Understanding the Effects of Roasting on Antioxidant Components of Coffee Brews by Coupling On‐line ABTS Assay to High Performance Size Exclusion Chromatography

Chemicals and reagents

5‐Caffeoylquinic acid (5‐CQA) (> 99%) was purchased from Acros Organic (Geel, Belgium), 5‐feruoyl quinic acid (5‐FQA) and 3,5‐di‐caffeoyl quinic acid (3,5‐Di‐CQA) from Chengdu Biopurify Phytochemicals Ltd (Chengdu, Sichuan, China), and gallic acid (GA), ABTS^•+, fluorescein, 2,2′‐azobis(2‐methylpropionamidine) dihydrochloride (AAPH), FC reagent, salts and solvents from Sigma‐Aldrich (Buchs SG, Switzerland).

The FC reagent for the assay was prepared by 10× dilution of the commercial 2 M (with respect to acid) FC reagent. ABTS^•+ solutions for the on‐line analysis were synthesised by oxidation with potassium persulphate (Re et al., 1999); equal volumes of 7 mM ABTS and 2.45 mM potassium persulphate solutions were mixed and the solution stored before use for at least 15 h at 4 °C. Fluorescein for the ORAC assay was prepared in a 0.37 μg/mL concentration and mixed with cooled AAPH solution. Standard solutions were prepared by dissolving the standards in water followed by appropriate dilution of the stock solution with water.

Coffee roasting

Two species of coffee [Coffea arabica (Arabica) from Costa Rica and C. canephora (Robusta) from Vietnam] were used in this study, and all samples were roasted in a 20 kg batch roaster (Roastmaster™20, Bühler AG, Uzwil, Switzerland). The roasting process was varied in order to produce a variety of roasted coffees covering the whole roasting range. Eight kilograms of green coffee beans were roasted in four profiles: fast (5–6 min), fast‐medium (8–9 min), medium‐slow (11–12 min) and slow (15–18 min), each to light, medium and dark roast degrees (see Tables 1 and 2). In order to obtain these samples, a large number of trial roasts were performed to establish highly reproducible roasting conditions (see Fig. 1 for the Arabica roasting trials; Robusta data not shown). Subsequently, five batches were roasted and the antioxidant analyses were performed on the three most similar ground samples as determined with a colour measuring instrument Colortest II (NeuhausNeotec GmbH, Reinbek, Germany). In addition, mass loss and the humidity content were analysed by gravimetric analysis.

Figure 1

Time–temperature profiles of the different roast trials (N = 3 for each roast degree and N = 9 for each roast profile) that were conducted in this roasting campaign withCoffea arabicacoffee. Coffee beans were roasted in four different roast profiles from fast, fast‐medium, medium‐slow to slow, as it is visible with increasing roasting time. Furthermore within each profile coffee beans were roasted to three roasting degrees (light, medium and dark) as it is visible with the increasing approximate temperature of the product. [Colour figure can be viewed at wileyonlinelibrary.com]

Coffee brew preparation

After storing for 12 days in a refrigerator at 4 °C, the coffee samples were ground with an espresso coffee grinder (Ditting KED 640, Bachenbülach, Switzerland) at a coarseness setting of seven. Samples were brewed in triplicate for the antioxidant assays using a 200 mL French press (Bodum, Triengen, Switzerland) in which 12 g of ground coffee powder was infused with approximately 200 mL [mean value 190.5 g ± 1 standard deviation (SD)] of hot water at 92 °C and stirred briefly. The coffee brew was then extracted for 4 min, before it was poured into 250 mL flasks (Schott, Germany) to let cool to room temperature. The extract was also diluted to adjust concentrations to the specific dynamic range of the antioxidant assays.

Antioxidant measurements

HPSEC coupled to on‐line ABTS antioxidant assay

Coffee brews were diluted with water (1:2.5), filtered (0.45 mm PET filters) and separated on a HPLC system (Agilent Series 1200, Santa Clara, CA, USA) using Supermultipore PW‐N HPSEC columns (TSKgel, Tosoh Bioscience, Stuttgart, Germany); the effluent was then coupled to the ABTS assay (scheme in Fig. 2) as described and validated by (Smrke et al., 2013). For size exclusion chromatography (SEC), all measurements were performed on a HPLC system (Agilent Series 1200, Santa Clara, CA, USA) with an eluent of 0.1 M ammonium phosphate buffer adjusted to pH 7 pumped isocratically at 0.4 mL/min, a sample injection volume of 5 μL, and a column oven temperature of 30 °C. The eluent was monitored at 210, 280, 325 and 405 nm using a diode array detector (DAD). An Ismatec ISM8273 peristaltic pump (IDEX Health and Science SA, Glattbrugg, Switzerland) was used for delivering the ABTS^•+ reagent with a flow rate of 0.12 mL/min to the effluent via a T‐element after the DAD. The effluent was then passes through a reaction coil at 60 °C in an additional HPLC column compartment, which produces a 30 s delay to adjust and control the reaction conditions. The formation of the reduced ABTS product was monitored with a MWD (multiple wavelength detector, Agilent Series 1100) at 734 nm. Initially, the ABTS^•+ concentration was manually set to get an initial baseline absorbance of 0.80 AU (absorbance unit) in order to ensure comparable starting conditions for all experiments. In each method the results are expressed relative to GA (in mg/L), which was used as a standard reference material.

Figure 2

Measurement of decolourisation of ABTS^•+ using HPSEC to analyse fast medium roasted Arabica coffee in three roast degrees (light, medium, dark). (a) Flow scheme of the post‐column derivatisation ABTS assay coupled to HPSEC, (b) chromatograms to illustrate the contribution of high molecular weight compounds, melanoidins (4–13 min) and low molecular weight compounds, chlorogenic acids (13–20 min) to the total antioxidant capacity (full integral) of the three roast degrees. [Colour figure can be viewed at wileyonlinelibrary.com]

Batch assays with flow injection analysis

For comparison, two validated assays, FC and ORAC, were employed. The FC assay used a flow injection analysis (FIA) method, while the ORAC assay used a stop flow sequential injection analysis (SIA) method (for details see Opitz et al., 2014). In both cases, the coffee extracts were diluted to adjust the concentration to the specific dynamic range of the assay (1:50 for FC and 1:400 for ORAC), filtered using 0.45 mm PET filters and analysed with a FIAlab 3200 (FiALab Instruments Inc., Bellevue, WA, USA) instrument. In each method the results are expressed as GA equivalents (mg GAE/L) relative to GA, which was used as a standard reference material.

CGA analysis

Green coffee brews were prepared in triplicate as described earlier, and analysed directly by HPLC‐MS (1200 series and 6130 series, Agilent Technologies, Santa Clara, CA, USA). CGAs were separated on a Luna (2) C18 column (3 mm × 150 mm, 3 μm; Phenomenex, Torrance, CA, USA) by running a gradient using eluent A (95% water and 5% acetonitrile, 0.1% FA) and eluent B (95% acetonitrile and 5% water, 0.1% FA), starting with 100% A, then to 98% A in 20 min, to 89% A in 15 min and to 35% A in 10 min, followed by a cleaning and regeneration cycle. Absorbance at 325 nm was used for quantification, and an external calibration curve was prepared using 5‐CQA, 5‐FQA and 3,5‐Di‐CQA as representatives of the three main CGA groups of caffeoyl‐quinic acids, feruoylquinic acids and di‐caffeoylquinic acids. CGA contents were also estimated by HPSEC for green and roasted coffee extracts by integration of area under the curve from 13 min to 20 min, taking the values recorded at 325 nm after calibrating with 5‐CQA.

Statistical analysis

Statistical data analyses were performed with R. One‐way analysis of variance (ANOVA) followed by a pairwise t‐test and Holm adjustment were used to compare the antioxidant values for all roast degrees among the roast profiles. For graphical visualisation of the data SigmaPlot 11.0 (Systat Software) was used.

Coffee species

For each of the three assays, the results from the Robusta beans were significantly higher than those from the corresponding samples from Arabica beans. Comparing the antioxidant values between both coffee species over the wide range of different roasts, GAE (in mg/L) antioxidant values obtained with the ABTS on‐line assay were on average 1.32 (± 0.07 SD) times higher for Robusta than those for the corresponding Arabica coffee, 1.2 (± 0.04 SD) times for the FC assay and 1.5 (± 0.13 SD) times for the ORAC assay. The assays approximately reflected the 1.5 times higher content of CGAs in green beans of Robusta where the concentrations were 3.09 ± 0.07 g/L in Arabica and 4.72 ± 0.3 g/L in Robusta and are consistent with other reports (e.g. Perrone et al., 2008). Besides the three major caffeoylquinic acid‐isomers (5‐CQA, 3‐CQA, 4‐CQA), three feruoylquinic acids (3‐FQA, 4‐FQA and 5‐FQA) and four di‐caffeoylquinic acids (3,4‐Di‐CQA co‐eluting with 1,5‐Di‐CQA, 3,5‐Di‐CQA, 4,5‐Di‐CQA) were included in the quantification. Quantification of CGAs based on HPSEC‐DAD integral of green Arabica coffee resulted in values of 2.92 ± 0.15 g/L, which is comparable to the HPLC‐DAD concentrations. Although green Robusta coffee contained higher absolute antioxidant values, it also experienced a more pronounced loss of antioxidant capacity during roasting. The range of values between coffees roasted at fast or slow speeds was usually higher for Robusta coffee, e.g. 560 GAE for dark/fast to 430 for dark/slow roasted Robusta coffee compared to 390 GAE for dark/fast to 330 GAE for dark/slow roasted Arabica coffee. This higher relative degradation in Robusta beans can be explained by their higher CGA content, which is prone to degradation during the roasting process. Similarly, the ORAC values for dark/slow roasted Robusta coffee were only around ~75% of the level for light/fast‐medium roasted coffee, whereas dark/slow roasted Arabica coffee was still ~90% of that of the dark/fast coffee.

Antioxidant assays

All three assays showed similar trends with coffee having lower antioxidant values when roasted to darker roasts, which is explainable by degradation of antioxidant active compounds (Vignoli et al., 2014).

HPSEC coupled to ABTS antioxidant assay

The flow scheme as well as chromatograms of light, medium and dark roast degrees of the fast‐medium roast profile are shown in Fig. 2 to illustrate the performance of HPSEC separation coupled to the ABTS assay. On‐line measurements resulted in an overall decrease of antioxidant capacity towards longer roasting times (Fig. 3(a) for Arabica and Fig. 4(a) for Robusta coffee) (Smrke et al., 2013; van der Werf et al., 2014). This decrease was significant regarding roast degree from light to dark roasts of both coffee species, but it was less prominent regarding roast profile, e.g. the light roasts of all four roasting speeds were not different for Arabica. Since Arabica coffee roasted to a light roast did not exceed a maximum temperature of 200 °C during roasting, which would indicate that only above that temperature degradation processes have an effect on ABTS^•+ antioxidant activity (Fig. 1). For Robusta coffee, light roasts differed in their antioxidant capacity and light/fast as well as light/slow coffees had higher antioxidant capacity than the medium roasts, although the roast profiles (time–temperature profiles) were different in order to obtain similar roast degrees to the Arabica coffees (see Tables 1 and 2). For other roast degrees, roasting speed affected antioxidant capacity, e.g. for Arabica coffee from around 390 GAE for dark/fast roast to 330 GAE for dark/slow roast (Figs 3(a) and 4(a)).

Adding the additional dimension of HPSEC enabled separate quantification of HMW (essentially melanoidins) and LMW (mainly CGAs) fractions. Assessing the contribution of HMW and LMW fractions to the total antioxidant capacity separately, it became obvious that the HMW fraction always increased with increasing darkness of the roast, despite the overall decrease of antioxidant capacity (Fig. 5(a) for Arabica and Fig. 5(b) for Robusta). This increase with roasting degree is based on increased melanoidin formation, and partly also because of CGA incorporation into melanoidins (e.g. Perrone et al., 2012), which was evident for each roasting profile, but with lesser magnitude from fast to slow. The steepest increase of HMW‐based antioxidant capacity was thereby seen within the fast roasting profile, which indicates that in addition to roast duration, heat load per unit time also increases melanoidin production possibly due to polymer expansion with higher CGA incorporation together with less degradation of the catechol moiety (Echavarría et al., 2012). However, after an initial increase from light to medium roast degree for the slow Robusta profile, antioxidant capacity of the HMW decreased again slightly towards slow/dark roast (Fig. 5(b)), suggesting that the melanoidins are also subject to degradation by heat, and that (some of) their components lose antioxidant activity during roasting. This increase in HMW melanoidin content already at early roast stages could explain why the light Robusta roasts differed between the four profiles, whereas no increase was visible for Arabica coffee (Fig. 5).

Figure 5

Measured antioxidant capacity as gallic acid equivalents (in mg/L) for (a)Coffea arabicaand (b)C. canephoracoffee beans as a function of roasting time and degree using the ABTS assay coupled on‐line to HPSEC. For comparison of high to low molecular weight compounds, high molecular weight compounds (H) are given on the left and low molecular weight compounds (L) on the right side of the figure. The roasting times varied according to the roasting profiles; which were fast: 5.27–6.21 min; fast‐medium: 8.08–9.10 min, medium‐slow: 11.03–11.43 min, and slow: 15.12–17.53 min. For each roast profile three roast degrees were obtained, the light roast denoted with a light brown colour, the medium roast with a medium brown colour and the dark roast with a dark brown colour. All roasts were roasted in triplicate and each roast brewed in triplicate (N = 9). Statistical differences were compared using ANOVA followed by a t‐test with Holm adjustment, different letters denote significant differences.

However, the most dominant effect of roasting is the steep decline in LMW compounds with increasing darkness level for all roast profiles and for both coffee species (Fig. 5(a) and 5(b)). This decrease is caused by a decrease in CGA concentrations that are either destroyed or partly incorporated into melanoidins, e.g. in Arabica from 1.4 g/L (light roast/fast profile) to 0.36 g/L (dark roast/slow profile) based on the HPSEC‐DAD integral at 325 nm. The observation of the lowest concentration of CGA in dark/slow roasts also indicates that lower temperatures over a longer time span have a higher impact on CGA degradation than short roasting at higher temperatures. Furthermore, LMW compounds are already prone to degradation at very early stages of the roasting process and the highest antioxidant activities are observed at even lighter roast degrees (Opitz et al., 2014). In addition to incorporation of CGAs in melanoidins, many other new compounds with antioxidant capacity are generated during roasting, especially structural isomers of CGAs (van der Werf et al., 2014).

Relationships between the assays

Increasing roasting time and darkness of roast clearly correlate with lower antioxidant capacities of the coffee brews, as shown for on‐line ABTS as well as FC assay and partly for the ORAC assay. The effect was most clearly visible with the FC assay, which also had the highest precision and lowest variation among replicates (Opitz et al., 2014). The differences observed with the ORAC assay were mostly only of a minor extent, but darker roasts of Robusta coffee showed clearly lower antioxidant values.

More importantly, coupling an antioxidant assay to HPSEC revealed that the relative contents of HMW and LMW compounds are dependent on both the darkness of the roast and speed of roasting, as well as on the coffee bean genetics. By integrating HMW and LMW compounds separately, opposing trends for melanoidins and CGAs during roasting were revealed and CGAs were contributing more to the total antioxidant capacity for the light roasts, whereas melanoidins contributed more at medium and dark roasts. Thus whereas CGAs are transformed or degraded during roasting under mild conditions and reduced by half for the slow/dark roast, melanoidins show increasing antioxidant capacity with darkness of roast; this increase is likely also due to incorporation of CGAs in the melanoidin backbone, but there are probably also contributions from other unknown mechanisms.

While the incorporation of CGAs in melanoidins has been confirmed by many studies, inconsistent results have been published to what extent melanoidins are contributing to total antioxidant capacity (Perrone et al., 2012). Hence, coupling chromatography to antioxidant assays made it possible to reliably quantify only part of the antioxidant mixture, thereby lowering the competitive effect of complex matrices in batch assays. Thus far, mainly reversed phase chromatography methods have been employed to detect individual compounds with antioxidant activity (e.g. De Smet et al., 2015). However, detection and quantification of individual CGAs with HPSEC was possible only to a limited extent due to much lower peak capacities in SEC.

Armed with such knowledge it is possible to adjust the roasting conditions to optimise the overall antioxidant capacity of coffee beverages, and the relative proportions of the major types of antioxidant molecules. As a result of the formation of the highest amounts of melanoidins and lowest amounts of CGA degradation, fast roasting at high temperatures results in a coffee with high antioxidant capacity and low degradation of active compounds, whilst the actual roast degree can be selected to satisfy the preference of the consumer. The speed of roasting does have substantial effects on coffee bean composition and increasing the heat transfer within short roasting times would presumably lead to even higher antioxidant values for fast roasted coffee, although extremely high gas flow rates increases the prospect of product oxidation (Pascual et al., 2002).