Understanding the stability, biological impact, and exposure markers of black and using an untargeted metabolomics approach

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

By

Matthew D. Teegarden, M.S.

Graduate Program in Food Science and Technology

The Ohio State University

2018

Dissertation Committee

Devin G. Peterson Ph.D., Advisor

Jessica L. Cooperstone Ph.D., Advisor

Steven K. Clinton M.D., Ph.D.

David M. Francis Ph.D.

1

Copyrighted by

Matthew Daniel Teegarden

2018

2

Abstract

Pre-clinical and clinical evidence suggest that dietary berries may help prevent

against the development of some chronic diseases such as cardiovascular disease and oral

cancer. Berries contain hundreds to thousands of that are thought to act

synergistically to produce a wide range of biological effects. While most of the research

with berries have been performed with fresh or minimally processed products, thermally

processed and stored berry products also represent important dietary sources. The overall

goal of this work is to build upon our understanding of the potential role of berry

phytochemicals in the prevention of chronic diseases by investigating how they exist in

foods and are metabolized by the body. We hypothesize that an untargeted metabolomics

approach will provide novel insights to this end.

The objective of these studies are to: profile the thermal processing-induced

changes in the profile of a black nectar beverage, investigate the

impact of storage on the chemistry and bioactivity of this nectar against premalignant oral

cancer cells, and profile the urinary metabolome of individuals consuming - based confections.

Thermal processing of black raspberries was shown to lower relative levels of some chemical features, while elevating proportionally more chemical features. Some of the elevated features were characterized as potential phenolic degradation products. The

ii stability of several classes of black raspberry (poly)phenolics was also demonstrated.

Storage of black raspberry nectar at 4 °C – 35 °C induced large amounts of chemical variation in the product, without markedly affecting its overall bioactivity. A model system that mimicked the chemistry of the stored nectar products indicted that black raspberry phytochemical degradation products may play a role in maintaining the bioactivity of the product. Untargeted metabolomics revealed a chemical signature of strawberry exposure in a free-living population following a low-anthocyanin background diet.

These studies support the use of untargeted metabolomics in berry research and provide novel insight that will be useful for the future development and evaluation of berry-based functional foods.

iii

Dedication

To my family and friends

iv

Acknowledgments

First, I would like to thank Dr. Jessica Cooperstone who has been one of my biggest

champions and greatest mentors, and her support has never wavered over the past six years. I would also like to thank Dr. Devin Peterson for his willingness to take on a new student before he even moved to Ohio State and for his support ever since. I also owe Dr.

Steven Schwartz a debt of gratitude for inspiring me to pursue a Ph.D.

I would also like to thank my committee members Dr. Steve Clinton and Dr. David

Francis for their advice and guidance throughout my degree.

A special thanks to Dr. Tom Knobloch for helping me learn cell culture, and to Dr. Ken

Riedl who taught me everything I know about mass spectrometry. Their technical

guidance helped me become a more confident scientist.

I am so grateful to have met an amazing bunch of students and labmates who passed

through the office of Parker 240. To the same end, I am also thankful for my family at

the Institute of Food Technologists Student Association. The comradery and support of

these friends has been invaluable over the past six years.

v

I would like to thank the United States Department of Agriculture for providing me with funding throughout my Ph.D. and Dr. Ken Lee for administering this funding, as well as his continual support throughout my degree. I would also like to thank Lisa and Dan

Wampler for their support of students involved in food and health research.

Finally, I would like to thank my family, whose love and support has been constant throughout my education.

vi

Vita

June 2008 ...... Sycamore High School, Cincinnati, OH

June 2012 ...... B.S. Food Science and Technology, Summa

cum Laude, The Ohio State University,

Columbus, OH

August 2012 to present ...... Graduate Research Associate, Department

of Food Science and Technology, The Ohio

State University, Columbus, OH

August 2012 to July 2013 ...... University Fellow, The Ohio State

University, Columbus, OH

August 2014 to July 2017 ...... USDA National Needs Fellow, The Ohio

State University, Columbus, OH

December 2014 ...... M.S. Food Science and Technology, The

Ohio State University, Columbus, OH

August 2017 to present ...... Lisa and Dan Wampler Endowed Fellow for

Food and Health Research, Department of

Food Science and Technology, The Ohio

State University, Columbus, OH

vii

Publications

Teegarden, M.D.; Campbell, A.; Cooperstone, J.L.; Tober, K.L.; Schwartz, S.J.; Oberyszyn, T.M. 25-hydroxyvitamin D3 and its C-3 epimer are elevated in the skin and serum of Skh-1 mice supplemented with vitamin D3. Mol. Nutr. Food Res. 2017; 61; 1700293.

Cooperstone, J.L.; Tober, K.L.; Riedl, K.M.; Teegarden, M.D.; Cichon, M.J.; Francis, D.M.; Schwartz, S.J.; Oberyszyn, T.M. Tomatoes protect against development of UV- induced keratinocyte carcinoma via metabolomic alterations. Scientific Reports 2017; 7:5106.

Ahn-Jarvis, J.H.; Teegarden, M.D.; Schwartz, S.J.; Lee, K; Vodovotz, Y. Modulating conversion of isoflavone glycosides to aglycones by crude beta-glycosidase extracts from almonds and processed soy. Food Chemistry. 2017; 237:685-692.

Obodai, M.; Mensah, D.L.N.; Fernandes, Â.; Kortei, N.K.; Dzomeku, M.; Teegarden, M.D.; Schwartz, S.J.; Barros, L.; Prempeh, J.; Takli, R.K.; Ferreira, I.C.F.R. Chemical Characterization and Antioxidant Potential of Wild Ganoderma Species from Ghana. Molecules. 2017; 22: 196.

Teegarden, M.D.; Riedl, K.M.; Schwartz, S.J. Chromatographic Separation of PTAD- derivatized 25-hydroxyvitamin D­3 and its epimer from human serum and murine skin. J. Chromatography B. 2015; 991:118-121

Fields of Study

Major Field: Food Science and Technology

viii

Table of Contents

Abstract ...... ii Dedication ...... iv Acknowledgments...... v Vita ...... vii List of Tables ...... xii List of Figures ...... xiii Chapter 1. Literature Review...... 1 1.1 Evidence for the role of berries in chronic diseases ...... 1 1.1.1 Black Raspberries ...... 2 1.1.2 Strawberries ...... 5 1.2 Berry Phytochemicals ...... 6 1.2.1 ...... 9 1.2.2 ...... 12 1.2.3 Phenolic acids ...... 14 1.2.4 Phenolic profile and content of Black Raspberries ...... 15 1.2.5 Phenolic profile and content of strawberries ...... 19 1.3 Absorption and Metabolism of Berry Phytochemicals ...... 21 1.3.1 Absorption and Metabolism of Flavonoids...... 22 1.3.2 Absorption and Metabolism of Procyanidins ...... 26 1.3.3 Absorption and Metabolism of Ellagitannins ...... 27 1.3.4 Absorption and Metabolism of Phenolic Acids ...... 31 1.4 Effects of Food Processing and Storage on Berry (Poly)phenols ...... 32 1.4.1 Processing Effects ...... 40 1.4.2 Effects of Storage ...... 44 1.4.3 Effects of Processing and Storage on Bioactivity ...... 46 ix

1.5 Untargeted Metabolomics ...... 47 1.5.1 Mass Spectrometry-based Metabolomics ...... 48 1.5.2 Metabolomics Applications in Berry Research ...... 56 1.6 Specific Aims ...... 58 Chapter 2. Profiling the impact of thermal processing on black raspberry phytochemicals using untargeted metabolomics...... 62 2.1 Abstract ...... 63 2.2 Introduction ...... 63 2.3 Materials and Methods ...... 65 2.3.1 Chemicals ...... 65 2.3.2 Nectar Production ...... 65 2.3.3 Determination of Total Monomeric Anthocyanins ...... 67 2.3.4 Sample Preparation for Untargeted Metabolomics ...... 67 2.3.5 UHPLC-QTOF-MS Metabolomics Data Acquisition ...... 68 2.3.6 Data Pre-processing and Analysis ...... 69 2.3.7 Compound Identification ...... 69 2.4 Results and Discussion ...... 70 2.4.1 Monomeric anthocyanin content...... 71 2.4.2 Untargeted metabolomics ...... 73 2.4.3 Identification of Features Elevated in Thermally Processed BRBs ...... 75 2.4.4 Identification of Features Stable to Thermal Processing ...... 81 2.4.5 Ellagitannins ...... 84 2.5 Conclusions ...... 87 Chapter 3. Storage conditions modulate the metabolomic profile of a black raspberry nectar with minimal impact on bioactivity ...... 88 3.1 Abstract ...... 89 3.2 Introduction ...... 90 3.3 Materials and Methods ...... 92 3.3.1 Chemicals ...... 92 3.3.2 Nectar Processing, Storage, and Sampling ...... 92 3.3.3 Sample Preparation and Analysis for Untargeted Metabolomics ...... 95 3.3.4 Data Pre-processing and Analysis for Untargeted Metabolomics ...... 96 3.3.5 Targeted Compound Analysis...... 99 x

3.3.6 Extract Preparation for Cell Study ...... 99 3.3.7 Cell Culture and Growth Inhibition Assay ...... 100 3.4 Results and Discussion ...... 101 3.4.1 Untargeted Metabolomics Revealed Large Chemical Variation with Elevated Storage Temperatures ...... 102 3.4.2 All Extracts of Stored BRB Nectar Inhibited SCC-83-01-82 Cell Growth Similarly ...... 109 3.4.3 C3R and its degradation product PA Equally Contribute to the Bioactivity of BRB Nectar ...... 113 3.5 Conclusions ...... 118 Chapter 4. Elucidating markers of strawberry consumption in smokers and non-smokers using untargeted metabolomics...... 119 4.1 Abstract ...... 120 4.2 Introduction ...... 120 4.3 Materials and Methods ...... 122 4.3.1 Chemicals ...... 122 4.3.2 Subjects and Study Design...... 123 4.3.3 Sample Preparation ...... 126 4.3.4 UHPLC-MS Data Acquisition ...... 126 4.3.5 Data Pre-processing ...... 127 4.3.6 Data Analysis ...... 128 4.3.7 Feature Selection and Identification ...... 129 4.4 Results and Discussion ...... 130 4.4.1 Multi-level Multivariate Data Analysis ...... 131 4.4.2 Identification of Features Elevated with Strawberry Consumption ...... 137 4.4.2 Identification of Features Elevated with Placebo Consumption ...... 145 4.5 Conclusions ...... 148 References ...... 149

xi

List of Tables

Table 1.1 BRB phenolics with reported quantitation in literature ...... 18

Table 1.2 Strawberry phenolics with reported quantitation from whole in literature 20

Table 1.3 Representative studies on berry (poly)phenols over food processing and storage ...... 35

Table 2.1 BRB nectar formulation ...... 66

Table 2.2 Identified features elevated in BRB nectar ...... 76

Table 2.3 Identified BRB phytochemicals with P > 0.05 or fold change < 2 ...... 83

Table 2.4 Identified ellagic acids (EA) and ellagitannins (ET) ...... 86

Table 3.1 BRB nectar beverage formulation ...... 94

Table 3.2 PLS model cross validation results ...... 106

Table 3.3 Level 1 and 2 identified compounds from list of features with VIP>1 across all four PLS models ...... 108

Table 3.4 Quantitative analysis of C3R and PA in nectar from t0 and 60 days at 35 °C 114

Table 4.1 Cross-validation summary of PLS-DA models on variation due to treatment 136

Table 4.2 Identified metabolites elevated in the strawberry study arm ...... 141

Table 4.3 Identified features elevated in the placebo study arm ...... 147

xii

List of Figures

Figure 1.1 Molecular structures of representative berry phytochemicals ...... 8

Figure 1.2 Molecular structures of ...... 30

Figure 1.3 Factors that affect levels of (poly)phenolic compounds in processed berry products ...... 34

Figure 2.1 Total monomeric anthocyanins (expressed as mg cyanidin-3-glucoside equivalents) in BRB powder and lyophilized nectar. A numeric correction was applied to the nectar values to account for other solids present. (*)Decrease in anthocyanin content was statistically significant (P<0.01)...... 72

Figure 2.2 Hierarchical clustering analysis on features that differ ≥2 fold between lyophilized BRB powder and nectar made from that powder (P < 0.05). Features were clustered using Euclidian distance metrics and Ward’s linkage rule...... 74

Figure 2.3 MS/MS spectra and fragment rationalization of protocatechuic acid (A) and phloroglucinaldehyde (B)...... 77

Figure 2.4 Proposed fragmentation pattern of 2-oxo-2-(2,4,6-trihydroxyphenyl) acetaldehyde...... 80

Figure 3.1 Summary of untargeted metabolomics data pre-treatment and analysis ...... 98

Figure 3.2 PCA scores plot of all samples colored by storage temperature and labeled according to length of storage...... 103

Figure 3.3 Heat map of molecular features with VIP>1 in PLS models for storage at 4–35 °C. Features were clustered using Euclidian distance metrics and Ward’s linkage rule. * Denotes potential Maillard-related sugar fragmentation-phenolic degradation products determined after derivatization with o-phenylenediamine...... 107

Figure 3.4 Growth inhibition of SCC-83-01-82 cells by extracts of BRB nectar stored at increasing temperatures. ANOVA terms for storage time, temperature, and their interaction were significant (P<0.01). Only significant differences within each storage temperature are denoted (*)...... 112

xiii

Figure 3.5 Averaged relative abundances of C3R and PA over time in each storage condition...... 114

Figure 3.6 Dose-response relationship between increasing levels of C3R (A) and PA (B) and growth inhibition of SCC-83-01-82 cells. (C) When the ratio of C3R and PA were varied in equimolar solutions (molarity on right y-axis), growth inhibition (left y-axis) was maintained (P = 0.092 for differences among treatments)...... 117

Figure 4.1 Crossover study design with 24 hr urine collection time points (triangles). Circled timepoints indicate those selected for further study using untargeted metabolomics...... 125

Figure 4.2 PCA of untransformed, filtered data collected in ESI+ mode. (A) Data without outliers removed. The same two outliers were observed in ESI- mode data collected on a different day. (B) Data with outliers removed, demonstrating a lack of separation achieved with traditional PCA...... 132

Figure 4.3 Multilevel PCA analysis of inter-individual variation (panel A: ESI+; panel B: ESI-) and variation due to treatment (panel C: ESI+; panel D: ESI-). NS= Nonsmoker, S= Smoker, Straw= Strawberry, Plac= Placebo...... 135

Figure 4.4 Example of level 1 metabolite identification. A: mass chromatogram demonstrating co-elution of a A glucuronide standard with peak in urine sample under identical analytical conditions. A slight retention time shift from the value reported in Table 4.2 because a new UHPLC column (same model) was used. B: fragmentation of glucuronide standard and peak detected in urine (collision energy: 40 eV). 142

Figure 4.5 Mass fragmentation spectra (collision energy = 40 mEV) of tentatively identified o-coumaric acid in urine (top) and an authentic standard of p-coumaric acid (bottom)...... 143

xiv

Chapter 1. Literature Review

1.1 Evidence for the role of berries in chronic diseases

Epidemiological evidence suggests a potential role for fruit and vegetable consumption in the prevention of chronic diseases including some cancers (Key 2011;

Norat et al. 2014) and cardiovascular disease (Aune et al. 2017). Despite challenges inherent to specifically assessing berry intake in epidemiological studies (e.g. limitations of food frequency questionnaires) data suggests berry consumption may also impact these diseases (Hjartåker et al. 2015; Basu et al. 2010). Here, we refer to berries by the culinary definition, which encompasses small, fleshy, and typically anthocyanin-rich that span several botanical fruit types including true berries (e.g. blueberries) and aggregate fruits (e.g. raspberries, blackberries, strawberries). Comprehensive reviews on clinical studies involving berries, or their phytochemical components, highlight a myriad of conditions that have been researched as potential targets for berry interventions including inflammation, cardiovascular health, cancer prevention, and diabetes among others (Yang & Kortesniemi 2015; Pojer et al. 2013; Rodriguez-Mateos, Heiss, et al.

2014; Joseph et al. 2014; Bishayee et al. 2016). Berries have received a large amount of attention for potential health benefits due to their rich content of antioxidant phytochemicals, which is estimated to be higher than many other fruits and vegetables

(Carlsen et al. 2010). Numerous studies have attempted to identify the putative bioactive

1

compounds in berries, with a great deal of attention paid to a select few classes of

compounds. Anthocyanins, (a type of ) which provide the vibrant red to blue

colors of most berries, have been extensively studied for their chemopreventative properties (Cooke et al. 2005; Stoner et al. 2010) and other health benefits (Pojer et al.

2013). Ellagitannins, found in high concentrations in the seeds of some berries, and their

related metabolites have also been pinned as potent bioactives (Ismail et al. 2016; Ross et

al. 2007). In reality, berries are complex mixtures of phytochemicals that are thought to

act in synergistic or additive manners (Seeram & Heber 2007; Liu 2003). Research on

berry products that allows for a greater appreciation of comprehensive chemical profiles

will allow for greater insight into potential health benefits of these fruits.

1.1.1 Black Raspberries

Interest in the cancer prevention properties of berries first came about due to their

high content of (Stoner et al. 2007). Specific interest in black raspberries

(BRBs) originated from their high contents of phenolic compounds, such as ellagic acid

and anthocyanins, in addition to the presence of and micronutrients (Harris

et al. 2001). BRBs are native to North America and include two species,

occidentalis and Rubus leucodermis (Kula & Krauze-Baranowska 2016; Hummer 2010).

Production of BRBs in the Unites States has been low, limited to under 2000 acres since

2007 (United States Department of Agriculture n.d.). This is likely due to their

susceptibility to disease and greater popularity of red raspberries (Kula & Krauze-

Baranowska 2016; Hummer 2010). Promising pre-clinical investigations have suggested

2

BRBs may play a preventative role in many cancers including skin, breast, oral,

esophageal, and colorectal as recently reviewed (Kula & Krauze-Baranowska 2016).

Clinical trials involving the impact of BRBs on oral, esophageal, and colorectal cancer

have also been recently reviewed, all of which suggest potential roles for BRBs in these

cancers (Kresty et al. 2016). For the purpose of the current review, focus will be given to the potential role of black raspberries in oral cancer.

Cancers of the oral cavity and pharynx are expected to exceed 50,000 new cases

and 10,000 deaths in the United States in 2018, ranking among the top ten sites for new

cancer diagnoses in men (Siegel et al. 2018). In fact, oral cancer is prevalent throughout

the world, with higher incidence observed in men and people in less developed regions

(Petersen et al. 2005; Ferlay et al. 2015). The majority of oral cancers are squamous cell

carcinomas (SCCs), and major risk factors include tobacco use, alcohol consumption,

human papillomavirus infection, and chronic periodontal disease (Bsoul et al. 2005).

Early interest in the ability of BRBs to prevent oral cancer was a result of their relatively

high content of ellagic acid (Stoner 2009). However, the oral cavity provides a unique

environment for exposure to high concentrations of many BRB phytochemicals, as

compared to other cancer sites.

In a growth inhibition screening panel utilizing the CAL-27 oral SCC cell line,

BRBs emerged as the most potent growth inhibitors compared to five other common

berry types (N. Seeram et al. 2006). Other work with pre-malignant and malignant oral

SCC lines have indicated that fractionated BRB extracts can impact levels of cell cycle-

related proteins (Han et al. 2005) and exhibit anti-angiogenic, pro-apoptotic, and anti-

3 oxidative properties (Rodrigo et al. 2006). In vivo animal studies further capitulate the capacity of BRBs to inhibit oral cancer. Topical application (Warner et al. 2014) and dietary supplementation of BRBs (Oghumu et al. 2017; Casto et al. 2002) resulted in decreased malignant and total (including pre-malignant) lesion incidence (animals with tumors/total animals) and multiplicity (number of tumors/animal) in chemically induced models of oral cancer. A recent animal study by Oghumu and colleagues showed that this treatment inhibited pro-inflammatory and anti-apoptotic biomarkers (Oghumu et al.

2017), two fundamental features of cancer (Hanahan & Weinberg 2011).

Clinical trials administering BRBs in the context of oral cancer have produced compelling results. Mallery and colleagues developed a bioadhesive gel with freeze- dried BRB powder to provide prolonged, direct contact between BRB phytochemicals and the oral mucosa (Mallery et al. 2007). A pilot study in patients with intraepithelial neoplasia (premalignant) lesions demonstrated that this treatment lowered histological grades of lesions in some subjects, lowered levels of a pro-inflammatory enzyme, and reduced the occurrence of a genetic event (loss of heterozygosity) associated with cancer development (Mallery et al. 2008; Shumway et al. 2008). This mucoadhesive gel was also evaluated in a placebo-controlled multicenter study, which further demonstrated the aforementioned reductions in loss of heterozygosity as well as significant decreases in the size and severity of premalignant lesions (Mallery et al. 2014). In a separate study, patients with confirmed oral SCC (malignant cancer) were treated with BRB-based troches. Following two weeks of treatment, anti-apoptotic and pro-inflammatory markers were significantly reduced (Knobloch et al. 2016), demonstrating in this application, that

4

BRB-based troches can also impact malignant cancer. Together these studies provide

convincing evidence of the role for BRBs in modulation of oral cancer, however further

work is needed to understand how to translate these pharmacological approaches to food-

based, dietary approaches.

1.1.2 Strawberries

Strawberries (Fragaria x ananassa) are an important berry crop worldwide, with

production in some quantity reported on all six inhabited continents. China and the

United States produce the most strawberries exceeding an average yearly production

(1994-2016) of 1.8 million metric tons and 1.0 million metric tons, respectively (Food

and Agriculture Organization of the United Nations n.d.). Likely due to their widespread

production, and thus potential importance to public health, strawberries have been

heavily researched for their potential health benefits. The in vitro, animal, and human

studies evaluating effects of strawberries on inflammation, cardiovascular disease,

metabolic syndrome, cancer, and neuroprotection have been previously reviewed

(Giampieri et al. 2015; Giampieri et al. 2014; Basu et al. 2014). Of the conditions

researched, some human clinical trial evidence for effects on cardiovascular disease and

cancer are particularly interesting. For instance, subjects fed 500 g of fresh strawberries/d for 30 d had lower urinary levels of oxidative stress biomarkers compared to immediately before or 15 d after the intervention (Alvarez-Suarez et al. 2014). With regard to cancer, Chen and colleagues demonstrated a reduction of premalignant esophageal dysplastic lesions, as well as several cancer-related biomarkers, with the

5

consumption of an aqueous slurry of freeze-dried strawberry powder (60 g/d) for 6

months (Chen et al. 2012). With compelling results such as these, strawberries remain an

active subject of research in terms of their potential role in health.

1.2 Berry Phytochemicals

Edible berries constitute a diverse range of lightly colored to deep purple, red, and

blue fruits, due to their contents of anthocyanins, a type of flavonoid. While berries do

have appreciable content of some micronutrients, such as vitamin C (United States

Department of Agriculture n.d.), they are also known for their high content of phenolic

and polyphenolic phytochemicals (Figure 1.1). These compounds are considered secondary metabolites, which serve to protect from environmental influences such as ultraviolet radiation and pathogen exposure among other functions (Seeram 2006). The

main classes of (poly)phenols in berries include flavonoids, condensed and hydrolysable

, phenolic acids, and stilbenoids (Seeram 2008). While this list may seem limited,

there is a great deal of structural diversity in each class of phenolic compounds based on

functional group substitution patterns, stereochemistry, degree of polymerization,

glycosylation with various sugar moieties, and conjugation with organic acids (Seeram

2006). These compounds are proposed to impart many of the health benefits associated

with berries. While attempts have been made to identify singular bioactive components,

the bioactivity of the whole berry fruit cannot necessarily be explained by one

phytochemical. For instance, rats showed equally lowered indices of esophageal cancer

progression when fed diets supplemented with freeze-dried BRBs, anthocyanin-rich BRB

6 extracts, or anthocyanin-deplete BRB extracts (Wang et al. 2009). Similarly, Paudel and colleagues noted a wide range of bioactivity of numerous BRB components in a cell model of colon cancer (Paudel et al. 2014). Thus, as the chemistry and potential bioactivity of each class of berry phenolic compound is briefly discussed below, it should be interpreted with this in mind. Additionally, in reviewing the health effects of these specific compounds, special attention was paid to those with in vivo evidence, including animal studies and human clinical trials.

7

Figure 1.1 Molecular structures of representative berry phytochemicals

8

1.2.1 Flavonoids

The flavonoids are a diverse set of over 4,000 polyphenolic compounds divided

into seven structural subclasses (Xiao & Ho 2017). The major classes of flavonoids

found in berries include anthocyanidins, flavonols, and flavan-3-ols (Del Rio et al. 2010;

United States Department of Agriculture 2013). In the actual berry fruit, these are found

conjugated to numerous sugar, and sugar-organic acid ester moieties primarily at the third carbon. The specific structures of the predominant berry flavonoid aglycones (United

States Department of Agriculture 2013), are shown in Figure 1.1.

Anthocyanins are highly pigmented compounds responsible for the red to dark purple and blue colors of berries. Their vibrant color in acidic conditions is a result of extended conjugation in their molecular structure (He & Giusti 2010). Generally, anthocyanins are found in higher concentrations than other flavonoids in edible berries, with levels ranging from tens to hundreds of mg/100 g fresh weight (fw; in aglycone equivalents) (United States Department of Agriculture 2013). A cadre of research exists on the potential bioactivity of anthocyanins in conditions from cancer, to cardiovascular disease and neurological conditions. These have recently been comprehensively reviewed (Li et al. 2017; De Pascual-Teresa 2014). Much of the research that specifically investigates anthocyanins, or anthocyanin-enriched food extracts, is limited to cell and animal studies. Nonetheless, these provide insight into possible mechanisms of action including inhibition of key cellular pathways and influence on genetic expression (Li et al. 2017). A recent analysis of randomized clinical trials evaluating the effects of anthocyanin and anthocyanin-rich extracts on cardiovascular disease demonstrated mixed

9

results, depending on the population studied (healthy vs. at risk for disease). When

considered together, the studies suggest that anthocyanins may lower elevated LDL

cholesterol towards more optimal levels (Wallace et al. 2016). Several of the

anthocyanin doses used in these studies provided clinically relevant levels of

anthocyanins, in similar concentrations to quantities found in some freeze-dried berry powders. Thus, clinical evidence for the efficacy of anthocyanins, at levels that can be achieved with foods, exists and warrants further study to decipher optimal target populations and mechanisms of action.

Flavonols are present in berries at concentrations typically under 30 mg aglycone equivalents/100 g fw (United States Department of Agriculture 2013; Häkkinen et al.

1999). The most common, and often predominant, flavonol aglycone in berries is quercetin. Numerous cell and animal studies have investigated quercetin for its potential role in immune modulation, exercise recovery, ageing, obesity, diabetes, cardiovascular disease, and cancer (Miles et al. 2014). Interestingly, the work with cancer has centered on the use of quercetin as a potential adjuvant to chemotherapy drugs, but limited animal work and no clinical work has been completed on that subject. With regard to clinical trials, minimal or no effects of quercetin supplementation were observed as they pertain to immune modulation and exercise recovery (Miles et al. 2014). A reasonably large number of clinical trials have been performed to investigate the potential cardioprotective role of quercetin. In an early clinical trial with healthy volunteers, no effect of quercetin supplementation on heart disease risk factors was observed, despite an approximately 23 fold increase in plasma quercetin levels (Conquer et al. 1998). Later studies that

10

recruited volunteers at higher risk for cardiovascular diseases noted significant decreases

in blood pressure (Egert et al. 2009; Edwards et al. 2007), suggesting that the potential

benefits of quercetin for cardiac health are limited to these populations. Similar

phenomenon were observed with regard to lowering of markers of oxidative stress

(plasma malondialdehyde) and inflammation (plasma cytokine ratios) (Boots et al. 2011).

While clinical results such as these are promising for diseased or at-risk populations, it

should be noted that doses of quercetin used in clinical trials are often more than what

can be reasonably achieved from foods alone.

The predominant flavan-3-ols found in berries are catechin and epicatechin with most berries containing less than 10 mg/100g (United States Department of Agriculture

2013). These compounds are found as aglycones or covalently bound to each other to form condensed tannins known as procyanidins. Various berries can contain procyanidins with >10 linked units. The predominant procyanidins in berries are B-type

(linked via one carbon-carbon bond), although cranberries uniquely contain A-type

(linked via one carbon-carbon and one ether linkage) (José Serrano et al. 2009; United

States Department of Agriculture 2004). The content of procyanidin dimers or larger in berries varies widely from below 20 mg/100 g fw in blackberries to >300 mg/100 g fw in cranberries and some blueberries (United States Department of Agriculture 2004).

Laboratory studies with catechin, epicatechin, and procyanidins have explored potential mechanisms by which these compounds, or foods/extracts rich in these compounds, may help prevent cardiovascular disease, cancer, diabetes, and lower oxidative stress (Li et al.

2015; José Serrano et al. 2009; Aron & Kennedy 2008). Human clinical trials with these

11 compounds are mostly limited to those on factors related to cardiovascular disease.

Reviews of these trials point towards potential benefits such as lowered blood pressure, vasodilation, and certain markers of oxidation (Hooper et al. 2012; Williamson &

Manach 2005). Notably, none of these clinical trials utilized berries as a source of flavan-3-ols or procyanidins. Instead, they used purified compounds, cocoa, or tea.

1.2.2 Ellagitannins

Ellagitannins are a type of hydrolysable (can be cleaved to smaller units by dilute acid) that consist of units esterified to sugar molecules.

Hydrolysis of ETs liberates monomeric hexahydroxydiphenic acid, which then spontaneously lactonizes to form ellagic acid (Clifford & Scalbert 2000; Landete 2011;

Josø Serrano et al. 2009). Although ellagic acid is also present in various conjugated forms (e.g. methylellagic acid, glycosides, etc.) in foods, estimation of content is often performed after complete hydrolysis of ellagitannins to ellagic acid and expressed as ellagic acid equivalents (Talcott & Krenek 2012; Landete 2011).

Ellagitannins and ellagic acid are found in foods such as nuts, , and barrel aged spirits and wine (including wine oaked with woodchips) due to contact with ellagitannin-rich woods (Landete 2011). Berries from the taxonomic family Rosaciae

(strawberries, raspberries, etc.) are the most important dietary source of ellagic acid and ellagitannins in the western diet (Landete 2011). The hydrolyzed content of ellagic acid and ellagitannins in berries ranges from 0 to approximately 1.8 g ellagic acid /100 g fw, with many berries containing an average of approximately 150 mg ellagic acid/100 g fw.

12

The seeds of berries contain much higher levels of ellagitannins and ellagic acid, ranging from 6.7 – 32 mg/g seed (Landete 2011).

ETs and ellagic acid have been well studied to understand any benefits they may concern with regard to inflammation, cardiovascular disease, gut health, and cancer as previously reviewed (Ismail et al. 2016; Landete 2011). Many of these studies utilize extracts, which represent a rich source of ellagitannins. Much of the in vitro work that has been done directly exposes various cell types (e.g. prostate cancer cells) to ellagitannins that they would never be exposed to in vivo. Similarly, the concentrations of ellagic acid used are often higher than those achieved in vivo. As will be discussed later, ellagic acid is poorly bioavailable and ellatitannins are not absorbed following consumption. However, microbial metabolites of these compounds are bioavailable and may be responsible for many of the health benefits associated with ellagitannins and ellagic acid (Landete 2011). This incongruence has caused discrepancies between in vitro and in vivo studies of ellagitannin/ellagic acid bioactivity (Landete 2011). Most clinical interventions have focused on potential ellagitannin benefits for cardiac health. Although results from these trials are mixed, they suggest consumption of ellagitannin-rich pomegranate juice may have some cardio protective effects (Basu & Penugonda 2009).

Animal studies with dietary ellagitannin or ellagic acid supplementation provide

suggestive evidence for a role of these compounds in cancers of the prostate (Seeram et

al. 2007; Bell & Hawthorne 2008), oral cavity (Anitha et al. 2013; Vidya Priyadarsini et al. 2012; Oghumu et al. 2017), esophagus (Mandal & Stoner 1990), liver (Hau et al.

2010), and lung (Khan et al. 2007). Due to these favorable results and their high

13 concentrations in berries, ellagitannins and ellagic acid are actively studied as potential health promoting berry phytochemicals.

1.2.3 Phenolic acids

Phenolic acids are ubiquitous plant compounds that contain at least one carboxylic acid moiety, and are inclusive of hydroxycinnamic and hydroxybenzoic acids.

These compounds serve a variety of functions in plants, such as participation in structural elements and defense mechanisms (Robbins 2003). Various patterns of hydroxylation and metholylation on the phenyl ring give rise to a multitude of structures, including the commonly occuring coumaric, caffeic, ferulic, protocatechuic, and vanillic acids (Mattila et al. 2006). Phenolic acids are typically found in plants esterified to other molecules including structural macromolecules (e.g. cellulose), sugars, and flavonoids, which creates unique challenges for their analysis (Robbins 2003). The contents of phenolic acids in berries range from 5-100 mg/100 g fw following extensive hydrolysis. The percentage of free phenolic acids also ranges from <10% free in raspberries to 90% in rowanberries, which may impact bioavailability (Mattila et al. 2006). A broad overview of potential bioactivities of phenolic acids, including antimicrobial, anticarcinogenic, neuroprotective, and antioxidative activity, in cell models has recently been completed

(Heleno et al. 2014). In vivo studies also support potential activities for phenolic acids in various cancers (Tanaka et al. 2011; Yang et al. 2015; Pei et al. 2016; Baskaran et al.

2010; El-Seedi et al. 2012), neuroprotection (Yabe et al. 2010), and cardiovascular health

(Park 2009; Luceri et al. 2007; El-Seedi et al. 2012). Of note, many of the doses of

14

phenolic acids used to produce effects in animal models are higher than normal dietary

levels (El-Seedi et al. 2012). Clinical evidence for the health effects of purified phenolic

acids are limited to dietary administration of coffee-derived caffeoylquinic acids. The

results of these studies suggest that acute consumption of these compounds may be

beneficial for vascular function (Ward et al. 2016; Mills et al. 2017).

1.2.4 Phenolic profile and content of Black Raspberries

BRBs contain a diverse array of phytochemicals, though most of the reported quantitative data is limited to levels of anthocyanins and a few other select phenolics

(Table 1.1). BRBs have been reported to contain the highest total levels of anthocyanins out of all Rubus berries, such as blackberries and red raspberries (Torre & Barritt 1977;

Wada & Ou 2002). The anthocyanin profile of BRBs mostly contains cyanidin-based

glycosides, but some glycosides of pelargonidin have also been reported (Tian et al.

2006; Kula et al. 2016; Krauze-Baranowska et al. 2014). The predominating anthocyanin

in BRBs is cyanidin-3-O-rutinoside, but relative levels of individual anthocyanins can

vary according to genetic background (Dossett et al. 2010) and crop year (Stoner 2009).

As can be seen from Table 1.1, total levels of anthocyanins can exceed 2% of the dry

weight of BRBs.

Ellagitannins also constitute a major class of phenolic in BRBs. Little work has

been done to comprehensively profile and quantify the ellagitannins present in these

fruits, but sanguiin H-6, sanguiin H-10, penduculagin, and a glucose isomer of bis-

hexahydroxydiphenic acid have been reported (Kula & Krauze-Baranowska 2016). The

15 average reported level of sanguiin H-6 is approximately 1500 mg/100g, or 1.5% of the dry weight of BRBs (Table 1.1), which is similar to levels reported for several varieties of red raspberries (Kula et al. 2016). Total fresh weight content of ellagitannins and ellagic acid in BRBs is approximately 90 mg/100 g fw in the fruit and 6.7 mg/g in the seeds (Landete 2011). Whole BRBs contain roughly equivalent amounts of ellagic acid bound in ETs and free ellagic acid, though considerable variation exists in the reported levels (Table 1.1) (Wada & Ou 2002). Given the potential biological importance of BRB ellagitannins, further work is needed to comprehensively understand the ellagitannin profile of these berries.

Other phenolic compounds quantified in BRBs include phenolic acids and derivatives of the flavonol, quercetin. The levels of these compounds are notably lower than those of anthocyanins and ellagitannins (Table 1.1). Qualitative analyses provide further insight into the phenolic profile of BRBs. Paudel and colleagues recently completed an overview of the non-anthocyanin phenolic compounds present in BRBs and reported the presence of organic acids: protocatechuic acid, salicylic acid, benzoic acid, coumaric acid, caffeic acid, and sinapic acid; stilbenoid: trans-piceid; and flavonoids: dihydrokaempferol, epicatechin, phloretin, myricetin. Aside from epicatechin, all of these compounds were found linked to various sugar molecules, mostly hexoses. The investigators also noted several methylellagic acid derivatives containing sugar and acyl moieties (Paudel et al. 2013). Kula and colleagues also comprehensively studied the phenolic profile of the ‘Litacz’ BRB varietal and additionally noted the presence of , catechin, B-type procyanidins, sanguiin H-10 (an ellagitannin), sanguiin H-2,

16 (an ellagitannin), and kaempferol (Kula et al. 2016). The results of these studies are in agreement with others (Kula & Krauze-Baranowska 2016). Though not thoroughly discussed here, BRBs also contain some quantity of vitamins, minerals, and the phytosterols ß-sitosterol and campesterol (Stoner 2009). Thus, although BRBs are known to contain high levels of anthocyanins and ellagitannins, they are complex mixtures of many phytochemicals.

17

Table 1.1 BRB phenolics with reported quantitation in literature

Compound a Mean ± SD b Min Max References Anthocyanins Cyd-3-O-glucoside 304.8 ±65.6 250.0 399.9 1,2,3 Cyd-3-O-sambubioside 299.5 ±219.5 56.0 576.9 1,2,3 Cyd-3-O-rutinoside 1997.9 ±386.8 1661.3 2538.6 1,2,3 Cyd-3-O-xylosylrutinoside 846.2 ±280.0 510.0 1195.2 1,2,3 Cyd-3,5-O-diglucoside 29.5 ±8.7 23.4 35.7 2,3 Cyd-3-O-sophoroside 66.2 ±19.6 52.3 80.0 2,3 Plg-3-O-glucoside 30.6 ±9.0 24.2 36.9 2,3 Plg-3-O-rutinoside 72.7 ±21.4 57.5 87.8 2,3

Other Phenolics Ellagic acid 127.1 ±91.5 29.50 225.00 1,2,3 Sanguiin H-6 1509.9 ±38.7 1482.50 1537.20 2,3 Ferulic acid 23.2 ±21.6 5.00 47.10 1 Coumaric acid 7.7 ±1.4 6.82 9.23 1 Chlorogenic acid 0.14 1 Quercetin 40.1 ±5.0 36.50 43.60 1 Quercetin-3-O-rutinoside 111 4 a Cyd = cyanidin; Plg = pelargonidin b mean of average levels reported in literature, measured as mg/100g dry weight Reference 1: (Stoner 2009) Reference 2: (Kula et al. 2016) Reference 3: (Krauze-Baranowska et al. 2014) Reference 4: (Wada & Ou 2002)

18

1.2.5 Phenolic profile and content of strawberries

Strawberries have been extensively studied for their phytochemical profile and

contents. Levels of quantified strawberry phenolics reported in the literature are

summarized (as mg/100g fresh weight) in Table 1.2. The anthocyanins in strawberries

are primarily pelargonidin-based and include both glycosides and acylated glycosides.

Anthocyanins are found in the flesh and achenes (seed-containing fruits) of strawberries,

with cyanidin-3-O-glucoside-malonate uniquely present in the achenes (Aaby et al.

2005). Contents of anthocyanins in strawberries can also be affected by factors related to

genetic background and growth conditions (Tulipani et al. 2011). This is also reflected in

the data from each source used to construct Table 1.2, and it should be noted that only the

mean values of quantitated phenolics, often across many cultivars, are summarized here.

Ellagitannins are present in both the flesh and achenes of strawberries (Aaby et al.

2005), and the profile is very similar to that observed in BRBs. Strawberries also contain

free ellagic acid and corresponding sugar conjugates that are not necessarily observed in

BRBs. Strawberries also contain notable levels of procyanidins that range in degree of

polymerization from 1 to >5 units (Gu et al. 2003). Other phenolics including flavonoids

and phenolic acids have also been reported, but at lower quantities (Table 1.2). Apart from their phenolic content, strawberries also contain appreciable amount of micronutrients, such as vitamin C and fiber (Basu et al. 2014).

19

Table 1.2 Strawberry phenolics with reported quantitation from whole fruit in literature

Compound a Mean ± SD b Min Max References Anthocyanins Cyd-3-O-glucoside 1.4 ±0.8 0.4 2.2 1,2,3,4 Cyd-3-O-malonylglucoside 0.1 1 Plg-3-O-glucoside 23.3 ±4.6 16.6 26.7 1,2,3,4 Plg-3-O-rutinoside 1.5 ±1.0 0.6 3.0 1,2,3,4 Plg-3-O-malonylglucoside 3.2 ±2.5 1.1 6.0 1,2,3 Plg-3-O-acetylglucoside 0.3 ±0.3 0.1 0.6 1,4 Plg-3-O-succinylglucoside 4.5 2

Flavonoids and Phenolic Acids Quercetin conjugates 1.4 ±0.4 1.1 1.8 1,2,4 Kaempferol conjugates 0.8 ±0.1 0.6 0.8 1,2,4 Catechin 3.5 ±1.5 2.4 4.5 1,2 Coumaric acid conjugates 3.2 ±1.9 1.9 5.4 1,2,4 Cinnamic acid conjugates 5.0 1 Gallic acid conjugates 11.5 1

Ellagic acid and Ellagitannins Sanguiin H-6 4.7 4 Lambertianin C 8.8 4 Agrimoniin 8.8 1 Other ellagitannins 2.3 ±0.5 1.8 2.8 1,2,4 Total Ellagitannins 14.1 ±3.6 11.6 16.7 1,4 Ellagic acid 2.3 ± 2.5 0.5 4.1 1,2 Ellagic acid conjugates 0.9 ±0.5 0.6 1.3 1,4

Procyanidins Dimers 9.1 1 Trimers 7.9 1 a Cyd = cyanidin; Plg = pelargonidin b mean of average levels reported in literature, measured as mg/100g fresh weight Reference 1: (Aaby et al. 2012) Reference 2: (Määttä-Riihinen et al. 2004) Reference 3: (Tulipani et al. 2008) Reference 4: (Buendía et al. 2010)

20

1.3 Absorption and Metabolism of Berry Phytochemicals

In order to impart their associated health benefits, berry phytochemicals must

theoretically be absorbed and circulated to relevant disease sites. The current consensus

for absorption and metabolism of important classes of berry phytochemicals will be

discussed below. In general, phenolic compounds, like those found in berries, are poorly

absorbed with considerable variation between individuals. Absorption of ingested

phenolics mostly occurs in the small intestines through a variety of mechanisms. Once

absorbed, compounds can be subjected to phase I/II metabolism in the enterocyte before

passing through the portal vein to the liver, another site of phase I/II metabolism.

Absorbed compounds are then circulated in the blood before filtration by the kidney and

excretion in the urine (Cassidy & Minihane 2017). This description is extremely generalized, and more specific aspects that pertain to each major class of berry phenolics

will be discussed below. Recent research has also highlighted the role of the gut

microbiome in the catabolism of to new products that are additionally

absorbed and may mediate bioactive effects (van Duynhoven et al. 2011). In fact,

catabolites of berry polyphenols have been proposed to mediate many of the health

benefits of berries because they are typically found at higher concentrations in the body

than they berry polyphenols themselves (Williamson & Clifford 2010). Thus, details

relevant to microbial metabolism of berry polyphenols will also be discussed below.

21

1.3.1 Absorption and Metabolism of Flavonoids

As discussed previously, flavonoids exist in plant-based foods primarily as

glycosides, with the exception of flavan-3-ols catechin and epicatechin. Aside from

anthocyanins, flavonoid glucosides are not observed in the blood or urine of humans,

indicating that cleavage of the sugar moiety must occur before absorption (Walle 2004).

Cleavage of flavonoid glycosides can occur at several sites. Walle and colleagues

demonstrated that this cleavage can begin in the oral cavity, mediated by bacteria and

shedded oral epithelial cells with β-glucosidase activity (Walle et al. 2005). As has been

thoroughly reviewed, cleavage in the small intestine is facilitated by intestinal brush

border lactase phloridzin hydrolase (LPH). There is also evidence to suggest that intact

flavonoid glycosides can be transported into intestinal enterocytes via the sodium-

dependent glucose transporter 1 (SGLT1) and subsequently cleaved by β-glucosidase

present in the cells. Flavonoid glycosides that are not cleaved in the small intestine (and

are thus not absorbed there) pass to the large intestine where they can be cleaved and

subsequently degraded by colonic bacteria (Rodriguez-Mateos, Vauzour, et al. 2014;

Walle 2004; Hollman 2004; Del Rio et al. 2013). The enzymes in the small intestines with β-glucosidic activity have been well characterized and act on a broad range of flavonoid and glucoside types. Small intestinal enzymes most efficiently cleave simple flavonoid glucosides (Németh et al. 2003; Day et al. 2000), which impacts their relative

absorption into human plasma. This principle was recently demonstrated in vivo by Lee

and colleagues who fed human subjects onion powder and powdered apple peel.

Although the total quercetin glycoside content was identical between the two treatments

22

(mass basis), the profile of the onion powder had predominantly simple glucosides, and

thus a greater absorption of quercetin was observed (Lee & Mitchell 2012). Earlier studies of a similar nature have been thoroughly reviewed (Hollman 2004). Once cleaved to the aglycone, flavonoids are subjected to phase II metabolism in the intestinal enterocyte or liver and then excreted primarily in the urine. It is also worth mentioning that small amounts of phase II metabolites can be effluxed back to the lumen from enterocytes and passed down to the colon, where they are subjected to microbial catabolism (Hollman 2004; Del Rio et al. 2013; Rodriguez-Mateos, Vauzour, et al. 2014;

Walle 2004).

Pharmacokinetic parameters for flavonoids indicate that they are quickly absorbed and excreted. The time to maximal plasma concentration for berry-relevant flavonoids

(anthocyanins, flavonols, and flavan-3-ols) is typically around 2 h or less, indicating that the small intestines represents an important site of absorption (Manach et al. 2005;

Hollman 2004). This timing may be delayed when flavonoids glycosylated to larger sugar moieties are delivered, indicating that the sugar moiety of these compounds may be cleaved by colonic bacteria with subsequent absorption in the large intestines (Hollman

2004). Nevertheless, data on the total urinary excretion of flavonoids provides insight into their relative bioavailabilities. Anthocyanins appear to exhibit extremely poor bioavailability with <2% of an administered dose typically recovered in urine. In some of these studies, anthocyanins were delivered via berries (Manach et al. 2005). Urine recovery of flavonols is generally <7% recovered (Manach et al. 2005). Flavan-3-ols are better absorbed with an average of 18.5% of a given dose recovered in urine (Manach et

23

al. 2005). Animal studies on tissue distribution of quercetin and pelargonidin indicate

that low levels of these flavonoids are found in various tissues following oral

administration. However, they are mostly concentrated in organs associated with their

absorption and excretion, and their presence is likely transient (Bieger et al. 2008; de

Boer et al. 2005; El Mohsen et al. 2006), in accordance with the quick absorption and

excretion observed in humans.

As alluded to earlier, flavonoids are subject to extensive degradation by microbes

present in the colon. Gut microbes are capable of cleaving sugar moieties that are not

substrates for human enzymes (e.g. rutinose), as well as modifying the resulting aglycone

flavonoid structure. Such modifications include ring fission, removal of hydroxyl,

methoxyl, and methyl functional groups, and α/β-oxidation (Williamson & Clifford

2010). Because the production and subsequent absorption of these catabolites must occur in the colon, their presence in plasma is delayed with peak concentrations reached more than 5 h after consumption (Williamson & Clifford 2010). The molecular structures of microbial catabolites vary, but they tend to be phenolic acids derived from the B-ring of the parent flavonoid (Williamson & Clifford 2010). For instance, the flavan-3-ol epicatechin may give rise to valeric-, propionic-, and acetic- phenolic acids, as well as hydroxybenzoic acids, all with 3-, and 3,4- hydroxylation patterns on the phenolic ring.

Human metabolism of these microbial products can also result in the formation of methylated derivatives, such as vanillic acid (from protocatechuic acid), and glycine conjugates, such as 3-hydroxyhippuruc acid (from 3-hydroxybenzoic acid) (Monagas et al. 2010), in addition to other phase II metabolites. It is also important to note that

24

considerable overlap in microbial degradation products exists between berry flavonoid

subclasses (Williamson & Clifford 2010), which presents a challenge in determining the

source of specific berry phytochemical metabolites in vivo. Nonetheless, these

metabolites can represent a significant portion of total flavonoid bioavailability. For

example when subjects were fed a bolus of quercetin-3-O-rutinoside (rutin), an average of only 0.6% was observed in urine as phase II quercetin metabolites. Conversely, 22% of the administered rutin was recovered in urine as microbial catabolites of quercetin

(Jaganath et al. 2006). Thus, these more bioavailable microbial metabolites may be partially responsible for the proposed biological effects of berry flavonoids because they are more bioavailable. Elucidating the biological activity of these compounds is an active area of research (Cassidy & Minihane 2017).

Significant work has been done to better understand the absorption and microbial degradation of anthocyanins in particular. Czank and colleagues recently completed a trial in which subjects consumed 500 mg of 5-13C labeled cyanidin-3-glucoside (Czank et

al. 2013). By following the concentration of 13C in feces, urine, breath, and blood the

authors were able to better determine the biological fate of anthocyanins. Importantly,

they confirmed that cyanidin-3-glucoside is partially degraded to protocatechuic acid and phloroglucinaldehyde in the intestine, likely because of the instability of anthocyanins at neutral pH. These products, as well as the parent cyanidin-3-glucoside are rapidly absorbed and excreted (Czank et al. 2013; De Ferrars et al. 2014). However, further degradation by gut microbiota also occurred (De Ferrars et al. 2014). Analysis of total

13C content of breath and urine demonstrated that the total bioavailability of cyanidin-3-

25

glucoside is actually more similar to that of other flavonoids when its degradation

products (chemical and microbial) and all their resulting human metabolites are

considered (Czank et al. 2013). Tracer studies such as these are rare but incredibly valuable to understanding the absorption and metabolism of phytochemicals.

1.3.2 Absorption and Metabolism of Procyanidins

As discussed earlier, procyanidins are a class condensed tannins composed primarily of individual units of catechin and epicatechin. As the absorption and metabolism of these monomeric flavan-3-ol units of procyanidins have already been discussed, this section will focus on procyanidins composed of ≥ 2 units. Some animal studies support the idea that procyanidins oligomers can be cleaved to smaller units in the stomach, but available human data does not support this finding (Zhang et al. 2016).

Absorption of procyanidins is limited to dimers, although they circulate at approximately

100 fold lower levels in blood than epicatechin despite equal dosing (mass basis) (Holt et al. 2002; Ottaviani et al. 2012). Importantly, consumption of procyanidins also does not elevate circulating levels of monomeric catechin or epicatechin (Ottaviani et al. 2012).

Intestinal models suggest procyanidin absorption is paracellular in nature (Zhang et al.

2016). In terms of kinetics, the absorption and excretion of procyanidin dimers is quick, with peak plasma levels at approximately 2 h (Holt et al. 2002; Ottaviani et al. 2012).

However, the bioavailability of procyanidins in the small intestine is low. A human study with ileostomy patients suggests that approximately 90% of ingested procyanidins reach the large intestine (Kahle et al. 2007). Once in the large intestine, procyanidins undergo

26

extensive degradation by gut bacteria (José Serrano et al. 2009; Zhang et al. 2016). The

main microbial metabolites of procyanidins include those discussed for epicatechin

(Monagas et al. 2010), as well as larger catabolites that partially preserve the oligomeric

structure of the procyanidin (Zhang et al. 2016). Many of the small molecular weight

metabolites have been observed in human urine after consumption of chocolate (Rios et

al. 2003), and the main metabolite detected in plasma is 3,4-dihydroxyphenylacetic acid

(and its phase II metabolites) (Saura-Calixto et al. 2010). Thus, although the bioavailability of intact procyanidins is low (dimers) or virtually zero (trimers and above), microbial metabolism may play an important role in the overall bioavailability of these compounds.

1.3.3 Absorption and Metabolism of Ellagitannins

Ellagitannins are large polyphenolic structures that, when hydrolyzed, form ellagic acid. There is no evidence for intact absorption of ellagitannins from the diet, but work with ileostomy patients has demonstrated that they can be cleaved in physiologic conditions in the small intestine to form ellagic acid (González-Barrio et al. 2010).

Ileostomy patients have essentially had the end of their small intestines diverted from the colon to a bag outside the body because of large intestinal dysfunction, allowing direct sampling of luminal contents. Even if more ellagic acid is released in the small intestine, the bioavailability of ellagic acid is exceedingly poor with <1% of a supplied dose, from raspberries or BRBs, excreted in urine (González-Barrio et al. 2010; Stoner et al. 2005).

The small amount of ellagic acid that is absorbed reaches maximal concentration in

27

plasma within approximately two hours and is rapidly excreted (Stoner et al. 2005).

Rather than absorbed in its native form, ellagic acid (free or from ellagitannins) is

catabolized by gut microbiota to dibenzopyran-6-one derivatives known as urolithins.

The bioavailability of these compounds is greater than their chemical parents. An average of almost 15% of a provided dose of ellagitannins/ellagic acid in pomegranate juice was recovered in urine as these metabolites (Cerdá et al. 2004). Due to their production by colonic bacteria, urolithins reach peak concentrations between 24-48 h and can persist up to 72 h (Landete 2011).

Identified urolithins include urolithin M-5, M-6, M-7, E, D, C, A, and B (Figure

1.2). These differ in number and position of hydroxyl groups on the dibenzopyran-6-one

backbone. The primary urolithins observed in plasma and urine are urolithin A, iso-

urolithin A, and , typically conjugated to a glucuronide moiety (Tomas-

Barberan et al. 2017). Large inter-individual variation has been observed in the type and amount of urolithins produced, leading some researchers to define “metabotypes” according to an individual’s urolithin profile (Tomas-Barberan et al. 2017). Some individuals are capable of producing all identified urolithins, while others do not produce isourolithin A, or urolithin B. In both cases, individuals can be classified as high and low excreters. On the other hand, there are individuals who do not produce urolithins at all.

Interestingly, some non-producers will begin to produce urolithins following high dosages of ellagitannins and ellagic acid over a long period of time (Tomas-Barberan et

al. 2017). Several animal studies have reported the presence of urolithins in tissues

involved with their absorption and excretion, but also in the heart, lung, brain, and

28

prostate (Tomas-Barberan et al. 2017). One mouse study evaluated the distribution of

urolithins in prostate, intestine, colon, liver, kidney, lung, plasma, and brain over time

following an oral dose of urolithin A. Fascinatingly, prostate levels of urolithin A and

methylurolithin A were as high as gastrointestinal tissue levels, indicating that these

metabolites can be highly concentrated in the mouse prostate (Seeram et al. 2007). The

presence of urolithins in the human prostate has also been confirmed following

consumption of pomegranate juice or walnuts (González-Sarrías et al. 2010).

The biological activity of urolithins has also been extensively studied in in vitro models, which has been thoroughly reviewed (Tomas-Barberan et al. 2017). Work with

animal models appears limited, though evidence exists for the potential activity of

urolithins to regulate aspects of colonic inflammation (Larrosa et al. 2010). Human

studies demonstrating biological activity of urolithins is also limited. González-Sarrías

and colleagues observed no changes in markers of cell proliferation in human prostate of

men with benign prostatic hyperplasia or prostate cancer, following a 3 d intervention

with pomegranate juice or walnuts, despite the presence of urolithins in the organ in a

subset of subjects (González-Sarrías et al. 2010). Thus, more work is needed to better

understand how urolithins may mediate the biological activity associated with the

consumption of ellagitannins and ellagic acid.

29

O

R3 O

R2 R4

R5 R1

Urolithin R1 R2 R3 R4 R5 M-5 OH OH OH OH OH M-6 OH H OH OH OH M-7 OH H OH H OH E OH OH OH H OH D OH OH OH OH H C OH H OH OH H A OH H OH H H Iso A OH H H OH H B OH H H H H

Figure 1.2 Molecular structures of urolithins

30

1.3.4 Absorption and Metabolism of Phenolic Acids

Although hydroxybenzoic and hydroxycinnamic acids are collectively referred to as phenolic acids, their absorption and bioavailability in humans are different.

Hydroxybenzoic acids are absorbed via paracellular transport between intestinal enterocytes (Lafay & Gil-Izquierdo 2008). The bioavailability of hydroxybenzoic acids in humans has mainly been assessed by studying gallic acid. Absorption and excretion of gallic acid is rapid, with peak plasma concentration of the compound and its metabolites occurring in approximately 1.5 h (Shahrzad et al. 2001). The average proportion of administered gallic acid recovered in urine (as gallic acid or metabolites) is approximately 37% (Lafay & Gil-Izquierdo 2008). Work with rat colonic microflora suggests that hydroxybenzoic acids are quite stable against modification by gut bacteria

(Serra et al. 2012). Thus ingested hydroxybenzoic acids display relatively high bioavailability but similar pharmacokinetic parameters compared to other berry polyphenols.

Hydroxycinnamic acids are also absorbed in a paracellular manner, but cell studies also suggest transport via the monocarboxylic acid transporter in the small intestine (Lafay & Gil-Izquierdo 2008). The absorption of hydroxycinnamic acids in humans has mainly been studied with caffeic, ferulic, and chlorogenic acids (caffeic acid- ester). In general, the absorption and subsequent excretion of these compounds is rapid, with peak plasma levels reached by 2 h following an administered dose (Lafay & Gil-Izquierdo 2008; Monteiro et al. 2007; Kern et al. 2003). Esterification of hydroxycinnamic acids, with sugar or quinic acid moieties, generally decreases the

31

absorbed amount, and notably caffeoyl-quinic esters are observed intact in human plasma

and urine (Lafay & Gil-Izquierdo 2008). Urinary excretion of hydroxycinnamic acids can vary, but levels from human studies suggest an average of approximately 10% of an administered dose. However, this range is quite wide with levels as low as 3.1% and up to 25% even 95% (one study) reported depending on food source (Lafay & Gil-Izquierdo

2008). Hydroxycinnamic acids are also metabolized by gut bacteria. For instance,

chlorogenic acids are cleaved to yield caffeic acid, which is then further degraded to

other phenolic acids (Lafay & Gil-Izquierdo 2008). Thus, the colon represents an

important site for absorption of chlorogenic acids via their derived metabolites

(Williamson & Clifford 2010). Notably, hydroxycinnamic and hydroxybenzoic acids are

also downstream products of microbial catabolism (and subsequent human metabolism)

of other phenolic molecules, such as anthocyanins (De Ferrars et al. 2014) and flavan-3-

ols (Monagas et al. 2010). Thus, these compounds are highly important to berry research,

not just because they are natively present in these fruits, but also because they are

products of microbial catabolism of other berry phytochemicals.

1.4 Effects of Food Processing and Storage on Berry (Poly)phenols

There are numerous factors that can affect the phytochemical profile of berries

and berry products. The berry cultivar, environment in which it is grown, and techniques

used to produce and harvest it can all influence the phenolic profile of the fruit itself

(Tiwari & Cummins 2013). However berries are not just consumed as fresh produce.

They are also consumed as processed products such as jams, jellies, syrups, and juices.

32

Thus, in order to assess how exposure to berry phenolics may impact human health, a better understanding of how factors related to food processing and storage affect the phenolic profiles of processed berry products is essential.

There are several reviews that thoroughly detail findings from studies on the effects of food processing and storage on levels of berry phenolics, especially as they pertain to anthocyanins, procyanidins, and ellagitannins (Francis 1989; Patras, Nigel P

Brunton, et al. 2010; Howard et al. 2012; Bakkalbaşi et al. 2009; Kårlund et al. 2014;

Amarowicz et al. 2009). To broadly summarize, the impact that food processing may have on berry phenolic profiles is highly dependent on the type of product that is being made, the process used to make it, and the conditions in which it is stored. There are numerous details that feed in to each of these factors, and while all three of them can independently determine final phenolic content, they are also interrelated (Figure 1.3).

Representative publications that illustrate the influence of these factors are detailed in

Table 1.3. Here, we will summarize the key aspects of food processing and storage that can affect the final (poly)phenol content of berry products. The interrelation of food processing, storage and the food matrix will be integrated throughout this discussion.

33

34

Figure 1.3 Factors that affect levels of (poly)phenolic compounds in processed berry products

34

Table 1.3 Representative studies on berry (poly)phenols over food processing and storage

Product Processing Analytes Major Findings Phytochemical Changes (White et al. 2011) Cranberry Juice 1. Blanch (95 °C, Anthocyanins • Anthocyanin loss in blanching, pressing • Procyanidins ↓ 43% – 51% 3 min) Flavonols (most), pasteurization (slight). • Flavonols: ↓ 33% – 56% 2. Pectinase (45 Procyanidins Glucosides most stable conjugates. • Anthocyanins: ↓ 47% – 61% °C, 1 h) • Flavonols more heat stable than 3. Press anthocyanins. Increase with blanching. 4. Clarify Significant loss after pressing. Stability 5. Pasteurization dependent on glycoside. Aglycone (90 °C, 10 content increased with processing. Low min) pH (2.6) of cranberry may encourage hydrolysis • Procyanidins increase with blanching. Loss with pressing.

35 (Amakura et al. 2000) Berry Jams 1. Fruit and Flavonols • Flavonol conjugates are generally stable • Caffeic acid: ↓ 64% sugar slowly (with and over processing of berries into jam. • Ellagic acid: ↓18% – ↑102% stirred with without Some evidence of glycoside hydrolysis • Myricetin: ↓28% – ↑111% “mild heat” hydrolysis), to aglycone. • Quercetin: ↓25% – ↑123% 10-15 min. Ellagic acid, • Ellagic acid stable to processing. • Kampferol: ↓28% – ↑115% Total Evidence for greater content of free Phenolics ellagic acid with processing. • Total phenolics: ↓3.5% – ↓23%

• Total phenolics values showed little if any numerical change.

35

Table 1.3 Continued Product Processing Analytes Major Findings Phytochemical Changes (Hager et al. 2010) Blackberry: Juice Ellagitannins • Thermal aspects of processing had no • Canned/Pureed juice, puree, 1. Blanch (95 °C, effect on total ellagitannin levels, o Processing: ↑ 140% – 160% canned in syrup, 3 min) though some potential evidence of o Storage: ↓ 28% (canned) canned in water 2. Pectinase(45 depolymerization was seen. • Juice °C, 1 h) • Heat increases extractability of o Processing: ↓ 55% – 73% 3. Press ellagitannins from fruit, but large losses o Storage: ↓ 43% 4. Clarify were seen when fruit was juiced. 5. Pasteurize • High MW ellagitannins hydrolyze to (heat to 90 °C) form ellagic acid over 6 months storage Canned (25 °C) 1. Boil 15 min • Some leaching of ellagitannins into Puree canning liquid over storage. 1. Blend 36 2. Heat to 90 °C 3. Add sugar 4. Heat to 93 °C 5. Fill and boil 15 min Storage 25 °C, 6 months

36

Table 1.3 Continued Product Processing Analytes Major Findings Phytochemical Changes (Hakkinen et al. 2000) Strawberry jam, 1. Strawberry Flavonol • Loss of flavonols with all domestic • Quercetin: Bilberry soup, Jam: “cooked” aglycones cooking methods o Processing: ↓ 18% (jam) – Crushed with sugar 15 (after acid • Crushing of fruit prior to cooking may 85% (juice) lingonberries, min hydrolysis) facilitate flavonol degradation (enzyme Storage: ↓ ND – ↑ 32% Black currant 2. Bilberry soup: o activity) • Kaempferol: juice, boiling 10 min • Juicing resulted in significant loss of Processing: ↓ 15% lingonberry 3. Lingonberries: flavonols mostly due to removal of fruit o Storage: ↓ ND juice crushed, skins o • Myricetin: refrigerated. • Storage effects varied by fruit and Filtered, flavonol analyte. Quercetin was most o Processing: ↓ 35% – 70% sweetened for stable, kaempferol and myricetin less (juices) juice (no heat) so. Impact of other food components o Storage: ↓ ND 4. Black currant (vitamin C) on oxidative degradation of 37 juice: cold flavonols is discussed. press or steam extracted (steamed 60 min) 5. Crowberry juice: cold press 6. Storage (-20 and 5 °C, 9 months)

37

Table 1.3 Continued Product Processing Analytes Major Findings Phytochemical Changes (Hager et al. 2008) Black raspberry: 1. See (Hager et Total • Anthocyanin levels decreased with all • Canned juice, puree, al. 2010) monomeric processes due to physical separation o Processing: ↓ 42% – 51% canned in syrup, anthocyanins (juices) and chemical/enzymatic o Storage: ↓ 75% – 76% canned in water degradation. Levels decreased markedly • Juice over storage. Processing: ↓ 69% – 73% • o % polymeric color was relatively Storage: ↓ 62% – 75% unaffected by processing, but increased o • Puree over storage. o Processing: ↓ 37% o Storage: ↓ 76% (Zafrilla et al. 2001) Red raspberry 1. Mix with Free and • Ellagic acid glycosides and flavonol • Quercetin glucoside:

38 jam pectin and glycosylated glycosides not highly affected by o Processing: ↓ 6% sugar. ellagic acid, processing into jam. 2.5-fold increase in o Storage: ↓ 40% 2. Heat at 78 °C quercetin free ellagic acid with processing. • Kaempferol glucoside: under vacuum glucoside, • Storage for 6 mo (20 °C) had greater Processing: ↓ 20% 15 min kaempferol o effect on analyte levels than processing. Storage: ↓ 50% 3. Heat to 92 °C glucoside • Half of flavonol glycosides lost, o • Free ellagic acid: and hot fill kaempferol glucoside more so than Processing: ↑ 2.5% 4. Store 20 °C, 6 quercetin glucoside. Loss happened o months mainly over first three months o Storage: No effect • Free ellagic acid increased over first • Ellagic acic glycosides: month of storage, then decreased to o Processing: ↓ 4% original levels. Ellagic acid glycosides o Storage: No large effect had minimal change over storage.

38

Table 1.3 Continued Product Processing Analytes Major Findings Phytochemical Changes (Brownmiller et al. 2009) Blueberry: juice, See (Hager et al. Procyanidins • High MW polymers more affected by • Canned puree, canned in 2010) all juice processing steps than mono- o Processing: ↓ 22% – 35% syrup, canned in di- and trimers. Attributed to ability of o Storage: ↓ 59% – 66% water high MW polymers to bind to cell wall • Juice components and greater thermal Processing: ↓ 77% – 81% instability. o Storage: ↓ 40% – 64% • Procyanidins lost in juice processing o • Puree because of removal (i.e. left in press cake) and thermal degradation. o Processing: ↓ 59% • Storage for 6 mo (25 °C) resulted in o Storage: ↓ 84% decreased levels of procyanidins, but less so than observed over juice

39 processing.

• Loss of procyanidins also seen in processing and storage of blueberry puree. In processing, high MW polymers more affected (binding to cell wall components, hydrolysis to low MW oligomers). In storage, mono- and dimers were more affected. • Better retention of procyanidins were seen in canned blueberries, with potential effects based on canning medium (syrup, water). Considerable loss over storage occurred.

39

1.4.1 Processing Effects

As seen in Table 1.3, the phenolic content of berry products differs based on the type of processing used to produce them. For instance, Hager and colleagues observed a marked decrease in the ellagitannin content of blackberries when processed in to juice, but an increase in content when the berries were canned (Hager et al. 2010). Key

differences in the processing used to make these two products easily explain this

discrepancy. Ellagitannins are concentrated in the seeds of berries, so mechanical

removal of the seeds in the juicing process (pressing) resulted in lower content in the

final product. The increase noted in canned products was hypothesized to be an effect of

depolymerization of large, unmeasured, ellagitannins or an enhanced ability to extract

ellagitannins from blackberry seeds (Hager et al. 2010). Other phenolics, such as

anthocyanins and procyanidins, can be lost when pressing berries into juice due to their

retention in the pomace, in the case of procyanidins, their affinity to bond with plant cell

wall components (Howard et al. 2012). In general, retention of phenolic compounds

throughout processing decreases as the number of required processing steps increases.

Thermal processes are a common step in the production of almost all berry

products. In broad terms, ellagitannins are rather heat-tolerant, while levels of

procyanidins, anthocyanins, and flavonols are reduced by thermal processing to differing

degrees. The magnitude of these losses are dependent on factors such as molecular

structure and pH. Low molecular weight procyanidins were better retained through

thermal processes than high molecular weight polymers in the production of various

blueberry products (Brownmiller et al. 2009). In a model system, quercetin-3-O-

40 rutinoside was found to be more stable than the quercetin aglycone when held at 100 °C, even when degradation of both was enhanced at pH 8 (Buchner et al. 2006). The stability of anthocyanins is similarly affected by molecular structure, as hydroxylation on the B- ring may decrease stability, and methoxylation and acylation increases stability. In addition, the type and location of conjugated sugar moieties have also been found to affect stability (Howard et al. 2012; Sadilova et al. 2006; Francis 1989). Anthocyanins are also notoriously affected by pH, as they exist in structural equilibrium mainly between the charged flavylium cation form (pH < 3), and the carbinol pseudo-base form

(approximately pH 4.5, colorless). The exact equilibrium of these two anthocyanin forms, and additionally the quinoidal base and chalcone forms, over at different pH values is dependent on molecular structure (Francis 1989; Schwartz et al. 2017). The stability of anthocyanins to thermal processing is enhanced at lower pH (Loypimai et al.

2016; Sadilova et al. 2007). However, as seen in some studies in Table 1.3, the loss of anthocyanins due to thermal treatment alone may be minor compared to losses due to other factors, such as removal of berry tissue (e.g. pressing into juice), or storage.

Other intrinsic and added food components can affect thermal degradation of phenolic compounds. For example, when several different types of berries were individually processed into jams, quercetin was lowered by up to 25% or increased by the same amount, suggesting differences in fruit composition may influence stability of the flavonol (Amakura et al. 2000). Tinsley and Bockain completed an extensive study on the influence of sugars, pH, oxygen, and amino acids in a model system of pelargonidin-

3-glucoside degradation. Sugar addition resulted in faster anthocyanin degradation,

41

primarily mediated through degradation of sugars to furfural-like products. Thus factors that encouraged the formation of furfural-like products, such as the presence of oxygen and amino acids, also resulted in faster degradation of pelargonidin-3-glucoside (Tinsley

& Bockian 1960). These studies illustrate the importance of considering the complete composition of a food product when evaluating factors that may modulate the thermal degradation of phenolic compounds.

During thermal processing, phenolic compounds can be degraded by a number of mechanisms. One common way among all phenolic compounds is enzymatic degradation via oxidase and peroxidase, but this can be limited by blanching berries (and thus inactivating enzymes) as an initial processing step (Howard et al. 2012). Chemical degradation of phenolics is hypothesized to be mostly oxidative in nature. The thermal degradation of anthocyanins has been extensively investigated and typically involves opening of the favylium C-ring, hydrolysis of glycosidic and acyl bonds, and eventual cleavage of the anthocyanidin to form lower molecular weight organic acids and aldehydes derived from the A- and B-rings (Patras, Nigel P Brunton, et al. 2010; Sadilova et al. 2006; Sadilova et al. 2007). While this degradation mechanism can be modulated by pH, this pathway is hypothesized to be the most relevant to berry products that typically have a pH value of approximately 3.5 (Howard et al. 2012; Sadilova et al.

2007). Additional degradation can occur via condensation of anthocyanins with their own cleavage products or other phenolic constituents present in the food being processed

(Sadilova et al. 2006; Patras, Nigel P Brunton, et al. 2010). The flavonol quercetin follows a similar pattern. Hydrolysis of quercetin glycosides to the aglycone over

42 thermal processing can occur, but is more likely at low pH (White et al. 2011). A model system of quercetin degradation characterized several possible oxidative products including chalcan-trione products from C-ring opening and cleavage products that are structurally related to both the A- and B-rings (Barnes et al. 2013). Thus flavonoids are susceptible to degradation over thermal processing, resulting in a variety of new phenolic products.

Mechanisms of procyanidin thermal degradation have not been as extensively elucidated. Pasteurization of cranberry juice did not cause a decrease in procyanidin levels (White et al. 2011), whereas pasteurization or thermal processing of different blueberry products resulted in decreased levels (Brownmiller et al. 2009). Cleavage of procyanidin polymers to smaller oligo/monomeric units and complexation with plant wall components have been suggested as potential degradation pathways (Howard et al. 2012).

Ellagitannins are rather stable to thermal processing, though minor levels of hydrolysis to ellagic acid may occur (Bakkalbaşi et al. 2009). When blackberry juice was pasteurized, no significant change in the contents of selected ellagitannins were observed (free ellagic acid was not quantified). However, when blackberries were thermally processed into canned goods, a significant increase in total ellagitannin content was observed, potentially due to thermally-assisted extraction of ellagitannins from blackberry seeds or cleavage of high molecular weight ellagitannins (Hager et al. 2010). Levels of free and glycosylated ellagic acid were also stable when raspberries were processed into jam

(Zafrilla et al. 2001). For both procyanidins and ellagitannins, more significant losses occur with other processing steps, especially removal of fruit tissue and seeds.

43

1.4.2 Effects of Storage

Many of the principles discussed with respect to thermal processing also hold true for the stability of berry phenolics over storage. In some cases, the loss of phenolic products can actually be more severe in storage than in processing (Howard et al. 2012).

As discussed with respect to food processing, the storage temperature, length of storage time, and several aspects of the food matrix all influence the extent of degradation. In general, storage at lower temperatures results in greater retention of berry phenolics

(Patras, Nigel P Brunton, et al. 2010; Amarowicz et al. 2009). As seen in Table 1.3, the chemical structure of the phenolic and type of berry product additionally influence the degree to which the phenolic profile changes over time. For instance, ellagitannins were more stable over 6 months of storage at 25 °C in canned blackberry products than non- clarified blackberry juice. The loss of ellagitannin content was potentially attributed to complexation with remnant plant cell wall components (non-clarified juice), while hydrolysis of high molecular weight ellagitannins to ellagic acid also occurred (Hager et al. 2010). When raspberry jam was stored at 20 °C, an increase in free ellagic acid content was observed for the first 30 days, but the content then decreased to initial levels, resulting in no net decrease in ellagic acid content over 6 months (Zafrilla et al. 2001).

Similar phenomenon has been observed in ellagitannin-containing juice products as well

(Bakkalbaşi et al. 2009).

Procyanidins are rather unstable to storage, as illustrated when processed blueberry products were stored at 25 °C for 60 days. The loss of procyanidin content in

44 these products was attributed to condensation with other phenolic components, complexation with cell wall components, and residual oxidative enzyme activity.

Interestingly, the loss in procyanidin content was proportionally less in juice than canned products, but in the final juice product, total procyanidin content is less than in canned fruit (because of the removal of berry pomace) (Brownmiller et al. 2009). The stability of flavonols, such as quercetin, over storage is not clear in the literature (Amarowicz et al. 2009). As demonstrated by Hakkinen and colleagues, levels of quercetin can increase, decrease, or remain stable over storage dependent on the domestic processing technique used (Hakkinen et al. 2000).

Anthocyanins are unstable over storage in processed berry products. As demonstrated by Hager and colleagues, total monomeric anthocyanins decreased by approximately 75% in all processed BRB products over 6 months of storage at 25 °C, which was a larger change in magnitude than was incurred over processing of canned/pureed BRB products. As opposed to degradation over processing, loss of monomeric anthocyanins over storage was accompanied by a significant increase in polymeric color value (Hager et al. 2008). This illustrates an important mechanism by which anthocyanins degrade over storage. In addition to the cleavage of anthocyanins discussed earlier, anthocyanins can condense with themselves or other phenolic compounds, such as procyanidins, to form brown polymers (Patras, Nigel P Brunton, et al. 2010; Francis 1989). This process may be facilitated by sugar degradation products created during thermal processing, like glyoxylic acid and furfural (Howard et al. 2012).

Other reactions can actually preserve the color imparted by anthocyanins by what is

45

known as co-pigmentation. Condensation of anthocyanins with phenolic acids in berry

juices has resulted in greater color stability via formation of pyranoanthocyanins, though

the phenolic acids were added to the juices at concentrations 10-fold higher than the

anthocyanins (Rein & Heinonen 2004; Rein et al. 2005). Non-covalent complexation of

anthocyanins with other flavonoids can also enhance anthocyanin stability, as can

interactions within the molecular structures of acylated anthocyanins (Francis 1989).

Despite these stabilization methods, anthocyanins are highly susceptible to degradation

over storage, especially at elevated temperatures, and their degradation results in the

formation of small molecular weight phenolic acids and high molecular weight phenolic

polymers.

1.4.3 Effects of Processing and Storage on Bioactivity

Although it is clear that both food processing and storage change the phenolic

profiles of berry products, few studies have investigated the impact of post-harvest

processing and storage factors on the bioactivity of berries. Schmidt and colleagues found

fractionated extracts from store-bought thermally processed blueberry products exhibited decreased anti-proliferative activity in murine liver cancer cells when compared to non- thermally treated samples (Schmidt et al. 2005). This study, however, only evaluated isolated berry components and did not control for variation due to horticultural factors

(e.g. cultivar, production year and environmental conditions), production techniques, or storage time. Another study found that anthocyanin extracts from blueberry juice exhibited slightly decreased anti-proliferative activity in human colorectal

46

adenocarcinoma cells when the juices were stored for 90 days at refrigerated and room

temperatures (Srivastava et al. 2007). Rodriguez-Mateos and colleagues conducted a randomized, controlled, crossover trial in which subjects were fed equivalent doses of freeze-dried blueberry powder as a freshly prepared beverage or baked into a muffin.

Despite some differences in food and plasma levels of blueberry phenolics (or metabolites thereof), the two treatments similarly affected flow mediated dilation, a measure of vascular endothelial function (Rodriguez-Mateos, del Pino-García, et al.

2014). There is a need for well-controlled studies that investigate the impact of not only

processing, but also storage, on the bioactivity of berry products with respect to their

comprehensive chemical profiles.

1.5 Untargeted Metabolomics

Metabolomics is an emerging field of chemical analysis that focuses on the

comprehensive profiling of as many small molecules as possible in a given system. This

is appealing because it allows for a more thorough and encompassing analysis of

chemical composition than traditional methods (Wishart 2008). As opposed to targeted

methodology, where assays are specifically developed and optimized to quantify known

compounds of interest, metabolomics takes a more untargeted approach. The

metabolomics workflow follows a general scheme. First, high resolution spectral data is

obtained from samples using analytical platforms like nuclear magnetic resonance

(NMR) and mass spectrometry (MS). Once the spectra is obtained, it is processed to

extract chemical features (i.e. compounds) and provide intensities of the hundreds to

47

thousands of detected chemical features across samples, resulting in a large,

multidimensional dataset. This dataset can then be analyzed using univariate or

multivariate statistics to understand which chemical features are significantly altered

between sample treatment conditions. Identification of these chemical features is then

attempted by comparison of spectral data to available databases and authentic standards

(Wishart 2008). As with other “omic” techniques, metabolomics is regarded as

hypothesis-generating, in that the objectives for metabolomics experiments may be a little

more general and exploratory in nature. As will be discussed at the end of this section,

this approach has potential application in studies aimed at better understanding the

phenolic profiles of berries and how their consumption may impact health.

1.5.1 Mass Spectrometry-based Metabolomics

As mentioned previously, there are several potential platforms that can be used to

execute metabolomics experiments, the two most common being NMR and MS. NMR

has several advantages over MS, including its quantitative nature, a large number of well-

developed databases for the identification of chemical compounds, and the ability to

generate identifications for unknown compounds (Wishart 2008). However this platform is not as sensitive as MS, as compounds below µM levels are often not detectable, limiting it’s applicability to study phytochemicals that exist in concentrations below this limit (Wishart 2008; Pan & Raftery 2007). Chromatographic separation is typically used in front of an MS detector to improve compound resolution, though direct injection MS is sometimes used. Gas chromatography is often used in front of MS as it is rather

48

reproducible, and has well-developed libraries to assist in compound identification.

However, this technology requires analytes to be volatile often via chemical

derivatization, which greatly complicates sample preparation and data analysis (Dunn &

Ellis 2005). Liquid chromatography coupled to MS (LC-MS) is well suited to

metabolomics because sample preparation is simple and it can cover a wide range of

analytes (Wishart 2008; Dunn & Ellis 2005). This technique can cover a larger proportion of the metabolome than other platforms (Wishart 2008). Because of this and its flexibility to a wide range of sample matrices, LC-MS-based platforms are widely used for metabolomics analyses.

Due to the untargeted nature of metabolomics experiments, special considerations must be taken to ensure that true differences between groups are due to the variables being investigated. Steps should be taken to ensure equal treatment of all samples to avoid unwanted variation that might affect eventual statistical analyses (Naz et al. 2014).

Additionally, extraction of samples should be simplified as much as possible while providing sufficient coverage of metabolites (Naz et al. 2014). Prior to running a set of samples, the run-order should be randomized. Pooled quality control samples, which consist of equal aliquots from all samples being analyzed, are interspersed throughout the sample set to monitor and correct for drifts in instrument performance. Internal standard mixes can also be used, but they may not accurately reflect the analysis of all features detected in the sample (Naz et al. 2014). Other considerations for optimizing metabolomics studies to maximize raw data quality have been comprehensively reviewed, but include a high level of experimental control, appropriate study design (i.e.

49

cross-over as preferred over parallel arm design), and sufficient analytical method

development (e.g. reduce ion suppression, increase chromatographic resolution),

(Scalbert et al. 2009; Naz et al. 2014).

Once raw spectral data is collected, it must be pre-processed into a form that can

be statistically analyzed. This includes picking and binning peaks (detected ions) in the

data and quantifying (integrating) them (Hendriks et al. 2011). For MS-based metabolomics, peak binning includes grouping all expected isotopes and adducts for a singular compound (e.g. +/- H, +Na, +K, M-2H, 2M-H) into a singular entity. Once peaks are binned, they must be aligned according to retention time across all samples, resulting in a large set of tabulated data that contains information on each molecular feature, including its calculated molecular weight, retention time, and relative abundance among samples (Katajamaa & Orešič 2007). This data can then be passed through quality assurance and normalization processes prior to statistical analyses.

Quality assurance in metabolomics is an iterative process that employs data analysis tools to assess the quality of the multidimensional data. In short, quality assurance procedures aim to remove low quality data points from analysis (Considine et al. 2018). Examples include the removal of features that appear in less than 75% of samples in at least one treatment group, features that are below a certain abundance value, and features that were highly variable across quality control samples. There is no set methodology for quality control within the metabolomics community, but it is recommended that protocols are thoroughly reported for each study (Considine et al.

2018). Data normalization can take many forms, but typically includes centering, scaling,

50 and transformation, as reviewed by van den Berg and colleagues (van den Berg et al.

2006). Centering offsets the measured abundance of each chemical feature to fluctuate around zero. Scaling methods (e.g. Pareto scaling and autoscaling) account for the large differences in abundances observed in each chemical feature that can skew statistical results and thus treat high and low abundance compounds as they are of equal importance in modeling approaches. Transformations, such as logarithmic transformation, can have a similar effect as scaling, but the aim is to reduce heteroscedasticity (dependency of variability on measurement value). There are several different methods for scaling and transformations, each with inherent advantages and disadvantages (van den Berg et al.

2006). The goal of data normalization in the end is to emphasize the variation caused by sample treatments thereby facilitating later statistical analyses (van den Berg et al. 2006).

A final point worth mentioning at this point in the workflow is the presence of missing values, which are likely to occur due to technological limitations. There are several approaches that can be used, such as imputation using various algorithms or removal of chemical features that contain missing values. The approach that is used can impact the outcome of statistical analyses and thus should be described when reporting results

(Hrydziuszko & Viant 2012). The result of the quality assurance process is a high quality dataset that is optimized for future statistical analyses.

The goal of data analysis in metabolomics experiments is to uncover chemical features that differ significantly between treatment conditions. A typical first step is surveying the dataset for overall trends using principal components analysis (PCA). PCA is a technique that reduces highly dimensional datasets (like those in metabolomics) to a

51

smaller set of new variables (components) that capture major sources of systematic

variation present among all the chemical features across all samples (Kemsley et al.

2007). This technique is also unsupervised, in that the model is not computed in a way to

maximize differences among specified groups. Partial least squares (PLS) analysis is

related to PCA in that the dimensionality of the data is also reduced to a set of new,

latent, variables, but this method is a supervised analysis. In PLS, categorical (PLS- discriminant analysis; PLS-DA) or continuous (PLS-regression; PLS-R) dependent variables are coded for each sample, and the algorithm calculates a model that maximizes the covariance between the independent variables (metabolomics dataset) and dependent variables (Kemsley et al. 2007). Using PLS, features can be ranked by their importance to the model (Chong & Jun 2005), which enhances interpretation and prioritization of features for identification in later steps in the metabolomics workflow. PLS can be prone to overfitting, wherein noise in the dataset contributes to the discriminating power of the models making them inaccurate and less robust when applied to separate datasets. This phenomenon can be evaluated using cross-validation techniques, which consist of consecutive rounds of excluding random samples from the dataset, re-calculation of the

PLS model, and evaluation of the model’s predictive ability using the excluded samples

(Kemsley et al. 2007; Westerhuis et al. 2008). The performance of PLS models is described using a number of statistical terms, mainly Q2, though other options exist

(Szymanska et al. 2012). There are no firm cutoffs for valid Q2 statistcs, but values > 0.4

(max of 1.0) have been suggested in some scenarios. Permutation tests can provide some

measure on the statistical significance of the Q2 measure (Worley & Powers 2012).

52

Modern software facilitates quick computation of PLS models, but they must be carefully

evaluated for performance prior to results interpretation.

PCA and PLS represent mainstays of multivariate analyses used in metabolomics,

but extensions of these methods as well as different advanced statistical methods (e.g.

random forests, artificial neural networks, etc.) can also be used in lieu or as

complements to PCA and PLS (Ronningen et al. 2018; Hendriks et al. 2011; Gromski et

al. 2015). One important extension of these methods for nutritional studies has been

described by Van Velzen and colleagues as multi-level PCA and PLS-DA (Van Velzen et

al. 2008). In studies where individuals are fed a particular item and perturbations to their

metabolome are subsequently investigated, metabolomic variation associated with the

consumption of specific foods can be easily overwhelmed by variation due to other

biological and extrinsic factors (Rezzi et al. 2007). When a study is designed as a cross-

over intervention, where each person serves as their own control, the variation in the

dataset can be split into between-subjects variation (inter-individual variation) and within-subject variation (variation due to treatment) and subsequently modeled using techniques such as PCA and PLS-DA (Van Velzen et al. 2008). This example demonstrates the importance of study design and large amount of flexibility in adapting multivariate analyses to situations that require special consideration to uncover interesting variation within metabolomics datasets.

Univariate analyses, such as t-tests and ANOVA, are also used in the analysis of metabolomics datasets. As opposed to multivariate analyses that consider all data points simultaneously, univariate analyses consider each chemical feature independently,

53

resulting in a large number of repeated statistical tests for a typical metabolomics dataset

(Saccenti et al. 2014). Because of this, measures must be taken to account and correct for

false positives that may arise due to chance (i.e. false discovery rate; type I error). There

are many options for false discovery rate corrections, such as Bonferroni and Benjamini-

Hochberg, but depending on the desired strictness, these methods also increase the

chance of false negatives to differing degrees (type II error) (Broadhurst & Kell 2006;

Saccenti et al. 2014). The results of univariate and multivariate analyses on the same

metabolomics dataset may differ, but it is difficult to conclude that one technique is

superior. In reality, the information obtained from both types of analyses may be

complementary and when combined may lead to greater insight into interesting features

in metabolomics datasets (Saccenti et al. 2014).

Up until this point in the metabolomics workflow, chemical features in the dataset

are typically only identifiable by their calculated accurate mass and retention time (for

LC-MS based metabolomics). Once chemical features are found that statistically differ between sample treatments, focus can be placed on identifying these features. This step represents one of the most significant bottlenecks in the metabolomics workflow

(Hendriks et al. 2011). The general process includes: determination of the elemental composition of a given analyte based on its accurate mass, searching accurate mass and molecular formulas against databases of known compounds, fragmentation of the analyte to gain substructural information, and eventually comparison to authentic standards to verify metabolite identity (Warwick B Dunn et al. 2013). Each of these steps has inherent challenges, which will be summarized below. In the end, the level of confidence in a

54

compound identification can be expressed at four different levels, as proposed by the

Metabolomics Standards Initiative. Level 1 identified compounds are those that match a chemical feature with an authentic standard using identical methodology. Level 2 represents putatively identified compounds that were not matched using authentic standards but exhibit similarities to previously published data. Level 3 identified compounds are identified by potential compound class based on previously published characteristics, and level 4 compounds are those that remain unidentified (Sumner et al.

2007). Although confidence in metabolite annotation can vary, features identified at all levels may be useful in describing and utilizing results of metabolomics experiments.

Determining the elemental compositions of chemical features is facilitated by the high resolution mass spectra collected in metabolomics experiments. The monoisotopic mass, isotope distribution, and fundamental rules of chemistry can be used to produce candidate molecular formulas for a given chemical feature, though there are typically several formulas that can be proposed from a singular monoisotopic mass (Rathahao-

Paris et al. 2016; Warwick B Dunn et al. 2013). The monoisotopic mass can also be used

to search compound databases such as the human metabolome database (HMDB)

(Wishart et al. 2013) and Metlin metabolite database (Guijas et al. 2018). These, often

freely available and web-based, tools will return potential matches based off of the

queried monoisotopic mass with a specified window of accuracy. In practice these

databases can provide several hits on a singular search, which again does not allow for

absolute identification but provides additional and valuable information. It should also

be noted that, while these databases are continuously updated and developed, they are not

55

all encompassing, and thus a query may return no search results. In either case, further

experimentation in which tandem mass spectrometry (MS/MS) is used to fragment

chemical features of interest is often used to gain additional information on chemical

substuctures. Fragmentation of molecules can lead to characteristic losses of phase II

metabolite conjugations (e.g. glucuronide, sulfate) (Levsen et al. 2005) and other

functional groups (Levsen et al. 2007) that can aid in structural characterization of

chemical features. In the case of phase II metabolite conjugates, re-searching the

aforementioned databases with the accurate mass of the aglycone may aid in obtaining

additional database hits. Mass fragmentation spectra can also be processed using a

number of computer programs to elucidate potential molecular structures and narrow in

on likely molecular formulae (Rathahao-Paris et al. 2016). As mentioned earlier,

comparison with an authenticated standard, if available, under identical analytical

conditions provides unequivocal identification of chemical features. Rathahao-Paris and colleagues provide a comprehensive framework for the compound identification process

(Rathahao-Paris et al. 2016), but in practice this is an iterative process that, at this time, lacks automation and is incredibly time consuming. However, once features are identified, the results of a metabolomics experiment can be interpreted in a greater context that is relevant to the experimental question at hand.

1.5.2 Metabolomics Applications in Berry Research

Metabolomics is a flexible research approach that is adaptable to many different research questions. In recent years, it has been used in a variety of applications in food

56 and nutrition sciences. Applications in food science range from authentication of food products to understanding the dynamics of food fermentation and processing (Rubert et al. 2015). In terms of nutrition, metabolomics has been used to understand metabolic perturbations associated with the intake of certain foods and for the discovery of biomarkers associated with dietary patterns or specific foods, which has the potential to improve upon dietary assessment methods (Brennan 2013). Importantly, this includes the use of metabolomics to understand the interactions between host and microbial metabolism of (poly)phenolic food components (Moco et al. 2012). Foods are complex mixtures of hundreds to thousands of phytochemicals, so as an analytical technique, metabolomics is uniquely situated to provide a more comprehensive picture of how food may impact human health from production through consumption.

With regard to the relationship between berries and health, metabolomics has recently been used in several applications. Paudel and colleagues used NMR-based metabolomics to understand how phytochemical constituents of BRBs vary according to cultivar and ripeness stage. They then applied these same extracts to HT-29 colon cancer cells and related the untargeted metabolomics dataset to the bioassay results using PLS.

Their findings illustrated that a number of phytochemicals in BRBs, including anthocyanins, organic acids, and flavonoids all contribute to the bioactivity of the fruit

(Paudel et al. 2014). Another study compared the metabolomic profiles of red and yellow raspberries, leading to the discovery of A-type procyanidins in yellow raspberries

(Carvalho et al. 2016). With regard to nutrition, a NMR-based metabolomics approach uncovered changes in serum lipid profiles of overweight women following consumption

57

of sea buckthorn berries or bilberries (Larmo et al. 2013). Another study used

metabolomics to discover biomarkers of berry intake. Cuparencu and colleagues fed

subjects a single test meal containing strawberries, sea buckthorns, or a control meal and

collected urine samples at set intervals over 24 hours, which were subsequently analyzed

using an LC-MS. PLS-DA was performed on data from each time point, and subsequent

identification of metabolites led to the discovery of biomarkers unique to each test meal

and one, catechin sulfate, that was common between them (Cuparencu et al. 2016). These

data demonstrate that metabolomics has a place in berry research and has the potential to

advance the field with novel findings.

1.6 Specific Aims

The overall objective of this work is to build upon our understanding of berry

phytochemicals in the prevention of chronic diseases, such as oral cancer by investigating

how these phytochemicals exist in foods and are metabolized by the body. We

hypothesize that untargeted metabolomics will be a suitable technique to profile changes

in berry foods over processing and storage, as well as aid in distinguishing those who

have been subjected to berry exposure.

This will be achieved by investigating: 1) thermal processing-induced changes in the phytochemical profile of a BRB nectar beverage, 2) the impact of storage on the chemical profile and bioactivity of this same beverage against premalignant oral cancer cells, and 3) the urinary metabolomic profiles of strawberry consumption (as compared to a placebo) in smokers and non-smokers.

58

BRBs have been heavily researched with regard to chemoprevention, as they contain high levels of specific phytochemicals of interest, namely anthocyanins and ellagitannins. The majority of research on BRBs has utilized minimally processed, freeze-dried berry powder, yet as available to consumers, BRBs are most typically found in thermally processed products including jams, jellies, and syrups among others.

Understanding key chemical differences mediated by food manufacturing techniques will enhance interpretation of clinical results from studies with processed BRB-containing functional foods. Additionally, knowledge of phytochemical changes over storage of

BRB nectar, and associated changes in bioactivity, will inform future research on strategies to modulate phytochemical changes that may impact the efficacy of berry products.

In stark contrast to other berry crops, strawberries are grown in great quantity throughout the world. According to the most recent data from the Food and Agriculture

Organization of the United Nations, worldwide strawberry production exceeded 9 million metric tons in 2016, more than other documented berries combined (Food and

Agriculture Organization of the United Nations n.d.) Understanding the urinary metabolic signature associated with strawberry consumption will provide insight on the bioavailability and metabolism of the potentially bioactive phytochemicals in this most important berry crop.

Aim 1: Use untargeted metabolomics to elucidate alterations in phytochemicals arising from the thermal processing of BRBs. 59

We will use UHPLC-QTOF-MS based metabolomics to obtain untargeted

phytochemical profiles of freeze-dried BRB powder and a thermally processed nectar

made from the same powder. We hypothesize that the process used to produce this

beverage will demonstrate stability of some phytochemicals, induce small changes in

some phytochemicals, such as ellagic acid, while other phytochemicals, such as

anthocyanins will degrade to form new phenolic products.

Aim 2: Profile phytochemical changes in BRB nectar over storage and assess how these

phytochemical changes alter the anti-proliferative activity of the product.

We will produce UHT pasteurized, aseptically filled BRB nectar, store it at

various temperatures, and sample it over 60 days of storage. UHPLC-QTOF-MS will be used to profile chemical changes across time and temperature conditions. These findings will be related to the anti-proliferative activity of the nectar against SCC-83 precancerous oral epithelial cells. We hypothesize that prolonged storage of the BRB nectar at elevated temperatures will result in large chemical changes and a decreased ability of the nectar to inhibit growth of SCC-83 cells.

Aim 3: Investigate the urinary metabolomic signatures of strawberry-consuming

individuals, including smokers, who are at high risk for developing oral cancer and other

diseases which berry consumption may benefit.

60

We will use samples acquired from a previously conducted clinical trial in which

smokers and non-smokers consumed strawberry-containing confections or a placebo confection for one week. The untargeted metabolomic profile of urine samples will be obtained using UHPLC-QTOF-MS, and multivariate analyses will be used to differentiate those consuming strawberries, including any differences that are due to smoking status. We hypothesize that this technique will allow us to understand unique contributions of strawberry consumption to urinary metabolite profiles, as well as uncover global differences in phytochemical metabolism between smokers and non- smokers.

61

Chapter 2. Profiling the impact of thermal processing on black raspberry phytochemicals using untargeted metabolomics

Matthew D. Teegarden1, Steven J. Schwartz1, Jessica L. Cooperstone1,2

1Department of Food Science and Technology, The Ohio State University, Columbus, OH, USA 2Department of Horticulture and Crop Sciences, The Ohio State University, Wooster, OH, USA

62

2.1 Abstract

Clinical and laboratory studies have implicated black raspberries (BRBs) and their associated phytochemicals in the modulation of several chronic diseases. Most research on the health benefits of BRBs is conducted using freeze-dried or otherwise minimally processed products, yet BRBs are typically consumed as thermally processed goods like jams and syrups. The objective of this work was to profile the chemical changes that result from thermal processing of BRB powder into a nectar beverage. Using an untargeted UHPLC-QTOF-MS metabolomics approach, key degradation products of anthocyanins were identified along with several other proposed phenolic degradants. The effects of processing on other key BRB compound groups, including ellagitannins, are also discussed. This work demonstrates the utility of an untargeted metabolomics approach in describing the chemistry of complex food systems and provides a foundation for future research on the impact of processing on BRB product bioactivity.

2.2 Introduction

Black raspberries (BRBs) are heavily researched for their anti-cancer properties.

In vitro models suggest BRBs may be active against a variety of cancer types (N. Seeram et al. 2006), while animal studies and clinical trials provide compelling evidence for the potential role of BRBs, and their phytochemical components, in preventing aerodigestive cancers (Bishayee et al. 2015). For example, a two week treatment with BRB-based troches reduced malignant tumor levels of hallmark biomarkers of cancer in oral cancer

63 patients (Knobloch et al. 2016). Studies such as this support further research on BRBs as part of a food-based approach for cancer prevention.

BRBs contain a wide array of phytochemicals including anthocyanins, flavonols, ellagitannins, and hydroxycinnamic acids, among others (Paudel et al. 2013). Work to understand which of these components contribute to the bioactivity of BRBs has shown that singular phytochemicals cannot explain the complete bioactivity of the fruit (Paudel et al. 2014; Wang et al. 2009). Thus, the complete phytochemical profile of the fruit is critical when conducting research on the potential health effects of BRBs.

In assessing BRB bioactivity, laboratory and clinical studies have historically used lyophilized BRB powder in treatments, though consumers do not typically encounter these berries in their freeze-dried or even fresh forms. Instead, BRBs are more commonly found incorporated into thermally processed products, such as jams and syrups. Knowledge of how thermal processing affects the phytochemical profile of BRB products is limited to a few select compounds (Gu et al. 2014; Hager et al. 2008), despite the biological importance of the whole phytochemical profile. Untargeted metabolomics is an analytical technique that aims to provide a comprehensive chemical profile of as many small molecules in a system as possible, which allows for a more thorough and encompassing analysis of molecular composition than traditional methods (Wishart

2008).

The objective of this study is to understand how thermal processing impacts the phytochemical profile of BRBs using an untargeted metabolomics approach. The product used in this work is a BRB nectar beverage, which has been previously described (Gu et

64

al. 2014). The nectar is an optimal product to study, as it contains the whole BRB fruit

and components typically incorporated in other BRB products, such as sugar and pectin,

and thus applicable to additional BRB products.

2.3 Materials and Methods

2.3.1 Chemicals

All chemicals used were obtained from Fisher Scientific (Pittsburgh, PA).

2.3.2 Nectar Production

Nectar was produced in 250 g batches (n = 3) according to the formulation described by Gu and colleagues, with slight modifications as shown in Table 2.1 (Gu et al. 2014). The BRB ( cv. Jewel) powder used in this work was acquired from Stoke’s Berry Farm (Wilmington, OH) as freeze-dried product and was produced from a single lot of berries. To manufacture the nectar, all components, except the BRB powder, were combined and heated to approximately 30 °C with constant stirring. Once the pectin was dissolved, the BRB powder was added and the product was heated to 95 °C with constant stirring until a final soluble solids content of 9 °Brix was achieved; these parameters were chosen to model the pasteurization procedure previously described for this product (Gu et al. 2014). Nectar was immediately flash frozen with

liquid nitrogen, lyophilized, and stored at -20 °C until analysis. A process blank, which

consisted of all nectar components without the BRB powder (i.e. water, sugar, corn

syrup, pectin), was also produced in the same manner as the BRB nectars.

65

Table 2.1 BRB nectar formulation

Ingredient Percent (wet basis) Water 89.9 Sucrose 3.0 Pectin 0.5 Corn syrup 1.0 BRB powder 5.6

66

2.3.3 Determination of Total Monomeric Anthocyanins

BRB powder and lyophilized nectar were extracted according to Hager and

colleagues with slight modification (Hager et al. 2008). Briefly, 100 mg was combined

with 2 mL of 60:37:3 methanol:water:formic acid and vortexed for 15 sec. Following

centrifugation at 4000 × g for 10 min, the supernatants were decanted and pellets

extracted twice more. Pooled extracts were diluted to 10 mL with 0.01% aqueous HCl,

and total monomeric anthocyanin content was determined as previously described (Giusti

& Wrolstad 2001). Results are expressed as mg cyanidin-3-glucoside equivalents/g powder, and levels in the lyophilized nectar were multiplied by a factor of 1.8 to account for additional dry ingredients in the nectar formulation.

2.3.4 Sample Preparation for Untargeted Metabolomics

Lyophilized nectar and BRB powder were extracted using identical protocols except 180 mg of nectar was extracted, compared to 100 mg of BRB powder, to account for other ingredients used in the nectar formulation. Briefly, 1 mL of 75% methanol in water, with 0.1% formic acid, was added to each sample. Samples were sonicated in an ice bath for 15 min, centrifuged at 21,130 x g for 2 min, and decanted into glass vials.

The resulting pellets were extracted twice more with the use of a probe sonicator (8 sec,

Branson Ultrasonics; Danbury, CT). Each nectar batch was extracted in triplicate (total n=9), and an equal number of BRB powder samples were extracted (n=9). The nectar process blank (no BRB added) was also extracted using this protocol. Extracts were centrifuged at 21,130 x g for 4 min and the supernatant immediately analyzed. A set of

67

quality control (QC) samples was produced by pooling equal volume aliquots from all

nectar and BRB powder extracts.

2.3.5 UHPLC-QTOF-MS Metabolomics Data Acquisition

BRB and nectar samples were randomized for run order, and QC samples were

positioned after every sixth injection. The use of QC samples allows for monitoring of

instrument stability over the sample set. Untargeted full-scan data was acquired using a

1290 Infinity II series UHPLC (Agilent, Santa Clara, CA) coupled to an Agilent iFunnel

6550 QTOF-MS. Samples were injected (3 μL) onto a 100 x 2.1 mm Agilent SB-Aq column (1.8 μm particle size) maintained at 50 °C. The mobile phase consisted of (A)

0.1% formic acid in water and (B) 0.1% formic acid in methanol at a flow rate of 0.6 mL/min. The mobile phase composition was as follows: 0-2 min, 0% B; 2-3 min increase

to 10% B; 3-8 min, increase to 40% B; 8-14 min increase to 100% B; 14-16 min, hold at

100% B; 16-18 min, immediate switch to 0% B for a total run time of 18 min. The

UHPLC was interfaced with the QTOF-MS with an ESI source operated in negative ion

mode. Relevant MS parameters were as follows: gas temp 150 °C, drying gas 18 L/min,

nebulizer 30 psig, sheath gas temp 350 °C, sheath gas flow 12 L/min, VCap 4000 V,

nozzle voltage 2000 V, acquisition mode was 2 GHz extended dynamic range with a

mass range of 50-1700 m/z.

68

2.3.6 Data Pre-processing and Analysis

Full scan UHPLC-QTOF-MS data was processed using the batch recursive

feature extraction algorithm in Agilent Profinder (B.06.00). This process bins mass

spectral features according to expected isotope patterns, adducts, and charge states, and

then aligns them across all samples. Feature groups that appeared in at least two samples

in either the nectar or BRB powder were retained and re-extracted across all samples. The recursive nature of this workflow ensures high quality data for statistical analysis. Further filtering of the data was performed in Agilent Mass Profiler Professional. First, features unique to the processing blank, which consisted of all nectar components except BRB powder, were removed from analysis. The analysis was then restricted to features with retention times between 1-12.5 min and a calculated neutral mass <1200 amu. Finally, features that were present in at least 66.6% of nectar or BRB samples, and those with a

CV<25% in either of these groups were retained for statistical analyses. All data was log2

transformed and median-centered prior to analysis. Differential analysis was performed

using an unpaired t-test (P < 0.05) with the Benjamini-Hochberg false discovery rate

multiple testing correction applied.

2.3.7 Compound Identification

Features that differed significantly between the BRB powder and nectar were

considered for identification if their average ion abundance was > 1.0 × 105 in either

sample group. Highly abundant features (abundance > 1 × 106) that differed < 2 times between BRB powder and nectar were also considered for identification. These

69

thresholds were used to ensure high data quality and sufficient signal for further

experimentation. Identification was achieved using a combination of MS/MS

fragmentation spectra, accurate mass, isotope analysis, and database matches when available. For MS/MS fragmentation studies, extracts were injected on the previously mentioned UHPLC-QTOF-MS system operated in targeted or auto MS/MS mode using collision energies of 10, 20, and 40 eV. Fragmentation patterns were compared to literature or curated spectra in the FooDB (Wishart et al. 2013) and Metlin

(metlin.scripps.edu) databases when available. CFM-ID (Allen et al. 2014) was used to

rationalize proposed molecular structures as needed.

2.4 Results and Discussion

In the present work, we compared the metabolomic profiles of lyophilized BRB

powder and a thermally processed nectar beverage made from that same powder. This

allows us a more holistic view on both phytochemicals that change, and those that do not

change, with thermal treatment. A metabolomic approach has been previously used to

understand chemical changes during BRB ripening (Kim et al. 2011), as well as the

potential bioactivity of BRB phytochemicals (Paudel et al. 2014; Jo et al. 2015). In an

effort to provide context for the practical translation of clinical research that supports

consumption of BRB phytochemicals, we used this approach to more comprehensively

characterize the effects of producing a beverage that incorporates the entire BRB fruit.

70

2.4.1 Monomeric anthocyanin content

Compared to other berries, BRBs are particularly concentrated in anthocyanins as marked by their dark pigmentation (Moyer et al. 2002). Because the UHPLC-QTOF-MS method used in this study was not optimized for anthocyanin analysis, a spectrophotometric approach was used to quantify these compounds. The total content of monomeric anthocyanins in nectar was approximately 80% of that in the BRB powder

(Figure 2.1), indicating anthocyanin loss due to heat treatment. Similar decreases in anthocyanin content have been described in previous work on BRB nectars (Gu et al.

2014). This is a higher level of anthocyanin retention than observed in other processed

BRB food products (Hager et al. 2008), however the fewer number of processing steps used here may account for the greater retention in this nectar product.

71

Figure 2.1 Total monomeric anthocyanins (expressed as mg cyanidin-3-glucoside equivalents) in BRB powder and lyophilized nectar. A numeric correction was applied to the nectar values to account for other solids present. (*)Decrease in anthocyanin content was statistically significant (P<0.01).

72

2.4.2 Untargeted metabolomics

Following pre-processing of the data, a total of 4411 molecular features were

evaluated in subsequent analyses. As part of the pre-processing procedure, features present in a “processing blank,” which consisted of all nectar ingredients minus the BRB powder, were removed from analysis. Thus, this final list of 4411 features does not include those native to non-BRB nectar ingredients, but it does contain features that correspond to BRB phytochemicals and any potential products of interactions between

BRB phytochemicals and other nectar components.

In order to focus on features that differ significantly between the BRB powder and nectar, an unpaired t-test and fold change analysis was performed on each molecular feature. Following manual review for peak quality, a total of 547 features were found to differ by at least two-fold (P<0.05) between the BRB powder and nectar. Unsupervised hierarchical clustering analysis was performed on these features according to their normalized abundances (Figure 2.2). Large differences in abundance are evident between the two sample groups. A majority of features were present at higher levels in the nectar than the BRB powder. Interestingly, 170 features were unique to the nectar while 40 were unique to the BRB powder. We hypothesize that the features unique to the BRB powder were present in the fruit but completely degraded over thermal processing, while those unique to the nectar represent degradation or reaction products.

73

74

Figure 2.2 Hierarchical clustering analysis on features that differ ≥2 fold between lyophilized BRB powder and nectar made from that powder (P < 0.05). Features were clustered using Euclidian distance metrics and Ward’s linkage rule.

74

2.4.3 Identification of Features Elevated in Thermally Processed BRBs

Given the large number of features elevated in the nectar, special attention was paid to annotating those that were most abundant with a fold change >3. A total of nine potential degradation or reaction products were annotated on the basis of their accurate mass, relative retention time and MS/MS fragmentation spectra (Table 2.2). Of these, three were assigned tentative identifications as anthocyanin-derived degradation products including cyanidin chalcone rutinoside, protocatechuic acid, and phloroglucinaldehyde.

Cyanidin chalcone rutinoside is a thermal degradation product of cyanidin-3-O- rutinoside, the major anthocyanin present in BRBs. This product results from the oxidative opening of the flavilium C-ring (Markakis & Jurd 1974). Similar chalcone glycosides have been observed in thermally processed strawberry and elderberry concentrates (Sadilova et al. 2007). Protocatechuic acid and phloroglucinaldehyde are well-characterized thermal degradation products of cyanidin that result from the liberation of the catecholic B-ring and trihydroxyphenolic A-ring, respectively (Sadilova et al. 2006; Sadilova et al. 2007). Protocatechuic acid can also originate from flavonoids such as quercetin (Buchner et al. 2006), which is also present in BRBs. Although isobaric, protocatechuic acid and phloroglucinaldehyde can be differentiated by their

MS/MS fragmentation patterns, Figure 2.3. Comparison of retention times of authentic standards confirmed the identities of protocatechuic acid and phloroglucinaldehyde. The rise in all three of these compounds is well explained by the simultaneous decrease in total monomeric anthocyanin content, and demonstrates the potential of metabolomics in monitoring phytochemical degradation in complex food mixtures.

75

Table 2.2 Identified features elevated in BRB nectar

RT Observed Major Chemical Fold Tentative ID Δ ppm (min) [M-H]- fragments formula changea 2.25 2-oxo-2-(2,4,6- 181.0144 153.0186 C8H6O5 -2.7 Nectar trihydroxyphenyl) 125.0237 Only acetaldehyde b 2.32 Protocatechuic acid 153.0192 109.0296 C7H6O4 -0.65 4.0 b 4.86 Phloroglucinaldehyde 153.0192 151.0034 C7H6O4 -0.65 6.0 125.0238 107.0137 4.88 Unknown phenolic 275.0552 165.0186 C14H12O6 -3.3 11.7 109.0294 5.39 Cyanidin chalcone 611.1600 303.0497 C27H32O16 -2.9 9.2 rutinoside 6.33 Unknown phenolic 319.0451 183.0287 C15H11O8 -2.2 8.4 153.0184 139.0390 109.0313 8.01 Unknown quercetin 565.1547 499.0492 C26H30O14 -1.8 7.4 derivative 323.0176 301.0340 257.0451 151.0401 8.45 Unknown phenolic 301.0340 165.0190 C15H10O7 -2.7 5.4 137.0240 165.0187c 137.0240 109.0290 8.85 Unknown phenolic 331.0441 221.0086 C16H12O8 -3.9 6.2 195.0294 193.0136 151.0181 109.0290 9.53 Unknown phenolic 313.0341 285.0403 C16H10O7 -2.2 10.7 109.0300 a Mean abundance in nectar vs BRB powder. b Identity confirmed with authentic standard c In-source fragment.

76

Figure 2.3 MS/MS spectra and fragment rationalization of protocatechuic acid (A) and phloroglucinaldehyde (B).

77

We were able to propose a structure for one additional feature, 2-oxo-2-(2,4,6- trihydroxyphenyl) acetaldehyde (Figure 2.4). The spectra of this compound contained successive neutral losses of 27.995 amu, which is associated with the loss of CO and characteristic of aldehyde functional groups (Levsen et al. 2007). Similar to phloroglucinaldehyde (Figure 2.3B), the product ion at m/z 125.0237 suggests this is a degradation product that originates from the A ring of a flavonoid parent structure. A carboxylic acid product with a nearly identical structure was found in a study on the thermal degradation of quercetin with an identical fragmentation pattern observed

(Barnes et al. 2013), further supporting the structure we have proposed here. Based on the proposed structure, this compound may be a degradation product of a flavonol-type molecule.

Additional potential degradation products were annotated based on characteristic fragments observed in MS/MS spectra. One was described as an unknown quercetin derivative because of the presence of the m/z 301.034 fragment and abundance of quercetin in BRBs. An additional five features were annotated as unknown phenolic compounds due to the presence of a 109.029 fragment, which corresponds to a [catechol-

H]- ion. This same ion has been noted in the fragmentation of quercetin degradation

products where a carboxyl group is directly attached to the catecholic B ring of the parent

structure (Barnes et al. 2013), like protocatechuic acid (Figure 2.3A). Thus, we hypothesize that these features are potential phenolic degradation products, or potential reaction products that result from an interaction between phenolic degradants.

78

The identification of chemical features in metabolomics experiments remains challenging despite advances in the tools available to aid in this task (Warwick B. Dunn et al. 2013). Spectral libraries are integral parts of the compound identification process, and efforts are underway to develop databases specific to foods, such as FooDB, a subset of the Human Metabolite Database (Wishart et al. 2013). While these databases can contain information thousands of phytochemicals, they do not tend to include information on their chemical degradation products, or interactions thereof. The curation of these types of compounds presents inherent difficulties, but represents a potential area for future growth. This is important because in some cases, bioactive phytochemicals may degrade to other bioactive molecules. For example, cyanidin-based anthocyanins degrade to protocatechuic acid, as described here, which has demonstrated chemopreventative activity against several different types of cancers in animal and cell studies (Tanaka et al.

2011). In fact, it has been proposed that degradation products of anthocyanins likely mediate their associated bioactivity (Kay et al. 2009). This may be partly explained by the ortho-dihydroxyphenyl chemical moiety shared between cyanidin and protocatechuic acid, which has been found to be critical for the biological properties of anthocyanins

(Hou et al. 2003; Hou et al. 2005). Thus, a greater understanding of compounds present in foods in the forms in which they are eaten (i.e. processed, stored) is relevant when considering health outcomes.

79

Figure 2.4 Proposed fragmentation pattern of 2-oxo-2-(2,4,6-trihydroxyphenyl) acetaldehyde.

80

2.4.4 Identification of Features Stable to Thermal Processing

Because of the interest in the potential bioactive properties of BRB phytochemicals, we chose to complete a survey of abundant features that were minimally influenced by processing BRB powder into nectar. Significant losses of key phenolic compounds in processed berry products can occur depending on the techniques that are employed (Howard et al. 2012). An auto-MS/MS approach, in which the most abundant ions at any given time are fragmented, was used to acquire additional spectral information for highly abundant compounds. Tentative identifications of features were assigned based off of characteristic fragmentation data reported in literature (Gu et al.

2014; Paudel et al. 2013; Gu et al. 2003; Määttä-Riihinen et al. 2004), and authentic standards were used to confirm some identities, as noted in Table 2.3. Many of the key non-anthocyanin compounds associated with BRBs were identified via this untargeted approach.

Quercetin is a flavonol ubiquitously found in edible berries (Häkkinen et al.

1999). It is typically found conjugated to various glycosyl groups, which have been previously described in BRBs (Gu et al. 2014; Paudel et al. 2013). The potential biological significance of quercetin in several diseases has been previously reviewed

(Erlund 2004). Decreases in levels of quercetin glycosides were observed with a modest, concurrent increase in the aglycone quercetin. The thermal stability of quercetin and its rutinoside are enhanced in low pH systems (Buchner et al. 2006), like that of the nectar beverage (pH = 3.7). Changes in the abundance of quercetin and its glycosides may be due to hydrolysis of the sugar moieties and/or oxidative processes (White et al. 2011;

81

Buchner et al. 2006). Similar to quercetin-based compounds, the glucoside of myricetin was minimally degraded by processing.

Gallic acid and salicylic acid (and its glucoside) are hydroxybenzoic acids, which have been previously reported in BRBs (Kula et al. 2016; Paudel et al. 2013). Coumaric acid, caffeic acid, and their glycosides are hydroxycinnamic acids that have also been previously reported in BRBs (Paudel et al. 2013). Phenolic acids have been extensively studied for their potential biological activities including antimicrobial and

antitumorigenic effects (Heleno et al. 2014).

B-type procyanidins are oligomers of catechin and/or epicatechin, and have been

previously documented in BRBs (Kula et al. 2016). The potential bioactive properties of

these compounds are thought to be primarily mediated through products of microbial

catabolism (Monagas et al. 2010). Similar to our findings, Brownmiller and colleagues

noted the stability of monomeric and dimeric procyanidins, relative to higher oligomers,

in a variety of processed blueberry products (Brownmiller et al. 2009). Increased loss of

procyanidins has also been associated with increased number and complexity of steps

involved in processing (Howard et al. 2012). Thus, like the relatively high retention of

total anthocyanins, the simplistic processing design likely contributed to retention of

procyanidins in the final product.

82

Table 2.3 Identified BRB phytochemicals with P > 0.05 or fold change < 2

RT Observed Qualifying Chemical Fold Tentative ID Δ ppm a (min) [M-H]- fragment formula change Flavonols 7.36 Quercetin 741.1864 301.0322 C32H38O20 -2.7 - xylorutinoside 7.97 Quercetin hexuronide 477.0668 301.0354 C21H18O13 -1.3 -1.09 b 8.10 Quercetin rutinoside 609.1447 301.0345 C27H30O16 -2.3 -1.08 b 8.22 Quercetin glucoside 463.0867 301.0351 C21H20O12 -3.2 -1.13 8.63 Myricetin glucoside 479.0813 317.0299 C21H20O13 -3.8 -1.12 b 9.60 Quercetin 301.0345 151.0032 C15H10O7 -3.0 1.15

Procyanidins 4.60 B-type procyanidin 577.134 289.0714 C30H26O12 -2.1 - b 4.63 Catechin 289.0714 245.0815 C15H14O6 -1.4 - 5.03 B-type procyanidin 577.134 289.0714 C30H26O12 -2.1 -1.23 b 5.28 Epicatechin 289.0715 245.0817 C15H14O6 -1.0 1.17

Phenolic acids b 1.40 Gallic acid 169.0146 125.0243 C7H6O5 2.4 1.92 3.06 Salicylic acid hexoside 299.0773 137.0246 C13H16O8 0.33 - 4.23 Caffeic acid hexoside 341.087 179.0352 C15H18O9 -2.3 -1.14 5.00 Coumaric acid hexoside 325.0923 163.0399 C15H18O8 -1.8 -1.24 b 5.11 Caffeic acid 179.0361 135.0452 C9H8O4 6.1 1.71 b 5.28 Salicylic acid 137.0247 C7H6O3 2.2 1.69 a Mean abundance in nectar vs BRB powder, fold change shown for compounds with P<0.05. b Identity confirmed with authentic standard

83

2.4.5 Ellagitannins

Ellagitannins are polymers that consist primarily of hexahydroxydiphenic acid units that, when hydrolyzed, lactonize to form ellagic acid (Quideau & Feldman 1996).

Ellagic acid and its related derivatives were tentiatively identified in BRB powder and nectar based on previous reports (Gu et al. 2014; Paudel et al. 2013). Similar to other compounds identified here, small decreases in ellagic acid and methylellagic acid glycosides were observed, concurrent with an increase in their aglycone forms likely due to hydrolysis over thermal processing.

Collision-induced dissociation of ellagitannins produces characteristic product ions corresponding to ellagic acid (Mullen et al. 2003). Several features produced this fragment (m/z 300.99) throughout the course of data analysis. Features consistent with ellagitannins were found to increase, decrease and remain constant with processing, as summarized in Table 2.4. The ellagitannins that decreased with processing may have depolymerized to smaller ellagitannin products, and thus contributed to the slight elevation of ellagic acid in the nectar. As ellagitannins are primarily concentrated in the seeds of BRBs (Howard et al. 2012), the increased levels of some ellagitannins in the nectar could have resulted from enhanced extraction during processing from thermal and shear induced destruction of cell walls. This increase could also be the result of hydrolysis of even larger ellagitannins. Structures as large as 3740 amu have been reported in Rubus fruits (Vrhovsek et al. 2006). In the current study, a mass cutoff of

1200 amu was applied, thus larger ellagitannins previously observed in BRBs are not included in this analysis. Interestingly, at least one ellagitannin, sanguiin-H4, was stable

84

over processing. Hager and colleagues observed minimal changes in total ellagitannin

content when blackberries were processed into various products containing the whole

fruit (Hager et al. 2010). Together with the modest increase in ellagic acid, our data

suggests that the ellagitannin profile of BRBs may shift over processing into nectar

without a large net decrease in total content.

While in vitro studies suggest ellagic acid and ellagitannins may be bioactive in

various assays, they are poorly bioavailable. Urolithins are bioavailable microbial

catabolites of ellagic acid and ellagitannins and are hypothesized to be at least partially

responsible for their associated bioactivity in vivo (Landete 2011). Truchado and

colleagues recently compared urinary exrection of urolithins after subjects were fed fresh

and processed strawberries. Despite differences in the ellagitannin/ellagic acid profiles of

the treatments, no change in urolithin excretion was observed (Truchado et al. 2012). We

speculate similar results would be observed after feeding processed BRB nectar,

indicating no impact on ellagitannin-associated bioactivity after thermal processing of

BRBs.

85

Table 2.4 Identified ellagic acids (EA) and ellagitannins (ET)

Ret. time Observed Qualifying Chemical Δ Fold Tentative ID a (min) [M-H]- fragment formula ppm change Ellagic acid derivatives 8.57 EA pentoside 433.0388 300.9983 C19H14O12 -5.8 -1.07 b 8.90 Ellagic acid 300.9978 229.0134 C14H6O8 -3.9 1.43 9.87 Methyl EA 315.0149 300.9940 C15H8O8 0.95 1.39 10.12 Acylated methyl 489.0667 315.0136 C22H18O13 -1.6 -1.1 EA pentoside 10.97 Malonyl methyl 531.0773 315.0136 C24H20O14 -1.3 -1.13 EA pentoside

Ellagitannins 3.92 Unknown ET 484.0263 (- 300.9995 C44H26O26 4.1 -17.1 2H) 3.92 Unknown ET 473.0354 (- 300.9991 C42H28O26 1.8 -8.3 2H) 4.45 Unknown ET 480.0432 (- 300.9970 C43H30O26 - -4.9 2H) 0.74 4.68 Sanguiin H4 633.0738 300.9985 C27H22O18 0.79 - 5.54 Unknown ET 587.0564 (- 300.9989 C55H36O30 -1.5 9.5 2H) 6.63 Unknown ET 587.0568 (- 300.9994 C55H36O30 - 5.2 2H) 0.85 7.34 Unknown ET 587.0575 (- 300.9984 C55H36O30 0.34 5.1 2H) a Mean abundance in nectar vs BRB powder, fold change shown for compounds with P<0.05. b Identity confirmed with authentic standard.

86

2.5 Conclusions

Here we have described the use of an untargeted metabolomics approach to

extract >4000 chemical features in BRB powder and nectar. Using this technique, we

profiled the global phytochemical changes that occur when producing a thermally-

processed nectar product from lyophilized BRB powder by broadly describing the differences and similarities between the two products. For instance, key BRB components such as quercetin, phenolic acids, and ellagic acid were relatively stable to processing, while a decrease in total anthocyanin content was observed concurrent with

large increases in degradation products. Since phytochemical profile is thought to be a

driver of bioactivity, the implications of thermal processing of BRB food products should

be considered in future studies using BRB-based functional foods. This work demonstrates the utility of an untargeted metabolomics approach in analyzing and understanding complex food systems.

87

Chapter 3. Storage conditions modulate the metabolomic profile of a black raspberry nectar with minimal impact on bioactivity

Matthew D. Teegarden1, Thomas J. Knobloch2, Christopher M. Weghorst2, Jessica L.

Cooperstone1, Devin G. Peterson1

1Department of Food Science and Technology, The Ohio State University, Columbus,

OH, USA

2College of Public Health, Division of Environmental Health Sciences, The Ohio State

University, Columbus, OH, US

88

3.1 Abstract

Pre-clincial and clinical studies suggest black raspberries (BRBs) may inhibit the

development of oral cancer. Lyophilized BRB powder is commonly used in these studies,

but processed BRB products are more often consumed. The objective of this work was to

understand how storage conditions influence the phytochemical profile and anti-

proliferative activity of a BRB nectar beverage. Untargeted UHPLC-Q-TOF-MS based metabolomics analyses demonstrated that large chemical variation was introduced by storage above -20 °C over 60 days. However, minimal change in anti-proliferative activity was observed when stored nectar extracts were applied to SCC-83-01-82 premalignant oral epithelial cells. As proof of concept, cyanidin-3-O-rutinoside and its degradation product, protocatechuic acid, were administered in different ratios maintaining an equimolar dose, and anti-proliferative activity was maintained. This study shows the utility of metabolomics to profile global chemical changes in foods, while demonstrating that isolated phytochemicals do not explain the complete bioactivity of a complex food product.

89

3.2 Introduction

Black raspberries (BRBs) are extensively studied for their cancer preventative

properties (Kula & Krauze-Baranowska 2016; Kresty et al. 2016). Their bioactivity has

been attributed to their rich phytochemical profile inclusive of anthocyanins,

ellagitannins, organic acids, and quercetin among other phenolic compounds (Stoner

2009). It has been hypothesized that these components elicit a complex series of biological responses that result in a net inhibition of cancer growth (Liu 2003). Because

of their high concentration of phenolic compounds, BRBs have been the subject of many

studies on food-based chemoprevention strategies.

Much of the interest in the chemopreventative properties of BRBs has focused on

oral cancer (Warner et al. 2014; Knobloch et al. 2016; Oghumu et al. 2017; Mallery et al.

2014; El-Bayoumy et al. 2017). The oral cavity presents unique opportunities for

chemoprevention through dietary means due to direct exposure of tissues to food

phytochemicals. Oral cancer is prevalent throughout the world, with higher incidence

observed in men and people in less developed regions (Petersen et al. 2005; Ferlay et al.

2015). The majority of oral cancers are squamous cell carcinomas (SCCs), and risk

factors include tobacco use, alcohol consumption, human papillomavirus infection, and

chronic periodontal disease (Bsoul et al. 2005). Pre-clinical models have consistently

shown reduction of oral SCC incidence and multiplicity using whole freeze-dried BRBs,

likely due to engagement of a number of biological mechanisms (Casto et al. 2011). A

human clinical trial with a BRB-based mucoadhesive gel demonstrated the ability of

BRBs to reduce the size and severity of precancerous oral lesions (Mallery et al. 2014),

90

while a relative reduction in the expression of molecular biomarkers indicative of SCC

was observed after patients were treated with BRB-based troches for two weeks

(Knobloch et al. 2016). These studies support a role for BRB-mediated efficacy in oral

cancer prevention strategies.

Most research using BRBs has been conducted with minimally processed,

lyophilized BRB powder. In practical terms, consumers mostly encounter BRBs after

they have been incorporated into shelf-stable food products and stored for varying lengths

of time, during which the phytochemical profile may be altered (Howard et al. 2012).

Research on the stability of the phytochemical profile in BRB-based food products is limited to a short defined list of compounds (Howard et al. 2012; Hager et al. 2008; Gu et al. 2014), while effects on the global phytochemical profile and bioactivity of these products are unknown. Metabolomics is an emerging approach to chemical analysis in which hundreds to thousands of compounds within a food system are profiled, with the potential to provide new insight into the relationship between food phytochemicals and health outcomes (Manach et al. 2009). The objective of the current study is to use an untargeted metabolomics approach to understand the global differences in the phytochemical profile of a BRB nectar beverage over storage time and temperature variations, and how these changes relate to the growth inhibition activity in an in vitro

oral premalignacy model.

91

3.3 Materials and Methods

3.3.1 Chemicals

All solvents were of HPLC-MS grade from Fischer Scientific (Pittsburgh, PA) unless otherwise noted. Cyanidin-3-O-rutinoside (C3R) and protocatechuic acid (PA) standards were from Sigma Aldrich (St. Louis, MO). Cell culture-grade dimethyl sulfoxide and water were also from Sigma Aldrich.

3.3.2 Nectar Processing, Storage, and Sampling

Nectar was prepared in the pilot plant facilities located at The Ohio State

University (Columbus, OH) using a formula similar to that described by Gu and colleagues, as shown in Table 3.1 (Gu et al. 2014). The BRB powder used was produced from whole BRBs (Rubus occidentalis cv. Jewel) harvested at Stokes Berry Farm

(Wilmington, OH). All components were combined in a high shear mixer for 20 min, and the nectar was subsequently pasteurized using a MicroThermics UHT/HTSTLab-25HV

Hybrid unit (MicroThermics, Inc., Raleigh, NC, USA). The processing specifications mirrored industry practices for pasteurization of this product by which the nectar was held at 100 °C (± 1.1 °C) for 23 sec, immediately cooled, and aseptically filled into pre- sterilized 50 mL conical centrifuge tubes.

Nectar was stored at -20 °C, 4 °C, 10 °C, 25 °C, or 35 °C for 60 d with samples

(n=4) removed from each condition at 5 d, 10 d, and subsequently in 10 d intervals. Two months has been described as an appropriate amount of time to study stability in foods intended for clinical trials, as time is often needed for subject recruitment and

92 intervention (Gu et al. 2014). At each time point the nectar was centrifuged (1000 x g) for

5 min, partitioned into smaller aliquots, and stored at -80 °C prior to use. Samples of freshly produced nectar were also stored at -80 °C at the time of production as a t0 sample. Aerobic plate counts and yeast and mold counts were obtained for the 35 °C incubated samples at each time point with 3M Petrifilm (3M Company; Maplewood,

MN), according to manufacturer instructions.

93

Table 3.1 BRB nectar beverage formulation

Ingredients % Wet Basis Water 89.9 Sucrose 3.0 Pectin 0.5 Corn Syrup 1.0 BRB Powder 5.6 TOTAL 100.0

94

3.3.3 Sample Preparation and Analysis for Untargeted Metabolomics

Aliquots (1 mL) of nectar were thawed in a room temperature water bath for 10

min. Once thawed, 750 µL were deposited into a glass vial followed by 2.25 mL of 0.1%

formic acid in methanol. The mixture was homogenized with a probe sonicator (Branson

Ultrasonics; Danbury, CT) for 10 sec and centrifuged for 5 min at 1000 x g (4 °C). The

supernatant was decanted into a glass vial, and the pellet was extracted twice more with 3

mL of 75% methanol in water with 0.1% formic acid. Aliquots (200 µL) were deposited into 4 mL glass vials, dried under a stream of nitrogen, and stored at -20 °C until analysis.

Dried aliquots were solubilized in 100 µL 25% methanol in water with 0.1% formic acid and vortexed for 15 sec. Samples were then centrifuged at 21,130 x g for 4 min (4 °C) and placed in the autosampler of a 1290 Infinity II series UHPLC (Agilent

Technologies, Santa Clara, CA) maintained at 4 °C until analysis. Samples were injected

(5 µL) onto a 2.1 x 100 mm, 1.8 µm Acquity HSS T3 column (Waters, Milford, MA) maintained at 40 °C. The mobile phase consisted of A: 0.1% formic acid in water and B:

0.1% formic acid in acetonitrile with a flow rate of 0.5 mL/min. The linear gradient program was as follows: 0% B held for 1 min, increased to 60% B over 5 min, increased to 100% B over 2 min and hold for 1.5 min, immediately switched to 0% B and held for 2 min for a total run time of 11.5 min.

Eluent was directed to an Agilent iFunnel 6550 QTOF-MS interfaced with an electrospray ionization (ESI) source operated in negative ion mode. The first minute of flow from the UHPLC was directed to waste. Relevant MS settings were as follows: gas 95 temp 150 °C, drying gas 18 L/min, nebulizer 30 psig, sheath gas temp 350 °C, sheath gas flow 12 L/min, VCap 4000 V, nozzle voltage 2000 V, acquisition mode was 2 GHz extended dynamic range with a mass range of 50–1700 m/z. Reference mass solution

(Agilent Technologies) was concurrently infused into the source via a dedicated sprayer for continual mass correction. Sample run order was randomized. Quality control samples, composed of equal portions of each nectar sample, were run every 10 samples to monitor instrument performance over the run time (data not shown).

3.3.4 Data Pre-processing and Analysis for Untargeted Metabolomics

Raw spectral data was processed using the batch recursive feature extraction algorithm in Profinder (B.08.00, Agilent Technologies). Mass spectral features were picked and binned according to expected isotope patterns, adducts, and charge states.

These molecular features were then aligned across all samples, and those appearing in less than three samples per time/temperature group were removed from further analysis.

The raw data was then searched against this assembled list in a targeted manner to improve the quality of the data used for multivariate analysis. Further data pre-processing was performed in Mass Profiler Professional (version 14.5, Agilent Technologies), including removal of features present in sample blanks. To remove low quality peaks from the data, an additional abundance filter was applied which required a minimum peak height of 5.0 x 104 in 75% of the samples in at least one time/temperature sample group.

Multivariate analyses, including principal component analysis and partial least squares regression (PLS), were executed in R (version 3.2.3) with the ropls package

96

(Thévenot et al. 2015). Data were log10 transformed and Pareto scaled prior to analysis.

PCA is a dimensional reduction technique that allows for analysis of multidimensional data in an easily-visualized space. PLS is a common multivariate modeling technique that builds off the dimensional reduction properties of PCA but in the framework of a linear regression (Kemsley et al. 2007). Separate PLS models were constructed for each storage condition. The X matrices were composed of features present in 75% of replicates from at least one time point in each storage condition, and the Y matrix was storage time.

Performance of the PLS models was assessed using 8-segment cross validation, and statistical significance of each model was determined using permutation tests (n = 100).

Features with a variable importance on projection value (VIP) ≥ 1 across all successful models were manually reviewed before further analysis. Similarly, features with a VIP ≥

1 in only the 35 °C samples were also manually reviewed for further analysis. VIP scores are estimates of the relative importance of a chemical feature to a given PLS model, and features with a score ≥ 1 are typically considered to be important in the model. A data pre-treatment and analysis summary is shown in Figure 3.1.

97

Figure 3.1 Summary of untargeted metabolomics data pre-treatment and analysis

98

3.3.5 Targeted Compound Analysis

Cyanidin-3-O-rutinoside (C3R) and protocatechuic acid (PA) were quantified in

the nectar samples from t0 and 60d at 35 °C. Extracts of BRB nectar were obtained as

described for the untargeted metabolomics workflow, reconstituted in 5 mL of 5%

aqueous formic acid, and filtered through a 0.22 μm nylon filter. Samples were then

injected (0.5 μL) into an Agilent 1290 Infinity II UHPLC coupled to an Agilent 6495

triple quadrupole MS equipped with an ESI source operated in positive and negative ion

modes. The mobile phase consisted of A: 5% aqueous formic acid and B: 5% formic acid

in acetonitrile. The column and gradient program were identical to that was described

here for untargeted analyses. MS parameters included gas temp: 150 °C, gas flow: 18

L/min, nebulizer: 45 psi, sheath gas heater: 375 °C, sheath gas flow: 12 L/min, capillary:

3000 V, fragmentor: 350. Quantitation was performed using standard curves constructed

from serial dilutions of authentic standards. The transitions used for each compound were

as follows: C3R: 595 [M+] 287 (CE = 17 V), PCA: 153 [M-H]- 109 (CE = 10).

3.3.6 Extract Preparation for Cell Study

Extraction of the nectar was scaled up from the procedure used in the untargeted metabolomics workflow to ensure sufficient extract mass. Briefly, nectar replicates were pooled and 1 mL aliquots were deposited into glass vials followed by 3 mL of 0.1% formic acid in methanol. The mixture was homogenized with a probe sonicator and centrifuged at 3220 × g (4 °C) for 7 min. The supernatant was decanted into a glass vial, and the pellet was extracted once with 75% aqueous methanol with 0.1% formic acid.

99

The pooled supernatants were dried using a Genevac EZ2 vacuum evaporation system

(SP Scientific; Ipswich, United Kingdom) set at 30 °C. Remaining water was removed by

lyophilization on a Labconco FreeZone 12 Plus system (Kansas City, MO). Nectar

extracts were reconstituted in 1:1 DMSO/water, sonicated for 15 sec, and diluted to a

concentration of 2 mg extract/mL in cell culture media.

3.3.7 Cell Culture and Growth Inhibition Assay

Premalignant human oral epithelial cells (SCC-83-01-82) were maintained in modified minimal essential medium (MEM) with 10% fetal bovine serum and 5% antibiotic/antimycotic solution including penicillin (10,000 U/mL), streptomycin (10,000

U/mL), and amphotericin B (25 μg/mL) as previously described (Han et al. 2005; Ding et al. 2009). The characteristics of this cell line have been previously described (Lee et al.

1997). Cell cultures were incubated at 37 °C in a 90% humidified environment with 5%

CO2 atmosphere.

Cells were seeded at a density of 1000 per well in 96-well plates. After 24 hr, the

media was replaced to deliver 200 μg extract/well or standards of C3G or PA at

concentrations ranging from 3–100 μg/mL using previous work with crude berry product

extracts and their isolated components as a guide (Bishayee et al. 2016). Control samples were composed of an equivalent amount of 1:1 DMSO/water diluted in MEM. All samples were incubated for 72 hr.

Growth inhibition was determined using a WST-1 assay (Roche; Pleasanton, CA) according to manufacturer instructions. Growth inhibition was calculated as 1 - ((Atrt-Atrt

100 blank)/(Acontrol-Acontrol blank)). Treatment blanks were made by incubating sterile media with the corresponding dose of nectar extracts or phytochemicals in identical conditions as the treated cells. Technical replicates were performed in quadruplicate, while biological replicates were performed in triplicate. Cytotoxic activity was evaluated using the

Clontech LDH Cytotoxicity Detection Kit (Mountain View, CA) according to manufacturer instructions. Data were analyzed using the generalized linear model procedure in SAS version 9.4. The data were fitted with an ANOVA model with terms corresponding to nectar incubation time, temperature, and their interaction with significance reported at P<0.05. Differences between treatments were assessed using

Tukey’s post hoc test with α = 0.05.

3.4 Results and Discussion

We report on the phytochemical stability of a BRB nectar over storage using targeted and untargeted metabolomics, and we relate these chemical changes to their bioactive properties on premalignant oral epithelial cell proliferation. The product was a viscous liquid with pH of 3.4 and soluble solids reading of 9 °Brix. Microbial growth observed during storage was below the limit of quantitation (data not shown), indicating that any chemical changes incurred over storage were not due to microbial metabolism.

Untargeted metabolomics has been used by others to understand the chemistry of foods in several applications including food authentication, effects of different production practices, the dynamics of fermentation processes, and recently, changes in flavor attributes during storage (Cevallos-Cevallos et al. 2009; Ronningen et al. 2018;

101

Ronningen & Peterson 2018). Here we use the technique to understand how the chemistry of BRB nectar, as impacted by storage, may relate to the biological activity of the product.

3.4.1 Untargeted Metabolomics Revealed Large Chemical Variation with Elevated Storage Temperatures

Full scan UHPLC-MS-QTOF data was acquired for all nectar samples. Following the extraction, alignment, binning, and filtering of peaks in the data, a total of 1,712 molecular features were considered for further analyses. Overall trends across the dataset

were visualized using principal component analysis (PCA). The scores plot in Figure 3.2

indicate that the samples stored at -20 °C were relatively stable over 60 days of storage as

demonstrated by their close clustering and proximity to the samples from t0. Samples

stored at higher temperatures for longer amounts of time were further separated from the

t0 samples along the first component, which explained 37.7% of the variation, suggesting

that considerable chemical variation was introduced with elevated temperature and time.

102

Figure 3.2 PCA scores plot of all samples colored by storage temperature and labeled according to length of storage.

103

Partial least squares regression (PLS) was used to further understand how

chemical profiles of BRB nectars stored at different temperatures changed over time. A

separate model for each storage temperature was generated in which relative feature

abundances were regressed against storage time, including t0 (Table 3.2). The model for

samples stored at -20 °C was of poor quality (Q2 = 0.165; P = 0.03), indicating that

storage time was not a strong predictor of chemical variation in these samples. This

further demonstrated the stability of BRB nectar stored at -20 °C for 60 d. The models for

samples stored at 4 °C–35 °C all had a Q2 > 0.9 (P = 0.01), which indicated good

performance of these models. We focused our analysis on features that were influenced

by storage time, regardless of storage temperature, by collating features with VIP ≥ 1

across all four PLS models. Following a manual data quality review, 73 features were

found to contribute significantly to all four models. Figure 3.3 displays the mean relative

abundance of these features at each time point across storage conditions. Features were

clustered using Euclidian distance metric and Ward’s linkage rule. The heat map

demonstrates that these significant features increased and decreased simultaneously at

storage temperatures above -20 °C. These relative changes in abundance appeared to be

more severe at 25 °C and 35 °C storage, as anticipated. The features in cluster A reflect a

pattern of formation by which features were not present at t0 and were created over time, more so at higher temperatures. At 35 °C, some of these features decreased in abundance before day 60, indicating further degradation of these generated compounds. Cluster B contains features that degraded over time, some which degraded after 20 days at 35 °C.

The features in cluster C increased in abundance continuously over time with elevated

104

storage temperatures. Many of these features were present at low levels in the t0 samples.

These data demonstrate that above -20 °C, the BRB nectar is a system in dynamic

chemical flux over 60 days of storage.

Tentative identifications were generated for some features based on plausible

database matches from FooDB (www.foodb.ca), a component of the human metabolome

database (Wishart et al. 2013). Identities were confirmed by authentic standards or by

collecting additional MS/MS fragmentation data and comparing to published values when

authentic standards were unavailable. These techniques correspond to identification

levels 1 and 2, respectively, as proposed by the Metabolomics Standard Initiative

(Sumner et al. 2007). Features that were identified using these methods are shown in

Table 3.3, all of which have been previously reported in BRBs (Kula et al. 2016).

Catechin and epicatechin are isomeric flavan-3-ols, and have been shown to degrade over storage in other products such as apple juice (Van Der Sluis et al. 2005), and a variety of blueberry products (Brownmiller et al. 2009). B-type procyanidins are oligomers of catechin and/or epicatechin linked by C-C bonds, and have also been reported to be unstable over storage (Brownmiller et al. 2009). The MS/MS fragmentation patterns of the B-type procyanidins from the current work closely resembled those reported previously (Gu et al. 2003). PA is a B-ring cleavage product of cyanidin-based anthocyanins that can form as a result of heating or storage, but is also present in fresh

BRBs (Stintzing & Carle 2004; Kula et al. 2016).

105

Table 3.2 PLS model cross validation results

Storage 2 2 2 2 Temperature R Y P R Y Q P Q RMSEE -20 °C 0.891 0.06 0.165 0.03 6.98 4 °C 0.998 0.01 0.948 0.01 1.04 10 °C 0.995 0.01 0.980 0.01 1.58 25 °C 0.998 0.01 0.919 0.01 0.997 35 °C 0.991 0.01 0.985 0.01 2.05

106

Figure 3.3 Heat map of molecular features with VIP>1 in PLS models for storage at 4– 35 °C. Features were clustered using Euclidian distance metrics and Ward’s linkage rule. * Denotes potential Maillard-related sugar fragmentation-phenolic degradation products determined after derivatization with o-phenylenediamine.

107

Table 3.3 Level 1 and 2 identified compounds from list of features with VIP>1 across all four PLS models

Molecular Retention Heat map Compound Name [M-H]- Δppm Formula time (min) cluster

1 Epicatechin C15H14O6 3.6 289.0718 0 B 1 Catechin C15H14O6 3.3 289.0718 0 B 2 B-type procyanidin dimer A C30H26O12 3.6 577.1344 1 B 2 B-type procyanidin dimer B C30H26O12 3.4 577.1398 8 B 1 Protocatechuic acid C7H6O4 2.9 153.0196 2 C

108

The lack of plausible database matches for many of the features in clusters A and

C in Figure 3.3 led us to hypothesize that these entities may be uncharacterized

degradation products of BRB components. The Maillard reaction is a prevalent reaction

between reducing sugars and amino acids that occurs over processing and storage of

foods. Intermediates in this reaction include reactive carbonyl species that can form

adducts with phenolic compounds, such as epicatechin, in food products (Totlani &

Peterson 2005). Kokkinidou and Peterson demonstrated that phenolic-reactive carbonyl

species adducts can be decomposed by derivatization with o-phenylenediamine

(Kokkinidou & Peterson 2013). When o-phenylenediamine was added to BRB nectar

extracts, the abundances of 7 features from clusters A and C in Figure 3.3 were

significantly or completely reduced, suggesting that these may be Maillard-related sugar

fragmentation-phenolic degradation products.

3.4.2 All Extracts of Stored BRB Nectar Inhibited SCC-83-01-82 Cell Growth Similarly

BRB nectar extracts were applied to SCC-83-01-82 premalignant oral epithelial cells to assess the effects of storage time and temperature on their cell growth activities.

Extracts of the t0 nectar samples inhibited cell growth 27.8 ± 2.8% with inhibition of the

stored samples shown in Figure 3.4. Data were evaluated using a two-way ANOVA model including terms for nectar storage time, temperature, and their interaction. The terms for time (P < 0.01), temperature (P < 0.01), and their interaction (P<0.0001) were significant. Few significant differences, however, were seen across time within any one storage condition, except for a 13% difference between samples stored for 10, 20, and 60 109

days at 35 °C (Figure 3.4). Non-significant trends emerged in the dataset, but were

inconsistent among storage conditions. For example, a non-significant decrease in cell

growth activity was observed for nectars stored at 10 °C, but this same trend was not

maintained with storage at 25 °C. Thus, the capacity of BRB nectar extracts to inhibit

SCC-83-01-82 cell growth after 72 hr of incubation was relatively unaffected by nectar

storage conditions, despite the large variation seen in the nectar chemical profiles.

Few studies have investigated the relationship between storage conditions of berry

products, their corresponding chemical profiles, and bioactivity. A decrease in total

anthocyanin content was observed over 60 days in blueberry juice produced from two

different cultivars and stored at 6 °C and 23 °C. When the anthocyanin fraction of the

juice was isolated and applied to HT-29 colorectal adenocarcinoma cells, the authors

observed significant decreases in anti-proliferative activity after 30–90 days of storage.

Only a slight decrease in anti-proliferative activity was noted in samples stored at 23 °C

for 60 days, but it was concluded that storage conditions influenced the anthocyanin

profiles and biological activities of the juices (Srivastava et al. 2007). The untargeted

metabolomics approach employed in the present work aims to elucidate the relationship

between the chemical profile and biological activity of a berry product in a more

comprehensive way. BRBs contain a complex mixture of phytochemicals, thus it is

unlikely that any single chemical component can account for the complete bioactivity of

the fruit. For example, feeding whole BRB powder, anthocyanin-rich BRB extract, or

anthocyanin-deplete extract all suppressed the growth of tumors to an identical amount in a rat model of esophageal cancer (Wang et al. 2009). Paudel and colleagues used NMR-

110 based metabolomics to understand the effects of BRB cultivar and degree of ripeness on bioactivity in HT-29 colon cancer cells. They observed a myriad of biologically active

BRB components including anthocyanins, other flavonoids, organic acids, and ellagic acid derivatives (Paudel et al. 2014). Our data support these findings in that we observed nominal changes in bioactivity of BRB nectar products with considerably different phytochemical profiles, further demonstrating that a number of phytochemicals are responsible for this bioactivity.

111

Figure 3.4 Growth inhibition of SCC-83-01-82 cells by extracts of BRB nectar stored at increasing temperatures. ANOVA terms for storage time, temperature, and their interaction were significant (P<0.01). Only significant differences within each storage temperature are denoted (*).

112

3.4.3 C3R and its degradation product PA Equally Contribute to the Bioactivity of BRB Nectar

Since the anti-proliferative activity of stored nectars was relatively unchanged

despite large changes in chemical profiles, we hypothesized that parent phytochemicals,

as well as their degradation products, both contribute to the bioactivity of the product due

to similarity in structural motifs. Anthocyanins constitute a large portion of the total

polyphenols of BRBs, with C3R as a predominant species (Rothwell et al. 2013; Tulio et

al. 2008; Torre & Barritt 1977). Given that PA is a reported degradation product of C3R

and was identified as an important feature in our PLS models, we focused on these two

compounds in a model system as proof-of-concept that parent compounds and their associated degradation products can be complementarily bioactive.

To understand how these two related compounds changed in the nectar over time,

we extracted information about their abundances from the untargeted UHPLC-MS-QTOF dataset (Figure 3.5). The relative change in abundance over time was greater at higher storage temperatures for both compounds, consistent with prior findings on anthocyanin degradation (Patras, Nigel P. Brunton, et al. 2010). C3R and PA were subsequently quantitated at t0 and 60 d of storage at 35 °C using UHPLC-MS/MS and authentic

standard curves (Table 3.4). Interestingly, C3R decreased by 13.1 nmol/mg, while PA

increased by 14.9 nmol/mg during storage, demonstrating that these two bioactive

compounds exchanged in near-equimolar amounts in the BRB nectar.

113

Figure 3.5 Averaged relative abundances of C3R and PA over time in each storage condition.

Table 3.4 Quantitative analysis of C3R and PA in nectar from t0 and 60 days at 35 °C

Time C3R PCA (d) (μg/mg extract) (μg/mg extract)

0 8.02 ± 0.20 3.27 × 10-2 ± 8.1 × 10-3 60 0.20 ± 5.7 × 10-4 2.33 ± 0.13

114

Independently, C3R and PA each inhibited the growth of SCC-83-01-82 cells in a dose-dependent manner (Figure 3.6A, B), with increasing concentrations corresponding with increased growth inhibition. The growth inhibition by C3R is similar to levels previously reported on cyanidin-3-O-glucoside isolated from strawberries, which inhibited the growth of CAL-27 malignant oral cancer cells by approximately 50% at a level of 100 μg/mL (222 μmol/L) (Zhang et al. 2008). The current work further validates the bioactivity of cyanidin-based anthocyanins to inhibit cell growth in human oral cell lines. The anticancer activity of PA against oral cancer has previously been demonstrated in animal models (Tanaka et al. 2011). While the concentrations we tested in vitro are higher than those found in the BRB nectar, our results show that SCC-83-01-82 cells respond to individual treatments of C3R or PA in a dose-dependent manner.

To demonstrate that BRB phytochemicals and their degradation products can each contribute to the biological activity of the nectar, we delivered doses of equal molarity but differing molar ratios of C3R:PA. The conditions used mirror the equimolar exchange of these two compounds observed in the nectar. As shown in Figure 3.6C, after a starting dose of 170 μmol/L, C3R was reduced by 25% in successive treatments, while in parallel the concentrations of PA were increased in 25% increments to a final treatment dose of

170 μM. A consistent level of growth inhibition was maintained across treatments (P =

0.092 for differences among treatments) despite differing molar ratios of C3R:PCA.

Consequently, our data demonstrates that the loss in bioactivity of a parent phytochemical (C3R) may be recovered by increased levels of their degradation products

(PA) (Figure 3.6C). Previous studies with other cancer models have found the ortho-

115

dihydroxyphenyl structural element of some anthocyanidins, such as cyanidin, to be

critical for anti-cancer properties of these compounds (Hou et al. 2003). Our data suggest

that this structural moiety, the main molecular structure maintained between C3R and

PA, may also play a role in suppressing the growth of SCC-83-01-82 cells. Partial

degradation of cyanidin-3-glucoside in cell culture media has been previously reported,

with PA as the primary degradation product (Kay et al. 2009). While this represents an

inherent limitation of studying anthocyanins in vitro, it further validates the idea that

phytochemical degradation products can maintain active chemical moieties, and thus

bioactivity. Although C3R and PA appeared to exchange in roughly equimolar amounts

in the nectar, anthocyanins can degrade via a multitude of mechanisms to form several

different products, while PA can also be a degradation product from other phenolics

(Patras, Nigel P. Brunton, et al. 2010). Thus, we speculate that this phenomenon of

degrading phytochemicals while maintaining active chemical moieties occurs on a larger

scale with other components of the nectar. And while not addressed in the current study,

it is also conceivable that the biochemical signaling and activation mechanisms

underlying the growth inhibition shifted with changing nectar chemical profiles. In

addition, our bioassay was an in vitro model with oral cells that can be directly exposed to BRB phytochemicals in vivo. Not addressed in this study is the impact that storage- induced changes in BRB phytochemicals affects their bioaccessibility and bioavailability in the remainder of the GI tract, which could have implications for their actions elsewhere in the body.

116

117

Figure 3.6 Dose-response relationship between increasing levels of C3R (A) and PA (B) and growth inhibition of SCC-83-01- 82 cells. (C) When the ratio of C3R and PA were varied in equimolar solutions (molarity on right y-axis), growth inhibition (left y-axis) was maintained (P = 0.092 for differences among treatments).

117

3.5 Conclusions

In conclusion, we investigated the impact of storage on the phytochemical stability and bioactivity of a BRB nectar product. Our data demonstrate that nectar stored at -20 °C is chemically stable over 60 days, but storage above this temperature introduces large amounts of chemical variation through a variety of mechanisms including cleavage of phenolic compounds and potential adduct formation with reactive carbonyl species.

Despite the large chemical variation observed using untargeted metabolomics, storage conditions had minimal impact on the ability of the nectar to differentially inhibit growth in premalignant oral epithelial cells. Exploration of this phenomenon in vitro supports our hypothesis that degradation products of bioactive phytochemicals also demonstrate bioactivity, allowing maintenance of growth inhibition capacity, through independent, cooperative, or redundant mechanisms. This work demonstrates that BRBs are a complex mixture of compounds with potential anticancer activities. Assigning functional activity to a single black raspberry compound or metabolite fails to explain and appreciate this fluidity, as different compounds increase and decrease with the dynamics of storage. It remains important to dissect these pleiotropic phytochemical bioactives to fully understand the health benefits and consequences of consuming BRBs and their components.

118

Chapter 4. Elucidating markers of strawberry consumption in smokers and non- smokers using untargeted metabolomics

Matthew D. Teegarden1, Jennifer H. Ahn-Jarvis1,2, Thomas J. Knobloch2, Christopher M.

Weghorst2, Yael Vodovotz1, Steven J. Schwartz1, and Jessica L. Cooperstone1,3

1Department of Food Science and Technology, The Ohio State University. Columbus,

OH, USA.

2College of Public Health, The Ohio State University. Columbus, OH, USA.

3Department of Horticulture and Crop Science, The Ohio State University. Columbus,

OH, USA.

119

4.1 Abstract

Laboratory and clinical evidence suggest dietary berries may be beneficial for health, including diseases for which cigarette smokers are at high risk. Our group recently conducted a clinical trial in which smokers and non-smokers consumed confections containing lyophilized strawberry powder (24 g powder/day) or a macronutrient-matched placebo. The objective of this work was to explore global differences in the untargeted urinary metabolomes of smokers and non-smokers from the strawberry and placebo arms of this trial. Using multilevel-multivariate analysis, we identified ellagitannin-related metabolites, dihydroxybenzoic acids, a hydroxycinnamic acid, and metabolites of a strawberry aroma compound as elevated by strawberry exposure. Metabolites of flavor compounds and colorants were elevated in the placebo arm. No major effect of smoking on phytochemical excretion was observed using this approach. These findings will inform future clinical trials with strawberry products with regard to potential markers of exposure. The implications of study design on observed metabolomic profiles are also discussed.

4.2 Introduction

Pre-clinical and clinical evidence suggest that dietary berries may play a role in the prevention of several diseases, including cancers of the aerodigestive tract (Stoner et al. 2008) and cardiovascular disease (Basu et al. 2010). The health promoting properties of berries are thought to be mediated by their diverse array of phenolic phytochemicals including (but not limited to) anthocyanins, hydroxycinnamic acids, ellagitannins, and

120

flavonols (Seeram 2008) and the intestinal and microbial metabolites of these compounds

(Del Rio et al. 2010). Research on specific compounds, such as anthocyanins (De Ferrars

et al. 2014), has expanded understanding of the fate of some berry phytochemicals once

they are consumed. However, to better discern how berry consumption may affect

certain diseases, further work is needed to understand the impact of whole berries on a

person’s metabolic profile.

Cigarette smokers represent a high-risk population for diseases that berries may impact, including several cancers and cardiovascular diseases (United States Department of Health and Human Services 2014). Epidemiological data suggest that smokers who regularly consume fruits and vegetables have a lower relative risk of cardiovascular disease (Hung et al. 2004) and a slightly decreased hazard ratio for incident cancer

(Boffetta et al. 2010) compared to smokers who eat fewer fruits and vegetables.

However, there is evidence to suggest that smokers exhibit either lower absorption or

faster turnover of nutrients and phytochemicals when compared to non-smokers,

including vitamin C, vitamin E, and carotenoids (Wei et al. 2001; Lykkesfeldt et al. 2000;

Kallner et al. 1981; Hammond et al. 1996; Bruno et al. 2006). Animal evidence also suggests that the microflora in the GI tract may be influenced by smoking status (Allais et al. 2015), which could lead to altered profiles of microbiome-derived metabolites of nutrients and phytochemicals in smokers. However, smokers are generally excluded from clinical trials due to the existence of possible confounding factors to study outcomes.

Our group recently conducted a clinical trial in which smokers and non-smokers consumed confections containing whole, freeze-dried strawberries (Fragaria x ananassa)

121 and berry-flavored placebo confections (unpublished data, Vodovotz Lab). By utilizing samples from this previously conducted clinical trial, we aim to understand the impact of strawberry consumption on the overall metabolic profiles of free-living individuals on a low anthocyanin background diet, as well as the potential impact of cigarette smoking on berry phytochemical metabolism. Here, we used untargeted metabolomics, an analytical technique which aims to profile, as comprehensively as possible, all the small molecules in a given sample. Various statistical techniques are then applied to understand chemical differences between treatment conditions (Manach et al. 2009). Metabolomics has previously been used in nutritional interventions to understand metabolic perturbations due to various interventions or to discover markers of exposure for specific foods and dietary patterns (Brennan 2013). The objective of this work was to use this technique to explore global differences in the urinary metabolomes of smokers and non-smokers consuming strawberries.

4.3 Materials and Methods

4.3.1 Chemicals

Optima LC/MS grade water, acetonitrile, and formic acid were from Fisher

Scientific (Pittsburgh, PA). Urolithin A standard was obtained from Chem Bridge (San

Diego, CA), 3,4-dihydroxyphenylvaleric acid was from Enamine (Kyiv, Ukraine), and all other standards and reagents were obtained from Sigma Aldrich (St. Louis, MO).

122

4.3.2 Subjects and Study Design

The study protocol was approved by The Ohio State University’s (OSU)

Biomedical Sciences IRB (2010H0073) and registered with ClinicalTrials.gov

(NCT01514552). Participants were recruited from the Columbus, OH area, and were either smokers (≥10 cigarettes/day with urine cotinine > 200 ng/mL) or nonsmokers with no prior history of tobacco use. Participants also had no significant co-morbidities, a body mass index between 20-35 kg/m2, and no prior history of intolerance or allergy to

strawberries, wheat, or corn. In addition, subjects had no oral lesions or periodontal

disease, no difficulty swallowing, and no history of antibiotic use within 6 months.

Informed consent was collected from all subjects, and all procedures were performed at the OSU Clinical Research Center.

The intervention was a single-blinded, randomized, placebo-controlled, crossover

clinical trial, outlined in Figure 4.1. Subjects were block randomized and age-matched

(smoker and non-smoker) to one of the two treatment sequences by the study coordinator.

Once enrolled, subjects were placed on a standardized vitamin regimen and instructed to

abstain from additional vitamin and usage over the course of the

study. Subjects were additionally instructed to follow an anthocyanin-free (and thus low-

ellagitannin diet) and were provided with a list of corresponding foods to avoid for 14

days leading up to, and during, each 7 day treatment period. Restricted foods included

essentially all raw and processed berry products and any other anthocyanin containing

fruits and vegetables. Adherence to the diets was high at 92±11% (unpublished data,

Vodovotz lab). Treatments consisted of gummy confections formulated with lyophilized

123

strawberry powder (strawberry treatment) or strawberry flavor and food coloring

(placebo) (Fisher et al. 2014). Subjects were instructed to consume 8 confections

interspersed throughout each treatment day. Total daily consumption of strawberry

powder via the confections was 24 g/d, equivalent to approximately 266 g of fresh

strawberries/d assuming a 91% moisture content (Wu et al. 2006).

Subjects were instructed to collect urine for 24 h and fast for 8 h immediately

leading up to each clinic visit. At each clinic visit, aliquots of urine were acidified with

5% formic acid, and stored at -80 °C until analysis. Smoking status was confirmed at

each clinic visit (COT One Step Quik Test; Boca Raton, FL). Adherence to study

protocols was determined using a combination of study agent consumption logs and two,

3 day diet records. A total of 12 smokers and 13 non-smokers completed the initial study.

The primary outcome of this study was to assess compliance and toxicity of the

strawberry and placebo confections, which will be described in detail elsewhere (Ahn-

Jarvis et al. 2014). The study sample size was calculated to achieve sufficient statistical power for this outcome. The work described herein addresses secondary outcomes related to understanding how consumption of the strawberry confection modulates urinary profiles of strawberry polyphenols, relative to a placebo. Urine from smokers (n =

9) and non-smokers (n = 10) with 100% adherence (self-reported) to the treatment protocol at the time points following each intervention arm was selected for untargeted metabolomics analysis.

124

Figure 4.1 Crossover study design with 24 hr urine collection time points (triangles). Circled timepoints indicate those selected for further study using untargeted metabolomics.

125

4.3.3 Sample Preparation

Urine was volumetrically normalized according to osmolality prior to analysis to

account for differences in urine concentration to improve sample discrimination for data

analysis (Chetwynd et al. 2016). Briefly, osmolality of all urine samples was measured using an Advanced Instruments 3300 model micro osmometer (Norwood, MA). The sample with the lowest osmolality was diluted three times with 0.1% aqueous formic acid. The remaining samples were then diluted with 0.1 aqueous formic acid to that corresponding osmolality. Samples were then centrifuged at 21,130 x g for 4 min prior to analysis. Sample preparation blanks were concurrently prepared with 0.1% aqueous formic acid. Quality control (QC) samples were prepared by combining equal volumes of each sample and prepared identically to the other samples.

4.3.4 UHPLC-MS Data Acquisition

Prior to all data acquisition, the system was equilibrated with consecutive

injections of a QC sample. Samples were randomized for run order and interspersed with

QC samples every six injections. Two sample preparation blanks were run following the final QC sample. Chromatographic conditions were similar to Want and colleagues

(Want et al. 2010) with some modification. Samples were injected (3 μL) into a 1290

Infinity UHPLC (Agilent Technologies; Santa Clara, CA) equipped with a Waters HSS-

T3 column (1.8 μm; 2.1x100 mm; Milford, MA). The linear gradient of 0.1% aqueous formic acid (A) and 0.1% formic acid in acetonitrile (B) running at 0.5 mL/min was programmed as follows: hold at 1% B for 1 min, 1% to 15% B over 4 min, 15% to 50% B

126

over 3 min, 50% to 95% B over 3 min, hold at 95% B for 1 min, immediately switch and

hold at initial conditions for 2 min for a total run time of 14 min.

Eluent from the UHPLC was directed to an iFunnel 6550 Q-TOF-MS (Agilent

Technologies). Electrospray ionization (ESI) was utilized in both positive (ESI+) and negative (ESI-) modes, with data acquired separately for each mode. The MS parameters were as follows: gas temperature, 150 °C; gas flow, 18 L/min; nebulizer, 30 psig; sheath gas temp, 350 °C; sheath gas flow, 12; VCap, 4000 V; nozzle voltage, 2000V. Spectra were recorded between 50-1700 m/z at a rate of 3 spectra/s. The instrument was operated in 2 GHz extended dynamic range mode with 20K resolution, and calibrated before each experiment using ESI-L solution (Agilent). Continuous mass correction was achieved

with simultaneous infusion of reference solution (Agilent) through a dedicated sprayer.

4.3.5 Data Pre-processing

Molecular features were detected in raw data using the batch recursive feature

extraction process in Profinder (B.07.00, Agilent). In short, mass spectral features,

inclusive of isotopes and expected adducts, were grouped into singular entities and

aligned by retention time across samples. Entities that were present in at least two files in

the strawberry, placebo, QC, or preparation blank sample groups were retained and re-

extracted in all samples in a targeted manner. This recursive process ensures high quality

and consistency of extracted data. Following extraction, data was imported into Mass

Profiler Professional (Agilent Technologies) for further processing. Briefly, entities

present in extraction blanks and those with an abundance less than 5000 counts were

127

removed from analysis. Entities with a coefficient of variation above 30% across the QC

samples were additionally excluded. Finally, entities present in less than 75% of samples

in at least one treatment group (strawberry or placebo/smoker or nonsmoker) were

removed from further analysis.

4.3.6 Data Analysis

Abundance values were log2 transformed and normalized to MS total useful

signal (Chetwynd et al. 2016) prior to analysis. Initial data surveys and univariate

analyses were performed in Mass Profiler Professional. To better ascertain the effects of

each intervention and capitalize on the paired data structure introduced by the crossover

study design, multilevel analysis was employed as described by Westerhuis and

colleagues (Westerhuis et al. 2010). The variation in the data was split into between-

subject and within-subject variation in Matlab (version R2016a, MathWorks, Natick,

MA) using the algorithm available at http://www.bdagroup.nl/. Each resulting dataset

was interquartile range (IQR) filtered (Xia et al. 2012) in R (version 3.2.3) using the

‘genefilter’ package. The split data was Pareto scaled and analyzed using principal

components analysis (PCA) and partial least squares discriminant analysis (PLS-DA) in

R with the package ‘ropls’(Thévenot et al. 2015). Performance of PLS-DA models were cross-validated by randomly dividing samples into seven segments and evaluated using

permutation testing (n=1000).

128

4.3.7 Feature Selection and Identification

Features with a variable importance on projection score >1 that also passed a paired t-test with Benjamini-Hochberg false discovery rate correction (α=0.05) were selected as candidates for identification. Each entity that passed this filtering was manually reviewed to ensure proper data extraction and integration. Co-eluting entities

were manually reviewed for chromatographic shape and cross-sample correlation.

Selected entities were subjected to MS/MS fragmentation on the aforementioned

UHPLC-QTOF-MS platform operated in targeted MS/MS mode. Fragmentation spectra

were collected at collision energies of 10, 20, and 40 V. To aid in compound

identification, the strawberry powder used in this study was extracted once with 75%

aqueous methanol. The extract was analyzed using the same UHPLC-QTOF-MS

method, but with constant collision energies of 0 and 15 V. This was used to confirm if

the presence of metabolites in urine were directly derived from the study agent.

Tentative compound identifications were assessed using the Metlin Metabolite

Database (Guijas et al. 2018) and Human Metabolome Database (Wishart et al. 2013).

Spectral annotation and molecular formula generation were achieved using MassHunter

Qualitative Analysis (Agilent Technologies). Candidate identifications were further

evaluated using Molecular Structure Correlator (Agilent Technologies) and CFM-ID

(Allen et al. 2014) for in-silico spectral comparison as needed. When possible, identities

were confirmed with chemical standards. To synthesize glucuronidated or sulfated

metabolites, S9 liver isolates were prepared from pig liver obtained from the Ohio State

Meat Lab (Columbus, OH) as previously described (Wu and McKown 2014). These

129 isolates were then used to produce metabolite standards according to the protocol described Cuparencu and colleages (Cuparencu et al. 2016).

4.4 Results and Discussion

All nineteen subjects that reported 100% compliance to the strawberry intervention was selected for further analysis using untargeted metabolomics, including

10 non-smokers and 9 smokers. Urine samples collected from subjects after the placebo and strawberry interventions were selected to compare strawberry vs. no-strawberry consumption within a more similar background diet. Additionally, the comparison of these two time points allows for evaluation of metabolomic changes in the context of a placebo-controlled dietary intervention. On initial data quality review using PCA, two outliers were apparent in the dataset, consistent across both ESI modes, which were run on separate days (Figure 4.2A). The samples from these subjects were removed from further analysis, resulting in a final dataset consisting of two time points from 9 non- smokers and 8 smokers.

A total of 29,827 ESI- and 22,349 ESI+ chemical entities were extracted from the untargeted metabolomics data. Following the removal of the sample preparation blank

(i.e. extraction excluding urine), filtering on percent coefficient of variation across QC samples, removal of features with abundance < 5000 units, and IQR filtering, a total of

2972 ESI- and 2984 ESI+ entities were considered for further analysis. This filtering protocol ensured chemical entities were consistently measured across samples, were of

130 reasonable abundance for further experimentation, and removed entities with low variability among the samples.

4.4.1 Multi-level Multivariate Data Analysis

Initial visualization of the dataset by PCA did not reveal clear clustering of sample groups by intervention type (strawberry vs. placebo) or smoking status (smoker vs. nonsmoker) (Figure 4.2B). Supervised modeling of the data using PLS-DA was also unsuccessful. This can be attributed to the fact that metabolomic variation associated with the consumption of specific foods is easily overwhelmed by variation due to other biological and extrinsic factors (Rezzi et al. 2007). For example, considerable inter- individual variation exists in the metabolism of polyphenols, like those present in strawberries. The gut microbiome, genetic factors, and environmental factors are all influential in how different individuals absorb, metabolize, and excrete these phytochemicals (Lampe & Chang 2007; van Duynhoven et al. 2011). Study design and statistical techniques can be used to overcome these challenges. By using a cross-over study design where each person serves as their own control, statistical techniques that take advantage of the paired nature of the data can be employed (Elbourne et al. 2002).

Van Velzen and colleagues have developed a multilevel approach to analyze paired data in metabolomics experiments. In their procedure, variation in the dataset is first split into between-subjects variation (inter-individual variation) and within-subject variation

(variation due to treatment) and subsequently modeled using techniques such as PCA and

PLS-DA (Van Velzen et al. 2008).

131

Figure 4.2 PCA of untransformed, filtered data collected in ESI+ mode. (A) Data without outliers removed. The same two outliers were observed in ESI- mode data collected on a different day. (B) Data with outliers removed, demonstrating a lack of separation achieved with traditional PCA.

132

By using a multilevel approach, we observed clear patterns in both the inter-

individual variation and treatment variation of our dataset. The PCA scores plots for

each variation term in each mode of analysis (ESI+ and ESI-) are shown in Figure 4.3.

From this unsupervised analysis, it is clear that a large source of inter-individual variation was due to smoking status, since smokers are separated from non-smokers on the first component (Figure 4.3A-B). The PCA scores plots of variation due to treatment show clear separation according to placebo or strawberry interventions along the first components, with no differentiation of smokers and non-smokers (Figure 4.3C-D).

Together these data suggest that apart from representing a large source of inter-individual variation, smoking status did not additionally influence the excretion patterns associated with the strawberry or placebo interventions. Previous work demonstrated that depletion of plasma vitamin C in smokers could be increased to levels of non-smokers with a 90 d vitamin supplement intervention (Lykkesfeldt et al. 2000). Additionally, Bruno and colleagues demonstrated that two weeks of vitamin C supplementation lowered fractional plasma disappearance rates of vitamin E in smokers to resemble those of non-smokers

(Bruno et al. 2006). Thus it is possible that, if smoking does influence the absorption, metabolism, and excretion of strawberry phytochemicals, this effect could have been normalized over the course of our intervention due to the study agent itself or the prescribed multivitamin supplement. These data suggest that greater experimental control is needed to discern potentially acute effects of smoking on phytochemical absorption and metabolism.

133

Further analyses focused on modeling the variation due to treatment (strawberry vs placebo) using PLS-DA. The summaries of model parameters and cross validation results are shown in Table 4.1. The models for both ESI modes performed well with Q2 values > 0.8. Over 700 entities achieved a variable importance on projection (VIP) score

>1 in each model, which indicates significant contribution to the PLS-DA model. To further prioritize entities for identification, we also performed a paired t-test (placebo vs. strawberry) with Benjamini-Hochberg multiple testing correction (P<0.05). Following manual review for peak quality and combining of entities redundant between both ESI modes, a total of 41 entities were prioritized for identification. Of these, 26 were elevated in urine from the strawberry treatment arm, and 15 were elevated in the placebo arm.

134

Figure 4.3 Multilevel PCA analysis of inter-individual variation (panel A: ESI+; panel B: ESI-) and variation due to treatment (panel C: ESI+; panel D: ESI-). NS= Nonsmoker, S= Smoker, Straw= Strawberry, Plac= Placebo.

135

Table 4.1 Cross-validation summary of PLS-DA models on variation due to treatment

ESI Features with Features with R2Y (P-value a) Q2 (P-value a) Mode VIP > 1 VIP>1, P<0.05b Negative 0.954 (0.001) 0.871 (0.001) 731 31 Positive 0.972 (0.001) 0.888 (0.001) 718 24 a Following a permutation testing (n = 1000) b Following a paired t-test with Benjamini-Hochberg multiple testing correction

136

4.4.2 Identification of Features Elevated with Strawberry Consumption

Identification of metabolites was achieved using a combination accurate mass, database searches, MS/MS fragmentation, and matching with authentic and synthesized standards when possible (Figure 4.4). The Metabolomics Standards Initiative has outlined four levels of compound identification, with level 1 representing absolute identification with standards and level 2 as putatively identified compounds. Level 3 corresponds to characterization related to a specific class of compounds, while level 4 compounds are unknown and not readily classified (Sumner et al. 2007). We identified a total of 6 metabolites at level 1 and 1 at level 3 (Table 4.2). Urolithin A glucuronide, and a lower level of free urolithin A were found to be elevated in the strawberry treatment arm. Urolithins are produced from ellagic acid and ellagitannins, like those found in strawberries (N. P. Seeram et al. 2006), by intestinal microflora. Because of the enhanced bioavailability of urolithins as compared to ellagic acid/ellagitannins, these products are thought to mediate the majority of bioactivity associated with the consumption of ellagitannins and ellagic acid (Landete 2011). Ellagitannins and ellagic acid have limited distribution in foods, essentially only present in berries, pomegranates, nuts, and barrel aged beverages (Bakkalbaşi et al. 2009). Urolithin A is the most prevalently produced urolithin, but there is considerable inter-individual variation in the type and quantity of urolithins produced following consumption of ellagitannin- containing foods (Tomas-Barberan et al. 2017). Here, we observed over a 50-fold difference between the highest and lowest urolithin A glucuronide producers. The large

137

fold change between the mean abundance in the placebo and strawberry arms

demonstrates urolithin A was specifically elevated in the strawberry arm.

4-Hydroxy-2,5-dimethyl-3(2H)-furanone and its glucuronide and sulfate

conjugates were elevated in urine from the strawberry arm but also detected in the

placebo arm. 4-hydroxy-2,5-dimethyl-3(2H)-furanone is a strawberry aroma compound

(Pickenhagen et al. 1981), which potentially explains its simultaneous appearance in

urine from the strawberry and strawberry-flavored placebo arm in our study. It is also found in other foods that were unrestricted over the course of the study, such as pineapples and tomatoes (Buttery et al. 2001; Tokitomo et al. 2005). These compounds were recently noted as markers of strawberry exposure using an untargeted metabolomics approach in the context of a highly controlled meal study with a single dose of strawberry puree (Cuparencu et al. 2016). The fact that both of these metabolites were noted in two independent studies indicates elevated levels of conjugates of this aroma compound, relative to a baseline measurement, may be part of a metabolomic pattern distinct for strawberry exposure.

We additionally identified two different dihydroxybenzoic acids elevated in the strawberry study arm. The identity of 2,6-dihydroxybenzoic acid was confirmed with an authentic standard. This compound has been previously reported in strawberries

(Gasperotti et al. 2015), and analysis of the strawberry powder used in the present study showed a peak consistent with that seen in the urine. This suggests that 2,6- dihydroxybenzoic acid is directly absorbed from the strawberry powder and at least partially excreted without further metabolism. The remaining dihydroxybenzoic acid was

138

initially hypothesized to be protocatechuic acid (3,4-dihydroxybenzoic acid), as this is a

common metabolite associated with consumption of phenolic-rich foods (Williamson &

Clifford 2010). However, comparison with an authentic standard refuted this identity due

to inconsistent retention times. Comparison with an authentic standard confirmed the

identity as 2,5-dihydroxybenzoic acid, which has been observed in urine following

strawberry consumption (Russell et al. 2009). An unknown hydroxycinnamic acid was

also observed. This feature did not co-elute with authentic standards o-, m-, or p- coumaric acid, but it did exhibit nearly identical MS/MS spectra and similar retention time to m- and p-coumaric acid, suggesting it may be a structural isomer of one of these hydroxycinnamic acids (Figure 4.5).

An additional 9 metabolites elevated in the strawberry arm were putatively characterized (level 4) as unknown products of phase II metabolism based on characteristic neutral losses consistent with glucuronic acid (-176.03 amu) and sulfate (-

79.96 amu) conjugates, as shown in Table 4.2. The unknown sulfate with m/z 353.0333 at

4.3 min produced a fragment consistent with the mass of the dihydrochalcone, phloretin, which has been previously reported in strawberries as a glucoside conjugate (Hilt et al.

2003). However, comparison with a synthesized standard of phloretin sulfate refuted this identity due to inconsistencies in retention time. The unknown glucuronide with m/z

385.115 produced a fragment consistent with the mass of 3,4-dihydroxyphenylvaleric acid, a microbial metabolite of catechin and epicatechin (Aura et al. 2008), but this was also refuted when tested against a synthesized standard when comparing retention times.

We also evaluated sinapyl alcohol glucuronide (an accurate mass match) as a potential

139

identification for this feature, but it was not a match for the same reason. To glean

additional information regarding origins of the unidentified metabolites in urine, we

analyzed the strawberry powder from the study using the same MS method with an

applied collision energy. This approach allows us to search for common parent

(aglycone) masses between the urinary metabolites and compounds in the study food.

From this analysis, we determined that 3 metabolites may not have been directly

absorbed from the strawberry treatment. We hypothesize that these metabolites may be

the result of additional microbial catabolism and/or human metabolism of strawberry

phytochemicals because we did not observe the corresponding aglycone in the strawberry

powder. Phytochemicals, like the phenolic compounds found in strawberries, can be subject to catabolism by gut microbiota prior to absorption (van Duynhoven et al. 2011).

These compounds, in addition to those absorbed intact, can be further modified by phase

I metabolic enzymes before conjugation by phase II enzymes.

140

Table 4.2 Identified metabolites elevated in the strawberry study arm

RT Chemical Δ Fold VIP Metabolite IDa Ions used for ID (min) formula ppm changeb IV - 1.2 Unknown sulfate 190.9662 [M-H] C5H4O6S 6.3 1.9 1.18 111.0089 -SO3 IV - 1.4 Unknown sulfate 208.9763 [M-H] C5H6O7S 3.3 2.0 1.37 129.0191 -SO3 - 2.8 4-hydroxy-2,5- 206.9974 [M-H] C6H8O6S 2.4 2.5 1.58 dimethyl-3(2H)- 127.0402 -SO3 furanone sulfate I - 3.6 4-hydroxy-2,5- 303.0722 [M-H] C12H16O9 0.0 2.3 1.56 dimethyl-3(2H)- 127.0400 -C6H8O6 furanone glucuronide I + 2 4.3 Unknown 381.0805 [M+H] C15H14N3O9 -0.8 2.1E 3.55 IV glucuronide* 205.0461 -C6H8O6 - 4 4.3 Unknown 385.115 [M-H] C17H22O10 3.9 1.3E 5.16 IV glucuronide 209.081 -C6H8O6 - 4.4 2,5- 153.0197 [M-H] C7H6O4 2.6 1.8 1.12 dihydroxybenzoic 109.0292 acid I - 4.5 Unknown 247.0283 [M-H] C9H12O6S 2.8 2.4 1.46 IV sulfate* 167.0710 -SO3 - 5.0 2,6- 153.0196 [M-H] C7H6O4 2.0 2.0 1.24 dihydroxybenzoic 135.0083 acid I 109.0291 IV - 3 5.8 Unknown sulfate 353.0333 [M-H] C15H14O8S 0.6 3.0E 4.68 273.0770 -SO3 - 5.9 Unknown 535.1790 [M-H] C26H32O12 -4.7 89.9 3.04 IV glucuronide* 359.1504 -C6H8O6 + 3 6.3 Urolithin A 405.0818 [M+H] C19H16O10 0.5 2.6E 5.01 I glucuronide 229.0494 -C6H8O6 - 6.5 Unknown 163.0406 [M-H] C9H8O3 3.1 2.7 1.63 hydroxycinnamic 119.0502 acid III 93.0345 IV - 7.3 Unknown sulfate 213.0227 [M-H] C9H10O4S 2.3 2.1 1.32 133.0653 -SO3 I - 3 7.6 Urolithin A 227.0361 [M-H] C13H8O4 4.8 4.1E 4.69 198.0324 a Level of identification: I= Level 1, III= Level 3, IV= Level 4 b Mean fold change compared to the opposite treatment *No peak corresponding to aglycone was observed in the strawberry powder used on study

141

Figure 4.4 Example of level 1 metabolite identification. A: mass chromatogram demonstrating co-elution of a urolithin A glucuronide standard with peak in urine sample under identical analytical conditions. A slight retention time shift from the value reported in Table 4.2 because a new UHPLC column (same model) was used. B: fragmentation of urolithin A glucuronide standard and peak detected in urine (collision energy: 40 eV).

142

Figure 4.5 Mass fragmentation spectra (collision energy = 40 mEV) of tentatively identified o-coumaric acid in urine (top) and an authentic standard of p-coumaric acid (bottom).

143

Comparison of metabolomic studies can be difficult due to a diverse set of factors

including study design and data analysis techniques. For instance, 24 hr urine collections

were analyzed in this study, but Cuparencu and colleagues noted that some low-abundant

strawberry exposure markers were only detected when urine was collected in shorter (2

h) intervals (Cuparencu et al. 2016). Additionally, subjects enrolled in the present study were more free-living, with a background diet designed to be low-anthocyanin, as compared to the previous investigation (Cuparencu et al. 2016), which controlled all food intake. This context has implications for the discovery of unique and specific strawberry exposure markers. Many of the key phytochemicals found in strawberries, such as conjugates of catechin, quercetin, and coumaric acid, are found in a variety of other foods

(Rothwell et al. 2013). In the absence of a background diet containing various other fruits and vegetables, many phytochemicals in strawberries may begin to appear as unique biomarkers. However, the ubiquity of these compounds in foods other than strawberries, including those that were not restricted by our protocol, could contribute to urinary levels of their related metabolites. This would then decrease the impact of strawberry exposure on levels of these urinary metabolites in studies with a less restricted background diet.

In the context of our study, strawberries uniquely contain high levels of pelargonidin-based anthocyanins, which would theoretically lead to unique metabolic products as previously observed (Cuparencu et al. 2016). Studies employing targeted analyses have shown that, following a single dose of strawberries, < 2.5% of the delivered pelargonidin is excreted in urine, mainly as phase II metabolites (Azzini et al.

144

2010). Instead of being directly absorbed, anthocyanins are additionally subjected to a complex series of catabolic reactions primarily mediated by gut microbiota. These culminate in the formation of various products, such as lower molecular weight phenolic acids, but these products are also common to other catabolized phenolic compounds

(Williamson & Clifford 2010; Serra et al. 2012). Thus, even these metabolites may not be elevated due to strawberry exposure in the context of a minimally controlled background diet over the course of 24 hours of food consumption (i.e. as seen in 24 hr urine collection), but they may be elevated relative to a control time point. As discussed previously (Cuparencu et al. 2016), this reinforces the usefulness of a cluster of markers, instead of a singular marker, for determining exposure to specific foods. Here, we describe a secondary analysis of a small, pilot study thus the addition of a validation cohort may enhance generalizability of these findings. Nonetheless, the metabolites discussed here provide further insight on the contributions of strawberries to the urinary metabolome.

4.4.2 Identification of Features Elevated with Placebo Consumption

We identified four metabolites elevated in the placebo arm of the study (Table

4.3). Three of these are hypothesized to originate from flavors included in the placebo product. Vanillyl alcohol glucuronide is a phase II metabolite of vanillyl alcohol. This compound is used for flavor in a variety of products, and is a characteristic component of vanilla aroma (Ranadive 1992). We also observed phase II metabolites of raspberry ketone, which appeared to be highly specific to the placebo dose. As the name suggests,

145

this compound is a characteristic aroma compound in raspberries (Borejsza-Wysocki et

al. 1992), which were restricted in the anthocyanin-free background diet used in this

study. This compound was likely included as a component of the berry flavor used in the

placebo formulation. We also identified 1-amino-2-methoxy-5-methyl-4- benzenesulfonic acid in urine from the placebo arm. This compound has been previously identified as a metabolite of the food colorant allura red (FD&C No. 40) (Parkinson &

Brown 1981), which is listed in the ingredient statement of the flavoring used in the placebo formulation. All four of these metabolites support the use of food additive metabolites as potential markers of placebo exposure, and may be useful as biomarkers of compliance to the placebo intervention in placebo-controlled studies.

146

Table 4.3 Identified features elevated in the placebo study arm

RT Chemical Δ Fold VIP Metabolite IDa Ions used for ID (min) formula ppm changeb + 3 0.9 1-amino-2- 218.0496 [M+H] C8H11NO4S 6.4 1.0E 4.32 methoxy-5-methyl- 203.0250 4-benzenesulfonic acid I - 3.2 Vanillyl alcohol 329.0884 [M-H] C14H18O9 1.8 3.0 1.55 glucuronide I 153.0557 - C6H8O6 - 4 5.5 Unknown 341.1236 [M-H] C16H22O8 -0.1 2.2E 5.22 glucuronideIV 165.0928 - C6H8O6 - 3 5.7 Raspberry ketone 339.1082 [M-H] C16H20O8 -0.9 1.3E 4.20 glucuronide I 163.0767 - C6H8O6 - 3 5.8 Raspberry ketone 243.0334 [M-H] C10H12O5S 0.4 8.1E 4.88 I sulfate 163.0763 -SO3 a Level of identification: I= Level 1, IV= Level 4 b Mean fold change compared to the opposite treatment

147

4.5 Conclusions

Here we have described the untargeted metabolomics analysis of urine from a placebo-controlled clinical trial in which smokers and non-smokers consumed strawberry-containing and placebo confections. By using this untargeted approach, we identified compounds that are absorbed, metabolized, and excreted differentially on each arm of the intervention. Identified markers of strawberry exposure compounds that are directly absorbed from the treatment food as well as microbial-derived metabolites of strawberry components. These included phase II metabolites of a strawberry aroma compound, dihydroxybenzoic acids, free and conjugated urolithin A, and a hydroxycinnamic acid. Placebo exposure markers were metabolites of flavorants and colorants used in the placebo formulation. While smoking habits were a significant source of inter-individual variation, no large effect of smoking was observed on the metabolomic profile associated with the strawberry and placebo interventions. These findings will enable us to better conduct similar trials in the future by serving as metabolomic signatures of exposure to both of these products. This study also reinforces the important influence that study design can have on outcomes of untargeted metabolomics analyses. Comparison of ours and another recent study investigating markers of strawberry exposure demonstrated the influence of background diet and timing of specimen collection observed metabolites. Thus together, these findings will inform future clinical trials with strawberry products with regard to potential markers of exposure and implications of study design on metabolomic profiles.

148

References

Aaby, K. et al., 2012. Phenolic compounds in strawberry (Fragaria x ananassa Duch.) fruits: Composition in 27 cultivars and changes during ripening. Food Chemistry, 132(1), pp.86–97. Aaby, K., Skrede, G. & Wrolstad, R.E., 2005. Phenolic composition and antioxidant activities in flesh and achenes of strawberries (fragaria ananassa). Journal of Agricultural and Food Chemistry, 53(10), pp.4032–4040. Ahn-Jarvis, J.H. et al., 2014. Abstract LB-246: Dietary strawberry phytochemical metabolism in saliva, urine, and genetic biomarkers in smokers and non-smokers. Cancer Research, 74(19), p.LB-246. Allais, L. et al., 2015. Chronic cigarette smoke exposure induces microbial and inflammatory shifts and mucin changes in the murine gut. Environmental microbiology. Available at: http://www.ncbi.nlm.nih.gov/pubmed/26033517. Allen, F. et al., 2014. CFM-ID: A web server for annotation, spectrum prediction and metabolite identification from tandem mass spectra. Nucleic Acids Research, 42(W1), pp.94–99. Alvarez-Suarez, J.M. et al., 2014. One-month strawberry-rich anthocyanin supplementation ameliorates cardiovascular risk, oxidative stress markers and platelet activation in humans. Journal of Nutritional Biochemistry, 25, pp.289–294. Amakura, Y. et al., 2000. Influence of jam processing on the radical scavenging activity and phenolic content in berries. Journal of Agricultural and Food Chemistry, 48(12), pp.6292–6297. Amarowicz, R. et al., 2009. Influence of postharvest processing and storage on the content of phenolic acids and flavonoids in foods. Molecular Nutrition and Food Research, 53(S2), pp.151–183. Anitha, P. et al., 2013. Ellagic acid coordinately attenuates Wnt/β-catenin and NF-κB signaling pathways to induce intrinsic apoptosis in an animal model of oral oncogenesis. European Journal of Nutrition, 52(1), pp.75–84. Aron, P.M. & Kennedy, J.A., 2008. Flavan-3-ols: Nature, occurrence and biological activity. Molecular Nutrition and Food Research, 52(1), pp.79–104. Aune, D. et al., 2017. Fruit and vegetable intake and the risk of cardiovascular disease, total cancer and all-cause mortality-A systematic review and dose-response meta- analysis of prospective studies. International Journal of Epidemiology, 46(3), 149

pp.1029–1056. Aura, A.M. et al., 2008. Microbial metabolism of catechin stereoisomers by human faecal microbiota: Comparison of targeted analysis and a non-targeted metabolomics method. Phytochemistry Letters, 1(1), pp.18–22. Azzini, E. et al., 2010. Bioavailability of strawberry antioxidants in human subjects. British Journal of Nutrition, 104(8), pp.1165–1173. Bakkalbaşi, E., Menteş, O. & Artik, N., 2009. Food ellagitannins-occurrence, effects of processing and storage. Critical reviews in food science and nutrition, 49(3), pp.283–98. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19093271 [Accessed November 11, 2014]. Barnes, J.S., Foss, F.W. & Schug, K.A., 2013. Thermally accelerated oxidative degradation of quercetin using continuous flow kinetic electrospray-ion trap-time of flight mass spectrometry. Journal of the American Society for Mass Spectrometry, 24(10), pp.1513–1522. Baskaran, N. et al., 2010. Chemopreventive potential of ferulic acid in 7,12- dimethylbenz[a]anthracene-induced mammary carcinogenesis in Sprague-Dawley rats. European Journal of Pharmacology, 637(1–3), pp.22–29. Available at: http://dx.doi.org/10.1016/j.ejphar.2010.03.054. Basu, A. et al., 2014. Strawberry As a Functional Food: An Evidence-Based Review. Critical Reviews in Food Science and Nutrition, 54(6), pp.790–806. Basu, A. & Penugonda, K., 2009. Pomegranate juice: A heart-healthy fruit juice. Nutrition Reviews, 67(1), pp.49–56. Basu, A., Rhone, M. & Lyons, T.J., 2010. Berries: emerging impact on cardiovascular health. Nutrition reviews, 68(3), pp.168–77. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3068482&tool=pmcentr ez&rendertype=abstract. Bell, C. & Hawthorne, S., 2008. Ellagic acid, pomegranate and prostate cancer -- a mini review. The Journal of pharmacy and pharmacology, 60(2), pp.139–144. van den Berg, R. a et al., 2006. Centering, scaling, and transformations: improving the biological information content of metabolomics data. BMC genomics, 7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16762068. Bieger, J. et al., 2008. Tissue distribution of quercetin in pigs after long-term dietary supplementation. The Journal of nutrition, 138(8), pp.1417–1420. Bishayee, A. et al., 2016. Potential Benefits of Edible Berries in the Management of Aerodigestive and Gastrointestinal Tract Cancers: Preclinical and Clinical Evidence. Critical Reviews in Food Science and Nutrition, 56(10), pp.1753–1775. Bishayee, A. et al., 2015. Potential Benefits of Edible Berries in the Management of Aerodigestive and Gastrointestinal Tract Cancers: Preclinical and Clinical Evidence.

150

Critical reviews in food science and nutrition, 8398(June 2016), p.0. de Boer, V.C.J. et al., 2005. Tissue distribution of quercetin in rats and pigs. The Journal of nutrition, 135(7), pp.1718–1725. Boffetta, P. et al., 2010. Fruit and vegetable intake and overall cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC). J Natl Cancer Inst, 102(8), pp.529–37. Boots, A.W. et al., 2011. Quercetin reduces markers of oxidative stress and inflammation in sarcoidosis. Clinical Nutrition, 30(4), pp.506–512. Available at: http://dx.doi.org/10.1016/j.clnu.2011.01.010. Borejsza-Wysocki, W. et al., 1992. (p-Hydroxyphenyl)butan-2-one Levels in Raspberries Determined by Chromatographic and Organoleptic Methods. Journal of Agricultural and Food Chemistry, 40(7), pp.1176–1177. Brennan, L., 2013. Metabolomics in nutrition research: current status and perspectives. Biochemical Society Transactions, 41(2), pp.670–673. Available at: http://biochemsoctrans.org/lookup/doi/10.1042/BST20120350. Broadhurst, D.I. & Kell, D.B., 2006. Statistical strategies for avoiding false discoveries in metabolomics and related experiments. Metabolomics, 2(4), pp.171–196. Brownmiller, C., Howard, L.R. & Prior, R.L., 2009. Processing and Storage Effects on Procyanidin Composition and Concentration of Processed Blueberry Products. Journal of Agricultural and Food Chemistry, 57, pp.1896–1902. Bruno, R.S. et al., 2006. Faster plasma vitamin E disappearance in smokers is normalized by vitamin C supplementation. Free Radical Biology and Medicine, 40(4), pp.689– 697. Bsoul, S., Huber, M.A. & Terezhalmy, G.T., 2005. Squamous Cell Carcinoma of the Oral Tissues : A Comprehensive Review for Oral Healthcare Providers. The Journal of Contemporary Dental Practice, 6(4), pp.1–16. Buchner, N. et al., 2006. Effect of thermal processing on the flavonols rutin and quercetin. Rapid Communications in Mass Spectrometry, 20, pp.3229–3235. Buendía, B. et al., 2010. HPLC-MS analysis of proanthocyanidin oligomers and other phenolics in 15 strawberry cultivars. Journal of Agricultural and Food Chemistry, 58(7), pp.3916–3926. Buttery, R.G. et al., 2001. Analysis of Furaneol in tomato using dynamic headspace sampling with sodium sulfate. Journal of Agricultural and Food Chemistry, 49(9), pp.4349–4351. Carlsen, M.H. et al., 2010. The total antioxidant content of more than 3100 foods, beverages, spices, herbs and supplements used worldwide. Nutrition Journal, 9(1), pp.1–11. Carvalho, E. et al., 2016. Discovery of A-type procyanidin dimers in yellow raspberries 151

by untargeted metabolomics and correlation based data analysis. Metabolomics, 12(9), p.144. Available at: http://link.springer.com/10.1007/s11306-016-1090-x. Cassidy, A. & Minihane, A.-M., 2017. The role of metabolism (and the microbiome) in defining the clinical efficacy of dietary flavonoids. American Journal of Clinical Nutrition, 105, pp.10–22. Casto, B.C. et al., 2002. Chemoprevention of oral cancer by black raspberries. Anticancer Research, 22(6C), pp.4005–4015. Casto, B.C., Knobloch, T.J. & Weghorst, C.M., 2011. Inhibition of Oral Cancer in Animal Models by Black Raspberries and Berry Components. In G. D. Stoner & N. Seeram, eds. Berries and Cancer Prevention. New York, NY: Springer Science+Business Media, pp. 189–207. Cerdá, B. et al., 2004. The potent in vitro antioxidant ellagitannins from pomegranate juice are metabolised into bioavailable but poor antioxidant hydroxy-6H- dibenzopyran-6-one derivatives by the colonic microflora of healthy humans. European Journal of Nutrition, 43(4), pp.205–220. Cevallos-Cevallos, J.M. et al., 2009. Metabolomic analysis in food science: a review. Trends in Food Science & Technology, 20(11–12), pp.557–566. Available at: http://linkinghub.elsevier.com/retrieve/pii/S092422440900226X [Accessed January 12, 2015]. Chen, T. et al., 2012. Randomized phase II trial of lyophilized strawberries in patients with dysplastic precancerous lesions of the esophagus. Cancer Prevention Research, 5(1), pp.41–50. Chetwynd, A.J. et al., 2016. Use of a pre-analysis osmolality normalisation method to correct for variable urine concentrations and for improved metabolomic analyses. Journal of Chromatography A, 1431, pp.103–110. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0021967315018543. Chong, I.G. & Jun, C.H., 2005. Performance of some variable selection methods when multicollinearity is present. Chemometrics and Intelligent Laboratory Systems, 78(1), pp.103–112. Clifford, M.N. & Scalbert, A., 2000. Review Ellagitannins – nature , occurrence and dietary burden. Journal of the Science of Food and Agriculture, 80(November 1999), pp.1118–1125. Conquer, J. a et al., 1998. Supplementation with quercetin markedly increases plasma quercetin concentration without effect on selected risk factors for heart disease in healthy subjects. The Journal of Nutrition, 128(October 1997), pp.593–597. Considine, E.C. et al., 2018. Critical review of reporting of the data analysis step in metabolomics. Metabolomics, 14(1), p.7. Available at: http://link.springer.com/10.1007/s11306-017-1299-3. Cooke, D. et al., 2005. Anthocyans from fruits and vegetables - Does bright colour signal 152

cancer chemopreventive activity? European Journal of Cancer, 41, pp.1931–1940. Cuparencu, C.S. et al., 2016. Identification of urinary biomarkers after consumption of sea buckthorn and strawberry, by untargeted LC–MS metabolomics: a meal study in adult men. Metabolomics, 12(2), p.31. Available at: http://link.springer.com/10.1007/s11306-015-0934-0. Czank, C. et al., 2013. Human metabolism and elimination of the anthocyanin, cyanidin- 3-glucoside: A 13C-tracer study. American Journal of Clinical Nutrition, 97(5), pp.995–1003. Day, A. et al., 2000. Dietary flavonoid and isoflavone glycosides are hydrolysed by the lactase site of lactase phloridzin hydrolase. FEBS Letters, 468, pp.166–170. Ding, H. et al., 2009. Selective induction of apoptosis of human oral cancer cell lines by avocado extracts via a ROS-mediated mechanism. Nutrition and Cancer, 61(3), pp.348–356. Dossett, M., Lee, J. & Finn, C.E., 2010. Variation in anthocyanins and total phenolics of black raspberry populations. Journal of Functional Foods, 2(4), pp.292–297. Dunn, W.B. et al., 2013. Mass appeal : metabolite identification in mass spectrometry- focused untargeted metabolomics. , pp.44–66. Dunn, W.B. et al., 2013. Mass appeal: Metabolite identification in mass spectrometry- focused untargeted metabolomics. Metabolomics, 9(SUPPL.1), pp.44–66. Dunn, W.B. & Ellis, D.I., 2005. Metabolomics: Current analytical platforms and methodologies. TrAC - Trends in Analytical Chemistry, 24(4), pp.285–294. van Duynhoven, J. et al., 2011. Metabolic fate of polyphenols in the human superorganism. Proceedings of the National Academy of Sciences, 108(Supplement_1), pp.4531–4538. Available at: http://www.pnas.org/cgi/doi/10.1073/pnas.1000098107. Edwards, R.L. et al., 2007. Quercetin Reduces Blood Pressure in Hypertensive Subjects. Journal of Nutrition, 137, pp.2405–2411. Egert, S. et al., 2009. Quercetin reduces systolic blood pressure and plasma oxidised low- density lipoprotein concentrations in overweight subjects with a high-cardiovascular disease risk phenotype: A double-blinded, placebo-controlled cross-over study. British Journal of Nutrition, 102(7), pp.1065–1074. El-Bayoumy, K. et al., 2017. Carcinogenesis of the oral cavity: Environmental causes and potential prevention by black raspberry. Chemical Research in Toxicology, 30(1), pp.126–144. El-Seedi, H.R. et al., 2012. Biosynthesis, natural sources, dietary intake, pharmacokinetic properties, and biological activities of hydroxycinnamic acids. Journal of Agricultural and Food Chemistry, 60(44), pp.10877–10895. Elbourne, D.R. et al., 2002. Meta-analyses involving cross-over trials: Methodological 153

issues. International Journal of Epidemiology, 31, pp.140–149. Erlund, I., 2004. Review of the flavonoids quercetin, hesperetin, and naringenin. Dietary sources, bioactivities, bioavailability, and epidemiology. Nutrition Research, 24(10), pp.851–874. Ferlay, J. et al., 2015. Cancer incidence and mortality worldwide: Sources, methods and major patterns in GLOBOCAN 2012. International Journal of Cancer, 136(5), pp.E359–E386. De Ferrars, R.M. et al., 2014. The pharmacokinetics of anthocyanins and their metabolites in humans. British Journal of Pharmacology, 171(13), pp.3268–3282. Fisher, E.L. et al., 2014. Assessment of physicochemical properties, dissolution kinetics and storage stability of a novel strawberry confection designed for delivery of chemopreventive agents. Food Structure, 1(2), pp.171–181. Available at: http://dx.doi.org/10.1016/j.foostr.2013.10.001. Food and Agriculture Organization of the United Nations, FAOSTAT. Available at: www.faostat3.fao.org [Accessed July 7, 2016]. Francis, F.J., 1989. Food colorants: Anthocyanins, Gasperotti, M. et al., 2015. Overall dietary polyphenol intake in a bowl of strawberries: The influence of Fragaria spp. in nutritional studies. Journal of Functional Foods, 18, pp.1057–1069. Giampieri, F. et al., 2015. Strawberry as a health promoter: an evidence based review. Food & function, 6(5), pp.1386–98. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25803191. Giampieri, F., Alvarez-Suarez, J.M. & Battino, M., 2014. Strawberry and human health: Effects beyond antioxidant activity. Journal of Agricultural and Food Chemistry, 62(18), pp.3867–3876. Giusti, M. & Wrolstad, R.E., 2001. Characterization and Measurement of Anthocyanins by UV-Visible Spectroscopy. In R. E. Wrolstad, ed. Current Protocols in Food Analytical Chemistry. New York: John Wiley. Available at: http://onlinelibrary.wiley.com/doi/10.1002/0471142913.faf0102s00/summary. González-Barrio, R. et al., 2010. Bioavailability of anthocyanins and ellagitannins following consumption of raspberries by healthy humans and subjects with an ileostomy. Journal of Agricultural and Food Chemistry, 58(7), pp.3933–3939. González-Sarrías, A. et al., 2010. Occurrence of urolithins, gut microbiota ellagic acid metabolites and proliferation markers expression response in the human prostate gland upon consumption of walnuts and pomegranate juice. Molecular Nutrition and Food Research, 54(3), pp.311–322. Gromski, P.S. et al., 2015. A tutorial review: Metabolomics and partial least squares- discriminant analysis - a marriage of convenience or a shotgun wedding. Analytica

154

Chimica Acta, 879, pp.10–23. Available at: http://dx.doi.org/10.1016/j.aca.2015.02.012. Gu, J. et al., 2014. Characterization of black raspberry functional food products for cancer prevention human clinical trials. J Agric Food Chem, 62, pp.3997–4006. Gu, L. et al., 2003. Screening of Foods Containing Proanthocyanidins and Their Structural Characterization Using LC-MS/MS and Thiolytic Degradation. Journal of Agricultural and Food Chemistry, 51(25), pp.7513–7521. Guijas, C. et al., 2018. METLIN: A Technology Platform for Identifying Knowns and Unknowns. Analytical Chemistry, (ePub ahead of print). Available at: http://pubs.acs.org/doi/10.1021/acs.analchem.7b04424. Hager, A. et al., 2008. Processing and Storage Effects on Monomeric Anthocyanins, Percent Polymeric Color, and Antioxidant Capacity of Processed Black Raspberry Products. Journal of Food Science, 73(6), pp.H134–H140. Available at: http://doi.wiley.com/10.1111/j.1750-3841.2008.00855.x [Accessed January 22, 2015]. Hager, T.J., Howard, L.R. & Prior, R.L., 2010. Processing and storage effects on the ellagitannin composition of processed blackberry products. Journal of Agricultural and Food Chemistry, 58, pp.11749–11754. Hakkinen, S.H. et al., 2000. Influence of Domestic Processing and Storage on Flavonol Contents in Berries. Journal of Agricultural and Food Chemistry, 48, pp.2960– 2965. Häkkinen, S.H. et al., 1999. Content of the flavonols quercetin, myricetin, and kaempferol in 25 edible berries. Journal of Agricultural and Food Chemistry, 47(6), pp.2274–2279. Hammond, B.R., Wooten, B.R. & Snodderly, D.M., 1996. Cigarette smoking and retinal carotenoids: Implications for age-related macular degeneration. Vision Research, 36(18), pp.3003–3009. Han, C. et al., 2005. Inhibition of the growth of premalignant and malignant human oral cell lines by extracts and components of black raspberries. Nutrition and cancer, 51(2), pp.207–217. Hanahan, D. & Weinberg, R. a, 2011. Hallmarks of cancer: the next generation. Cell, 144(5), pp.646–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21376230 [Accessed July 12, 2012]. Harris, G.K. et al., 2001. Effects of lyophilized black raspberries on azoxymethane- induced colon cancer and 8-hydroxy-2’-deoxyguanosine levels in the Fischer 344 rat. Nutrition and cancer, 40(2), pp.125–133. Hau, D.K.P. et al., 2010. In vivo anti-tumour activity of on Hep3B hepatocellular carcinoma. Phytomedicine, 18(1), pp.11–15. Available at: http://dx.doi.org/10.1016/j.phymed.2010.09.001. 155

He, J. & Giusti, M.M., 2010. Anthocyanins: Natural Colorants with Health-Promoting Properties. Annual Review of Food Science and Technology, 1(1), pp.163–187. Available at: http://www.annualreviews.org/doi/abs/10.1146/annurev.food.080708.100754. Heleno, S.A. et al., 2014. Bioactivity of phenolic acids : metabolites versus parent compounds : a review Centro de Investigação de Montanha , Escola Superior Agrária , Campus de Santa. Food Chemistry, 173, pp.501–513. Available at: http://dx.doi.org/10.1016/j.foodchem.2014.10.057. Hendriks, M.M.W.B.W.B. et al., 2011. Data-processing strategies for metabolomics studies. TrAC - Trends in Analytical Chemistry, 30(10), pp.1685–1698. Available at: http://dx.doi.org/10.1016/j.trac.2011.04.019. Hilt, P. et al., 2003. Detection of phloridzin in strawberries (Fragaria x ananassa Duch.) by HPLC-PDA-MS/MS and NMR spectroscopy. Journal of Agricultural and Food Chemistry, 51(10), pp.2896–2899. Hjartåker, A. et al., 2015. Consumption of berries, fruits and vegetables and mortality among 10,000 Norwegian men followed for four decades. European Journal of Nutrition, 54(4), pp.599–608. Hollman, P.C.H., 2004. Absorption, bioavailability, and metabolism of flavonoids. Pharmaceutical Biology, 42(SUPPL.), pp.74–83. Holt, R.R. et al., 2002. Procyanidin dimer B2 [epicatechin-(4B-8)-epicatechin] in human plasma after the consumption of a flavanol-rich cocoa. American Journal of Clinical Nutrition, 76, pp.798–804. Hooper, L. et al., 2012. Effects of chocolate , cocoa , and flavan-3-ols on cardiovascular health : a systematic review and meta-analysis of randomized trials 1 – 3. The American journal of clinical nutrition, 95(February), pp.740–751. Hou, D.-X. et al., 2003. Anthocyanidins inhibit activator protein 1 activity and cell transformation: structure-activity relationship and molecular mechanisms. Carcinogenesis, 25(1), pp.29–36. Available at: https://academic.oup.com/carcin/article-lookup/doi/10.1093/carcin/bgg184. Hou, D.X. et al., 2005. Anthocyanidins inhibit cyclooxygenase-2 expression in LPS- evoked macrophages: Structure-activity relationship and molecular mechanisms involved. Biochemical Pharmacology, 70(3), pp.417–425. Howard, L.R. et al., 2012. Processing and storage Effect on berry polyphenols: Challenges and implications for bioactive properties. Journal of Agricultural and Food Chemistry, 60(27), pp.6678–6693. Hrydziuszko, O. & Viant, M.R., 2012. Missing values in mass spectrometry based metabolomics: An undervalued step in the data processing pipeline. Metabolomics, 8, pp.161–174. Hummer, K.E., 2010. Rubus pharmacology: Antiquity to the present. HortScience, 156

45(11), pp.1587–1591. Hung, H.-C. et al., 2004. Fruit and Vegetable Intake and Risk of Major Chronic Disease. JNCI Journal of the National Cancer Institute, 96(21), pp.1577–1584. Available at: http://jnci.oxfordjournals.org/cgi/doi/10.1093/jnci/djh296. Ismail, T. et al., 2016. Ellagitannins in Cancer Chemoprevention and Therapy. Toxins, 8(5), p.151. Available at: http://www.mdpi.com/2072-6651/8/5/151. Jaganath, I.B. et al., 2006. The relative contribution of the small and large intestine to the absorption and metabolism of rutin in man. Free Radical Research, 40(10), pp.1035–1046. Jo, Y.-H. et al., 2015. Metabolomic Analysis Reveals Cyanidins in Black Raspberry as Candidates for Suppression of Lipopolysaccharide-Induced Inflammation in Murine Macrophages. Journal of Agricultural and Food Chemistry, 63(22), pp.5449–5458. Available at: http://pubs.acs.org/doi/abs/10.1021/acs.jafc.5b00560. Joseph, S. V., Edirisinghe, I. & Burton-Freeman, B.M., 2014. Berries: Anti-inflammatory effects in humans. Journal of Agricultural and Food Chemistry, 62(18), pp.3886– 3903. Kahle, K. et al., 2007. Polyphenols are intensively metabolized in the human gastrointestinal tract after apple juice consumption. Journal of Agricultural and Food Chemistry, 55(26), pp.10605–10614. Kallner, A., Hartmann, D. & Hornig, D.H. (1981)., 1981. On the requirement of ascorbic acid in man: steady state turnover and body pool in smokers. Am J Clin Nutr, 34, pp.1347–1355. Kårlund, A. et al., 2014. The Impact of Harvesting, Storage and Processing Factors on Health-Promoting Phytochemicals in Berries and Fruits. Processes, 2(3), pp.596– 624. Available at: http://www.mdpi.com/2227-9717/2/3/596/ [Accessed December 4, 2014]. Katajamaa, M. & Orešič, M., 2007. Data processing for mass spectrometry-based metabolomics. Journal of Chromatography A, 1158, pp.318–328. Kay, C.D., Kroon, P.A. & Cassidy, A., 2009. The bioactivity of dietary anthocyanins is likely to be mediated by their degradation products. Molecular Nutrition and Food Research, 53(SUPPL. 1), pp.92–101. Kemsley, E.K. et al., 2007. Multivariate techniques and their application in nutrition: A metabolomics case study. British Journal of Nutrition, 98(1), pp.1–14. Kern, S.M. et al., 2003. Absorption of hydroxycinnamates in humans after high-bran cereal consumption. Journal of Agricultural and Food Chemistry, 51(20), pp.6050– 6055. Key, T.J., 2011. Fruit and vegetables and cancer risk. British Journal of Cancer, 104(1), pp.6–11. Available at: http://dx.doi.org/10.1038/sj.bjc.6606032.

157

Khan, N. et al., 2007. Pomegranate fruit extract inhibits prosurvival pathways in human A549 lung carcinoma cells and tumor growth in athymic nude mice. Carcinogenesis, 28(1), pp.163–173. Kim, H.S. et al., 2011. Biochemical monitoring of black raspberry ( Miquel) fruits according to maturation stage by 1H NMR using multiple solvent systems. Food Research International, 44(7), pp.1977–1987. Available at: http://dx.doi.org/10.1016/j.foodres.2011.01.023. Knobloch, T.J. et al., 2016. Suppression of Proinflammatory and Prosurvival Biomarkers in Oral Cancer Patients Consuming a Black Raspberry Phytochemical-Rich Troche. Cancer Prevention Research, 9(2), pp.159–171. Available at: http://cancerpreventionresearch.aacrjournals.org/cgi/doi/10.1158/1940-6207.CAPR- 15-0187. Kokkinidou, S. & Peterson, D.G., 2013. Response surface methodology as optimization strategy for reduction of reactive carbonyl species in foods by means of phenolic chemistry. Food & Function, 4(7), p.1093. Available at: http://xlink.rsc.org/?DOI=c3fo60032g. Krauze-Baranowska, M. et al., 2014. The antimicrobial activity of fruits from some cultivar varieties of Rubus idaeus and Rubus occidentalis. Food and Function, 5(10), pp.2536–2541. Available at: http://xlink.rsc.org/?DOI=C4FO00129J. Kresty, L.A., Mallery, S.R. & Stoner, G.D., 2016. Black raspberries in cancer clinical trials: Past, present and future. Journal of Berry Research, 6(2), pp.251–261. Kula, M. et al., 2016. Phenolic composition of fruits from different cultivars of red and black raspberries grown in Poland. Journal of Food Composition and Analysis, 52, pp.74–82. Kula, M. & Krauze-Baranowska, M., 2016. Rubus occidentalis: The black raspberry—its potential in the prevention of cancer. Nutrition and Cancer, 68(1), pp.18–28. Available at: https://doi.org/10.1080/01635581.2016.1115095. Lafay, S. & Gil-Izquierdo, A., 2008. Bioavailability of phenolic acids. Phytochemistry Reviews, 7(2), pp.301–311. Lampe, J.W. & Chang, J.-L., 2007. Interindividual differences in phytochemical metabolism and disposition. Seminars in Cancer Biology, 17(5), pp.347–353. Landete, J.M., 2011. Ellagitannins, ellagic acid and their derived metabolites: A review about source, metabolism, functions and health. Food Research International, 44(5), pp.1150–1160. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0963996911002572 [Accessed November 5, 2014]. Larmo, P.S. et al., 2013. Effects of sea buckthorn and bilberry on serum metabolites differ according to baseline metabolic profiles in overweight women: A randomized crossover trial. American Journal of Clinical Nutrition, 98(4), pp.941–951. 158

Larrosa, M. et al., 2010. Anti-inflammatory properties of a pomegranate extract and its metabolite urolithin-A in a colitis rat model and the effect of colon inflammation on phenolic metabolism. Journal of Nutritional Biochemistry, 21(8), pp.717–725. Available at: http://dx.doi.org/10.1016/j.jnutbio.2009.04.012. Lee, H. et al., 1997. Ineffectiveness of the presence of H-ras / p53 combination of mutations in squamous cell carcinoma cells to induce a conversion of a nontumorigenic to a tumorigenic phenotype. Medical Biochemistry, 13, pp.419–434. Lee, J. & Mitchell, A.E., 2012. Pharmacokinetics of quercetin absorption from apples and onions in healthy humans. Journal of Agricultural and Food Chemistry, 60(15), pp.3874–3881. Levsen, K. et al., 2007. Even-electron ions: a systematic study of the neutral species lost in the dissociation of quasi-molecular ions. Journal of Mass Spectrometry, 42, pp.1024–1044. Levsen, K. et al., 2005. Structure elucidation of phase II metabolites by tandem mass spectrometry: An overview. Journal of Chromatography A, 1067(1–2), pp.55–72. Li, D. et al., 2017. Health benefits of anthocyanins and molecular mechanisms: Update from recent decade. Critical Reviews in Food Science and Nutrition, 57(8), pp.1729–1741. Available at: http://dx.doi.org/10.1080/10408398.2015.1030064. Li, S. et al., 2015. Efficacy of procyanidins against in vivo cellular oxidative damage: A systematic review and meta-analysis. PLoS ONE, 10(10), pp.1–14. Liu, R.H., 2003. Health benefits of fruit and vegetables are from additive and synergistic combinations of phytochemicals. American Journal of Clinical Nutrition, 78(suppl), p.517S–520S. Loypimai, P., Moongngarm, A. & Chottanom, P., 2016. Thermal and pH degradation kinetics of anthocyanins in natural food colorant prepared from black rice bran. Journal of Food Science and Technology, 53(1), pp.461–470. Luceri, C. et al., 2007. p-Coumaric acid, a common dietary phenol, inhibits platelet activity in vitro and in vivo. British Journal of Nutrition, 97(3), pp.458–463. Lykkesfeldt, J. et al., 2000. Ascorbate is depleted by smoking and repleted by moderate supplementation: a study in male smokers and nonsmokers with matched dietary antioxidant intakes. The American journal of clinical nutrition, 71(2), pp.530–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10648268. Määttä-Riihinen, K.R., Kamal-Eldin, A. & Törrönen, A.R., 2004. Identification and quantification of phenolic compounds in berries of Fragaria and Rubus species (family ). Journal of Agricultural and Food Chemistry, 52(20), pp.6178– 6187. Mallery, S.R. et al., 2007. Formulation and In-Vitro and In-Vivo Evaluation of a Mucoadhesive Gel Containing Freeze Dried Black Raspberries: Implications for Oral Cancer Chemoprevention. Pharmaceutical Research, 24(4), pp.728–737. 159

Mallery, S.R. et al., 2008. Topical application of a bioadhesive black raspberry gel modulates gene expression and reduces cyclooxygenase 2 protein in human premalignant oral lesions. Cancer Research, 68(12), pp.4945–4957. Mallery, S.R. et al., 2014. Topical application of a mucoadhesive freeze-dried black raspberry gel induces clinical and histologic regression and reduces loss of heterozygosity events in premalignant oral intraepithelial lesions: results from a multicentered, placebo-controlled clin. Clinical cancer research : an official journal of the American Association for Cancer Research, 20(7), pp.1910–24. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24486592 [Accessed January 22, 2015]. Manach, C. et al., 2005. Bioavailability and bioefficacy of polyphenols in humans. I. Review of 93 bioavailability studies. The American Journal of Clinical Nutrition, 81, p.243S–255S. Manach, C. et al., 2009. The complex links between dietary phytochemicals and human health deciphered by metabolomics. Molecular Nutrition and Food Research, 53(10), pp.1303–1315. Mandal, S. & Stoner, G.O., 1990. Inhibition of N-nitrosobenzylmethylamine-induced esophageal tumorigenesis in rats by ellagic acid. Carcinogenesis, 11, pp.55–61. Markakis, P. & Jurd, L., 1974. Anthocyanins and their stability in foods. Critical Reviews in Food Technology, 4(4), pp.437–456. Mattila, P., Hellström, J. & Törrönen, R., 2006. Phenolic acids in berries, fruits, and beverages. Journal of Agricultural and Food Chemistry, 54(19), pp.7193–7199. Miles, S.L., Mcfarland, M. & Niles, R.M., 2014. Molecular and physiological actions of quercetin: Need for clinical trials to assess its benefits in human disease. Nutrition Reviews, 72(11), pp.720–734. Mills, C.E. et al., 2017. Mediation of coffee-induced improvements in human vascular function by chlorogenic acids and its metabolites: Two randomized, controlled, crossover intervention trials. Clinical Nutrition, 36, pp.1520–1529. Moco, S., Martin, F.P.J. & Rezzi, S., 2012. Metabolomics view on gut microbiome modulation by polyphenol-rich foods. Journal of Proteome Research, 11(10), pp.4781–4790. El Mohsen, M.A. et al., 2006. Absorption, tissue distribution and excretion of pelargonidin and its metabolites following oral administration to rats. British Journal of Nutrition, 95(1), p.51. Available at: http://www.journals.cambridge.org/abstract_S0007114506000079. Monagas, M. et al., 2010. Insights into the metabolism and microbial biotransformation of dietary flavan-3-ols and the bioactivity of their metabolites. Food & function, 1(3), pp.233–253. Monteiro, M. et al., 2007. Chlorogenic acid compounds from coffee are differentially absorbed and metabolized in humans. The Journal of nutrition, 137(10), pp.2196– 160

2201. Moyer, R.A. et al., 2002. Anthocyanins, phenolics, and antioxidant capacity in diverse small fruits: Vaccinium, Rubus, and Ribes. Journal of Agricultural and Food Chemistry, 50(3), pp.519–525. Mullen, W. et al., 2003. Analysis of ellagitannins and conjugates of ellagic acid and quercetin in raspberry fruits by LC–MSn. Phytochemistry, 64(2), pp.617–624. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0031942203002814 [Accessed January 5, 2015]. Naz, S. et al., 2014. Method validation strategies involved in non-targeted metabolomics. Journal of Chromatography A, 1353, pp.99–105. Available at: http://dx.doi.org/10.1016/j.chroma.2014.04.071. Németh, K. et al., 2003. Deglycosylation by small intestinal epithelial cell β-glucosidases is a critical step in the absorption and metabolism of dietary flavonoid glycosides in humans. European Journal of Nutrition, 42(1), pp.29–42. Norat, T. et al., 2014. Fruits and Vegetables: Updatring the Epidemiologic Evidence for the WCRF/AICR Lifestyle Recommendations for Cancer Prevention. In V. Zappia et al., eds. Advances in Nutrition and Cancer. Cancer Treatment and Research. Springer, Berlin, Heidelberg, pp. 35–50. Oghumu, S. et al., 2017. Inhibition of Pro-inflammatory and Anti-apoptotic Biomarkers during Experimental Oral Cancer Chemoprevention by Dietary Black Raspberries. Frontiers in Immunology, 8(October). Available at: http://journal.frontiersin.org/article/10.3389/fimmu.2017.01325/full. Ottaviani, J.I. et al., 2012. Intake of dietary procyanidins does not contribute to the pool of circulating flavanols in humans. American Journal of Clinical Nutrition, 95(4), pp.851–858. Pan, Z. & Raftery, D., 2007. Comparing and combining NMR spectroscopy and mass spectrometry in metabolomics. Analytical and Bioanalytical Chemistry, 387(2), pp.525–527. Park, J.B., 2009. 5-Caffeoylquinic acid and caffeic acid orally administered suppress P- selectin expression on mouse platelets. Journal of Nutritional Biochemistry, 20(10), pp.800–805. Parkinson, T.M. & Brown, J.P., 1981. Metabolic fate of food colorants. Ann. Rev. Nutr., 1, pp.175–205. Available at: http://www.ncbi.nlm.nih.gov/pubmed/6764715. De Pascual-Teresa, S., 2014. Molecular mechanisms involved in the cardiovascular and neuroprotective effects of anthocyanins. Archives of Biochemistry and Biophysics, 559, pp.68–74. Available at: http://dx.doi.org/10.1016/j.abb.2014.04.012. Patras, A., Brunton, N.P., et al., 2010. Effect of thermal processing on anthocyanin stability in foods ; mechanisms and kinetics of degradation. , 21.

161

Patras, A., Brunton, N.P., et al., 2010. Effect of thermal processing on anthocyanin stability in foods; mechanisms and kinetics of degradation. Trends in Food Science and Technology, 21(1), pp.3–11. Paudel, L. et al., 2014. NMR-Based Metabolomic Investigation of Bioactivity of Chemical Constituents in Black Raspberry ( Rubus occidentalis L.) Fruit Extracts. Journal of agricultural and food chemistry, 62, pp.1989–1998. Paudel, L. et al., 2013. Nonanthocyanin secondary metabolites of black raspberry (Rubus occidentalis L.) Fruits: Identification by HPLC-DAD, NMR, HPLC-ESI-MS, and ESI-MS/MS Analyses. Journal of Agricultural and Food Chemistry, 61, pp.12032– 12043. Pei, K. et al., 2016. p-Coumaric acid and its conjugates: Dietary sources, pharmacokinetic properties and biological activities. Journal of the Science of Food and Agriculture, 96(9), pp.2952–2962. Petersen, P.E. et al., 2005. The global burden of oral diseases and risks to oral health. Bulletin of the World Health Organization, 83(9), pp.661–669. Pickenhagen, W. et al., 1981. Estimation of 2,5‐dimethyl‐ 4‐ hydroxy‐ 3(2H)‐ furanone (FURANEOLA®) in cultivated and wild strawberries, pineapples and mangoes. Journal of the Science of Food and Agriculture, 32(11), pp.1132–1134. Pojer, E. et al., 2013. The case for anthocyanin consumption to promote human health: A review. Comprehensive Reviews in Food Science and Food Safety, 12(5), pp.483– 508. Quideau, S. & Feldman, K.S., 1996. Ellagitannin Chemistry. Chemical Reviews, 96(1), pp.475–504. Available at: http://pubs.acs.org/doi/abs/10.1021/cr940716a. Ranadive, A.S., 1992. Vanillin and Related Flavor Compounds in Vanilla Extracts Made from Beans of Various Global Origins. Journal of Agricultural and Food Chemistry, 40(10), pp.1922–1924. Rathahao-Paris, E. et al., 2016. High resolution mass spectrometry for structural identification of metabolites in metabolomics. Metabolomics, 12(1), p.10. Available at: http://link.springer.com/10.1007/s11306-015-0882-8. Rein, M.J. et al., 2005. Identification of novel pyranoanthocyanins in berry juices. European Food Research and Technology, 220(3–4), pp.239–244. Rein, M.J. & Heinonen, M., 2004. Stability and Enhancement of Berry Juice Color. Journal of Agricultural and Food Chemistry, 52(10), pp.3106–3114. Rezzi, S. et al., 2007. Nutritional metabonomics: Applications and perspectives. Journal of Proteome Research, 6(2), pp.513–525. Del Rio, D. et al., 2013. Dietary (Poly)phenolics in Human Health: Structures, Bioavailability, and Evidence of Protective Effects Against Chronic Diseases. Antioxidants & Redox Signaling, 18(14), pp.1818–1892. Available at:

162

http://online.liebertpub.com/doi/abs/10.1089/ars.2012.4581. Del Rio, D. et al., 2010. Polyphenols and health: What compounds are involved? Nutrition, Metabolism and Cardiovascular Diseases, 20(1), pp.1–6. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0939475309001379. Rios, L.Y. et al., 2003. Chocolate intake increases urinary excretion of polyphenol- derived phenolic acids in healthy human subjects. The American journal of clinical nutrition, 77(4), pp.912–918. Robbins, R.J., 2003. Phenolic acids in foods: An overview of analytical methodology. Journal of Agricultural and Food Chemistry, 51(10), pp.2866–2887. Rodrigo, K. a et al., 2006. Suppression of the tumorigenic phenotype in human oral squamous cell carcinoma cells by an ethanol extract derived from freeze-dried black raspberries. Nutrition and cancer, 54(1), pp.58–68. Rodriguez-Mateos, A., Heiss, C., et al., 2014. Berry (poly)phenols and cardiovascular health. Journal of Agricultural and Food Chemistry, 62(18), pp.3842–3851. Rodriguez-Mateos, A., Vauzour, D., et al., 2014. Bioavailability, bioactivity and impact on health of dietary flavonoids and related compounds: an update. Archives of Toxicology, 88(10), pp.1803–1853. Rodriguez-Mateos, A., del Pino-García, R., et al., 2014. Impact of processing on the bioavailability and vascular effects of blueberry (poly)phenols. Molecular Nutrition and Food Research, 58(10), pp.1952–1961. Ronningen, I. et al., 2018. Identification and Validation of Sensory-Active Compounds from Data-Driven Research: A Flavoromics Approach. Journal of Agricultural and Food Chemistry, ePub ahead. Available at: http://pubs.acs.org/doi/abs/10.1021/acs.jafc.7b00093. Ronningen, I. & Peterson, D.G., 2018. Identification of Aging-Associated Food Quality Changes in Citrus Products Using Untargeted Chemical Profiling. Journal of Agricultural and Food Chemistry, 66, pp.682–688. Available at: http://pubs.acs.org/doi/10.1021/acs.jafc.7b04450. Ross, H. a., McDougall, G.J. & Stewart, D., 2007. Antiproliferative activity is predominantly associated with ellagitannins in raspberry extracts. Phytochemistry, 68, pp.218–228. Rothwell, J.A. et al., 2013. Phenol-Explorer 3.0: A major update of the Phenol-Explorer database to incorporate data on the effects of food processing on polyphenol content. Database, 2013, pp.1–8. Rubert, J., Zachariasova, M. & Hajslova, J., 2015. Advances in high-resolution mass spectrometry based on metabolomics studies for food – a review. Food Additives & Contaminants: Part A, 32(10), pp.1685–1708. Available at: http://www.tandfonline.com/doi/full/10.1080/19440049.2015.1084539.

163

Russell, W.R. et al., 2009. Selective bio-availability of phenolic acids from Scottish strawberries. Molecular Nutrition and Food Research, 53(SUPPL. 1), pp.85–91. Saccenti, E. et al., 2014. Reflections on univariate and multivariate analysis of metabolomics data. Metabolomics, 10(3), pp.361–374. Sadilova, E., Carle, R. & Stintzing, F.C., 2007. Thermal degradation of anthocyanins and its impact on color and in vitro antioxidant capacity. Molecular Nutrition and Food Research, 51(12), pp.1461–1471. Sadilova, E., Stintzing, F.C. & Carle, R., 2006. Thermal degradation of acylated and nonacylated anthocyanins. Journal of Food Science, 71(8). Saura-Calixto, F. et al., 2010. Proanthocyanidin metabolites associated with dietary fibre from in vitro colonic fermentation and proanthocyanidin metabolites in human plasma. Molecular Nutrition and Food Research, 54(7), pp.939–946. Scalbert, A. et al., 2009. Mass-spectrometry-based metabolomics: Limitations and recommendations for future progress with particular focus on nutrition research. Metabolomics, 5(4), pp.435–458. Schmidt, B.., Erdman, J.. & Lila, M.., 2005. Effects of Food Processing on Blueberry Antiproliferation and Antioxidant Activity. Sensory and Nutritive Qualities of Food, 70(6), pp.389–394. Schwartz, S.J. et al., 2017. Colorants. In S. Damodaran & K. L. Parkin, eds. Fennema’s Food Chemistry. Boca Raton, FL: CRC Press, pp. 681–752. Seeram, N. et al., 2006. Blackberry , Black Raspberry , Blueberry , Cranberry , Red Raspberry , and Strawberry Extracts Inhibit Growth and Stimulate Apoptosis of Human Cancer Cells In Vitro. Journal of agricultural and food chemistry, 54, pp.9329–9339. Seeram, N.P., 2008. Berry fruits: Compositional elements, biochemical activities, and the impact of their intake on human health, performance, and disease. Journal of Agricultural and Food Chemistry, 56, pp.627–629. Seeram, N.P., 2006. Bioactive Polyphenols from Foods and Dietary Supplements: Challenges and Opportunities. In M. Wang et al., eds. Herbs: Challenges in Chemistry and Biology. American Chemical Society, pp. 25–38. Available at: http://dx.doi.org/10.1021/bk-2006- 0925.ch003%5Cnhttp://pubs.acs.org/doi/abs/10.1021/bk-2006-0925.ch003. Seeram, N.P. et al., 2006. Identification of phenolic compounds in strawberries by liquid chromatography electrospray ionization mass spectroscopy. Food Chemistry, 97(1), pp.1–11. Seeram, N.P. et al., 2007. -derived metabolites inhibit prostate cancer growth and localize to the mouse prostate gland. Journal of Agricultural and Food Chemistry, 55(19), pp.7732–7737.

164

Seeram, N.P. & Heber, D., 2007. Impact of berry phytochemicals on human health: Effects beyond antioxidation. In Antioxidant Measurement and Applications. American Chemical Society, pp. 326–336. Available at: http://dx.doi.org/10.1021/bk-2007- 0956.ch021%5Cnhttp://pubs.acs.org/doi/book/10.1021/bk-2007-0956. Serra, A. et al., 2012. Metabolic pathways of the colonic metabolism of flavonoids (flavonols, flavones and flavanones) and phenolic acids. Food Chemistry, 130(2), pp.383–393. Serrano, J. et al., 2009. Review Tannins : Current knowledge of food sources , intake , bioavailability and biological effects. , pp.310–329. Serrano, J. et al., 2009. Tannins: Current knowledge of food sources, intake, bioavailability and biological effects. Molecular Nutrition and Food Research, 53(SUPPL. 2), pp.310–329. Shahrzad, S. et al., 2001. Pharmacokinetics of Gallic Acid and Its Relative Bioavailability from Tea in Healthy Humans. J. Nutr., 131(4), pp.1207–1210. Available at: http://jn.nutrition.org/content/131/4/1207.long. Shumway, B.S. et al., 2008. Effects of a Topically Applied Bioadhesive Berry Gel on Loss of Heterozygosity Indices in Premalignant Oral Lesions. Clinical Cancer Research, 14(8), pp.2421–2430. Siegel, R.L., Miller, K.D. & Jemal, A., 2018. Cancer statistics, 2018. CA: A Cancer Journal for Clinicians, 68(1), pp.7–30. Available at: http://doi.wiley.com/10.3322/caac.21442. Van Der Sluis, A.A., Dekker, M. & Van Boekel, M.A.J.S., 2005. Activity and concentration of polyphenolic antioxidants in apple juice. 3. Stability during storage. Journal of Agricultural and Food Chemistry, 53(4), pp.1073–1080. Srivastava, A. et al., 2007. Effect of storage conditions on the biological activity of phenolic compounds of blueberry extract packed in glass bottles. Journal of Agricultural and Food Chemistry, 55(7), pp.2705–2713. Stintzing, F.C. & Carle, R., 2004. Functional properties of anthocyanins and in plants, food, and in human nutrition. Trends in Food Science and Technology, 15(1), pp.19–38. Stoner, G.D. et al., 2010. Bioactive Compounds and Cancer. In J. A. Milner & D. F. Romagnolo, eds. Bioactive Compounds and Cancer. New York, NY: Humana Press, pp. 703–723. Available at: http://link.springer.com/10.1007/978-1-60761-627-6 [Accessed January 22, 2015]. Stoner, G.D. et al., 2007. Cancer prevention with freeze-dried berries and berry components. Seminars in Cancer Biology, 17(5), pp.403–410. Stoner, G.D., 2009. Foodstuffs for preventing cancer: the preclinical and clinical development of berries. Cancer prevention research, 2(3), pp.187–94. Available at: 165

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2769015&tool=pmcentr ez&rendertype=abstract [Accessed January 22, 2015]. Stoner, G.D. et al., 2005. Pharmacokinetics of anthocyanins and ellagic acid in healthy volunteers fed freeze-dried black raspberries daily for 7 days. Journal of clinical pharmacology, 45, pp.1153–1164. Stoner, G.D., Wang, L.-S. & Casto, B.C., 2008. Laboratory and clinical studies of cancer chemoprevention by antioxidants in berries. Carcinogenesis, 29(9), pp.1665–74. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3246882&tool=pmcentr ez&rendertype=abstract [Accessed January 22, 2015]. Sumner, L.W. et al., 2007. Proposed minimum reporting standards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). Metabolomics, 3(3), pp.211–221. Szymanska, E. et al., 2012. Double-check: Validation of diagnostic statistics for PLS-DA models in metabolomics studies. Metabolomics, 8, pp.3–16. Talcott, S.T. & Krenek, K. a, 2012. Analysis Methods of Ellagitannins. In Z. Xu & L. R. Howard, eds. Analysis of Antioxidant-Rich Phytochemicals. John Wiley & Sons, pp. 181–205. Tanaka, T., Tanaka, T. & Tanaka, M., 2011. Potential Cancer Chemopreventive Activity of Protocatechuic Acid. Journal of Experimental and Clinical Medicine, 3(1), pp.27–33. Thévenot, E.A. et al., 2015. Analysis of the Human Adult Urinary Metabolome Variations with Age, Body Mass Index, and Gender by Implementing a Comprehensive Workflow for Univariate and OPLS Statistical Analyses. Journal of Proteome Research, 14(8), pp.3322–3335. Tian, Q. et al., 2006. Characterization of a new anthocyanin in black raspberries (Rubus occidentalis) by liquid chromatography electrospray ionization tandem mass spectrometry. Food Chemistry, 94(3), pp.465–468. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0308814605000981 [Accessed January 6, 2015]. Tinsley, I.J. & Bockian, A.H., 1960. Some effects of sugars on the breakdown of pelargonidin-3-glucoside in model systems at 90 C. Journal of Food Science, 25(2), pp.161–173. Tiwari, U. & Cummins, E., 2013. Factors influencing levels of phytochemicals in selected fruit and vegetables during pre- and post-harvest food processing operations. Food Research International, 50(2), pp.497–506. Available at: http://dx.doi.org/10.1016/j.foodres.2011.09.007. Tokitomo, Y. et al., 2005. Odor-Active Constituents in Fresh Pineapple ( Ananas comosus [L.] Merr.) by Quantitative and Sensory Evaluation. Bioscience, 166

Biotechnology, and Biochemistry, 69(7), pp.1323–1330. Available at: http://www.tandfonline.com/doi/full/10.1271/bbb.69.1323. Tomas-Barberan, F.A. et al., 2017. Urolithins, the rescue of “old” metabolites to understand a “new” concept: Metabotypes as a nexus among phenolic metabolism, microbiota dysbiosis, and host health status. Molecular Nutrition and Food Research, 61(1). Torre, L.C. & Barritt, B.H., 1977. Quantitative Evaluation of Rubus Fruit Anthocyanin Pigments. Journal of Food Science, 42(2), pp.488–490. Totlani, V.M. & Peterson, D.G., 2005. Reactivity of epicatechin in aqueous glycine and glucose Maillard reaction models: Quenching of C 2, C 3, and C 4 sugar fragments. Journal of Agricultural and Food Chemistry, 53(10), pp.4130–4135. Truchado, P. et al., 2012. Strawberry Processing Does Not Affect the Production and Urinary Excretion of Urolithins, Ellagic Acid Metabolites, in Humans. Journal of Agricultural and Food Chemistry, 60, pp.5749–5754. Tulio, A.Z. et al., 2008. Cyanidin 3-rutinoside and cyanidin 3-xylosylrutinoside as primary phenolic antioxidants in black raspberry. Journal of Agricultural and Food Chemistry, 56(6), pp.1880–1888. Tulipani, S. et al., 2008. Antioxidants , Phenolic Compounds , and Nutritional Quality of Different Strawberry Genotypes Quality of Different Strawberry Genotypes. Journal of Agricultural and Food Chemistry, 56, pp.696–704. Tulipani, S. et al., 2011. Influence of environmental and genetic factors on health-related compounds in strawberry. Food Chemistry, 124(3), pp.906–913. United States Department of Agriculture, National Agricultural Statistics Service. Available at: www.nass.usda.gov. United States Department of Agriculture, 2013. USDA Database for the Flavonoid Content of Selected Foods, Release 3.1. Available at: http://www.ars.usda.gov/News/docs.htm?docid=6231. United States Department of Agriculture, 2004. USDA Database for the Proanthocyanidin Content of Selected Foods, Beltsville, MD. Available at: https://www.ars.usda.gov/ARSUserFiles/80400525/data/pa/pa.pdf. United States Department of Agriculture, USDA National Nutrient Database for Standard Reference. Available at: http://ndb.nal.usda.gov/. United States Department of Health and Human Services, 2014. The Health Consequences of Smoking—50 Years of Progress: a Report of the Surgeon General, Atlanta, GA. Van Velzen, E.J.J. et al., 2008. Multilevel data analysis of a crossover designed human nutritional intervention study. Journal of Proteome Research, 7(10), pp.4483–4491. Vidya Priyadarsini, R. et al., 2012. Gene expression signature of DMBA-induced hamster 167

buccal pouch carcinomas: modulation by chlorophyllin and ellagic acid. PloS one, 7(4), pp.1–9. Vrhovsek, U. et al., 2006. Concentration and mean degree of polymerization of Rubus ellagitannins evaluated by optimized acid methanolysis. Journal of Agricultural and Food Chemistry, 54(12), pp.4469–4475. Wada, L. & Ou, B., 2002. Antioxidant activity and phenolic content of Oregon caneberries. Journal of Agricultural and Food Chemistry, 50(12), pp.3495–3500. Wallace, T.C., Slavin, M. & Frankenfeld, C.L., 2016. Systematic review of anthocyanins and markers of cardiovascular disease. Nutrients, 8(1), pp.1–13. Walle, T., 2004. Absorption and metabolism of flavonoids. Free Radical Biology and Medicine, 36(7), pp.829–837. Walle, T. et al., 2005. Flavonoid glucosides are hydrolyzed and thus activated in the oral cavity in humans. The Journal of nutrition, 135(1), pp.48–52. Wang, L. et al., 2009. Anthocyanins in Black Raspberries Prevent Esophageal Tumors in Rats. Cancer prevention research, 2(1), pp.84–93. Want, E.J. et al., 2010. Global metabolic profiling procedures for urine using UPLC-MS. Nature protocols, 5(6), pp.1005–1018. Available at: http://dx.doi.org/10.1038/nprot.2010.50. Ward, N.C. et al., 2016. Acute effects of chlorogenic acids on endothelial function and blood pressure in healthy men and women. Food & Function, 7, pp.2197–2203. Available at: www.rsc.org/foodfunction. Warner, B.M. et al., 2014. Chemoprevention of oral cancer by topical application of black raspberries on high at-risk mucosa. Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology, 118(6), pp.674–683. Available at: http://dx.doi.org/10.1016/j.oooo.2014.09.005. Wei, W., Kim, Y. & Boudreau, N., 2001. Association of smoking with serum and dietary levels of antioxidants in adults: NHANES III, 1988-1994. American journal of public health, 91(2), pp.258–64. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1446535&tool=pmcentr ez&rendertype=abstract. Westerhuis, J.A. et al., 2008. Assessment of PLSDA cross validation. Metabolomics, 4(1), pp.81–89. Westerhuis, J.A. et al., 2010. Multivariate paired data analysis: Multilevel PLSDA versus OPLSDA. Metabolomics, 6(1), pp.119–128. White, B.L., Howard, L.R. & Prior, R.L., 2011. Impact of different stages of juice processing on the anthocyanin, flavonol, and procyanidin contents of cranberries. Journal of Agricultural and Food Chemistry, 59(9), pp.4692–4698. Williamson, G. & Clifford, M.N., 2010. Colonic metabolites of berry polyphenols: The 168

missing link to biological activity? British Journal of Nutrition, 104(SUPPL.3), pp.48–66. Williamson, G. & Manach, C., 2005. Bioavailability and bioefficacy of polyphenols in humans. II. Review of 93 intervention studies. Am J Clin Nutr, 81, p.243S–255S. Wishart, D.S. et al., 2013. HMDB 3.0-The Human Metabolome Database in 2013. Nucleic Acids Research, 41(D1), pp.801–807. Wishart, D.S., 2008. Metabolomics: applications to food science and nutrition research. Trends in Food Science & Technology, 19(9), pp.482–493. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0924224408000770 [Accessed October 16, 2014]. Worley, B. & Powers, R., 2012. Multivariate Analysis in Metabolomics. Current Metabolomics, 1(1), pp.92–107. Available at: http://www.eurekaselect.com/openurl/content.php?genre=article&issn=2213- 235X&volume=1&issue=1&spage=92. Wu, W.-N. & McKown, L.A., 2014. In Vitro Drug Metabolite Profiling Using Hepatic S9 and Human Liver Microsomes. In Z. Yan & G. W. Caldwell, eds. Methods in Pharmacology and Toxicology. Totowa, NJ: Humana Press Inc., pp. 163–184. Wu, X. et al., 2006. Concentrations of anthocyanins in common foods in the United States and estimation of normal consumption. Journal of Agricultural and Food Chemistry, 54, pp.4069–4075. Xia, J. et al., 2012. MetaboAnalyst 2.0-a comprehensive server for metabolomic data analysis. Nucleic Acids Research, 40(W1), pp.127–133. Xiao, H. & Ho, C.-T., 2017. Bioactive Food Components. In S. Damodaran & K. L. Parkin, eds. Fennema’s Food Chemistry. Boca Raton, FL: CRC Press, pp. 865–903. Yabe, T. et al., 2010. Ferulic acid induces neural progenitor cell proliferation in vitro and in vivo. Neuroscience, 165(2), pp.515–524. Available at: http://dx.doi.org/10.1016/j.neuroscience.2009.10.023. Yang, B. & Kortesniemi, M., 2015. Clinical evidence on potential health benefits of berries. Current Opinion in Food Science, 2, pp.36–42. Available at: http://dx.doi.org/10.1016/j.cofs.2015.01.002. Yang, G.W., Jiang, J.S. & Lu, W.Q., 2015. Ferulic acid exerts anti-angiogenic and anti- tumor activity by targeting fibroblast growth factor receptor 1-mediated angiogenesis. International Journal of Molecular Sciences, 16(10), pp.24011– 24031. Zafrilla, P., Ferreres, F. & Tomás-Barberán, F.A., 2001. Effect of processing and storage on the antioxidant ellagic acid derivatives and flavonoids of red raspberry (Rubus idaeus) jams. Journal of Agricultural and Food Chemistry, 49(8), pp.3651–3655. Zhang, L. et al., 2016. The absorption, distribution, metabolism and excretion of

169

procyanidins. Food and Function, 7(3), pp.1273–1281. Available at: http://xlink.rsc.org/?DOI=C5FO01244A. Zhang, Y.J. et al., 2008. Isolation and identification of strawberry phenolics with antioxidant and human cancer cell anti proliferative properties. Journal of Agricultural and Food Chemistry, 56, pp.670–675.

170