cacao L).—powder cans

• Grape Skins (Vitis sp).—Cotton Candy, Summer Royal, Autumm King, Crimson Seedless, Thompson Seedless, and Flame Seedless.

• Black chokeberry fruit (Aronia melanocarpa Michx).

C. Apparatus

• Bruker UltraFlex^® III MALDI TOF/TOF mass spectrometer equipped with a SmartBeam^TM laser (Billerica, MA, USA). MALDI-TOF MS conditions: ion source 1 (25.0 kV), ion source 2 (20.6kV), lens voltage (9.5 kV), reflector 1 (26.5 kV), reflector 2 (14.41 kV), pulsed ion extraction (130 ns), detector gain (2.64×; 1548V), preamplifier filter bandwidth (small), and digitizer sampling frequency (100 Hz). Spectra are the sum of different locations in each well, accumulating a total of 2000 shots with deflection set at 900 Da. FlexControl, and Flex Analysis software (Bruker Daltonik GmbH, Bremen, Germany) are used for data acquisition and data processing, respectively.

• Balance, analytical.—Minimum weighing capacity of 0.01 mg.

• Bench-top Dewar flask.

• High speed blender

• Ultrasonic bath.

• Centrifuge.

• Rotary vacuum evaporator.

• Centrifuge tubes.—50 mL HDPE plastic with screw caps.

• Pipetters.—with disposable tips.

• Strong cation resin tips PureSpeed IEX 1 mL/20 µL (Rainin, Oakland, CA, USA).

D. Reagents

• 2,5-dihydrobenzoic acid (DHB)—Sigma-Aldrich (St. Louis, Missouri, USA).

• 2-propanol—Sigma-Aldrich (St. Louis, Missouri, USA).

• Acetone—Fisher Scientific (Fair Lawn, New Jersey, USA).

• Bradykinin—Sigma-Aldrich (St. Louis, Missouri, USA).

• Cesium trifluoroacetate—Sigma-Aldrich (St. Louis, Missouri, USA).

• Ethanol—Decon Laboratories, Inc. (King of Prussia, Pennsylvania, USA).

• Glucagon—Sigma-Aldrich (St. Louis, Missouri, USA).

• Liquid nitrogen

• Methanol—Fisher Scientific (Fair Lawn, New Jersey, USA).

• Sephadex LH-20™—GE Healthcare, 18–111 µm (Uppsala, Sweden).

E. Reagent Preparations

• Acetone solution.— (1) 70% v/v. Place 700 mL acetone into 1 L volumetric flask, dilute to volume with deionized water. (2) 80% v/v. Place 800 mL acetone into 1 L volumetric flask, dilute to volume with deionized water.

• Ethanol: methanol solution.—1:1 v/v. Place 500 mL ethanol into 1 L volumetric flask, dilute to volume with methanol.

• 2,5-dihydroxybenzoic acid (DHB) solution.—1.30 M in methanol. Dissolve 400 mg of 2,5-dihydroxybenzoic acid in methanol and dilute to 2 mL with methanol.

  Note: Prepare ≤ 1 h before use.

• Cesium trifluoroacetate solution.— (1) Stock solution.—5 millimolar (mM) in methanol. Dissolve 2.46 mg of cesium trifluoroacetate in methanol and dilute to 2 mL. (2) Working solution.—Dispense 50 µL stock solution into 1 mL volumetric flask, dilute to volume with methanol

• Bradykinin solution.— (1) Stock solution.—100 micromolar (µM) in water. Dissolve 0.21 mg of bradykinin in water and bring to 2 mL. (2) Working solution.—Dispense 40 µL stock solution into 1 mL volumetric flask, bring to final volume with water.

• Glucagon solution.— (1) Stock solution.—100 micromolar (µM) in water. Dissolve 0.69 mg of glucagon in water and dilute to 2 mL. (2) Working solution.—Dispense 40 µL stock solution into 1 mL volumetric flask, dilute to water.

F. Column Preparation

• Chromatographic tube.—Use 25 mm id and 250 length chromatographic tube with medium porosity fritted glass.

• Preparation, wash, and activation of Sephadex LH-20^TM resin.—weigh 25 g Sephadex LH-20^TM resin into 600 mL beaker and add 200 mL deionized water. Mix by swirling vigorously. Let settle and decant upper layer. Wash slurry with three times with 100 mL of deionized water, decanting upper layer each time.

• Preparation of column.—Add an aqueous slurry of LH-20^TM to 10 cm bed depth (wet resin). Wash column with 100 mL deionized water and equilibrate the column with 100 mL ethanol. Maintain head of 1 cm liquid on column throughout assay.

G. Sample Preparation

• Whole fruits.—Cut the whole fruits into small pieces of approximately 1 cm³. Weigh 200 g of the pieces of fruits and transfer into a 2 L bench-top Dewar flask. Add 1 L liquid nitrogen into the 2 L bench-top Dewar flask (Caution: perform under good ventilation and wear insulated gloves and face shield to protect against liquid nitrogen). Decant the frozen pieces of fruits into a blender and grind the sample thoroughly. Let evaporate the remaining liquid nitrogen from the sample powder.

• Capsules.—Open at least 25 capsules and collect the powder. Discard the capsule shells.

• Powders.—Preparation is not required.

• Grape Skins.—Peel by hand 0.5 kg of grapes. Weight 40 g of the grape skins and transfer into a 1 L bench-top Dewar flask. Add 300 mL liquid nitrogen into the 1 L bench-top Dewar flask. (Caution: perform under good ventilation and wear insulated gloves and face shield to protect against liquid nitrogen). Decant the frozen grape skins into a blender and grind the sample thoroughly. Let evaporate the remaining liquid nitrogen from the sample powder.

H. Sample Extraction

Weigh 20 g of the powder into a 250 mL Erlenmeyer flask. Add 100 mL 70% (v/v) acetone solution into the 250 mL Erlenmeyer flask. Swirl the Erlenmeyer flask and sonicate for 15 min. Decant the extract into two 50 mL centrifuge tube. Centrifuge the extract at 1800 g for 10 min at 15°C and pooled together the supernatant from both tubes. Evaporate the supernatant to dryness by rotary vacuum evaporation at 35°C. Solubilize the extract in 10 mL of ethanol. Load 5 mL of the ethanolic extract onto the prepared chromatographic column. Sequentially elute the column with 150 mL ethanol, 150 mL ethanol: methanol solution, and 200 mL 80% (v/v) acetone solution. Evaporate the acetone fraction to dryness using a rotary evaporator at 35°C. Solubilize the PAC extract in 1.5 mL methanol.

I. Sample Deionization, DHB Solution Deionization, Deionized DHB Solution/Cesium Trifluoroacetate Solution, and Bradykinin/Glucagon Solution Procedures

• Sample deionization.—Pipette 500 µL of the PAC extract into a 1.5 mL transparent chromatography vial. Pipette the 500 µL of the PAC extract into the strong cation resin tip PureSpeed IEX 1 mL/20 µL. Maintain the tip inside the 1.5 mL transparent chromatography vial and mix the PAC extract by pipetting a minimum of 5 times up and down in the pipette tip. Dispense the deionized sample into a new 1.5 mL transparent chromatography vial.

• DHB solution deionization.—Pipette 500 µL of the DHB solution into a 1.5 mL transparent chromatography vial. Pipette the 500 µL of the DHB solution into the strong cation resin tip PureSpeed IEX 1 mL/20 µL. Maintain the tip inside the 1.5 mL transparent chromatography vial and mix the DHB solution by pipetting a minimum of 5 times up and down in the pipette tip. Dispense the deionized DHB solution into a new 1.5 mL transparent chromatography vial.

• Deionized DHB solution/cesium trifluoroacetate solution.—Mix 400 µL deionized DHB solution with 400 µL cesium trifluoroacetate working solution.

• Bradykinin/glucagon solution.—Mix 400 µL bradykinin working solution with 400 µL glucagon working solution

J. MALDI-TOF MS Spectra

• MALDI-TOF MS stainless steel target plate cleaning procedure.—Place the MALDI-TOF MS stainless steel target plate into a suitable container and pour in enough 2-propanol to cover the target surface. Sonicate the MALDI-TOF MS stainless steel target plate for 15 min. Remove the MALDI-TOF MS stainless steel target plate and rinse it thoroughly under deionized water. Rinse the MALDI-TOF MS stainless steel target plate with ethanol. Let the MALDI-TOF MS stainless steel target plate dry completely at room temperature. Store the clean MALDI-TOF MS stainless steel target plate in the container provided.

• Bradykinin/glucagon external standard procedure.—Pipette 0.5 µL of the bradykinin/glucagon solution onto five wells of the MALDI-TOF MS stainless steel target plate. Let the bradykinin/glucagon solution dry completely at room temperature. Overlay each bradykinin/glucagon solution position with 0.6 µL deionized DHB solution/cesium trifluoroacetate solution by pipetting a minimum of 10 times up and down in the pipette tip. Let the mix bradykinin/glucagon solution and deionized DHB/cesium trifluoroacetate solution dry completely at room temperature.

• Sample Preparation procedure.—Pipette 0.5 µL of the deionized sample (0.05–0.5 mg c-PAC equivalents) onto 5 wells of the MALDI-TOF MS stainless steel target plate. Let the deionized sample dry completely at room temperature. Overlay each deionized sample position with 0.6 µL deionized DHB solution/cesium trifluoroacetate solution by pipetting a minimum of 10 times up and down in the pipette tip. Let the mix deionized sample and deionized DHB/cesium trifluoroacetate solution dry completely at room temperature. The sample MALDI-TOF MS stainless steel target plate is now ready for use.

  Notes: (1).—For the quantification of c-PAC, refer to the Inter-laboratory validation of 4-(dimethylamino)cinnamaldehyde (DMAC) assay using cranberry proanthocyanidin standard for quantification of soluble proanthocyanidins in cranberry foods and dietary supplements, First Action Method: 2019.06 (15). (2). Use a new pipette tip to add the deionized DHB solution/cesium trifluoroacetate solution to each well of the MALDI-TOF MS stainless steel target plate.

K. Calculations

• Deconvolution of overlapping isotope patterns of A- to B-type interflavan bonds in PAC.—The absolute intensity (ai) of each PAC peak in the positive reflectron mode MALDI-TOF MS spectra data set is obtained using mMass software. Spectral data are excluded from deconvolution analysis when the signal-to-noise (S/N) ratio is lower than 3. The percentage of A- to B-type interflavan bonds in PAC is solved using the following algebraic matrix: $A^{-1}\times \mathrm{ }b=c,\mathrm{ }\mathrm{where}\mathrm{ }\mathrm{A}-1\mathrm{ }\mathrm{represents}\mathrm{ }\mathrm{the}\mathrm{ }\mathrm{inverse}\mathrm{ }\mathrm{coefficient}\mathrm{ }\mathrm{matrix}\mathrm{of}\mathrm{ }\mathrm{the}\mathrm{ }\mathrm{relative}\mathrm{ }\mathrm{intensity}\mathrm{ }\left(\mathrm{ri}\right)\mathrm{ }\mathrm{of}\mathrm{ }\mathrm{the}\mathrm{ }\mathrm{isotope}\mathrm{ }\mathrm{patterns}\mathrm{for}\mathrm{ }\mathrm{each}\mathrm{ }\mathrm{DP},\mathrm{ }\mathrm{b}\mathrm{ }\mathrm{represents}\mathrm{ }\mathrm{the}\mathrm{ }\mathrm{constant}\mathrm{ }\mathrm{matrix}\mathrm{ }\mathrm{of}\mathrm{ }\mathrm{ai}\mathrm{from}\mathrm{the}\mathrm{ }\mathrm{spectra},\mathrm{ }\mathrm{and}\mathrm{ }\mathrm{c}\mathrm{ }\mathrm{represents}\mathrm{ }\mathrm{the}\mathrm{ }\mathrm{variable}\mathrm{ }\mathrm{matrix}\mathrm{of}\mathrm{linear}\mathrm{ }\mathrm{combinations}\mathrm{ }\mathrm{that}\mathrm{ }\mathrm{solve}\mathrm{ }\mathrm{the}\mathrm{ }\mathrm{simultaneous}\mathrm{equation}.\mathrm{ }\mathrm{The}\mathrm{ }\mathrm{linear}\mathrm{ }\mathrm{combinations}\mathrm{ }\mathrm{obtained}\mathrm{ }\mathrm{from}{\mathrm{ }A}^{-1}\times b$ are divided by the sum of all possible iterations and multiplied by 100 to obtain the percentage of A- to B-type interflavan bonds.

Results and Discussion

MALDI-TOF MS spectra of PAC show masses [M+Cs]⁺ that correspond to cesium adducts of (epi)catechin oligomeric units, which can be explained according to the equation m/z = 290 + 288d − 2 A + c, where 290 represents the molecular weight of the terminal (epi)catechin unit, d is the DP, A is the number of A-type interflavan bonds, and c is the atomic weight of cesium cation (8). Analysis of the MALDI-TOF MS spectra showed that, for the 71 samples (30 cranberry fruits, 5 cranberry products, 25 apples, 6 grape skins, 1 black chokeberry, and 4 cocoa powders), the predominant mass at each DP (Δ288 amu) is representative of a PAC structure with (epi)catechin oligomeric units that have either predominantly A-type or B-type interflavan bonds (Figures 1–4) (4). The molecular weight difference between A-type and B-type interflavan bonds is due to the formation of an ether bond that results in the loss of two hydrogen atoms (Δ2 amu) (8, 17, 18). Visual inspection of MALDI-TOF MS spectra suggests that cranberry fruit, cranberry products, apple, black chokeberry, and cocoa powders have mainly (epi)catechin oligomeric units with structural variation in the nature of the interflavan bonds (Figures 1–3). In contrast, visual inspection of MALDI-TOF MS spectra of grape skins presented additional peaks that correspond to galloyl groups (Figure 4) (17).

Figure 1.

MALDI-TOF MS spectra in positive reflectron mode for cranberry fruit (A), which show (epi)catechin oligomeric units [M+Cs⁺] from trimers to undecamers.

Figure 2.

MALDI-TOF MS spectra in positive reflectron mode for apple (B), which show (epi)catechin oligomeric units [M+Cs⁺] from trimers to undecamers.

Figure 3.

MALDI-TOF MS spectra in positive reflectron mode for black chokeberry (C), which show (epi)catechin oligomeric units [M+Cs⁺] from trimers to undecamers.

Figure 4.

MALDI-TOF MS spectra in positive reflectron mode for grape skin (D), which show (epi)catechin oligomeric units [M+Cs⁺] from trimers to undecamers.

Visual inspection alone of the MALDI-TOF MS spectra is sufficient to differentiate samples such as grape skins due to the presence of the additional peaks that are galloyl units (17). A detailed inspection of the isotope patterns of MALDI-TOF MS spectra for all samples revealed that for cranberry fruit and cranberry products, the most predominant (epi)catechin oligomer units contain at least 1A-type bond, with peaks corresponding to 0A-type, 2A-type, and 3A-type interflavan bonds also present (Figure 5A). In contrast, spectra from the other botanical sources revealed that the most predominant linkage corresponds to 0A-type, with 1A-type interflavan bonds in lesser abundance (Figure 5B–D). As shown in Figure 5, the overlapping peaks for the hexamer cesium adducts suggest that peak patterns differ only in the A- and B-type interflavan bonds (e.g., for the hexamer cesium adducts, the isotope patterns are visible at m/z 1857.3 [3A-type : 2B-type], m/z 1859.3 [2A-type : 3B-type], m/z 1861.3 [1A-type : 4B-type], and m/z 1863.3 [0A-type : 5B-type]).

Figure 5.

MALDI-TOF MS spectra in positive reflectron mode for cranberry (A), apple (B), black chokeberry (C), and grape skin (D), which show the overlapping isotope patterns of PAC hexamers with cesium adducts.

A- and B-type interflavan bonds can be identified visually by the isotope patterns of MALDI-TOF MS spectra. However, the identification and authentication were confirmed by deconvolution of overlapping isotope patterns (18). The ability to deconvolute overlapping isotope patterns for each individual oligomer of PAC allowed for the estimation of the percentage of A- to B-type interflavan bonds for all samples. Median values from the deconvoluted data indicated that 95.6% of PAC from cranberry fruit and cranberry products contain at least one A-type interflavan bond and that 20.8% of PAC from cranberries and cranberry products contain two or more A-type interflavan bonds. In contrast, median values from the deconvoluted data indicated that only 6.6% of PAC from the other botanical sources contain at least one A-type interflavan bond and that only 0.5% of PAC from the other botanical sources contain two or more A-type interflavan bonds.

Deconvoluted MALDI-TOF MS spectra data are presented as a bar graph, providing a visual representation of A-type and B-type “PAC fingerprints” (Figure 6). Previous research reports similar profiles for the percentage of A-type interflavan bonds in cranberries (2–4, 19, 20). Moreover, our findings indicated that the established method determining relative percentages of A-type to B-type interflavan bonds is valid not only for cranberry fruit, but also for cranberry products (Figure 6A and B). Typically, in cranberry fruit and cranberry products, as the DP increases, the percentage of 2A-type and 3-type interflavan bonds increases, whereas the percentage of 1A-type interflavan bonds decreases (18, 19). In contrast, in the other botanical sources, as the DP increases, the percentage of B-type interflavan bonds is either consistently high or increases (>90%) (Figure 3C and D) (4).

Figure 6.

Percentage of A- and B-type interflavan bonds in cranberry (A), cranberry products (B), apple (C), and grape skin (D) from trimers to octamers.

The existing literature has not confirmed the presence of A-type interflavan bonds in apples, black chokeberries, grape skins, and cocoa powders (21, 22). Thus, although the MALDI-TOF MS spectra detected masses above our threshold S/N ratio for the overlapping of isotope patterns that correspond to 1 A-type or 2 A-type interflavan bonds, these masses may result from the presence of low quantities of A-type bonds (not previously described) and/or polyphenol oxidase (PPO) (4, 23, 24). PPO modifies the hydroxyl groups of the flavan units (B-ring) and produces quinones that alter the PAC structure, resulting in compounds that are 2 amu lighter than the all B-type PAC oligomeric series. This finding has been reported for MALDI-TOF MS analysis of apples and apple juice (4, 24). Overall, MALDI-TOF MS spectra and the deconvolution method were able to identify the presence of at least 1 A-type interflavan bond in cranberry fruit and cranberry products with a probability of 95.6% and 95% confidence. In contrast, MALDI-TOF MS spectra and the deconvolution method were able to confirm the absence of A-type interflavan bonds (all B-type) in the other botanical sources with a probability of 93.4% and 95% confidence.

MALDI-TOF MS has some drawbacks such as low shot-to-shot reproducibility and dependence on the sample preparation method. For instance, each laser shot ablates a layer of the sample/matrix, producing variation in the shot-by-shot spectrum. In addition, the position of the laser shot on the sample/matrix leads to spectral variation (25). In order to prevent these limitations, we used the adaptation of the dried method and controlled the inter- and intra-well variability reported by Feliciano et al. (2012) (18). The inter-well variability was decreased by spotting the sample onto 5 wells of the MALDI-TOF MS stainless target plate, while intra-well variability was decreased by shooting 2000 times in 10 different locations in each well of the MALDI-TOF MS stainless target plate. Once the sample/matrix deposition, and inter- and intra-well variability were addressed, precise ai were obtained for each m/z of interest. Deconvolution by matrix algebra was applied to ai for each m/z of interest and this data set was used to evaluate repeatability (18). By definition repeatability refers to the degree of agreement of results when conditions are maintained as constant as possible with the same analyst, reagents, equipment, and instruments performed within a short period of time (26), In order to evaluate repeatability, each sample analyzed in triplicate was evaluated at different times during the same day. Our results indicate that the coefficient of variation across PAC with DP3-8 was <5.0, demonstrating that our method is repeatable. Previously published research from our laboratory showed similar coefficient of variation across PAC with DP3-8 when c-PAC and apple PAC were extracted using 70% (v/v) acetone solution and when c-PAC were extracted using supercritical fluid extraction (3, 4). Because results from our previous research and the results reported here have a coefficient of variation of less than 5%, the “intermediate or large” within-laboratory precision was also satisfied (26). In addition, our previous study on the deconvolution of overlapping isotope patterns of MALDI-TOF MS to determine the percentage of A- to B-type interflavan bonds in c-PAC reported that when eleven ratios of procyanidins A2 and B2 (commercial dimers) were analyzed, the observed percentages were within 3.6% of the predicted percentages, which suggest that the method is accurate (18). Also, the coefficient of determination was >0.995, which suggest linearity. Finally, the standard deviation among series was ≤3.5 and the standard error of the mean was <0.8, which suggest that the method is precise(18).

MALDI-TOF MS spectra of PAC suggests that four optimized parameters described in Section C, I, and J resulted in high-resolution spectra. First, because the MALDI-TOF is a desorption-ionization technique that requires mixing the matrix solution with the sample, the matrix selection and matrix: sample ratio are crucial, as the S/N ratio differs depending on the matrix (27). Both DHB and trans-3-indoleacrylic acid (t-IAA) matrices have been used for the detection of PAC by MALDI-TOF MS (8). However, DHB was chosen because it is soluble in the same solvent as the sample, provides the widest mass range, and produces a suitable S/N ratio. Second, the PAC cationization with sodium [M+Na]⁺ and potassium [M + K]⁺ was addressed by deionizing both the matrix and the sample using a strong cation resin tip and adding cesium trifluoroacetate to the matrix solution. This produced PAC with cesium [M+Cs]⁺ adduct ions, which solved the problem of distinguishing between (a) the formation of both [M+Na]⁺ and [M + K]⁺ adduct ions from one species and (b) the presence of two species with an additional hydroxyl group (8). The molecular weight difference by one additional hydroxyl group substitution is equal to the atomic mass of oxygen (15.9949 amu) which is similar to the difference between the monoisotope of Na⁺ (22.9900 amu) and the monoisotope of K⁺ (39.0980 amu) which is 15.9739 amu (8). Third, the pulsed ion extraction was changed from default values to correct the energy dispersion of the ions leaving the source with the same m/z ratio. This modification produced sharper peaks because the ions with the same m/z ratio reached the detector at the same time. For MALDI-TOF MS spectra of PAC, the optimal pulsed ion extraction was 130 ns at a detector gain of 1548 V. Fourth, the laser power used to desorb and ionize the sample also affected the MALDI-TOF MS spectra. Low laser powers are desired over high laser powers because the latter can fragment the sample, which are not desirable in MALDI-TOF MS spectra (27). In addition, high laser powers are detrimental to the resolution of MALDI-TOF MS spectra. For MALDI-TOF MS spectra of PAC, the optimal laser power was between 45 and 60%. These parameters, as well as sample preparation and the location of the laser on the target, were the most influential parameters for obtaining high-resolution PAC spectra.

The deconvoluted MALDI-TOF MS spectra data of the 71 samples (30 cranberry fruits, 5 cranberry products, 25 apples, 6 grape skins, 1 black chokeberry, and 4 cocoa powders) were classified by principal component analysis (PCA) (Figure 7). PCA was used to reduce the dimensionality of the data and distinguish between samples with A- and B-type interflavan bonds. The first two components (PC1 and PC2) accounted for 90.50% of the total variance. PC1 (x-axis) compares the relative percentage of all B-type bonds (no A-type bonds) to 1A-type bond across PAC with DP3-8 and contributes 86.1% of the total variance. The samples with 1A-type interflavan bond were grouped along the positive values of PC1, whereas the samples with all B-type interflavan bonds were grouped along the negative values of PC1. A clear separation between samples with A-type interflavan bonds and samples with B-type interflavan bonds was obtained. Thus, the PCA score plot can be divided into two zones: (a) the cranberry fruit and cranberry products on the right and (b) the other botanical sources on the left. PC2 (y-axis) compares the relative percentage of 1A-type bond to 2A-type bonds and contributes 4.4% of the total variance. As expected, there is greater variability in the cranberry products in this dimension because cranberry PAC have been reported to contain one or more A-type interflavan bonds at each DP. In the case of cranberry fruits, this variability may be due to differences in variety and/or harvest date. In the case of the cranberry products, this variability may be due differences in cranberry fruit, harvest, post harvest storage, and/or processing (extraction, milling, and encapsulation, among others). Confidence ellipses with a 95% confidence interval are included for each set of samples in the PCA score plot; no outlier was detected. Overall, these results show that PCA correctly classifies samples as either A-type or B-type interflavan bonds using the deconvoluted MALDI-TOF MS spectra data.

Figure 7.

PCA score plot based on the deconvoluted MALDI-TOF MS spectra data, which show the PC1 (x-axis) and PC2 (y-axis) that accounted for 90.50% of the total variance.

Conclusion

The results of this study show that deconvolution of MALDI-TOF MS spectra distinguishes the A-type PAC in cranberry fruit and cranberry products from other botanical sources containing mostly B-type interflavan bonds with a probability greater than 90% and a confidence of 95%. In addition, information obtained by MALDI-TOF MS spectra allows for other aspects of the data set to be inspected and used as a metric of authentication (e.g., galloyl units in grape skins). The results also demonstrate that the deconvoluted MALDI-TOF MS spectra data in combination with PCA allowed a better understanding of the chemical profile of PAC. Therefore, PCA show that classification models can be built accurately based on the nature of the interflavan bonds (A- versus B-type) for cranberries and cranberries products and other botanical sources. Finally, the results of this study showed that MALDI-TOF MS is a powerful tool for structural characterization and identification of PAC.

Conflict of interest

The authors declare the following competing financial interest(s): Christian G. Krueger and Jess D. Reed have ownership interest in Complete Phytochemical Solutions, LLC, and acknowledge their affiliation with this company.

Acknowledgment

The authors acknowledge financial support from the Ministry of Science, Technology and Telecommunications (MICITT), Innovation and Human Capital Program for Competitiveness (PINN), and the National Council for Scientific and Technology Research (CONICIT) (Grant number PED-056-2015-I) at Costa Rica. The Authors acknowledge use of the Bruker UltraFlex^® III MALDI TOF/TOF MS supported by NIH thought the University of Wisconsin, Department of Chemistry, Mass Spectrometry facility (NCRR 1S10RR024601-01).