Chapter 1: Bioactive Compounds of Berries: Chemistry and Analytical Methods of Detection

Due to the diversity of (poly)phenol structures and their localization within a fruit morphology, and those modified as a result of human or microbial enzymes, multiple extraction and separation techniques have been developed and reported,^1–3  including spectrophotometry, gas chromatography, liquid chromatography, thin-layer chromatography and capillary electrophoresis. Similarly, there are a diverse array of techniques utilized for identification and quantitation, including flame ionization, thermal conductivity, near-infrared, mass, ultraviolet–visible (UV–Vis) and fluorescence spectrometry, light scattering, refractive index, as well as nuclear magnetic resonance (NMR) spectroscopy. No single technique is ideal for resolving, identifying and quantitating all subclasses of berry (poly)phenols, which comprise low-molecular-weight phenols (i.e., benzoic acids) to large polymers (i.e., procyanidins, hydrolyzable tannins, etc.). Dozens of textbooks, book chapters and review articles over the years have been devoted to flavonoid/(poly)phenol characterization in plants and animals, notably The Systematic Identification of Flavonoids,⁴  The Flavonoids: Advances in Research since 1980,⁵  a series of articles in Current Protocols in Food and Analytical Chemistry,^6–8  and Methods in Polyphenol Analysis.⁹  The textbook, Methods in Polyphenol Analysis⁹  describes in considerable detail methods for polyphenol extraction, identification and quantification across food and biological matrices. The text covers (poly)phenol extraction from foods, biological fluids and tissues, liquid chromatography–mass spectrometry (LC–MS) identification practices, high pressure/high-performance liquid chromatography (HPLC) with coulometric detection, application of liquid chromatography–nuclear magnetic resonance spectroscopy (LC–NMR) in structure elucidation and purification and quantitation of phenolic, (poly)phenols and procyanidins. More modern, notable texts and reviews include Photocatalysis – Fundamentals, Materials and Potential;¹⁰  Analytical Methods of Phenolic Compounds,¹¹  and Phenolics in Foods.¹²  As (poly)phenols have been covered extensively in a number of previous and recent reviews and textbooks, the present chapter aims to provide a brief summary of current practices, with a focus on modern techniques, tips for best reporting practices, and assuring qualitative and quantitative accuracy, which are not covered in detail in previous texts. Practical methodological considerations (i.e., advice and tips) are provided in detail as supplementary material (Appendix).

The most consumed berries (fresh-raw) in the USA are strawberry, blueberry and raspberry (44%, 29% and 15%, respectively), followed by blackberry and other, which include cranberry (9% and 3.4%, respectively) (https://www.statista.com/statistics/191512/fresh-berry-category-share-in-2011/).¹³  The primary (i.e., found in highest abundance) biologically active compounds reported in these berries include sugars, fiber, anthocyanins, hydroxycinnamic acids, vitamin C, flavonols, hydroxybenzoic acids, flavan-3-ols, ellagitannins, procyanidins and stilbenes (listed in order of relative abundance; Table 1.1). Sugar and fiber are by far the most prominent nutritive components and are generally reported between 12–15 g per 100 g and 2–5 g per 100 g (respectively), while the remaining non-nutrient (poly)phenols are reported between 0.1 and 170 mg per 100 g, or collectively between 95 and 300 mg per 100 g of fresh weight (http://phenol-explorer.eu; http://www.foodb.ca; https://fdc.nal.usda.gov/, 10/10/2020).^14,15  Some berries, such as strawberry and raspberry, contain relatively high amounts of vitamin C (60 and 25 mg per 100 g, respectively). Even though the focus of this chapter is on non-nutrient berry (poly)phenols, sugar, fiber and vitamin C should not be discounted when considering the broader array of constituents within a berry or when developing methods of extraction, as they provide considerable matrix components which can impact extraction efficiency, chromatographic performance, compatibility with selected detection systems and detection limits. In the present chapter, we focus on analytical considerations for the most abundant (poly)phenols (Table 1.1) within the most consumed berries, including anthocyanins, hydroxycinnamic acids, flavonols, hydroxybenzoic acids, flavan-3-ols and their polymers, procyanidins and ellagitannins; and their known biological (human, animal and microbial) metabolites (Table 1.2).

In these most highly consumed berries, anthocyanins are generally reported highest in blackberries and blueberries, while hydroxycinnamic acids are highest in blueberries, and flavonols in blueberries and cranberries (http://phenol-explorer.eu; http://www.foodb.ca; https://fdc.nal.usda.gov/, 10/10/2020).^14,15  Further, hydroxybenzoic acids are found in highest abundance in blackberries and cranberries, flavan-3-ols in blackberries, ellagitannins in raspberries and procyanidins in strawberries. Although numerous reported bioactivities of berries and their phytochemical constituents exist, including cardio-metabolic, anti-diabetic, anti-inflammatory, cognitive performance and anti-cancer activities,¹⁶  there is currently no direct evidence to suggest any single phytochemical is responsible for the health effects observed following berry consumption.

Anthocyanins, flavonols and flavan-3-ols are subclasses of flavonoids, which share the same basic structural characteristics, each having a base structure containing two linked aromatic rings (A-ring and B-ring) bridged by a heterocycle containing three carbons and one oxygen (C-ring), with each aromatic ring containing at least one aromatic hydroxyl (see Scheme 1.1). Anthocyanins, flavonols and flavan-3-ols are further divided into subclasses based on functional groups on the C-ring and hydroxylation and methoxylation patterns of the B-ring. The primary distinguishing feature separating these subclasses is C-ring unsaturation, where anthocyanins have double bonds between carbon 1–2 and 3–4, while flavonols contain a 2–3 double bond and flavan-3-ols contain no C-ring unsaturation. Other distinguishing characteristics include anthocyanins having a positively charged oxygen (at acidic pH) in their heterocyclic C-ring (C-1), while flavonols contain a C-4 ketone. Further, anthocyanins and flavonols are found in nature glycosylated, while flavan-3-ols are most commonly non-glycosylated, remaining distributed in aglycone form.¹⁷  These unique structural differences, i.e., unsaturation, charge state, glycosylation and hydroxylation, afford differential analytical consideration as they impact stability, recovery and detection characteristics.

Phenolic acids, including hydroxybenzoic and hydroxycinnamic acids (see Scheme 1.2), contain a six-carbon atom (C₆) phenyl ring and differ by the number of carbons on their side chain. Hydroxybenzoic acids are denoted C₁, having only a single carbon side chain, while cinnamic acids are denoted C₃ phenolics, having a three-carbon side chain. Both being acids, their side-chain terminates with a carboxyl group. Cinnamic acids have a future distinguishing feature, as their side-chain has a C1–C2 double bond (unsaturation). Together these features provide the descriptive nomenclature C₆–C₁ or C₆–C₃ (hydroxybenzoic and hydroxycinnamic acids, respectively).^17,18 

Gallotannins are composed of a central polyol, typically quinic acid or glucose, with phenolic acids (typically gallic acid) esterified at various positions and in various chain lengths (see Scheme 1.3). Hydrolyzable tannins are referred to as such due to their ability to be hydrolyzed by moderate levels of acid or base, whereas condensed tannins (see below) require stronger nucleophilic reagents for hydrolysis.¹⁹ 

Procyanidins are non-hydrolyzable (i.e., condensed) tannins composed of flavan-3-ol monomer subunits linked by C–C inter-flavan bonds (see Scheme 1.4). In berries, the flavan-3-ol monomers are primarily (+)-catechin and (−)-epicatechin. Procyanidins are further subdivided into A and B types. B-type procyanidins are the simplest, with C4–C6 inter-flavan bonds. A-type procyanidins are more complex, with C4–C8 and C2–O–C7 linkages. B-type procyanidins are widely distributed in berries, with A-type procyanidins primarily found in some Vaccinium species, particularly cranberries.^20,21 

Phenolic metabolites, products of microbial catabolism of larger polyphenol and/or polyphenol polymers, are generally annotated as C6 structures, signifying they have a six-carbon ring (see Scheme 1.5). These benzene derivative C6 rings, in their simplest forms, are bound with hydroxyl (OH) or methoxy (OCH₃) subgroups forming structures like dihydroxybenzene (catechol) or 1,2-dihydroxy-4-methylbenzene (4-methylcatechol or homocatechol). Microbes can cleave these rings forming even smaller structures such as muconates, oxalocrotonates and short-chain fatty acids; however, these low-molecular-weight metabolites often escape detection methods such as electrospray ionization-mass spectrometry (ESI-MS) and UV–Vis. Most frequently, phenolic metabolites of berry (poly)phenols are identified as phenolic acids, C6 rings containing carbon side chains of varying lengths, terminating with a carboxyl (COOH) group. Side chains are generally annotated C1, C2, C3, depending on the number of carbons in the chain. Together this nomenclature is denoted C6–C1, C6–C2, C6–C3, etc.²²  These side chains contain double bonds and hydroxyl groups and provide chromophores or ions providing detection using common ESI-MS and UV–Vis-detectors. The nomenclature of microbial metabolites differs by positioning of the double bonds (or site of unsaturation) and oxygen in the side chain. For example, an unsaturated C3 side chain with a double bond between C2 and C3 is referred to as a cinnamic acid.^17,18  Phenolic acids can be derived from most polyphenol forms; however, some microbial metabolites are unique to a specific subclass. For example, hydroxyphenylacetic, (hydroxyphenyl)propanoic and benzoic acid can be derived broadly from polyphenols, while urolithins are specifically derived from the catabolism of ellagitannins. Similarly, gamma-valerolactones are unique to flavan-3-ols. Urolithins differ from many berry-derived microbial metabolites as they possess two phenyl rings and lack a side chain. Phenolic metabolites derived from the microbial catabolism of berry (poly)phenols can be absorbed and found in the circulation as simple phenolics or as conjugated host metabolites, undergoing phase II conjugation in the intestine, liver and tissues, forming glucuronide, sulfate, methyl and/or glycine conjugates.¹⁶  As with their precursor (poly)phenols, differences in conjugation, unsaturation, hydroxylation and methoxylation afford differential analytical considerations.^16–21 

Traditionally, polyphenols in berry research were characterized and quantified using HPLC. As polyphenols are chromophores absorbing at wavelengths as high as 450–500 nM (and above), multi-wavelength detectors such as diode-array (DAD) or UV–Vis spectrophotometry were commonly used. Many reviews focusing on anthocyanin characterization (detailing packing materials, acid modifiers, mobile phase compositions, fragmentation patterns, etc.) have been previously published.^23–25  Originally, techniques focused on HPLC using single or multi-visual wavelength or UV–Vis–diode array detectors capable of collecting in-line spectra. These techniques are still useful today for the tentative identification and quantitation of anthocyanins and, in many cases, are more sensitive than MS due to the fact that anthocyanins possess a high-intensity chromophore at acidic pH. However, the most utilized modern instruments for (poly)phenol analysis is ultra-performance liquid chromatography/ultra-high-pressure liquid chromatography (UPLC/UHPLC) coupled with mass spectrometry (MS). Gas chromatography–mass spectrometry (GC–MS) is less commonly used for polyphenols due to their polarity and poor volatility but is likely to gain (or re-gain) popularity for characterizing low-molecular-weight microbial metabolites, which often lack a significant chromophore for spectrophotometric characterization or ionize poorly using ESI-MS. To date, any number of MS platforms have proven effective at characterizing polyphenols from berries and their metabolites, including triple quadrupole mass spectrometer (QqQ), time of flight (TOF) and ion trap; each having its own strengths and weaknesses (as discussed below). While fluorescence is generally less suitable for monomeric (poly)phenol analysis, larger polymers such as procyanidins are most commonly detected using fluorescence spectrometry.^26,27 

Procyanidins present unique challenges for analysis due to their immense (almost limitless) structural variation (i.e., numerous combinations of possible monomer subunits, linkages and degrees of polymerization), which leads to poor resolution performance of common separation methods, poor analytical sensitivity (with sensitivity inversely proportional to molecular weight) and dearth of available analytical standards.^26,28  Due to these factors, most research has focused on the smaller, more easily characterized procyanidins (dimers, trimers, etc.). Larger procyanidin oligomers and polymers may be identified by more complex analytical techniques such as matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF-MS).²⁹  However, even these methods have significant limitations, and thus the vast majority of complex procyanidins are characterized by somewhat non-specific techniques. These include normal-phase liquid chromatography (LC), hydrophilic–lipophilic interaction chromatography (HILIC),³⁰  or methods that degrade the polymers and quantify the resulting monomers to identify end-units and monomer ratios (such as thiolysis or phluoroglucinolysis).²⁷ 

LC remains the most commonly used separation technique for the characterization of anthocyanins and broader (poly)phenolics, including microbial and host metabolites in fruits, vegetables and biological fluids.³¹  Extensive lists of columns, LC conditions and MS platforms have previously been published.³¹  In the past 15 years, ultra-performance liquid chromatography (UPLC) has emerged as an eventual replacement for HPLC. UPLC does not differ theoretically from HPLC but rather capitalizes on column and LC platforms with higher optimal operational pressures and sub 2 μm particle size for stationary phases that enable higher levels of resolution. By affording column pressures in excess of 15 000 psi, higher flow rates can be achieved even with reductions in particle sizes without compromising column resolution. This results in improved efficiency and overall resolution due to reduced diffusion and shorter run times.

Many modes and configurations of mass spectrometers are reported for (poly)phenol analysis from berries. These include ionization modes such as ESI and MALDI coupled to detectors such as QqQ, TOF, quadrupole time of flight (QTOF) and ion trap (such as Orbitrap); while ion mobility (IM) MS is becoming more popular in recent years.³¹  As (poly)phenols have adequate ionization in negative mode ESI, the negative mode is often used for their detection due to reduced background noise to achieve higher S/N ratios, leading to lower limits of detection and better sensitivity compared to similar approaches in positive mode ionization. Anthocyanins, due to their flavylium cations at acid pH, lend themselves more to positive mode ESI; however, most flavonoids will ionize effectively in either positive or negative mode. Some have even reported negative mode ESI can be useful for anthocyanin characterization, as this often leads to higher sensitivity and more characteristic ions [M – [2H](–); [M – 2H  + H(2)O](–)], adducts (resulting from acid modifiers)and doubly charged ions, which may be useful in characterizing anthocyanins relative to other (poly)phenols in complex mixtures.³²  To date, the most common instrument configuration for anthocyanin analysis is ESI with QqQ or TOF MS platforms.³¹ 

Triple quadrupole mass spectrometry (QqQ MS) is a highly versatile platform, excelling in situations with broad-spectrum analytical objectives, but is not ideal for qualitative characterization as they provide only nominal mass (nominal mass resolution is insufficient for definitive exact mass conformation required for qualitative identification). The optimal (and most common) operation mode is using multiple reaction monitoring (MRM) or scheduled multiple reaction monitoring (sMRM) MS detection for its selectivity and sensitivity. The main draw of QqQ MS is that it can analyze several hundred mass transitions in a run, therefore proving ideal for the analysis of complex mixtures such as plant or animal tissue extracts. QqQ MS can provide full scan acquisition but is more commonly used for product ion scanning, i.e., parent, daughter or neutral loss scanning functions, which are useful in the characterization of plant and human metabolite conjugates. The QqQ MS has a wide range of applications due to its scan speed and various fragmentation profiling capabilities. QqQ-linear ion trap (Q-Trap) is an enhanced QqQ, where the final quadrupole can function as a virtual ion trap (or is replaced with a linear ion trap), and is capable of collision-induced dissociation (CID) fragmentation profiling, providing MS(n) scanning capabilities and often high mass resolution or accurate mass. Q-Trap data-dependent acquisition modes allow for selected/multiple reaction monitoring (SRM/MRM) and triggered enhanced product ion (EPI) MS/MS scanning for structural elucidation, making it ideal for screening applications.³³ 

The TOF platforms offer theoretically unlimited mass range scanning, which is useful for characterizing large compounds such as hydrolyzable tannins and procyanidins that exceed the typical mass ranges of QqQ platforms. TOF instruments also provide high-resolution mass accuracy required for molecular formula determination/confirmation, allowing discrimination of compounds with similar but different molecular masses (isobaric ions having the same unit mass with different exact masses). Hybrid quadrupole time-of-flight MS (Q-TOF) is often the chosen instrument for metabolite screening (plant or animal) and is preferred for their full scan sensitivity, mass accuracy and high data-acquisition rates. These instruments are also commonly used in metabolomics as they provide detection of a broad range of analytes (known and unknown) in complex mixtures. The advantage of Q-TOF over TOF instruments is the ability to obtain more detailed structural elucidation information, providing MS(E) fragmentation data.³³ 

Ion traps such as Orbitrap are high-resolution ion trap mass analyzers often used in proteomics and untargeted metabolomics, due to their ability to scan tens of thousands of MS features with ultra-high resolution and reliable mass accuracy. LTQ Orbitrap is a hybrid high-resolution mass analyzer combined with a linear ion trap, thus providing accurate mass and having the capacity for MS(n) scanning functionality, providing added functionality and versatility for structural elucidation. Traps and TOF have similar functionality but have one drawback relative to QqQ MS, as they are less compatible with high-throughput rapid chromatography applications, such as high-throughput quantitative metabolite screening in complex mixtures.³³ 

Accurate quantitation requires sound sample preparation techniques that maximize recovery of the analytes in question while optimizing the reproducibility and robustness of the method. When separating (poly)phenols from fruits such as berries, or biological matrices, there are many matrix components to consider, including sugars, proteins, carbohydrates, lipids and fibers, which can negatively impact extraction efficiency and quantitative accuracy. In a systematic review published by Singh et al. (2020)³⁴  summarizing LC–MS methods for the identification and quantification of anthocyanins in fruits and vegetables, the authors found that extraction and sample preparation techniques were the most critical factors in establishing quantitative accuracy and consistency. Many methods of extraction can be applied to berries or berry-derived products, including refluxing, heating, supercritical extraction, etc.; however, the most frequently utilized techniques in laboratory environments are liquid–liquid extraction (LLE), solid–liquid extraction (SLE) and solid-phase extraction (SPE). Each method has its advantages and disadvantages. Struck et al. (2016)³⁵  reviewed common processing and chemical analysis techniques (Table 1.3) for extracting and assessing (poly)phenols in berry pomace and containing details on a wide variety of berries, including blueberry, strawberry, raspberry, blackberry, chokeberry, black currant. Multiple processing variables were explored across published works; the most common extraction conditions for low-molecular-weight (poly)phenols in berries involved the use of organic solvents such as methanol, ethanol or acetonitrile, water and an acid modifier (often formic acid or citric acid if using MS).^31,35 

Many reviews on methods for the extraction of berry anthocyanins in plants, juices, wines and biological fluids have been published over the years. Originally, techniques focused on extraction using LLE, SLE or large open columns packed with sorbent materials such as Amberlite XAD resin or prepacked SPE single cartridges.^2,25,31  More recently, Kafkas et al.¹  provided a review of advanced methods for the analysis of phenolics in fruits and reported low-molecular-weight phenolics and higher-molecular-weight polyphenols (such as flavonoids) were generally extracted using the same extraction media (i.e., packing materials and sorbents) and conditions, while larger polymers required alternative methods. For example, hydrolyzable tannins (widely distributed in berries),³⁶  due to their structural complexity, require techniques such as methanolysis, which is used to break the hydrolyzable tannins down into their polyol and phenolic acid subunits, for analytical extraction and characterization of subunit ratios.^37–39 

Primary sample preparation methods are critical to establishing the best conditions for extraction and subsequent analysis from berry components. Primary extraction from fresh tissues, frozen tissues and freeze-dried powders are widely reported. In all cases, adequate homogenization of samples is critical to ensure a representative sample for the cellular disruption that will facilitate the release of phenolics for subsequent solvent extraction. Some common techniques include, but are not limited to, physical homogenization, heat treatment, ultrasonication and microwave-assisted extraction.^40,41 

For cells, tissues and organs, it is critical to lyse cells, ensuring the release of intracellular compounds. This is typically performed using a lysis buffer (such as RIPA: Radioimmunoprecipitation Assay Buffer), in combination with fine homogenization, a French press, ultrasonication, or magnetic bead disruption. Enzymatic digestion may be used for particularly difficult matrices, such as bone or connective tissues. Subsequently, protein–polyphenol bonds are disrupted using reagents such as phosphoric acid, and proteins and phospholipids are precipitated using organic solvents (acetonitrile, methanol) and zinc sulfate. Centrifugation or filtration is then used to remove the insoluble material, followed by re-extraction/washing with the solvent of choice. Extracellular components from urine or cell culture media can often be treated using “dilute-and-shoot” procedures; however, some form of sample “cleanup” is recommended (i.e., SPE, LLE) and often followed by concentration. For biological fluids, the process is similar, although cell lysis and disruption are typically omitted due to the absence of cells in these materials.

LLE and SLE have a long history of use in polyphenol extraction from plants due to their often simple and rapid nature, in addition to their ability to be scaled up for use on large masses of materials. Such methods use alcohols, acetone, hexane, ethyl acetate, diethyl ether, chloroform and often require the addition of varying proportions of water in order to capture more polar phenolic compounds such as highly hydroxylated flavonoids, hydroxybenzoic and hydroxycinnamic acids and larger procyanidins. Also, some of the solvents used in LLE and SLE are inherently dangerous to work with and require consideration. Although these techniques are useful for bulk extraction, they often have higher variation in quantitative analysis when using low sample volumes. Additional techniques are often required to remove solvent residues from final mixtures prior to analysis or consumption as extracts. One benefit of these procedures is that they rapidly separate polar polyphenols from other unwanted matrix components such as oils and waxes. More modern techniques include supercritical fluid extraction, which benefits from being solvent free and requires relatively no additional sample cleanup.^2,42 

Polarity is the number one consideration when establishing an extraction solvent or suitable chromatography column and solvent system. Polyphenols can be challenging to universally extract with any single solvent due to their varying polarities. For example, procyanidins are considerably more polar than their monomer subunits and these co-exist in a single botanical matrix-like cocoa, cinnamon and fruits, including grape. Catechin and epicatechin can be extracted effectively with methanol, ethyl acetate and diethyl ether, whereas procyanidins composed of these monomers are most efficiently extracted with mixtures of acetone, water and acetic acid. As such, complex extraction schemes are often reported in the literature, with fractions often requiring combination prior to final preparation for analyses. The introduction of water to extraction solvents is a common practice but presents challenges for subsequent concentration, often eliminating nitrogen evaporation or vacuum evaporation as practical options (where large volumes are at play), necessitating the use of a freeze dryer. Thus, care should be taken to select the extraction solvent that provides the best combination of efficiency while facilitating the desired post-extraction processes.⁴³ 

A common sample preparation technique used in berry research is SPE, which is useful in removing complex matrix components from plant and animal tissues that may not be fractionated during LLE while retaining aqueous polar (poly)phenols. SPE as a “cleanup” technique is preferred in quantitative workflows as its easily scalable, uses relatively low amounts of solvent, comes in a large number of sorbent bed materials, is relatively rapid, highly reproducible, can be used to concentrate samples, and results in final samples which are free of instrument interfering matrix components. Ninety-six well micro-elution SPE has proven particularly useful in biological matrix cleanup as it is amenable to low volumes of biological fluids and facilitates high-throughput sample processing. SPE can also be used for solid tissue, generally following dilution, homogenization and centrifugation steps. If non-compatible solvents are required for tissue homogenization and LLE, samples can be dried and re-dissolved in appropriate dilute loading solvents after centrifugation and concentration. Polyphenol stability can often be achieved by adjusting pH using acids such as formic acid or hydrochloric acid (HCl).⁴⁴ 

There are numerous columns and stationary phases on the market supplied by any number of manufactures, which provide appropriate qualities for polyphenol and phenolic metabolite separation. Most suppliers offer column selection guides, application notes or apps, which are useful in identifying an appropriate stationary phase. In 2010, Ignat et al. published a review on methods for characterizing (poly)phenolic compounds in fruits and vegetables,²  providing summaries of reported extraction and LC conditions. Here the most commonly reported methods of separation included the use of standard C18 or modified C18 columns and mixtures of acetonitrile, water and formic acid. More recently, Singh et al.³⁴  published a systematic review of LC–MS methods for the identification and quantification of anthocyanins in fruits and vegetables. The review summarized many types of column stationary phase, mobile phase, acid modifiers and detection techniques, citing reverse-phase C18 as the most common stationary phase and acetonitrile or methanol, and water with formic acid as mobile phase components. Other common stationary phases used in polyphenol research are reverse-phase octadecyl modified C18 (i.e., polar-embedded/end-capped), phenyl-hexyl modified, pentafluorophenyl propyl modified (fluorinated silica/PFP) and ammonium sulfonic acid modified (HILIC) columns. PFP columns are particularly useful for retaining low-molecular-weight phenolics, which may otherwise not be sufficiently retained using C18 columns. For establishing simple methods involving a few analytes, application notes provided by the manufacturers are generally useful; however, when choosing a specialized column/stationary phase for complex mixtures and requiring separation of a broad-spectrum of structures of varying size and polarities, compromises in resolution (peak shape vs. retention) may be required. HILIC (typically diol) columns are becoming more common, particularly where required to resolve highly polar polymers such as procyanidins. HILIC and associated solvent systems are also more compatible with ESI than traditional normal-phase solvent systems, which are optimal for atmospheric pressure chemical ionization (APCI). UPLC-HILIC columns facilitate rapid separations compatible with ESI and reduce run time considerably from traditional reverse phase columns (e.g., from 1 h to a few minutes).³⁰ 

While samples should be filtered or otherwise clarified of particulate matter through centrifugation prior to loading into the LC system, the use of guard columns is standard practice and generally paired to the primary column as recommended by the manufacturer. A guard column is used to protect the primary column and prolong its lifespan, which is dependent on the stationary phase, mobile phase, pore size and flow/pressure utilized. Guard column integrity is generally monitored by a % increase in pressure from standard operating pressures and/or retention time drift. Loss of peak shape can also be an indication of failure of the guard column and should be checked prior to the complete replacement of the analytical column. It is generally accepted that peak retention time drift should not exceed a 1.5% coefficient of variation (CV) across multiple injections.⁴⁵ 

The decision of what column length, internal diameter (ID) and pore size to utilize for a given method is largely made based on acceptable run-time and separation or resolution requirements. C18 phases are generally superior at retaining larger, more polar polyphenols while maintaining optimal peak shape; however, retention of smaller and less polar analytes, such as microbial metabolites, can be compromised. Alternatively, a modified silica phase such as PFP is capable of retaining low-molecular-weight phenolic metabolites with higher effectiveness; however, peak shapes of larger, more polar structures are often compromised. The impact of peak width or resolution on qualitative or quantitative data is dependent on the detection method utilized and may not be an important consideration for mass spectrometers with high duty cycle-time. In such cases, having a unique analyte mass, fragmentation profile and an appropriate number of data points captured is often more important than resolution. Generally, for reproducible quantitative MS data, 10–20 data points per peak are recommended.⁴⁶  This value generally refers to peaks of appropriate shape and size; for peaks larger than 10 s and/or abnormally shaped peaks, a proportionately greater number of data points should be captured.

The most common technique for the separation of low-molecular-weight (poly)phenols is reverse-phase LC. Most (poly)phenol chromatography is reported using a mixture of formic acid and acetonitrile or methanol, although other acids such as acetic acid, trifluoroacetic acid (TFA) or HCl are not uncommon when using colorimetric detection, TFA or HCl are generally avoided when using MS detectors, do to their ion suppressing qualities. Formic acid is typically reported in the mobile phase between 1 and 5% for colorimetric detection or 0.1–1.5% for MS. Optimizing the acid modifier (type and concentration) is important for polyphenols and phenolics, as it affects both chromatographic separation (stationary phase-specific) and ionization. Generally, a compromise is required between peak width or shape and peak intensity when using MS, as acids are known to influence ionization efficiency and can serve as ion suppressors. This is less important for UV–Vis detectors, which remain less impacted and often benefit from improvements in peak width and height. This requires significant method development when transitioning older methodology based on LC–UV to newer LC–MS and UPLC–MS platforms. When optimizing for complex mixtures of precursor (poly)phenols, their microbial metabolites and biological conjugates, the compromise between peak width and intensity should be made in consideration with the limit of detection, limit of quantitation and integration thresholds, which are relative to the signal-to-noise within a given matrix. Higher acid mobile phase concentrations are generally required to resolve berry anthocyanins but will often reduce the detection limit of low-molecular-weight phenolic metabolites when using MS as a detection method.⁴⁷ 

Reverse-phase separation is most commonly used for (poly)phenol monomers. However, normal-phase separation and HILIC (“aqueous normal-phase”)separations are particularly well-suited for the resolution of highly polar polymers such as procyanidins.^26,30,48  Normal-phase and HILIC solvent systems are not optimal for ESI interfaces (HILIC is better, but still not ideal), and thus post-column ionization enhancers such as ammonium formate are often used to improve ionization.^30,37  There are some negative consequences with the introduction of post-column buffers, such as source contamination/fouling and solubility with organic solvents, which can often lead to blockages in LC lines, injector ports, columns, etc.

The principles for method development remain constant regardless of (poly)phenol or berry bioactive targets, whether developing a method for the extraction of components from foods, tissues or liquids. The key is formulating a method development strategy with predefined endpoints, constraints and expectations, with consideration for the continuum from method development, to optimization and validation. In defining the goals of any analysis, it is important to understand, there is no “one-size-fits-all” strategy, particularly when developing a method for the characterization of a diversity of analytes from a complex matrix. Prioritizing the most important deliverable/outcomes, and acknowledging that sacrifices will likely have to be made on other parameters, is a fundamental principle of method development. Method optimization is generally a process used to fine-tune a method, either improving precision and accuracy or assay efficiency (time, materials use, etc.). Method optimization generally follows the same principles of method development: establish defined endpoints with predefined constraints and expectations and execute the strategy. A list of common tips for method development and optimization is provided below.

Precision is the assessment of how close measurement results are to each other. Precision can be established by providing evidence of variability in assay sensitivity and specificity, such as providing CV of calculated concentrations of internal standards or quality control (QC) samples, regression slopes, limits of quantitation, retention time, ion-pairing ratios, peak quality parameters, etc. Acceptable % CV for quantitation is generally considered below 15% within the expected range for a given analyte, and as much as 20% CV is considered acceptable at the lower limit of quantitation (LLOQ) and HLQ.^49–51  Assay and instrument performance are generally monitored by including blanks (and/or double blanks) and QC samples inserted randomly in a run at 10–20% of injection frequency.^49–51 

The Horwitz ratio (HorRat) is a normalized performance parameter often used as a measure of inter-laboratory precision as a function of analyte concentration. It can, however, be applied to assess the relative reproducibility of a method within a laboratory and allows for comparison across a range of metabolites. It is calculated as a ratio of the Relative Standard Deviation for an analyte response, RSDr, to that calculated from by the equation (PRSDr = 2C^−0.15) from the mean concentration, C, found or added, expressed as a mass fraction.⁵²  HorRat values can be calculated for individual phenolic species to assess the relative precision of the broader method. Ranges of 0.5–2.0 typically suggest high reproducibility for individual metabolites. However, it is not uncommon for HorRat values to be below 0.5 for species at the lower end of the calibration range of the method.

Accuracy is the assessment of how close a test result is to the expected value and has relevance to both qualitative and quantitative analyses. Qualitative accuracy can be established by providing evidence of variation in retention time, accurate mass, ion painting ratios, MS/MS fragmentation, etc. Quantitative accuracy can be established by providing evidence of regression linearity, actual vs. returned concentrations, limits of detection and quantitation, etc. Where available, stable isotope dilution can be employed for optimal reproducibility and accuracy. In this quantification technique, stable (¹⁵N, ¹³C and/or ²H) isotopes of each analyte are used as the internal standard in each sample.⁵³  These isotopes typically have identical elution and ionization properties and are thus optimal internal standards. Using stable isotopes as internal standards is problematic for many polyphenols, as only a small number of these compounds are available in the labeled form. Labeled isotopes are much more common for nutrients, endogenous and drug metabolites, and thus may be of limited utility for many berry compounds. However, when possible, the use of stable isotopes is considered a “gold standard” for quantitation.

The accuracy of a method can also be established through the application of extraction methods to standard reference materials (SRMs) for which the expected value is known. These are available from the National Institute of Standards and Technology (NIST) for a limited array of polyphenol-rich foods, including baking chocolate, cocoa extract, green tea leaves and extracts where a subset of phenolic metabolites have been quantified and can serve as the gold standard. However, the availability of validated SRM for fruits, including berry fruits, remains limited to non-existent.

Qualitative accuracy is important when interpreting studies reporting the identification of analytes (compounds, metabolites, etc.) using untargeted analysis techniques (e.g., untargeted metabolomics) or in situations where reference standards are not available. The difficulty in interpreting qualitative accuracy is that a significant proportion of studies in the literature do not sufficiently describe how analytes were identified. This is particularly the case for conjugates of secondary plant metabolites (i.e., sugar conjugation or acetylation) or conjugates of human phase-2 metabolism (i.e., glucuronide and sulfate), which are most often not commercially available as reference standards. In these situations, where reference standards are not available to assign structural identity (i.e., provide retention time, fragmentation profiling, etc.), identification should be reported as “putative”. Assigning an identification based solely on mass, UV–Vis or infrared (IR) spectrometry features is of particular concern with larger molecules and/or compounds having significant numbers of isomers. When it is necessary, investigators reporting qualitative data should describe the identification as an “annotation” or “putative identification”, and ideally report other possible isomers matching the molecular formula.

The qualitative accuracy of many studies appears inherently low, particularly as there are no standardized practices for reporting MS methodologies, making it difficult to compare findings reported in the literature.¹⁶  Consequently, it is likely that misclassified and falsely identified metabolites are abundant in the literature and propagated in online metabolomics databases, as highlighted in many publications.^16,54,55  Unfortunately, most researchers in the (poly)polyphenolic field report positive identification based on a single assignment (one precursor mass or precursor to product ion (MS² ) match), following low-resolution nominal mass scanning (i.e., one decimal place accuracy). This is problematic due to the fact that numerous compounds can have similar, or the same, integer/unit masses (isobaric ions). Significant mass resolution (down to 5–6 decimal places) and mass accuracy (>5 ppm difference between measured mass and assigned molecular formula mass)are required for confidence in a molecular formula determination. Furthermore, many investigators neglect to factor in the differences between monoisotopic exact mass (measured by MS)vs. average molecular weight (factoring in average isotope abundances) when interpreting mass spectra. This can be extremely problematic for large compounds with numerous carbons or other elements presenting significant isotopic variability. Likewise, compounds with identical molecular formulae but distinct structures will have the same exact mass. Moreover, as there are few reference standards available for confirmation or MS optimization of biological conjugates, false-positive identification is likely, particularly as signal quality (e.g., peak width, number of data points captured, signal to noise, etc.) is not reported in the majority of studies. Structures such as hydroxylated cyclic and polycyclic structures (i.e., polyphenols) are also prone to multiple gas phase artifacts, including water elimination (loss −17), hydrogen elimination (− 1), radical fragmentation (loss −15, −14), Retro-Diels–Alder (RDA) reactions (C-ring electron rearrangement)(+or −2), gas-phase dimers, or combinations thereof (such as loss −13 or −12).^56–58  Finally, in-source fragmentation following soft-ionization is often observed for sugars and phase II conjugates, particularly sulfate and glucuronide conjugates.

Developing a quantitatively accurate assay generally begins with establishing a linear regression derived from a serial dilution of reference standards in a background matrix representative of the samples requiring analysis. Assay linearity and its precision are established by performing a number of calibration or regression curve analyses (i.e., number of replicates), with accuracy established based on acceptable % CV. Assay linearity essentially defines assay quality relative to quantitative accuracy and is particularly important in establishing reproducibility at the lowest and highest accurate concentrations for the assay. Assay sensitivity is the ability to accurately detect small changes in concentration. Sensitivity is often related to the ability to detect changes at the LLOQ and is influenced by signal/noise and limit of detection but technically applies to any concentration reflected by the standard curve regression. Sensitivity can be negatively impacted by many factors, including carryover and matrix effects.⁵⁹ 

In berry research, semi-quantitative or relative quantitative assays are readily used and require discussion with respect to establishing assay performance. Berries are frequently consumed as juices (including cranberry, blackcurrant, etc.) and many rapid/simple analytical techniques have been established for characterizing total (poly)phenols in these often relatively “clean” fluid matrices. These rapid and cost-effective techniques have the greatest utility in industry for monitoring within-process or batch variability; however, they generally lack the analytical specificity required for academic laboratory research. For example, the Folin–Ciocalteu (F–C), Lowenthal permanganate (L-P), 4-dimethylaminocinnamaldehyde (DMAC) and the bovine serum albumin (BSA) precipitation assays are useful in establishing relative differences between treatments, but when evaluated for absolute qualitative and quantitative accuracy, they lack specificity and sensitivity. As exemplified in a recent publication by Ma et al. recently (2017),⁶⁰  the authors explored the accuracy of such methods and found the methods provided differential accuracy for quantifying various subclasses of polyphenols and absolute quantitative values for total (poly)phenols were not comparable across the assays. Consistency within a given assay was suitable. However, each assay had a considerably different linear range, indicating the assays have the greatest utility for studying changes to total (poly)phenol content within a common matrix (i.e., juice) and assay, but limited capacity to characterize and quantify individual (poly)phenol contents within and across assays.

When interpreting evidence from existing studies, it is often not apparent how metabolites are quantified. In most situations, reference standards do not exist. In such cases, quantitation should be described as “putative”, as surrogate reference standard curves are used for quantitation, which is highly inaccurate in most instances. This practice of using surrogate analytes assumes the response factor of one analyte is similar to another. A response factor is typically defined in spectrometry as the ratio between a signal intensity or area (or area: intensity ratio) and the known quantity of the analyte. For chromatographic techniques which use colorimetric detectors, this is less of an issue, particularly for compounds having similar chromophores; for MS detection, which uses voltage to create a mass-to-charge ratio, response factors for seemingly similarly structured compounds can be extremely dissimilar. Most evidence indicates MS response-factors are highly variable between chemical species, and the use of surrogate regression curves leads to significant over and underestimation.⁶¹  Investigators reporting quantitative data for compounds that do not have commercially available references standards should identify these instances as “putative quantitation”. This is particularly true for conjugated host metabolites of phenolic species. In validating methodology for flavan-3-ol metabolites in plasma, authors have reported greater than 100-fold difference in quantitative concentrations established using response factors derived from the regression lines of authentic catechin, epicatechin and their corresponding biosynthetically generated and authenticated glucuronides and O-methyl-glucuronide conjugates.^61,62  Considering the lack of commercial availability for most (poly)phenol metabolites, estimations are commonly established using aglycone forms, which is likely to lead to even greater inaccuracies. A list of common tips for achieving accuracy is provided below.

Method validation is the process of establishing the suitability of an assay for its intended use. Methods for quantification of berry phenolics and metabolites in food and biological matrices should follow standardized validation approaches. One approach, adopted by the AOAC Association of Official Analytical Chemists in collaboration with NIH-Office of dietary supplements and the Food and Drug Administration, is outlined in the Guidelines for Single-Laboratory Validation of Chemical Methods for Dietary Supplements and Botanicals.^63,64  In the simplest sense, this guidance document provides a framework from which to establish the performance characteristics of any method within a lab. Performance, in this case, includes the parameters already discussed above, including selectivity, repeatability, precision and accuracy. It may not be necessary for every laboratory assay to satisfy all of the validation criteria outlined in the guidelines for dietary supplements and botanicals protocols, and many other guidelines exist;^{50,51,63,65–67}  however, quality assurance requires some level of validation for every assay. Below summarizes some of the more commonplace validation processes used in single user, single lab assay validation.

The robustness or ruggedness of an overall method is defined by the extent to which select factors, such as environment, storage, extraction condition or otherwise, impact the accuracy and precision of a method. It is recommended to establish robustness during method development and refinement or optimization, as well as assessment of final assay validation. Importantly, robustness should be re-established when scaling a method or translating one method for use with different sample matrices, such as applying a method used for fruit on biological tissue. Traditional approaches commonly explored one critical variable at a time (e.g., sample weight, extraction time or solvent volumes/temperature). However, it is possible to determine the impact of multiple factors at once using methods defined by Youden & Steiner (1975).⁶⁸ 

Recovery/extraction efficiency is the fraction of an analyte (or analytes) that can be recovered within a given process. Often presented as % recovery, there are many factors that lead to poor or inconsistent recovery efficiency. These most often relate to matrix interference and lack of solvent or solute compatibility with process stages. Sample matrix effects can affect both analyte recovery and potentially analytical response. Therefore, it is appropriate to conduct calibrations using analyte-free matrices or “background blanks” as well as to rely on internal standards to define matrix interactions/effects. This is most feasible for some sample matrices such as blood plasma/serum and urine, where synthetic or blank matrices are commercially available or can be generated within one's laboratory. However, for many food matrices, especially whole foods such as fruits, juices and derived products, it is not always possible to establish comparable matrix blanks. Also, as discussed previously, the lack of common SRMs for berries, derived food composites and metabolites continues to limit the ability to adjust calibrations and improve the accuracy of new and existing methods. In the absence of such materials, assessment of (poly)phenolic extraction recovery and matrix effects is often difficult. Some laboratories or companies produce their own SRMs as a composite of multiple batches and store them under conditions that will prevent sample degradation. On rare occasions, companies or laboratories may register their SRMs with NIST and provide them as NIST reference materials for public and commercial access. Establishing NIST reference materials and protocols require multi-lab ring trials to establish performance, as recently established for production and analysis of RM 8403 cocoa flavanol extract (NIST SP #260-207).⁶⁹ 

Method transfer is the process of applying a method to a new process for which it was not originally intended when conceived, developed, optimized and/or validated. Methods transfer does not always require complete re-development but should include tests for performance and often requires re-optimization when assay quality parameters fail to meet acceptable precision and accuracy criteria. The most common method transfers reported in (poly)phenol analysis are scaling of extraction to accommodate higher throughput and lower sample size and transfer of instrumentation, such as moving from HPLC to UPLC method or photon (wavelength/frequency) detector to mass detector. Aside from re-establishing assay performance as outlined above, there are other practical considerations that should be considered when transferring methods for (poly)phenol analysis of plant or biological matrices. A list of common tips for more efficient methods of transfer is provided below. Here we use the example of transferring an HPLC-photon detection method (e.g., wavelength/frequency detector) to a UPLC–MS detection method.

Traditionally, approaches to characterize phenolic compounds in foods and biological samples were based on targeted methodology; however, untargeted metabolomic-based approaches are becoming more commonplace. Targeted approaches, as the name suggests, are methods/workflows focused on expected or known analytes, contain limited numbers of spectral features relative to untargeted metabolomic assays, and are more amenable to quantitative or semi-quantitative data analysis.⁷⁰  Metabolomic approaches can be completely untargeted, which is essentially a qualitative technique, or broad-spectrum, which can be qualitative, relative-quantitative or even quantitative, depending on the study design, use of reference/calibration standards and data analysis strategy. Targeted approaches tend to use QqQ platforms, while untargeted approaches most often use high-resolution MS systems (HRMS) such as TOF and ion traps (such as Orbitrap) or hybrids (e.g., Triple-TOF, LIT Orbitrap). Both targeted and untargeted approaches can be conducted using similar and varied analytical platforms, although UPLC-coupled HRMS is most commonly used for (poly)phenolic metabolomic analysis. There is an advantage of using HRMS over low-resolution QqQ MS, as it provides accurate/exact mass, which can be used to differentiate compounds with the same nominal mass, thus greatly reducing possible numbers of isomers. This is important for untargeted analyses, where the presence of multiple compounds with similar nominal masses is much more likely than in targeted analyses. Further, accurate mass fragmentation provides valuable information for structural conformation. HRMS platforms can still be used for targeted quantitative approaches and are often as or more sensitive; however, as they devote considerable cycle time for mass resolution, they are limited by the number of analytes that can be analyzed with quantitative accuracy within a given run. The major advantage of HRMS is it can scan for tens of thousands of signals within a given run, and in many instances, identification can be confirmed without the use of reference standards, particularly if MS(2) or MS(3) CID data is available.⁷¹ 

Strategies exist for drawing quantitative conclusions for non-targeted techniques; however, these should be considered relative-quantitation over absolute quantitation and should be interpreted with care, as ionization efficiency is not consistent across compounds or concentrations, and the use of proportional abundance or surrogate reference standard regression equations can lead to significant under- or overestimation (as discussed above). Techniques such as isotope dilution and radio-labeling have attempted to improve quantitative accuracy; however, this cannot be applied to the abundance of spectral features in an untargeted dataset and generally focus on a few targeted features such as in the study of specific metabolic pathways.⁷² 

Sample preparation methods also differ in untargeted metabolomics relative to targeted approaches. Targeted and quantitative approaches optimize sample preparation procedures to maximize recovery, whereas untargeted qualitative approaches optimize for a maximum number of features (and are thus more “generic”). Untargeted sample preparation method development procedures are generally less laborious than targeted approaches and focus on protein precipitation (solvent and centrifugation) and optimization of concentration (often dilution) to maximize the number of spectral features while minimizing matrix effects (background noise and ion suppression). The advantage of this approach is that it limits the loss of many compound classes, which may not be compatible with extraction process/matrices as used in SPE, LLE or SLE. A limitation of untargeted sample preparation techniques (Table 1.4) is it may result in increased contamination of instruments with non-compatible analytes and leads to increased abundance of adducts and artifacts. The use of standardized protocols, internal standards and quality control and pooled samples are often used to validate assay reproducibility and account for batch effects.⁷¹ 

Most commonly used metabolomic approaches to characterize (poly)phenolic compounds in foods have used a profiling approach which essentially matches precursor ions in captured datasets to databases or spectral libraries, using monoisotopic accurate mass and molecular formula. One drawback of this approach is that the exact masses of precursors can match many isomeric compounds. Furthermore, in-source fragmentation of conjugates and gas-phase artifacts can increase the potential for false-positive identification of precursor structures. Additional techniques such as product-ion scanning [MS(2) or MS(3)] provide further structural information and are often required for positive identification.⁷¹  Although online databases capture significant spectral data (including accurate mass, MS/MS spectra, adducts, etc.) for the characterization of plant-based compounds, data for phenolic compounds from processed foods and human metabolism is limited. Compound discovery platforms and workflows aim to reduce the number of unannotated signals in untargeted datasets and are important for discovery and hypothesis generation, but annotated structures still require conformation via structural elucidation experimentation or referee standards. Confirmatory analysis using MS(2), MS(3), MS(n) or MS(E) techniques has some limitations. Drawbacks of MS(n) scanning includes the design of suitable MS(n) experiments requires advanced knowledge; voltage ramping is required for fragmentation of sugars or biological phase II conjugates (glucuronides and sulfate) relative to base structure fragmentation; datasets acquired have greater complexity and structural identification often requires manual interpretation for conformation. Newly emerging techniques in HRMS, including developments in IM MS, software, machine learning and online databases, will undoubtedly reduce some of the limitations of untargeted techniques (Table 1.4), leading to increased use of untargeted and broad-spectrum metabolomics in (poly)phenol research.

Without prior knowledge or evidence of the voltages required to produce predicted or expected fragments from published works (or found in online databases such as HMDB, PubChem, etc.), structural elucidation from low-resolution techniques such as MS/MS is difficult. Further, the use of single mass in selected ion monitoring (SIM) or full scan analysis lends a high probability of false-positive identification. Further, as polyphenol conjugates suffer from in-source fragmentation, misidentification of precursor structures is likely in situations where reference standards are not available. In such circumstances, compounds should be reported as putatively identified or annotated, and interpretation should be considered uncertain until further confirmatory analysis using MS(n) and high-resolution techniques is performed.

Structural elucidation requires understanding the instrumentation environment, such as voltage and chemical ionization, fragmentation, and impact of the matrix within the source and mass detector, all of which complicate data interpretation. Without awareness of these potential limitations, there is a high likelihood of spectral data misinterpretation.⁷³  For example, not all peaks in an MS spectrum originate from the compounds of interest and are as likely to result from the chemical and physical background environment (i.e., artifacts). Also, as the MS signal is spread over different isotopes, fragments and adducts, it is inherently difficult to adequately characterize product ions that can originate from the loss of part of a molecule or molecular rearrangement. Therefore, structure elucidation with MS often leads to the partial assignment of a structure rather than fully identified structures.

The level of confidence in qualitative or quantitative data is dependent on the details captured and reported. Therefore, it is important to capture multiple chromatographic (i.e., retention time, peak width, retention time peak drift, and resolution) and spectroscopic (signal stability, accurate mass, fragmentation pattern) features as many of these can be verified by others or through use of online data repositories. This is particularly important when reporting novel structures or in situations where no reference standards are available, as identity cannot be established/confirmed based on a single spectral feature.⁷³ 

Recommendations on reporting requirements for flavonoids in research have been previously published⁵⁵  and focused primarily on the design and reporting of nutritional intervention studies but also touched on research challenges, limitations and inappropriate analytical methodologies. The themes described in this chapter are in-line with the 2015 recommendations; however, the more modern focus on the metabolome and microbiome in berry and (poly)phenol research adds further complexities which need to be considered, as discussed herein. Building on this past initiative, a 2020 special article¹⁸  addressed issues with required reporting of (poly)phenol microbial and human metabolite nomenclature. Appropriate and consistent chemical description brings us one step closer to best reporting practices. However, best practice guidelines for reporting analytical methodology should be the focus of future position/policy papers, particularly considering the widespread use of MS for identification and quantitation. Without adopting consistent practices and reporting criteria, the replication of experimental findings will be difficult and incorporation of nutrition intervention findings into databases and metabolomic initiatives will be hindered.

• Platform: Make and model of both flow-system (i.e., U/HPLC) and detector (i.e., DAD or MS), and in the case of MS, source ionization interface (ESI API, etc.), ionization mode (positive or negative) and scanning type, for example, SIM, SRM/MRM, sMRM, etc.

• Voltage: Reporting collision energy is useful when replicating a method for the same instrument and is a common element captured in online MS and metabolomic data repositories and databases.

• Gradient: Mobile phase assay conditions, mobile phase composition, flow rate and gradient program.

• Column: Column manufacturer, make/packing material, particle size, pore size, ID and length and column oven temperature.

• HPLC/UPLC-auto-sampler: Temperature, injection volume and method run-time.

• Materials/reference standards: Chemical name and manufacturer of all reference standards (ideally product number or CAS numbers).

• Qualitative data analysis: Qualitative reference standard details, retention time, accurate or MS(n) m/z, or indicate “putative identification/annotation” where no reference standard was available.

• Quantitative data analysis: Include a linear range of reference standard curves and reference standard matrix composition. Reference standard curve linearity and integration methodology is useful for replication of experiments or verification of assay results.

• Assay Precision and accuracy (i.e., quality): % CV for internal standards either in a blank or matched matrix, frequency of assay blanks and quality control samples. Extraction efficiency details provide confidence in assay quality. LOD, LLOQ, HLQ and % CV of regression slopes provide further confirmation of data quality for studies comprising multiple batches of samples run over long periods of time.

Due to the diversity of (poly)phenol structures, their distribution and pre- and post-absorptive metabolism, substantial analytical considerations are required when developing or utilizing assays for berry phytochemicals. Multiple extractions and analytical techniques suitable for characterization and quantitation are presented herein. Sound method development, optimization, validation and reporting practices are required to capture the impact of environment, processing and metabolism on berry phytochemicals. Many texts have been devoted to providing guidance for such an initiative. Understanding best analytical practices, and most importantly, reporting practices will allow for the effective advancement of berry research. However, findings from studies can only be verified when they report details required to establish study or experiment quality and identify limitations. Identifying or realizing methodological limitations should be considered a strength, not a weakness, as all science is limited by the capacities of the present technology/instrumentation and available funding. Knowing the limitations of the analytical instrumentation used for identification and quantitation is required to prevent over-interpretation of study findings. Reporting key methodological details and process limitations will allow for efficient/effective progression of the berry health research field.

IM

Ion mobility

LC–NMR

liquid chromatography–NMR

MALDI-TOF

matrix-assisted laser desorption/ionization-time of flight

CV

coefficient of variation

ESI-MS

electrospray ionization mass spectrometry

HPLC

high pressure/high-performance liquid chromatography

HILIC

hydrophilic–lipophilic interaction chromatography

LC–MS

liquid chromatography–mass spectrometry

LLE

liquid–liquid extraction

MRM

multiple reaction monitoring

NMR

nuclear magnetic resonance spectroscopy

QTOF

quadrupole time of flight

sMRM

scheduled multiple reaction monitoring

SPE

solid-phase extraction

SLE

solid–liquid extraction

TFA

trifluoroacetic acid

QqQ

triple quadrupole mass spectrometer

UPLC/UHPLC

ultra-performance liquid chromatography/ultra-high-pressure liquid chromatography

• Counterintuitively, lower injection volumes typically provide more consistent data, particularly when transferring methods from HPLC to UPLC or to columns of lower ID and pore size. Less volume on-column lends to reduced peak width, less impact of the sample diluent, higher chromatographic resolution, more consistent peak integration and greater peak height and intensity.

• For modern UPLC columns, optimizing methodology for relatively high and consistent pressure leads to more reliable chromatographic resolution, better peak shape (less tailing and fronting), reduced peak width, greater MS peak intensity and reduced retention time drift (i.e., greater precision/reproducibility).

• Caution should be taken when characterizing structures using any detection technique, particularly in instances where chemical reference standards are unavailable (aside from perhaps NMR) or where compounds cannot be appropriately chromatographically resolved. In such cases, a single spectral feature (e.g., exact mass, wavelength or mass transition) cannot be used to definitively identify a structure; multiple scanning experiments or platforms are required for structural conformation in these instances.

• Solubility is influenced by extraction time and temperature.

• Extraction efficiency is dependent on solute-to-solvent ratio, particle size, bonding properties of matrix components, pH and temperature.

• Sequential extractions with apolar solvents (defatting) should be used prior to full LLE or SLE to minimize potential interferences from lipid-soluble components.

• During any extraction process, one should be aware that many phenolic compounds are prone to oxidation and even condensation. Shorter extraction times at lower temperatures (<50 °C)are recommended, as is the use of antioxidants such as ascorbic acid.

• Extraction of low-molecular-weight (poly)phenols is often more efficient in solvents such as methanol, while higher-molecular-weight polymers generally require less polar solvents such as acetone.⁷⁴ 

The use of:

• reference standards spiked into SPE samples and solutions at loading and post-elution allows for rapid method development and validation.

• nitrogen or argon positive pressure manifolds can provide faster methods over the use of vacuum manifolds, in addition to improving sample stability, recovery and reducing variability (i.e., from evaporation, sample loss, column obstruction, etc.).

• 96-well SPE along with loading plates and multi-channel pipettes reduces variability over single column/cartridge methods.

• high-quality 96-well septum/sealing mats reduce evaporation and allow for direct transfer to instrument auto-samplers for rapid analysis.

• LLE, in combination with SPE as a final cleanup, has the advantage of limiting the abundance of known MS ion suppressors in sample eluents, thereby enhancing the quality of the final LC–MS analysis of samples.

• A review of published literature methods to establish variables that differ helps establish typical ranges in variables utilized, allowing for testing of parameters across the reported ranges, such as low, medium and high concentrations, times, temperatures, pH, etc.

• Test variables in a predefined order and stick to a predetermined plan. A second method development experiment can refine variables that were less than effective.

• Have an a priori plan for judging findings, including a hierarchy of what variables will be compared and what outcomes are primary vs. secondary (i.e., absolute recovery, precision, accuracy, etc.).

• The use at least three replicates of every variation at minimum to establish % CV (5–7 replicates are commonly recommended).

• The use of 96 well plates allows for making multiple comparisons across method parameters in a single method development experiment.

• Retain solvent phases or SPE eluent which would normally be discarded until results are analyzed as these can be used for troubleshooting if findings are ambiguous.

• Overcompensate for extraction solvent volume, number of replicate extractions or column washes or incubation time (temperature dependent), etc., as these can be refined at the method optimization stage.

• It is recommended to establish method validation after a method has been developed and optimized. If replicates and internal standards are used and collected at every stage, along with sufficient matrix blanks, data will be available to establish a number of method validation parameters from the development and validation experiments, thus streamlining further validation efforts if required.

• Having defined quality control/assurance (QC/QA) samples (standards, extracts, etc.) and blanks to run before, during, and after critical runs can facilitate early detection of performance issues (i.e., loss of sensitivity, retention time shifts, carryover, etc.) when running unknown samples, or large continuous runs of samples.

• When attempting to characterize a biological conjugate using MS, and no reference standard exists, for example, cyanidin-3-glucuronide, optimize fragmentation voltages for the aglycone (i.e., cyanidin), and a comparable glucoside (i.e., cyanidin-3-glucoside) as voltages required to fragment the precursor and major product (i.e., aglycone) are often similar.

• Compounds putatively identified or compounds which do not have an available reference standard can still be optimized for voltage and sMRM parameters (i.e., transitions, time windows, integration thresholds, etc.). Voltage ramping experiments can be conducted to establish optimized collision energies and automated or manual flow injection analysis experiments using concentrated and dilute samples allow for establishing sMRM parameters, linearity and matrix effects.

• Relative quantitation of compounds putatively identified or compounds which do not have available reference standards can still be established by assigning regression equations from compounds of high structural similarity or by using published response factors if available.

• Internal and external standards are highly recommended during all method development, optimization and validation procedures. Ideally, unique internal standards would be added at various stages of the process. Internal standards are added at a constant concentration to all samples, including blanks and calibration standards for correction of analyte loss or performance drift. Internal standards are also important to prove or improve the precision of an assay and aid in method development and method/instrument troubleshooting. Internal standards are particularly useful in establishing losses during assay processes, such as extraction, evaporation, centrifugation, etc. An external standard is similar to an internal standard but is run alone to account for instrument performance, matrix effects or to generate standard reference regression curves. Often, internal and external standard responses are compared to establish method and instrument compatibility, drift and for troubleshooting and establishing inter- and intra-user and -laboratory validation.

Optimize:

• for minimal injection volume to prevent column “overload”, and provide improved peak shape, reduced carryover, and prevent ion suppression

• matrix “cleanup” procedures to remove ion suppressors from the background matrix

• sample composition (aqueous-to-organic ratio) to improve atomization and volatility

• source parameters such as gas, temperature, and x- and y-axis to prevent excessive ion entrance into Q0

• peak shape by altering pressure trace (i.e., flow) to prevent large dips in system pressure known to broaden peak shape and cause fronting and tailing

• integration thresholds suitably above background signal/noise to avoid integration of background signals and irregular fronting and tailing peaks

• if using low injection volumes, sample organic solvent composition has a limited impact on chromatography

• clean source cone regularly to prevent ion contamination

• change capillary frequently to prevent arching or spray pattern degradation

• avoid excessive reference standard curve ranges by establishing regressions that only capture expected concentrations to improve reproducibility at limits of detection (LLOQ or HLOQ)

• Most parameters of precision and accuracy required for assay validation can be established in a single 96-well plate experiment, by running multiple blanks, reference standards and calibrations. It is recommended to separate reference standards by blanks, along with replicate injections of analytes at the assay lower quality control (LQC), medium quality control (MQC) and high quality control (HQC) limits.

• How a reference standard is dissolved can impact its assay performance. Therefore, it is recommended to always look up known solubility prior to dissolving a reference standard, particularly if it is of low quantity and high expense.

• When dissolving reference standards for MS use, it is important to use the highest quality MS grade solvents (including water) to prevent instrument and sample contamination.

• Primary stock solutions of reference standards can often be dissolved in dimethyl sulfoxide (DMSO) to prolong stability.

• If solubility dictates DMSO or water be used in high abundance in primary stock solutions, secondary stocks should dilute using more volatile solvents, such as acetonitrile or methanol, as water and DMSO can negatively impact declustering and ionization in ESI-MS; particularly DMSO, where it is recommended that final working concentrations remain below 5% (ideally <1–2% for quantitative analysis). High water abundance is less problematic, particularly if not injected directly into the source. If syringe infusion is being used for compound optimization, increasing the source temperature and declustering potential will improve atomization and ionization.

• Matrix-matched blanks and reference standard calibration curves (i.e., standard curves or serial dilutions) are required for quantitative mass spectrometry due to matrix suppression of ionization. Furthermore, and similar to colorimetric detection, the background matrix of a sample negatively impacts background noise, peak integration and, therefore, accurate quantitation.

• The gold standard method for establishing a suitable assay matrix involves the use of a pool of all samples to be analyzed for background matrix effects. In situations where this is not possible, a pooled subset of baseline samples is often utilized or a commercial matrix (i.e., pooled multi-donor serum, plasma, urine, tissue, etc.) or universally accepted synthetic matrix.^50,75,76 

• Quantitative accuracy in mass spectroscopy cannot be established if a matched-matrix is not utilized or without significant matrix validation, such as first establishing no matrix effects exist and the sample matrix has dilution integrity.

• Investigators reporting quantitative data for compounds that do not have commercially available references standards should identify these instances as “putative quantitation” (or semi-quantitative or relative quantitation).

• In the absence of commercially available SRM, investigators should consider establishing an “in-house” reference food or biological materials by pooling of materials, which can be used routinely as control samples to monitor changes in extraction efficiencies and/or LC/LC–MS performance.

• Percent recovery and matrix effects (described above) should be determined for each phenolic species assayed. However, the absence of analytical standards for some phenolic metabolites and many microbial and host metabolites precludes this. In these cases, it is recommended that one use recovery and matrix effects from the most structurally similar phenolic species where authenticated standards are available.

• Certified (poly)phenolic standards should be used to establish recovery and may be added to the analyte-free matrix (when available) at 1–2 times the concentration range anticipated in the test products.

• If a (poly)phenolic-free matrix is not available, such as often in the case of berries, a standard addition method can be used to establish accurate recovery.

• Recovery and matrix effects can be determined by plotting the response against actual concentrations of the reference standards, using pre- and post-extraction spiking with internal standards, and comparing peak areas, heights or area-to-intensity ratios of analytes within a blank and true sample matrix.

• It is important to consider acid/pH modifiers when transitioning from HPLC-UV–Vis to MS detection. pH modifiers are often used in the HPLC-UV–Vis mobile phase as an organic modifier to improve peak shape and chromatographic separation. This often affects chemical interactions between the stationary and mobile phase, which is particularly apparent for anthocyanin analysis.

• Mobile phase pH modifiers can have a significant impact on ionization efficiency (providing a charged state) and result in adduct formation, hampering identification and quantitation.

• It is most common to use formic acid as a volatile organic modifier in (poly)phenol research as it improves peak shape while having minimal impact on ionization efficiency relative to other acids such as TFA and HCl, which are well-known ion suppressors, especially in positive mode.

• 1–5% formic acid by volume in the mobile phase will generally provide narrow chromatography peak widths of 0.3–0.5 min (concentration dependent) and lends to optimal chromatography of polyphenols using UV–Vis detection methods but generally leads to ion suppression and loss of sensitivity using MS detectors.

• Formic acid levels of 0.1–1% will lead to increased sensitivity in MS but often at the detriment of peak width, particularly for anthocyanins. This can often be overcome by moving to a higher flow/pressure system or narrow bore column, which can narrow peak width (injection volume dependent). This dichotomy between impacts of acid on resolution and sensitivity is often referred to as the “acid effect” and is particularly prevalent for anthocyanins.

• It is important to note the concentrations of acid in the mobile phase in publications, as it can impact LOD and LLOQ, and potentially quantitatively accuracy.

• When transitioning from detection methods such as HPLC-UV–Vis to MS, it is often assumed that MS will provide a higher level of qualitative and quantitative certainty. However, MS comes with an added level of uncertainty, particularly with respect to polyphenols, which undergo many reactions (gas-phase artifacts) in the high-energy gas-phase of the MS, such as molecular rearrangement, Retro-Diels–Alder reactions, quasimolecular ion formation, radical formation, etc. Because of these confounders, misidentification of precursor polyphenols and their phase II metabolites is assumed to be highly prevalent when using a single transition or spectral feature for identification.

CDK was supported by the USDA National Institute of Food and Agriculture (Hatch/Kay-Colin; 1011757).