Food Chemistry 185 (2015) 135–144

Contents lists available at ScienceDirect

Food Chemistry

journal homepage: www.elsevier .com/locate / foodchem

Chlorogenic acid–arabinose hybrid domains in coffee melanoidins:Evidences from a model system

http://dx.doi.org/10.1016/j.foodchem.2015.03.0860308-8146/� 2015 Published by Elsevier Ltd.

⇑ Corresponding author. Tel.: +351 234 370 698; fax: +351 234 370 084.E-mail address: mrd@ua.pt (M.R.M. Domingues).

Ana S.P. Moreira a, Manuel A. Coimbra a, Fernando M. Nunes b, Cláudia P. Passos a, Sónia A.O. Santos c,Armando J.D. Silvestre c, André M.N. Silva d, Maria Rangel e, M. Rosário M. Domingues a,⇑a QOPNA, Department of Chemistry, University of Aveiro, 3810-193 Aveiro, Portugalb CQ-VR, Chemistry Research Centre, Department of Chemistry, University of Trás-os-Montes e Alto Douro, 5001-801 Vila Real, Portugalc CICECO, Department of Chemistry, University of Aveiro, 3810-193 Aveiro, Portugald UCIBIO/REQUIMTE, Department of Chemistry and Biochemistry, Faculty of Sciences, University of Porto, 4169-007 Porto, Portugale UCIBIO/REQUIMTE, Instituto de Ciências Biomédicas de Abel Salazar, University of Porto, 4050-313 Porto, Portugal

a r t i c l e i n f o

Article history:Received 12 December 2014Received in revised form 13 February 2015Accepted 17 March 2015Available online 28 March 2015

Keywords:RoastingPolysaccharidesArabinogalactansCaffeoylquinic acidTransglycosylationMass spectrometry

a b s t r a c t

Arabinose from arabinogalactan side chains was hypothesized as a possible binding site for chlorogenicacids in coffee melanoidins. To investigate this hypothesis, a mixture of 5-O-caffeoylquinic acid (5-CQA),the most abundant chlorogenic acid in green coffee beans, and (a1 ? 5)-L-arabinotriose, structurallyrelated to arabinogalactan side chains, was submitted to dry thermal treatments. The compounds formedduring thermal processing were identified by electrospray ionization mass spectrometry (ESI-MS) andcharacterized by tandem MS (ESI-MSn). Compounds composed by one or two CQAs covalently linked withpentose (Pent) residues (1–12) were identified, along with compounds bearing a sugar moiety but com-posed exclusively by the quinic or caffeic acid moiety of CQAs. The presence of isomers was demonstratedby liquid chromatography online coupled to ESI-MS and ESI-MSn. Pent1–2CQA were identified in coffeesamples. These results give evidence for a diversity of chlorogenic acid–arabinose hybrids formed duringroasting, opening new perspectives for their identification in melanoidin structures.

� 2015 Published by Elsevier Ltd.

1. Introduction

Melanoidins are formed during thermal processing of severalfood products, such as coffee, bakery products, malt and beef.Due to the uncertainties about their structures, they are genericallydefined as high molecular weight nitrogenous brown-colored com-pounds generated in the final stage of the Maillard reaction. Also,they are usually quantified by difference, subtracting from the totalthe percentage of known compounds. Using this criterion, theywere estimated to account for up to 25% (w/w) of roasted coffeebeans dry weight (Moreira, Nunes, Domingues, & Coimbra, 2012).Coffee brews, prepared by hot water extraction from ground androasted coffee beans, are considered one of the main sources ofmelanoidins in human diet (Fogliano & Morales, 2011). Several bio-logical activities have been associated to coffee melanoidins(Moreira et al., 2012), but more work, namely on their structuralcharacterization, is needed to better understand their healtheffects.

Since at least the 1960s (Maier, Diemair, & Ganssmann, 1968),several attempts have been made to elucidate the structure of cof-fee brew melanoidins. In recent years, their structural diversity hasbeen evidenced with the purification of different melanoidin pop-ulations by applying chromatographic separation techniques andother specific isolation procedures. On the other hand, thechemical characterization of these purified melanoidin populationshas given increasing evidences that polysaccharides, proteins, andchlorogenic acids are involved in the formation of coffee melanoi-dins (Bekedam, De Laat, Schols, Van Boekel, & Smit, 2007;Bekedam, Loots, Schols, Van Boekel, & Smit, 2008; Bekedam,Schols, Van Boekel, & Smit, 2008; Coelho et al., 2014; Gniechwitzet al., 2008; Nunes & Coimbra, 2007, 2010). However, it is stillunclear how these different constituents (or their derivatives) arelinked in the melanoidin structures.

Chlorogenic acids consist of a quinic acid (QA) moiety esterifiedto one or more trans-cinnamic acids, such as caffeic, p-coumaric,and ferulic acids (Clifford, 2000). In green coffee beans, the mostabundant chlorogenic acid is 5-O-caffeoylquinic acid (5-CQA)(IUPAC, 1976), a caffeic acid (CA) ester (Moon & Shibamoto,2009; Perrone, Farah, Donangelo, de Paulis, & Martin, 2008). Thepresence of covalently-linked chlorogenic acid derivatives in coffee

136 A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144

melanoidin fractions was first demonstrated by using alkalinefusion, an efficient method to release condensed phenolic struc-tures (Nunes & Coimbra, 2007; Takenaka et al., 2005). The incor-poration of covalently-linked chlorogenic acid derivatives,namely ester-linked phenolic and QA moieties, as well as the pres-ence of intact chlorogenic acids incorporated via CA moietythrough mainly non-ester linkages, was corroborated by subse-quent studies (Bekedam, Loots, et al., 2008; Bekedam, Schols,et al., 2008; Coelho et al., 2014; Perrone, Farah, & Donangelo,2012).

Galactomannans and type II arabinogalactans, the most abun-dant polysaccharides in green coffee beans (Bradbury & Halliday,1990), were also identified in coffee melanoidin fractions(Bekedam, Schols, van Boekel, & Smit, 2006; Bekedam et al.,2007; Nunes & Coimbra, 2007; Passos et al., 2014). In particular,arabinose from arabinogalactan side chains was hypothesized asa possible binding site for chlorogenic acid derivatives (Bekedam,Schols, et al., 2008). This hypothesis was proposed based on pre-vious studies demonstrating that the arabinose is quite susceptibleto degradation induced during roasting (Oosterveld, Voragen, &Schols, 2003; Redgwell, Trovato, Curti, & Fischer, 2002; Totlani &Peterson, 2007). However, no evidences have been reported ofthe presence of chlorogenic acids covalently linked to the arabi-nose side chains of the arabinogalactans incorporated in coffeemelanoidin structures. In order to investigate this hypothesis, anequimolar mixture of 5-CQA and (a1 ? 5)-L-arabinotriose (Ara3),an oligosaccharide structurally related to arabinose side chains ofarabinogalactans, was submitted to dry thermal treatments, mim-icking coffee roasting conditions. The compounds formed duringthermal processing were identified by direct electrospray ioniza-tion mass spectrometry (ESI-MS) analysis. The identification ofthese compounds was confirmed by determination of elementalcompositions using high resolution MS and their fragmentationpattern under tandem MS (ESI-MSn). High-performance liquidchromatography (HPLC) with photodiode array detection (PDA)online coupled to ESI-MS and ESI-MSn was also used to investigatethe presence of structures having the same elemental composition(isomers) and thus not able to be differentiated by direct MS analy-sis. In order to support the formation of chlorogenic acid–arabinosehybrid structures during coffee roasting, fractions previouslyrecovered from spent coffee grounds (SCG) were also analyzed byESI-MS and ESI-MSn.

2. Materials and methods

2.1. Materials

5-O-caffeoylquinic acid (5-CQA), having a purity P95%, wasobtained from Sigma (St. Louis, MO, USA). (a1 ? 5)-L-arabinotriose(Ara3), having a purity P95%, was purchased from Megazyme(County Wicklow, Ireland). Ultrapure water was obtained from aMilliQ water purification system (Millipore, Billerica, MA, USA).Other solvents used were of HPLC grade.

2.2. Preparation of the mixture

Equimolar amounts of Ara3 (19.3 mg) and 5-CQA (16.5 mg)were dissolved in 200 mL of ultrapure water. After freeze-drying,the solid mixture was powdered with an agate mortar and pestleand then stored in a desiccator containing P2O5 until use.

2.3. Thermal treatments

The 5-CQA and its mixture with Ara3 were submitted to differ-ent temperature programs using a TGA-50 thermogravimetric

analyzer (Shimadzu, Kyoto, Japan), operating with a controlledair flow of 20 mL/min and a heating rate of 10 �C/min. To studytheir thermal stability, samples of 2–4 mg were submitted to atemperature program from ambient temperature to 600 �C. Tostudy the roasting-induced products, samples of the 5-CQA andthe mixture (5–8 mg) were also subjected to the following thermaltreatments: heating from room temperature to 175 �C (175T1),heating from room temperature to 175 �C with additional 30 minat this temperature (175T2), and heating from room temperatureto 200 �C (200T1). The roasted samples were recovered, weighed,and dissolved in ultrapure water (5 mg/mL). To facilitate theirdissolution, they were stirred at 37 �C for 3 h, and then kept frozenat�20 �C until MS analysis. Solutions (1 mg/mL in ultrapure water)of the unroasted samples (T0) were prepared and stored under thesame conditions.

2.4. ESI-MS conditions

For all the ESI-MS analyses, samples in water were diluted inmethanol. As detailed below, three different mass spectrometerswere used. The quadrupole-time-of-flight (Q-TOF) spectrometerprovided positive ion MS spectra with a better signal/noise ratiothan those acquired using the linear ion trap (LIT), whereas thespectrometer which combines the linear ion quadrupole (LTQ)and Orbitrap mass analyzer was used due to its analytical perfor-mance in terms of resolution and mass accuracy. The full scanMS spectra were recovered in the m/z range of 100–1500 (or2000 using LTQ-Orbitrap instrument).

2.4.1. Q-TOF conditionsFor both untreated and thermally treated mixtures, positive ion

ESI-MS and ESI-MS/MS spectra were acquired using a Q-TOF 2hybrid instrument (Micromass, Manchester, UK). The flow ratewas set at 10 lL/min. The cone voltage was set at 35 V, and thecapillary voltage at 3 kV. The source temperature was adjusted to80 �C, and the desolvation temperature was 150 �C. The MS/MSspectra were obtained using argon as the collision gas, and the col-lision energy used was set between 30 and 43 eV. Data acquisitionand processing were carried out using MassLynx 4 data system(version 4.0).

2.4.2. LIT conditionsFor both untreated and thermally treated samples, negative ion

ESI-MS and ESI-MSn spectra were acquired on a LXQ linear ion trap(LIT) instrument (ThermoFinnigan, San Jose, CA, USA) using the fol-lowing conditions: spray voltage, 4.7 kV; capillary temperature,275 �C; capillary voltage, �22 V; tube lens voltage, �45 V.Samples were introduced into the source at 8 lL/min. Nitrogenwas used as nebulizing and drying gas. In the ESI-MSn experiments,the collision energy was set between 14 and 21 (arbitrary units).Data acquisition and processing were carried out using Xcaliburdata system (version 2.0).

2.4.3. LTQ-Orbitrap conditionsNegative ion ESI-MS spectra of the untreated mixture (T0) and

after its heating to 175 �C (175T1) were also acquired using aLTQ-Orbitrap XL mass spectrometer (ThermoFisher Scientific,Germany) controlled by the LTQ Tune Plus 2.5.5 software and theXcalibur 2.1.0 for data processing. The operating conditions wereas follows: sheath gas flow, 5 (arbitrary units); spray voltage,2.8 kV; capillary temperature, 275 �C; capillary voltage, �35 V;and tube lens voltage, �200 V.

A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144 137

2.5. HPLC-PDA-ESI-MS conditions

The separation of compounds in the mixture heated to 175 �C(175T1) was carried out on a Thermo Scientific Hypersil Gold RPC18 column (100 mm � 2.1 mm, 1.9 lm particle size) at controlledtemperature (15 �C) using a HPLC system equipped with an Accelaautosampler (set at 16 �C), an Accela 600 LC pump, and an Accela80 Hz PDA detector. The roasted mixture was dissolved in water:-methanol (50:50; v/v) and a volume of 10 lL was introduced intothe column, using a flow rate of 400 lL/min. The mobile phasesconsisted of water:acetonitrile (99:1, v/v) (A) and acetonitrile (B),both with 0.1% of formic acid. The percentage of B was kept at3% from 0 to 5 min, then reached 12% from 5 to 14 min, 12.8% from14 to 14.50 min, and 100% from 14.5 to 16 min. Before the follow-ing run, the column was re-equilibrated by decreasing the percent-age of B from 100% to 3% during 4.5 min, and held it at 3% for3.5 min. Double online detection was carried out in the PDA detec-tor, at 280 and 325 nm, and UV spectra were also recorded from210 to 600 nm. The HPLC was coupled to a LCQ Fleet ion trap massspectrometer (ThermoFinnigan, San Jose, CA, USA), equipped withan ESI source and operating in negative mode. The flow rate ofnitrogen sheath and auxiliary gas were 40 and 5 (arbitrary units),respectively. The spray voltage was 5 kV and the capillary tem-perature, 300 �C. The capillary and tune lens voltages were set at�28 V and �115 V, respectively. The CID-MSn experiments wereperformed on mass-selected precursor ions in the range of m/z100–2000. The isolation width of precursor ions was 1.0 massunits. The scan time was equal to 100 ms and the collision energywas optimized between 15 and 45 (arbitrary units), using heliumas collision gas. The data acquisition and processing were carriedout using Xcalibur 2.1.0.

2.6. Spent coffee grounds fractions

Spent coffee grounds (SCG) samples, previously dried at 105 �Cfor 8 h, were suspended in water in 1 g portion to 10 mL of water ina total volume of 70 mL in each one of 6 individual containers.Microwave irradiation was performed with a MicroSYNTHLabstation (Milestone Inc.), using operating conditions similar tothe ones previously described (Passos & Coimbra, 2013; Passos,Moreira, Domingues, Evtuguin, & Coimbra, 2014). The fractionsrecovered after a third microwave assisted extraction (MAE3) at170 �C for 5 min and a fourth microwave assisted extraction(MAE4) at 200 �C for 2 min were centrifuged at 15,000 rpm for20 min at 4 �C. The supernatant solution was filtered using MNGF-3 glass fiber filter, frozen, freeze-dried, and stored under ananhydrous atmosphere.

Each fraction was dissolved in the minimum amount of water,stirring during 10 min at room temperature, and then absoluteethanol was added to reach an aqueous solution containing 75%ethanol (v/v). The solution was then centrifuged at 15,000 rpmfor 10 min at 4 �C and the precipitated material recovered(PptEt). To check the presence of chlorogenic acid–arabinose struc-tures, samples (2 mg) of MAE3_PptEt and MAE4_PptEt were dis-solved in 1 mL of ultrapure water and then kept frozen at �20 �C.Before MS analysis, they were filtered using 0.2 lm syringe filtersand the filtrates were diluted in methanol. Negative ion ESI-MSand ESI-MSn spectra were acquired on the LIT mass spectrometer,using the operating conditions described above.

3. Results and discussion

3.1. Thermal stability of the 5-CQA and mixture with Ara3

To optimize the thermal conditions used in this work, the ther-mal stability of the 5-O-caffeoylquinic acid (5-CQA; for simplicity,

also abbreviated in this work as CQA) and mixture with(a1 ? 5)-L-arabinotriose (Ara3) was investigated. The respectivethermogravimetric (TG) and first derivative thermogravimetric(DTG) curves are shown as Supplementary material (Fig. S1),together with that previously obtained for Ara3 (Moreira,Coimbra, Nunes, & Domingues, 2013). Considering that the loss ofweight until around 100 �C is due to the loss of adsorbed water, itcan be observed that 5-CQA is thermally stable until around200 �C, as reported in previous studies (Owusu-Ware, Chowdhry,Leharne, & Antonijevic, 2013; Sharma, Fisher, & Hajaligol, 2002).More specifically, under the conditions used in this work, the firstdecomposition process of 5-CQA has a peak temperature at234 �C. Distinctly, the degradation of the Ara3 and the mixturebegins below 200 �C. In fact, a huge diversity of new compoundswas identified when Ara3 was individually heated from room tem-perature to 200 �C (Moreira et al., 2013).

In this work, the model mixture containing Ara3 and 5-CQA wasalso heated from room temperature to 200 �C (200T1). Aiming toobtain a lower degradation extent and, if possible, to identify inter-mediate degradation products, dry thermal treatments at a lowertemperature were also performed. The mixture was heated to175 �C and maintained at this temperature for two different peri-ods: 0 (175T1) and 30 min (175T2). The total mass loss percentageswere 5.5% for 175T1, 9.9% for 175T2, and 8.1% for 200T1. The solidmixture of 5-CQA and Ara3 (a yellowish-white powder) acquired adark brown coloration and appearance of a brittle caramel whensubmitted to the thermal treatments at 175 �C and 200 �C, similarto Ara3 when individually submitted at 200 �C. The caramel result-ing from the treatment 175T1 had a slightly less intense color thanthose from 175T2 to 200T1, suggesting that the lowest degradationwas promoted by the treatment 175T1, as corroborated by the low-est total mass loss. No visual color change was observed when the5-CQA was individually submitted to the same thermal treatments,corroborating that the development of the brown coloration duringdry thermal processing of the mixture resulted from the trans-formation of sugar moieties.

3.2. Identification of chlorogenic acid–arabinose hybrids and otherstructures

To evaluate the reactivity between Ara3 and 5-CQA when themodel mixture was submitted to the different thermal treatments,both untreated and thermally treated samples, completely sol-ubilized in water, were analyzed by ESI-MS. Under ESI-MS condi-tions, neutral oligosaccharides ionize preferentially in positivemode as [M+Na]+ ions (Moreira et al., 2013; Zaia, 2004), whereaschlorogenic acids ionize preferentially in negative mode as[M�H]� ions due to the presence of the carboxylic acid group(Clifford, Johnston, Knight, & Kuhnert, 2003). For this reason, inorder to obtain a clearer picture of the different compounds formedduring thermal processing, ESI-MS spectra were acquired in bothnegative and positive ion modes.

The negative ion ESI-MS spectrum of the untreated mixtureshowed as base peak the ion at m/z 353, attributed to [CQA-H]�,and the second most abundant ion at m/z 413, attributed to[Ara3-H]� (Fig. 1A). The ions observed at m/z 767 and 827 wereattributed to [CQA+Ara3-H]� and [2Ara3-H]�, respectively.Independently of the thermal treatment, the ion at m/z 353([CQA-H]�) remained as the base peak in the negative ESI-MS spec-tra of the thermally treated mixtures (Fig. 1B–D). Also, several newions, not observed in the ESI-MS spectrum of the untreated mix-ture (Fig. 1A), were identified. These ions, summarized in Table 1with the indication of the m/z value and the proposed assignment,were assigned as [M�H]� ions of hybrid compounds, derived fromAra3 and 5-CQA, and 5-CQA derivatives not bearing a sugar moietyformed during thermal processing of the model mixture, which

400 600 800 1000 1200 1400m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

353.0

485.1

617.1354.0 749.1486.1 881.2 1013.2 1145.2 1277.3 1409.2

400 600 800 1000 1200 1400m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

353.0

485.1

617.1354.0 749.1486.1 821.1 953.1 1085.2 1217.2 1349.3

400 600 800 1000 1200 1400m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

353.0

485.1

617.1354.0 749.1486.1 881.2 1013.2 1145.2 1277.3 1409.2

A

B

C

D

400 600 800 1000 1200 1400m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

353.0

413.0

767.0459.0 641.0 827.0

[CQA-H]-

[Ara3-H]-

[CQ

A+A

ra3-

H]-

[2A

ra3-

H]-

**

420 440 460 480 500m/z

0

20

40

485.1

486.1440.3 487.1449.1425.2 431.1 455.2 467.1 497.2

*

420 440 460 480 500m/z

0

20

40

485.1

486.1467.1 487.1440.3 449.1 476.1455.2425.2 497.2431.1

* *

420 440 460 480 500m/z

0

20

40

485.1

486.1

467.1 487.1449.1440.3 455.2425.2 476.2 497.2431.1* *

Fig. 1. Negative ion ESI-LIT-MS spectra of the (A) untreated mixture and afterthermal treatments: (B) 175T1, (C) 175T2, and (D) 200T1. Ions marked with anasterisk (*) are attributed to solvent impurities.

Table 1Summary of the [M�H]� ions identified in the negative ESI-MS spectra of the thermally treaassignment.

n 1 2 3 4 5

PentnCQA series[PentnCQA-H]� 485 617 749 881 1013[PentnCQA-H2O-H]� 467 599 731 863 995[PentnCQA-2H2O-H]� 449 581 713 845 977[PentnCQA-3H2O-H]� 431 563 695 827 959

Pentn(CQA)2 series[Pentn(CQA)2-H]� 821 953 1085 1217 1349[Pentn(CQA)2-H2O-H]� 803 935

PentnQA series[PentnQA-H]� 323 455 587 719 851[PentnQA-H2O-H]� 305 437 569

PentnCA series[PentnCA-H]� 443 575[PentnCA-H2O-H]� 425 557

CQA and derivatives without a sugar moiety[(CQA)n-H]� 353 689[(CQA)n-H2O-H]� 335[(CQA)nQA-H]� 527[(CQA)nCA-H]� 515[(CQA)nCA-H2O-H]� 497

a Ions observed exclusively in the ESI-MS spectrum acquired on the LTQ-Orbitrap inacquired on the LIT mass spectrometer in the m/z range 150–1500.

138 A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144

will be described in detail later. The assignment of each ion wassupported by the elemental composition obtained from high res-olution and exact mass measurements using a LTQ-Orbitrap hybridmass spectrometer (Supplementary Table S1). Note that the sugarmoiety of each hybrid compound is composed by pentose (Pent)units. According to the sugar and glycosidic linkage analyses per-formed when Ara3 was individually submitted to dry thermaltreatments at 200 �C (Moreira et al., 2013, 2014), the Pent unitsare mainly arabinose, although other Pent units, formed by isomer-ization, oxidation and decarboxylation reactions, were identified.

As result of the different ionization preferences of 5-CQA andAra3, the positive ion ESI-MS spectrum of the untreated mixture(Fig. 2A), in contrast to the corresponding negative ion ESI-MSspectrum (Fig. 1A), showed as base peak the ion at m/z 437([Ara3+Na]+), and the second most abundant ion at m/z 377([CQA+Na]+). Also, the [M+H]+ and [M+K]+ ions of Ara3 (m/z 415and 453) and [CQA+H]+ (m/z 355) were observed, but with relativeabundances lower than 3.5%. In respect to the positive ion ESI-MSspectra of the thermally treated mixtures, it is important to notethat, despite the ionization of oligosaccharides occurs pref-erentially in positive ion mode, the abundance of the ion at m/z437 decreased and the ion at m/z 377 became the base peak afterthe thermal treatments, corroborating that the Ara3 was moredegraded than 5-CQA, as expected considering the respectivethermogravimetric (TG) curves (Supplementary Fig. S1). As anexample, the positive ion ESI-MS spectrum of the mixture heatedat 175 �C for 30 min (175T2) is shown in Fig. 2B. As previouslydescribed for thermally treated Ara3 (Moreira et al., 2013), the ionsobserved at m/z 305 and 569 correspond to [Pent2+Na]+ and[Pent4+Na]+, supporting the occurrence of depolymerization andpolymerization (transglycosylation) reactions, respectively. Theions at m/z 287, 419, 551, and 683 correspond to [M+Na]+ ions ofdehydrated Pent oligosaccharides (Pentn-H2O, n = 2–5) (Moreiraet al., 2013). In accordance with the respective ESI-MS spectrumacquired in negative ion mode (Fig. 1C and Table 1), the ions atm/z 509, 641, and 773 are attributable to [M+Na]+ ions ofPentnCQA (n = 1–3) hybrid compounds, which will be describedbelow. The ion observed at m/z 731 correspond to [2CQA+Na]+, in

ted mixtures of Ara3 and 5-CQA with the indication of the m/z value and the proposed

6 7 8 9 10 11 12

1145 1277 1409 1541a 1673a 1805a 1937a

1127 1259 1391 1523a 1655a

1109 12411091

1481

the m/z range 150–2000. The other ions were also observed in the ESI-MS spectra

m/z300 400 500 600 700 800

%

0

100 437.1

377.1353.3382.3

438.1530.3

m/z300 400 500 600 700 800

%

0

100 377.1

353.3287.1

305.1

381.3

509.1419.1437.1 641.2551.2 731.2 773.2

*

*

[CQA+Na]+

[Ara3+Na]+A

B

*

*

683.

2

*

*

*

569.

2

Fig. 2. Positive ion ESI-QTOF-MS spectra of the (A) untreated mixture and (B) themixture heated at 175 �C for 30 min (175T2). Ions marked with an asterisk (*) areattributed to solvent impurities.

A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144 139

accordance with other noncovalently-linked dimers observed inthe negative ESI-MS.

In order to confirm the proposed assignments and gain addi-tional information about their structures, the hybrid compoundsand 5-CQA derivatives not bearing a sugar moiety formed duringthermal processing of the model mixture were characterized byESI-MSn. For their characterization, the negative ion mode was pre-ferred because they ionize better in negative than in positive mode.

Likewise [Ara3+Na]+ ions (Moreira et al., 2013), the ESI-MS2

fragmentation of [Ara3-H]� ions (m/z 413) (SupplementaryFig. S2A) yielded product ions resulting from glycosidic cleavages,loss of water, and cross-ring cleavages with neutral losses ofC2H4O2 (�60 Da) and C3H6O3 (�90 Da). The ESI-MS2 spectrum of[M�H]� ions of 5-CQA (m/z 353) (Fig. S2B), in accordance with pre-vious studies (Clifford et al., 2003; Fang, Yu, & Prior, 2002), showedas base peak the product ion at m/z 191, corresponding to [QA-H]�,and the product ion at m/z 179 with a low relative abundance,corresponding to [CA-H]�. Note that the negative charge is pref-erentially retained at the QA moiety due the existence of a free car-boxyl group (–COOH). Also, it is possible to observe the productions at m/z 173 and 161, formed by loss of CA (�180 Da) and QA(�192 Da) moieties, respectively. Similar to the nomenclature usedfor the product ions of oligosaccharides resulting from glycosidiccleavages (Moreira et al., 2013), and avoiding the confusion withthe dehydration induced by dry thermal processing, these productions are assigned as deprotonated acid residues, respectively,[QAres-H]� and [CAres-H]�, instead as [QA-H2O-H]� and [CA-H2O-H]� designations used in previous studies by other authors(Clifford et al., 2003; Fang et al., 2002). The product ions observedat m/z 309 (�44 Da) and 135 (�218 Da) were respectively formedby loss of CO2 and by combined loss of the QAres (�174 Da) and CO2

(�44 Da). The knowledge of the typical fragmentation pathways ofAra3 and 5-CQA was essential to understand the fragmentationpattern of the new ions identified after thermal processing of themixture, corresponding to hybrid compounds, derived from Ara3

and 5-CQA, and 5-CQA derivatives not bearing a sugar moiety. Allthese ions and their fragmentation patterns are described in thefollowing sections.

3.2.1. PentnCQA hybridsSeveral hybrid compounds composed by a CQA covalently

linked with pentose (Pent) units were identified in the negative

ESI-MS spectra of the thermally treated mixtures, corroboratingthe hypothesis of linkages between chlorogenic acids and arabi-nose in coffee melanoidin structures. For all treated mixtures, theone with the highest relative abundance was observed at m/z485, corresponding to [M�H]� of a compound formed by the reac-tion of a Pent and a CQA molecule with the release of a water mole-cule, assigned as [PentCQA-H]�. As part of the same ion series werealso observed the ions at m/z 617, 749, 881, 1013, 1145, 1277, 1409,1541, 1673, 1805 and 1937, assigned as [PentnCQA-H]� ions, n = 2–12. The PentnCQA (n = 1–12) compounds were also observed as[M-2H]2� ions at m/z 242, 308, 374, 440, 506, 572, 638, 704, 770,836, 902 and 968, but having lower relative abundance than thecorresponding [M�H]� ions. The positive ESI-MS spectra of the ther-mally treated mixtures, as previously mentioned, showed the ionswith m/z 509, 641, and 773, assigned as [PentnCQA+Na]+ ions(n = 1–3). The identification of PentnCQA compounds bearing a lower(n = 2) and higher (n = 4–12) number of sugar units than that of theoligosaccharide (Ara3) in the starting mixture also corroborates theoccurrence of depolymerization and polymerization (transglycosyla-tion) reactions.

All the ESI-MS2 spectra acquired from ions assigned as[PentnCQA-H]� support the presence of covalently linked Pent toCQA moieties, allowing to confirm the proposed assignments. Asexample, the ESI-MS2 spectra of Pent1–3CQA are shown in Fig. 3.

Fig. 3A shows the ESI-MS2 spectrum of the ion observed at m/z485, attributed to [PentCQA-H]�. The product ion at m/z 353,formed by loss of a Pentres and attributed to [CQA-H]�, confirmsthe presence of a CQA linked to a Pent. The product ion at m/z323 (base peak), formed by loss of a CAres and attributed to[PentQA-H]�, suggests that the Pent unit was linked to the QA moi-ety. However, the product ion at m/z 293, formed by loss of a QAand attributed to [(PentCA)res-H]�, suggests the presence of otherstructures with the Pent unit linked to the CA moiety.

Similar to the [PentCQA-H]� ions, the ESI-MS2 fragmentation of[M�H]� ions of Pent2CQA (m/z 617, Fig. 3B) and Pent3CQA (m/z749, Fig. 3D) did not yield products ions formed by loss of CQAor (CQA)res, not allowing to confirm the presence of structures withthe Pent residues linked together. For these compounds, the com-plementary study of the fragmentation of the correspondent[M+Na]+ ions allowed to obtain additional structural information.The ESI-MS/MS fragmentation of [Pent2CQA+Na]+ (m/z 641) and[Pent3CQA+Na]+ (m/z 773) (Supplementary Fig. S3) yielded, respec-tively, product ions at m/z 305 ([Pent2+Na]+) and 437([Pent3+Na]+), confirming the presence of structures with thePent units linked together. The ESI-MS2 fragmentation of [Pent2–

3CQA-H]� ions (Fig. 3B and D) also produced product ions formedby loss of CAres and QA, suggesting the coexistence of isomers hav-ing the sugar moiety (Pent2 or Pent3) linked either to the QA or CAmoiety of the CQA. In fact, the presence of isomers for eachPentnCQA (n = 1–12) compound was expected, considering the dif-ferent linkage possibilities, even in the simplest structure(PentCQA). In this case, considering that the anomeric oxygen ofthe Pent is involved in a glycosidic linkage, there are five freehydroxyl groups (three in QA and two in CA) as possible bindingsites in the CQA. Also, a- and b-anomers may be formed.Moreover, the free carboxylic acid group of the QA moiety maybe involved in the formation of an ester linkage.

3.2.2. Pentn(CQA)2 hybridsIn the negative ESI-MS spectra of the thermally treated mix-

tures, another ion series was observed at m/z 821, 953, 1085,1217, 1349 and 1481 (Table 1), attributed to [M�H]� ions of com-pounds bearing two CQAs covalently linked with a variable num-ber of Pent units (1–6), with the release of a water molecule foreach linkage ([Pentn(CQA)2-H]�, n = 1–6).

200 250 300 350 400 450 500m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

323.1

485.2

353.0293.0190.9 395.1 425.0 467.1

-H2O

[M-H]-

-Pentres[QA-H]- -60 Da-90 Da

A

-QA

-CAres

300 400 500 600 700m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

587.3

749.3

557.2

455.2323.1 617.2 731.3353.0 485.2 659.2293.0

[M-H]-

575.

2-Q

A res

569.

2-C

A

-CAres

-QA

-Pentres -H2O-90

Da

[Pen

tCQ

A-H

]-

[Pen

t 2QA-

H]-

[CQA-H]-[Pen

tQA-

H]-

[(Pen

tCA)

res-H

]-

D

200 300 400 500 600m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

455.2

617.3425.2353.0 485.0190.9 323.1 599.2527.2275.1

[M-H]-

[(Pen

tCA)

res-H

2O-H

]-

[Pen

tQA-

H]-

[QA-H]- -90Da -H2O[CQA-H]- -Pentres-QA

-CAresB

150 200 250 300 350 400 450m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

305.0

352.9

190.9275.0 467.3449.1178.9 377.0335.1257.0

-H2O [M-H]--90 Da

[QA-H]-

[CA-H]-

-CAres

-QA

C

-(QA+H2O)

-(Pent-H2O)res

-(Pen

t-H2O

)/Pen

t res

293.

0-Q

A res

/(QA-

H2O

)

Fig. 3. ESI-MS2 spectra of the ions observed at m/z (A) 485 ([PentCQA-H]�), (B) 617 ([Pent2CQA-H]�), (C) 467 ([PentCQA-H2O-H]�) and (D) 749 ([Pent3CQA-H]�), acquiredfrom the mixture heated to 175 �C (175T1).

140 A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144

All the ESI-MS2 spectra acquired from ions assigned as[Pentn(CQA)2-H]� support the presence of two CQA moleculesand one or more Pent units. As an example, the ESI-MS2 spectrumof the ion observed at m/z 821, attributed to [Pent(CQA)2-H]�, isshown in Supplementary (Fig. S4A). The product ion at m/z 689,formed by loss of a Pentres, suggests that the two CQA molecules(or their derivatives) are linked together. On the other hand, the pro-duct ion at m/z 323, attributed to [PentQA-H]�, suggests that thePent unit is linked to a QA moiety.

3.2.3. PentnQA and PentnCA hybridsAlso, [M�H]� ions of compounds composed exclusively by qui-

nic (QA) or caffeic (CA) acid moieties, derived from a CQA, cova-lently linked with a variable number of Pent units were identifiedin the negative ESI-MS spectra of the thermally treated mixturesand assigned as [PentnQA-H]� (n = 1–5) and [PentnCA-H]� (n = 2–3), respectively (Table 1).

All the ESI-MS2 spectra acquired from ions assigned as[PentnQA-H]� and [PentnCA-H]� support the presence of one or morePent units linked to a QA or a CA moiety, respectively. For the com-pounds bearing two or more Pent units (n P 2), the observation ofproduct ions attributed to [Pentn-H]� and [Pentn–1Pentres-H]� cor-roborated the presence of structures having all the Pent unitslinked together.

3.2.4. Dehydrated derivatives of PentnCQA, Pentn(CQA)2, PentnQA andPentnCA

For all the aforementioned series, [M�H]� ions of dehydratedderivatives resulting from the loss of another water molecule werealso identified in the negative ESI-MS spectra of the thermally trea-ted mixtures, assigned as [PentnCQA-H2O-H]� (n = 1–10),[Pentn(CQA)2-H2O-H]� (n = 1–2), [PentnQA-H2O-H]� (n = 1–3), and[PentnCA-H2O-H]� (n = 2–3). Also, [M�H]� ions of PentnCQAderivatives resulting from the loss of two and three additionalwater molecules were identified, assigned as [PentnCQA-2H2O-H]� (n = 1–7) and [PentnCQA-3H2O-H]� (n = 1–6), respectively(Table 1).

As an example of a dehydrated derivative of PentnCQA com-pounds, the ESI-MS2 spectrum of the ion observed at m/z 467,attributed to [PentCQA-H2O-H]�, is shown in Fig. 3C. The production at m/z 353, with a difference of 114 Da (132–18) from the

precursor ion, suggests that the dehydration induced by thermalprocessing occurred at the Pent moiety. Considering the loss ofwater at the Pent moiety, the product ions at m/z 335 (�132 Da)and 293 (�174 Da) can be identified as resulting from the loss of(Pent-H2O) and QAres, respectively. However, these product ionscan also result, respectively, from the loss of Pentres and (QA-H2O), and therefore the coexistence of other structures bearingan intact Pent and a dehydrated QA moiety cannot be completelyexcluded.

Similarly, the ESI-MS2 spectrum of the ion observed at m/z 803(Supplementary Fig. S4B), attributed to [Pent(CQA)2-H2O-H]�,showed the product ion at m/z 689 (�114 Da), suggesting the pres-ence of a dehydrated Pent moiety. However, the product ion at m/z485, with a difference of 318 Da (336–18) from the precursor ion,formed by loss of (CQA-H2O)res, suggests the coexistence of otherstructures bearing an intact Pent and a dehydrated CQA. All theESI-MS2 spectra acquired from ions assigned as [PentnQA-H2O-H]� and [PentnCA-H2O-H]� also showed a product ion with a differ-ence of 114 Da from the precursor ion, corroborating the presenceof structures bearing a dehydrated Pent unit.

3.2.5. CQA derivatives without a sugar moietyAfter thermal processing (175T1, 175T2 and 200T1) of the

model mixture, the MS2 fragmentation pattern of the ion observedat m/z 353 ([CQA-H]�) was not changed. After coffee roasting at230 �C for 5–6 min, it was observed the decrease of 5-CQA content,while the content of isomers, namely 3-CQA and 4-CQA, increased(Farah, de Paulis, Trugo, & Martin, 2005). Considering that [M�H]�

ions of 3-, 4- and 5-CQAs produce distinct ESI-MS2 spectra (Cliffordet al., 2003; Fang et al., 2002), changes in the fragmentation pat-tern of the ion at m/z 353 could be indicative of 5-CQA isomeriza-tion, not observed in this study.

In accordance with previous studies reporting the dehydrationof CQAs during coffee roasting (Farah et al., 2005; Jaiswal, Matei,Golon, Witt, & Kuhnert, 2012; Jaiswal, Matei, Subedi, & Kuhnert,2014), a dehydrated derivative of CQA was observed at m/z 335([CQA-H2O-H]�) in the negative ESI-MS spectra of the thermallytreated mixtures. The respective ESI-MS2 spectrum is shown inSupplementary Fig. S5. The product ion at m/z 179, with a differ-ence of 156 Da (174–18) from the precursor ion, formed by lossof (QA-H2O)res and attributed to [CA-H]�, corroborates that the loss

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e A

bund

ance

484.9

530.6

531.6 800.2 934.2656.4

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e A

bund

ance

706.8

352.9

729.0

420.9 730.0 914.0 1104.8

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

552.8

484.9

881.0617.0 749.0882.1816.9 1013.0684.7

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e A

bund

ance

552.9

484.9

617.0881.1749.1

1013.1750.2353.0

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e A

bund

ance

749.0881.0617.0

485.1

816.5552.8 684.71013.1

948.61025.5440.7378.7 1145.1

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e A

bund

ance

552.8484.9

881.0553.9 749.11013.0682.1

200 400 600 800 1000 1200m/z

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

485.0

617.0

953.0806.3749.1 1004.2467.0 486.0 674.2335.0 1049.3

1148.4440.5

B - RT 8.5 min

6 8 10 12 14 16 18Time (min)

0

500000

1000000

1500000

uAU

11.2

16.112.5 14.1

14.3

8.5

NL:ChaAna_MeH2O00umin

C - RT 11.2 min

A

[PentCQA+HCOOH-H]-

[PentCQA-H]-

[CQA+HCOOH+Na-2H]-

[CQA-H]-

[2CQA-H]-

[2CQA+Na-2H]-

[PentCQA-H]-

[PentCQA+HCOOH+Na-2H]-D - RT 12.5 min

E - RT 12.9 min

F - RT 13.3 min

G - RT 14.1 min

H - RT 16.1 min

Fig. 4. (A) HPLC-UV chromatogram recorded at 325 nm obtained from mixture heated to 175 �C (175T1), and (B–H) HPLC–ESI-MS spectra associated with the major peakswith retention times (RTs) between 8.5 and 16.1 min.

A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144 141

of the water molecule occurred at the QA moiety. This is also cor-roborated with the product ions at m/z 173, 161 (base peak), and135, attributed to [QA-H2O-H]�, [CAres-H]�, and [CA-CO2-H]�,respectively. According to previous studies, the loss of a singlewater molecule from the QA moiety of CQAs during coffee roastingcan produce either caffeoyl-1,5-quinides (lactones, abbreviated asCQLs), (Farah et al., 2005), or caffeoylshikimic acids (CSAs)(Jaiswal et al., 2012, 2014). Considering the ESI-MS2 fragmentationreported for both CQLs and CSAs, (Jaiswal, Matei, Ullrich, &Kuhnert, 2011), the possible coexistence of CQLs and CSAs formedduring thermal processing of the mixture cannot be excluded.

Other compounds derived from CQA, not bearing a sugar moi-ety, were also identified in the negative ESI-MS spectra of the ther-mally treated mixtures as [M�H]� ions at m/z 689, 527, 515, and497, assigned as [(CQA)2-H]�, [(CQA)QA-H]�, [(CQA)CA-H]�, and[(CQA)CA-H2O-H]�, respectively (Table 1). In accordance with theproposed assignments, the ESI-MS2 spectra of the ions observedat m/z 515, 527 and 689 (Supplementary Fig. S6) support the pres-ence of a CQA covalently linked with a CA, a QA or another CQAmolecule, respectively. However, the product ion observed at m/z395 (�132 Da) in the ESI-MS2 spectrum of the ion at m/z 527,attributed to [(CQA)QA-H]�, suggests the coexistence of [Pent4-H2O-H]� precursor ions, which is corroborated by the observation

of the corresponding [M+Na]+ ions (m/z 551) in the positive ionESI-MS spectra of the thermally treated mixtures (Fig. 2B). TheESI-MSn spectra (n = 2–3) acquired from the ion observed at m/z497, attributed to [(CQA)CA-H2O-H]�, corroborate the loss of awater molecule at a QA moiety, not excluding the possibility ofthe formation of either a lactone or a shikimic acid moiety(Supplementary Fig. S7).

The [M�H]� ions of CQA derivatives without a sugar moietywere also identified in samples of only 5-CQA submitted to thethermal treatments 175T1, 175T2 and 200T1 (data not shown).As in the ESI-MS spectra of the thermally treated mixtures, theseions showed a low relative abundance (61.5%) and the ion at m/z353 ([CQA-H]�) remained as the base peak after thermal process-ing, corroborating the thermal stability of 5-CQA until around200 �C, as evidenced by TG analysis.

3.3. Differentiation of PentCQA isomers

To unveil possible isomeric structures, in particular of PentnCQAcompounds, the mixture heated to 175 �C (175T1) was further ana-lyzed by HPLC-PDA-ESI-MS and HPLC-PDA-ESI-MSn. The Ara3 thatdid not react and other oligosaccharides formed during the thermalprocessing of the mixture were not retained by the C18 column,

150 200 250 300 350 400 450 500m/z

0

50

100 352.9

395.0466.7424.5

150 200 250 300 350 400 450 500m/z

0

50

100 352.9

323.0190.9 395.1 466.6

B - RT 8.6 min

C - RT 11.6 min

150 200 250 300 350 400 450 500m/z

0

50

100 323.0

0.3539.091 395.0

D - RT 12.6 min

150 200 250 300 350 400 450 500m/z

0

50

100 352.9293.0 323.0191.0

178.9 416.8 484.5353.5

E - RT 13.0 min

150 200 250 300 350 400 450 500m/z

0

50

100 352.9190.9293.0

311.0178.9 395.1220.4

F - RT 13.4 min

150 200 250 300 350 400 450 500m/z

0

50

100 323.0

172.9 310.1 388.9203.0

G - RT 14.2 min

150 200 250 300 350 400 450 500m/z

0

50

100 352.9

395.0 424.8191.0 293.2 323.2 455.2

H - RT 16.1 min

6 8 10 12 14 16 18Time (min)

0

20

40

60

80

100

Rel

ativ

e Ab

unda

nce

NL:m/zMSAnaMeH2Oul-m

8.6 11.6

14.2

16.1

12.6 13.0

13.4

A

-H2O

-Pentres

-60 Da-90 Da

[QA-H]-

-H2O

-Pentres

-90 Da-CAres

[QA-H]- -Pentres -90 Da

-CAres -Pentres

-60 Da-90 Da-30 Da

[QA-H]--CAres-QA

[M-H]-

[QA-H]-

[CA-H]-

-Pentres

-CAres-QA

[QA-H]-

[CA-H]-

-Pentres

-QA

-QAres

[QAres-H]-

-CAres

-90 Da

Fig. 5. Differentiation of PentCQA isomers in the mixture heated to 175 �C (175T1) by HPLC–ESI-MS and HPLC–ESI-MS2: (A) reconstructed ion chromatogram of the ion withm/z 485 ([PentCQA-H]�), and (B–H) the respective HPLC–ESI-MS2 spectra acquired at different retention times (RTs).

142 A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144

confirming the covalent linkage between CQA and sugar moietiesof the PentnCQA hybrid compounds, which were retained by thecolumn.

Fig. 4A shows the HPLC-UV chromatogram recorded at 325 nm,a characteristic absorption wavelength of CQAs. According to thischromatogram, 5-CQA that did not react and the compounds bear-ing a CQA moiety formed during the thermal processing of the mix-ture, including the PentnCQA compounds, eluted between 7.5 and16.5 min. On the other hand, the HPLC–MS spectra associated withthe major chromatogram peaks (Fig. 4B–H) show that the chro-matographic separation of each PentnCQA compound having a dis-tinct number of Pent units (n = 1–12) was not achieved, butisomers of these compounds were separated, eluting at differentretention times (RTs). However, it was not possible to achieve aperfect separation of all the isomeric structures of each PentnCQAcompound. Since a more reliable separation of the isomers wasobtained for the simplest hybrid compound (PentCQA), observedas [M�H]� at m/z 485, the respective reconstructed ion chro-matogram (RIC) is shown in Fig. 5A. According to this chro-matogram, PentCQA eluted in seven major peaks with RTsranging from 8.6 to 16.1 min. This suggests the presence of at leastseven isomeric structures, as corroborated by the distinct HPLC–MS2 spectra obtained at each RT (Fig. 5B–H). As there are no stan-dards available, it was not possible to identify the specific isomersgiving rise to these HPLC–MS/MS spectra. Nevertheless, theabsence of the product ion formed by loss of QA (m/z 293) in theMS2 spectra obtained at RTs 11.6 (C), 12.6 (D) and 14.2 (G) minsuggests the presence of isomers having the Pent linked to theCQA through the QA moiety. On other hand, the absence of the

product ion formed by loss of CAres (m/z 323) in the MS2 spectrumobtained at RT 13.4 min (F) suggests an isomer having the Pentlinked to the CA moiety of the CQA. In fact, the possible reactionof the anomeric carbonyl group of the Pent with any one of the fivefree hydroxyl groups in the CQA, giving rise to a- and/or b-anom-ers, as well as the possible reactivity of the acid group of the QAmoiety justify the diversity of isomers formed.

3.4. Identification of Pent1–2CQA in fractions recovered from spentcoffee grounds

In order to validate the strategy used to identify possible hybridstructures formed from chlorogenic acids and arabinose sidechains of arabinogalactans during coffee roasting, fractions recov-ered from spent coffee grounds (SCG) were analyzed by ESI-MS.In both negative ion ESI-MS spectra of MAE3_PptEt andMAE4_PptEt fractions, the most abundant ions were observed atm/z 191 and 353, attributed to [QA-H]� and [CQA-H]�, respec-tively. They also showed, although with a low relative abundance,the ions at m/z 485 and 617, attributed to [Pent1–2CQA-H]�, as wellas the ion at m/z 335, attributed to [CQA-H2O-H]�. These assign-ments were corroborated with respective negative ESI-MSn

(n = 2–3) spectra, showing the typical product ions identified fromthe fragmentation of the ions with the same m/z value identifiedafter thermal processing of the model mixture. Also in accordancewith the data obtained from the thermal treated mixture, thePent1–2CQA compounds identified in the SCG fractions may havebeen formed during coffee roasting. Accordingly, this type of

A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144 143

compounds may have also been incorporated into the coffee mel-anoidin structures, also formed during roasting.

4. Conclusions

The dry thermal processing of a model mixture composed byequimolar amounts of arabinotriose (Ara3) and 5-O-caffeoylquinicacid (5-CQA) promoted the formation of several hybrid compoundscomposed by one or two CQAs covalently linked with a variablenumber of pentose residues (mainly arabinose), corroboratingthe hypothesis of arabinose from arabinogalactan side chains as apossible binding site for chlorogenic acid derivatives in coffee mel-anoidin structures. The further analysis by HPLC–MS and HPLC–MSn allowed demonstrating the presence of isomeric hybrid struc-tures, namely PentCQA isomers. These results highlight the struc-tural complexity of compounds that can be formed betweenchlorogenic acids and carbohydrates during coffee roasting. Also,the formation of these chlorogenic acid–carbohydrate hybridstructures with functionalization of the carbohydrate moiety byinclusion of carboxylic groups can increase their reactivity andconstitute the starting point for the incorporation of carbohydratesin coffee melanoidins through the reaction of the chlorogenic acidspresent.

The identification of PentnCQA compounds from the model mix-ture, as well as the knowledge of their fragmentation pattern underESI-MSn conditions, made possible their identification in fractionsrecovered from spent coffee grounds, opening new perspectivesfor their identification in coffee melanoidin structures, but also inmelanoidins from other sources.

The presence of covalently linked chlorogenic acids to the mel-anoidin structures may contribute to their antioxidant activity.Having this in mind, the roasting of oligosaccharides or polysac-charides used as functional ingredients in the presence of chloro-genic acids may be used as a method to improve the antioxidantactivity of food products. Future work is needed to assess biologicalactivities of hybrid compounds formed from oligo- or polysaccha-rides and chlorogenic acids, as well as studies with synthetic stan-dards are needed to identify the specific fragmentation pattern ofthe different isomers formed.

Acknowledgments

Thanks are due to Fundação para a Ciência e a Tecnologia (FCT,Portugal), European Union, QREN, FEDER, and COMPETE for fund-ing the QOPNA research unit (Project PEst-C/QUI/UI0062/2013;FCOMP-01-0124-FEDER-037296), CICECO (Pest-C/CTM/LA0011/2013 and FCOMP-01-0124-FEDER-037271), Projects NORTE-07-0162-FEDER-000048, NORTE-07-0124-FEDER-000066/67 andPEst-C/EQB/LA0006/2011, and RNEM (REDE/1504/REM/2005 thatconcerns the Portuguese Mass Spectrometry Network). Thanksare also due to FCT for the Grants of Ana Moreira (SFRH/BD/80553/2011) and Cláudia Passos (SFRH/BDP/65718/2009).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.foodchem.2015.03.086.

References

Bekedam, E. K., De Laat, M. P. F. C., Schols, H. A., Van Boekel, M. A. J. S., & Smit, G.(2007). Arabinogalactan proteins are incorporated in negatively charged coffeebrew melanoidins. Journal of Agricultural and Food Chemistry, 55, 761–768.

Bekedam, E. K., Loots, M. J., Schols, H. A., Van Boekel, M. A. J. S., & Smit, G. (2008).Roasting effects on formation mechanisms of coffee brew melanoidins. Journalof Agricultural and Food Chemistry, 56, 7138–7145.

Bekedam, E. K., Schols, H. A., van Boekel, M. A. J. S., & Smit, G. (2006). High molecularweight melanoidins from coffee brew. Journal of Agricultural and Food Chemistry,54, 7658–7666.

Bekedam, E. K., Schols, H. A., Van Boekel, M. A. J. S., & Smit, G. (2008). Incorporationof chlorogenic acids in coffee brew melanoidins. Journal of Agricultural and FoodChemistry, 56, 2055–2063.

Bradbury, A. G. W., & Halliday, D. J. (1990). Chemical structures of green coffee beanpolysaccharides. Journal of Agricultural and Food Chemistry, 38, 389–392.

Clifford, M. N. (2000). Chlorogenic acids and other cinnamates – Nature, occurrence,dietary burden, absorption and metabolism. Journal of the Science of Food andAgriculture, 80, 1033–1043.

Clifford, M. N., Johnston, K. L., Knight, S., & Kuhnert, N. (2003). Hierarchical schemefor LC–MSn identification of chlorogenic acids. Journal of Agricultural and FoodChemistry, 51, 2900–2911.

Coelho, C., Ribeiro, M., Cruz, A. C. S., Domingues, M. R. M., Coimbra, M. A., Bunzel, M.,et al. (2014). Nature of phenolic compounds in coffee melanoidins. Journal ofAgricultural and Food Chemistry, 62, 7843–7853.

Fang, N., Yu, S., & Prior, R. L. (2002). LC/MS/MS characterization of phenolicconstituents in dried plums. Journal of Agricultural and Food Chemistry, 50,3579–3585.

Farah, A., de Paulis, T., Trugo, L. C., & Martin, P. R. (2005). Effect of roasting on theformation of chlorogenic acid lactones in coffee. Journal of Agricultural and FoodChemistry, 53, 1505–1513.

Fogliano, V., & Morales, F. J. (2011). Estimation of dietary intake of melanoidins fromcoffee and bread. Food & Function, 2, 117–123.

Gniechwitz, D., Reichardt, N., Ralph, J., Blaut, M., Steinhart, H., & Bunzel, M. (2008).Isolation and characterisation of a coffee melanoidin fraction. Journal of theScience of Food and Agriculture, 88, 2153–2160.

IUPAC (1976). Nomenclature of cyclitols. Biochemical Journal, 153, 23–31.Jaiswal, R., Matei, M. F., Golon, A., Witt, M., & Kuhnert, N. (2012). Understanding the

fate of chlorogenic acids in coffee roasting using mass spectrometry basedtargeted and non-targeted analytical strategies. Food & Function, 3, 976–984.

Jaiswal, R., Matei, M. F., Subedi, P., & Kuhnert, N. (2014). Does roasted coffee containchlorogenic acid lactones or/and cinnamoylshikimate esters? Food ResearchInternational, 61, 214–227.

Jaiswal, R., Matei, M. F., Ullrich, F., & Kuhnert, N. (2011). How to distinguish betweencinnamoylshikimate esters and chlorogenic acid lactones by liquidchromatography–tandem mass spectrometry. Journal of Mass Spectrometry, 46,933–942.

Maier, H. G., Diemair, W., & Ganssmann, J. (1968). Isolation and characterization ofbrown roast compounds of coffee (in German). Zeitschrift fürLebensmitteluntersuchung und -Forschung A, 137, 287–292.

Moon, J.-K., & Shibamoto, T. (2009). Role of roasting conditions in the profile ofvolatile flavor chemicals formed from coffee beans. Journal of Agricultural andFood Chemistry, 57, 5823–5831.

Moreira, A. S. P., Coimbra, M. A., Nunes, F. M., & Domingues, M. R. M. (2013).Roasting-induced changes in arabinotriose, a model of coffee arabinogalactanside chains. Food Chemistry, 138, 2291–2299.

Moreira, A. S. P., Nunes, F. M., Domingues, M. R., & Coimbra, M. A. (2012). Coffeemelanoidins: Structures, mechanisms of formation and potential healthimpacts. Food & Function, 3, 903–915.

Moreira, A. S. P., Simões, J., Pereira, A. T., Passos, C. P., Nunes, F. M., Domingues, M. R.M., et al. (2014). Transglycosylation reactions between galactomannans andarabinogalactans during dry thermal treatment. Carbohydrate Polymers, 112,48–55.

Nunes, F. M., & Coimbra, M. A. (2007). Melanoidins from coffee infusions.Fractionation, chemical characterization, and effect of the degree of roast.Journal of Agricultural and Food Chemistry, 55, 3967–3977.

Nunes, F. M., & Coimbra, M. A. (2010). Role of hydroxycinnamates in coffeemelanoidin formation. Phytochemistry Reviews, 9, 171–185.

Oosterveld, A., Voragen, A. G. J., & Schols, H. A. (2003). Effect of roasting on thecarbohydrate composition of Coffea arabica beans. Carbohydrate Polymers, 54,183–192.

Owusu-Ware, S. K., Chowdhry, B. Z., Leharne, S. A., & Antonijevic, M. D. (2013).Quantitative analysis of overlapping processes in the non-isothermaldecomposition of chlorogenic acid by peak fitting. Thermochimica Acta, 565,27–33.

Passos, C. P., Cepeda, M. R., Ferreira, S. S., Nunes, F. M., Evtuguin, D. V., Madureira, P.,et al. (2014). Influence of molecular weight on in vitro immunostimulatoryproperties of instant coffee. Food Chemistry, 161, 60–66.

Passos, C. P., & Coimbra, M. A. (2013). Microwave superheated water extraction ofpolysaccharides from spent coffee grounds. Carbohydrate Polymers, 94,626–633.

Passos, C. P., Moreira, A. S. P., Domingues, M. R. M., Evtuguin, D. V., & Coimbra, M. A.(2014). Sequential microwave superheated water extraction of mannans fromspent coffee grounds. Carbohydrate Polymers, 103, 333–338.

Perrone, D., Farah, A., & Donangelo, C. M. (2012). Influence of coffee roasting on theincorporation of phenolic compounds into melanoidins and their relationshipwith antioxidant activity of the brew. Journal of Agricultural and Food Chemistry,60, 4265–4275.

Perrone, D., Farah, A., Donangelo, C. M., de Paulis, T., & Martin, P. R. (2008).Comprehensive analysis of major and minor chlorogenic acids and lactones ineconomically relevant Brazilian coffee cultivars. Food Chemistry, 106, 859–867.

Redgwell, R. J., Trovato, V., Curti, D., & Fischer, M. (2002). Effect of roasting ondegradation and structural features of polysaccharides in Arabica coffee beans.Carbohydrate Research, 337, 421–431.

144 A.S.P. Moreira et al. / Food Chemistry 185 (2015) 135–144

Sharma, R. K., Fisher, T. S., & Hajaligol, M. R. (2002). Effect of reaction conditions onpyrolysis of chlorogenic acid. Journal of Analytical and Applied Pyrolysis, 62,281–296.

Takenaka, M., Sato, N., Asakawa, H., Wen, X., Murata, M., & Homma, S. (2005).Characterization of a metal-chelating substance in coffee. Bioscience,Biotechnology, and Biochemistry, 69, 26–30.

Totlani, V. M., & Peterson, D. G. (2007). Influence of epicatechin reactions on themechanisms of Maillard product formation in low moisture model systems.Journal of Agricultural and Food Chemistry, 55, 414–420.

Zaia, J. (2004). Mass spectrometry of oligosaccharides. Mass Spectrometry Reviews,23, 161–227.