Z LebensmUnters Forsch (1994) 199:17-21

Zeitschrift for

ee eee ee 9 Spfinger-Verlag 1994

Original paper HPLC Analysis of chlorogenic acid lactones in roasted coffee Carolin Bennat, Ulrich H. Engelhardt, Andrea Kiehne, Frank-Michael Wirries, Hans Gerhard Maier Institut ftir Lebensmittelchemie der TechnischenUniversit~t, Schleinitzstrasse 20, D-38106 Braunschweig, Germany Received October 13, 1993; revised version December 12, 1993

HPLC-Untersuchungen in R6stkaffee

fiber

Chlorogens/iurelactone

Zusammenfassung. Es wird die Identifiziernng yon 3- und 4-Caffeoyl-'y-chinid (CQL) in R6stkaffee beschrieben. Die Lactone wurden mit Hilfe eines isokratischen RP-HPLC-Systems und Dioden-Array-Detektion yon den fibrigen Chlorogensfiuren (CGA) getrennt und als Kaffees~iure-Derivate identifiziert. Thermospray-LC-MS-Spektren (Discharge- und Puffer-Ionisierung) zeigten das Quasi-Molektfl-Ion und charakteristische Fragmente (z. B. Kaffees~iure, Chinid). Die Lactone wurden mittels Polyamid-Sfiulenchromatographie gefolgt von semipr~iparativer RP-HPLC aus dem R6stkaffee isoliert. Die Identifiziernng erfolgte durch FTIR- und NMR-Spektroskopie. Es scheint ein Gleichgewicht zwischen den beiden Lactonen zu bestehen. Sowohl die Isolierung des 3- als auch die des 4-Caffeoylchinids liefert unter den experimentellen Bedingungen ein Gemisch aus den Substanzen. Die Gehalte der Lactone in Kaffeeproben des Handels reichten yon 1.5 bis 3.5 g/kg bezogen auf Trockenmasse.

Abstract. The identification of 3- (3-CQL) and 4-caffeoylquinic acid-~-lactone (4-CQL) in roasted coffee is described. The lactones were separated from chlorogenic acids (CQA) by an isocratic reversed-phase (RP)-HPLC system and identified as caffeic acid derivatives using their ultraviolet spectra obtained on the fly with an diode array detector. Thermospray liquid-chromatography mass-spectrometry spectra (discharge ionisation, buffer ionisation) showed the quasi molecule ion and characteristic fragments (e.g. caffeic acid, quinide). The lactones were isolated from coffee by polyamide column chromatography followed by RP-HPLC on a semipreparative scale, and structures assigned using Fourier-transform infrared spectroscopy and various nuclear magnetic resonance techniques. There seems to be an equilibrium between the two lactones: isolation of 3-CQL and 4-CQL under the experimental conditions used yielded both lactones in the final product. The content of the caffeoylquinides in commercial coffee samples ranged from 1.5 to 3.5 g/kg dry matter.

Correpondence to: H. G. Maier

Introduction Chlorogenic acids (CGA) are well known as major constituents of green coffee beans [1]. Besides mono- and dicaffeoylquinic acids a number of compounds containing cumaroyl and feruloyl moieties have been identified [2-4]. The pattern of minor CGA-like compounds have been claimed to correlate with the geographical origin of the beans [5]. The content of chlorogenic acids decreases during the roasting process. Quinic acid and the corresponding lactones [6] have been identified as degradation products, while the amount of free caffeic acid was negligible, probably due to oxidation/polymerisation. Four lactones of chlorogenic acid(s) have been detected in roasted coffee using gas chromatography/mass spectrometry (GC/MS) of the trimethylsilyl (TMS) derivatives [7]. No further information was given and the lactones were not assigned to peaks in the corresponding HPLC traces. More recently, the occurrence of feruloylquinic acid lactones has been described [8]. In this study the isolation, identification and quantification of 3- (3-CQL) and 4-caffeoylquinic acid lactone (4-CQL) is described.

Materials and methods Materials Indian robusta coffeebeans were roasted for 8 rain in a Nescaf6Rohkaffeer6ster (Siemens). This corresponds to a medium degree of roasting. Samples of different degrees of roasting were prepared from green arabica coffeebeans (Columbia). Commercialsamples were obtained from a local retailer. All samples were ground and the 0.5-mm sieve fraction was used for analysis. Chlorogenicacid (5-caffeoylquinicacid) was obtained from Fluka (Buchs).

Extraction (and decaffeination) A modified German standard procedure [9] was used. Coffee (5 g) was extracted with 200 ml of 70% methanol at 75-80~ C in a soxhlet apparatus for 90 min, concentrated to about 80 ml under reduced pressure, transferred into a 100-mlflask and made up to volume.

18 Extracts for HPLC-diode array detection (DAD) and HPLC-thermospray (TSP)-MS analyses were decaffeinated. For quantification, 0.25-0.5 g coffee was extracted with 100 ml of 70% methanol at 80 ~ C for 1 h under reflux, concentrated to 40-50 ml using a rotary evaporator, transferred into a 50-ml flask and made up to volume.

Identification Infrared (1R)spectroscopy. Spectra were obtained using an IFS 66 Fourier transform IR (FTIR) spectrometer (Bruker, Karlsruhe, Germany), wave-number range 700-4,800 cm-L The data system used was PC AT 386 with the Opus software version 1.4. Samples were analysed as KBr compacts.

Polyamide column chromatography Polyamide MN-SC-6 (Macherey & Nagel, Dfiren, Germany) was suspended in methanol and allowed to settle overnight. A glass column (22 x 3.0 cm i.d.) was filled with polyamide up to 16 cm, conditioned with methanol and water (250 ml each), then 50 ml of the coffee extract was applied, washed with 750 ml water and the lactones eluted with 500 ml methanol. The methanol was removed using a rotary evaporator and the residue taken up in 50 ml water.

HPLC-DAD analysis The apparatus used was the Gold Solvent Delivery Module M 116 (Beckman Instruments, San Ramon, USA), gradient mixer M 250 B (Gynkotek, Germering), injection valve Altex 210A (Beckman) with a 50-1xl loop, stainless steel column (250 mm x 4.6 mm i.d.) with octadecyl silane (ODS)-Hypersil 5 ~m (Shandon, Astmoor, UK); a Pye Unicam PU 4021 Multichannel Detector (Philips, Cambridge, UK), wavelength range 190-390 nm (UV); and the data system was IBM PC XT 286 with PU 6003 Diode Array Detection System and PU 6000 Integration Software (Philips). All eluents used were of HPLC grade (Baker, Gross-Gerau, Germany). The eluent was 2% acetic acid/methanol (75:25, v/v) at a flow rate of 1 ml/min (corresponding pressure 15 MPa). The column had to be rinsed with methanol after each run to remove di-CQA.

HPLC-TSP-MS HPLC. The apparatus used was the solvent delivery module 116 (Beckman), gradient controller M 250B (Gynkotek), ERC-3512 degasser (ERMA, Alteglofsheim, Germany); injection valve Rbeodyne 7125 (Berkeley, USA) with a 20-txl loop; UV detector Lichrograph L-4000/4200 (Merck Hitachi) with a high-pressure-flow cell; wavelength 324 nm. UV detector signals were transferred to the MS software (user trace) and (for quantification) to a computing integrator. For the HPLC conditions see above. In order to buffer ionization, 0.1 M ammonium acetate solution in 10% methanol was added at a flow rate of 0.5 ml/min using a second pump (between column and UV detector).

Intelface. A TSP-1 interface with control module was used with settings as follows: vaporizer 80~ C (discharge) 97~ (buffer); source block temperature 280 ~ C; repeller 50 V. The discharge mode was 850 V and the corresponding current 50 p~A.

Mass spectrometer. An SSQ 710 with TSP ion source and DEC station 5000]33 (Digital) with ICIS 2-software (Finnigan MAT, Bremen, Germany) was used. The settings were as follows: scan range: m/z 150-450 at 2 s per scan.

Preparative HPLC The equipment used comprised a Solvent Delivery Module M 114 (Beckmann), injection valve Rheodyne 7125 (Berkeley) with a 2000-1xl loop, stainless steel column (250 x 8 mm i.d.) with ODS-Hypersil 5 ~m, Pye Unicam PU 4025 UV detector (Philips), set at 324 nm and recorder LS-4 (Linseis, Selb, Germany). The eluent was water/methanol (75 : 25, v/v) at a flow rate of 3.3 ml/min (corresponding pressure 13.5 MPa). The methanol was removed using a rotary evaporator and finally the residue was freeze dried.

Nuclear magnetic resonance (NMR)spectroscopy. A WM 400 (Bruker) apparatus was used for 1-dimensional ~H-analysis and 2-dimensional H,H-COSY analysis, at a frequency of 400.13 MHz. Bruker Standard Software was used and CD3OD was the solvent. The reference signals for 1H were ~ = 3.35 ppm for CHD2OD, and 8 = 4.80 for HDO.

Quantification Determination was carried out by HPLC with UV detection at 324 nm and five-point calibration. 5-CQA was used as a calibration standard. The results were converted appropriate to the molecular masses.

Detection limit Using solutions of 5-CQA the detection limit (5-fold baseline noise) was determined to be beuer than 50 Ixg/1.

Results and discussion

Identification S e v e r a l H P L C systems h a v e b e e n e m p l o y e d for the determination of the m a i n C G A in c o f f e e [e.g. 10, 11]. T h e systems had to be o p t i m i z e d for the i n v e s t i g a t i o n of m i n o r C G A - l i k e compounds. F i g u r e 1 a s h o w s the separation of C Q A in roasted c o f f e e using an isocratic system consisting o f 2 % acetic a c i d / m e t h anol (75 : 25, v/v). T h e d i - C Q A w e r e not eluted under these conditions. Part o f the c h r o m a t o g r a m is s h o w n in Fig. 1 b. T h e m i n o r peaks 1 - 7 w e r e e x a m i n e d using H P L C - D A D . T h e spectra of peaks 4 and 9 c o u l d be assigned as p - c o u m a r i c acid derivatives. Peaks 7 and 10 s h o w U V spectra corresponding to derivatives o f ferulic acid. T h e other peaks (5, 6 and 8) are identified as caffeic acid derivatives. T h e latter attracted attention b e c a u s e they w e r e not o b s e r v e d in extracts o f g r e e n coffee. F o r additional i n f o r m a t i o n the coffee extract was analy s e d by T S P - L C - M S . F i g u r e 2 shows the mass spectra o f p e a k 6 and 5 - C Q A obtained in the discharge ionization m o d e . T h e difference b e t w e e n the signals o f the quasi m o l e cule ions in Fig. 2 a and Fig. 2 b was 18 mass units, w h i c h is typical for the e l i m i n a t i o n of water. T h e f r a g m e n t m / z = 337 appears under appropriate conditions also in the mass spectra of C Q A . T h e other f r a g m e n t s that o c c u r in the spectra derived f r o m caffeic acid and quinic acid. W h e r e a s caffeic acid is present in the f o r m o f its quasi m o l e c u l e ion m / z = 181, p r o t o n a t e d quinic acid m / z = 193 is absent in Fig. 2 a. T h e m o s t intensive f r a g m e n t o b s e r v e d was the [M + H] + o f quinide ( m / z = 175). T h e m a s s spectra o f p e a k 6 and 8 w e r e identical. It was c o n c l u d e d f r o m the T S P - H P L C - M S results that

19 0.5000"

2.5000

a

2.0000

0.4000'

i.5000

0.3000

t.0000

0.2000

0.5000

0. iO00

0.0000

~b

S Time

( mins

10 8

il3

0.0000

i~ )

Time

( mins

)

Fig. 1 a, b. HPLC separation of chlorogenic acids from a roasted coffee, a 1, 3-CQA; 2, 4-CQA; 3, 5-CQA. b Shows a part of a; for details see the text

8 I00

E+ 04

337

"

" 3.52

80"

60"

163

181

40"

20"

157

19~

(r

I

L.....i[.J,,~.ol,.,I ................................ 150

,

','.,...E.,:.._'..,.-'..- I............ L ... 3,' ......................... ................. L ............... 200 250 300

9

350

.-,............

?

400

E+ 05

355 I00

-

- 1.26

80-

60"

193

40-

2o-

1~

I

175

Fig. 2 a, b. Thermospray mass spectra of a 3-caffeoylquinide (peak 6) and b 5-caffeoylquinic acid (peak 3) obtained in the discharge ionization mode

3~7

/ 137

I

207

225

33~

L 150

200

250

300

350

both peaks could be assigned to caffeoylquinides. In experiments with the buffer ionization mode these conclusions were confirmed. For the IR and N M R spectroscopy the peaks of interest had to be isolated from the coffee brews. The extracts were cleaned up by p o l y a m i d e column chromatography and peaks

400

6 and 8 were isolated by preparative HPLC. The purity of the individual fractions was checked using analytical HPLCDAD. The pooled eluates were concentrated to remove methanol before the final freeze-drying step. It was not possible to isolate the individual lactones. During the concentration process isomerisation occurred, as observed previously [12].

20 Table 1. ~H-Nuclearmagnetic resonance (NMR) data of the caffeoylquinides (400 MHz)-g[ppm] Position a

3-Caffeoyl--/-quinide

4-Caffeoyl-'y-quinide

C-2-H~ C-2-Heq C-3-H.. C-4-Heq C-5-H,~ C-6-H.x C-6-Heq C-2'-H C-5'-H C-6'-H C-7'-H C-8'-H

2.12 (dd) 2.19 (m) 4.96 (ddd) 4.32 (dd) 4.79 (dd) 2.60 (d) 2.34 (ddd) 7.09 (d) 6.82 (d) 7.00 (dd) 7.66 (d) 6.34 (d)

2.00 (dd) 2.20 (m) 4.01 (ddd) 5.34 (dd) 4.91 (dd) 2.43 (m) 2.43 (m) 7.11 (d) 6.83 (d) 7.03 (dd) 7.68 (d) 6.40 (d)

O

O

2

OH

I

0

j~ O

8"

"

"

Fig. 3. Structure of 4-caffeoyl-3J-quinide

" See Fig. 3 /

Table 2. ~H-NMRcoupling constants J [Hz] of 3- and 4-caffeoylquinide (CQL) Coupling"

3-CQL

4-CQL

C-2-H~x, C-2-Heq C-2-Hax, C-3-Hax C-2-H~q,C-3-H,• C-2-Heq, C-6-Heq C-3-Hax, C-4-H~q C-4-H~q, C-5-Heq C-5-Heq, C-6-H,• C-5-Heq, C-6-H~q C-6-Ha~, C-6-Heq C-2'-H, C-6'-H C-5'-H, C-6'-H C-7'-H, C-8'-H

-11.6 11.6 6.8 2.6 4.3 4.6 < 2 6.0 -11.6 1.7 8.2 15.9

-11.8 11.7 6.5 _b 4.9 4.4 < 2 5.7 _b 1.8 8.1 15.9

/

// /

1,8- / /

/

1,6. / / / /

// 1.4////

1.2./ 1/ l./i / //

0.8"/ / / 0.6 . / / / / 0.4. / / -/

~-CQ

0.2,/ / / 0.79

10.53

2.80

1.11

5.72 ORV [~l

13.08

See Fig. 3 b Unresolved signal

Fig.4. Content of 3- (3-CQL) and 4-caffeoyl-3~-quinide(4-CQL) in roasted coffee as a function of the organic roasting loss (ORV). The resuits are calculated for green coffee; d.m., dry matter

Under experimental conditions a ratio of about 6 0 : 4 0 for peaks 6 and 8 resulted. This behaviour will be the subject of further research. The FTIR and N M R experiments were carried out with a 60 : 40 mixture of the compounds. The FTIR spectra showed an intensive signal at 1793 cm -1, which is typical for ",/-lactones [13]. The data obtained from NMR spectroscopy are given in Tables 1 and 2. The assignment was achieved by

H,H-COSY. The results confirmed the previous conclusions and gave additional information about the structure. The major component (peak 6) is 3-caffeoylquinide, whereas peak 8 was identified as 4-caffeoylquinide. The structure of 4-caffeoylquinide is given in Fig. 3. Because of the 60: 40 ratio, the substances could be assigned to the peaks in the UV chromatogram. The expected elution order was confirmed (3-CQL before 4-CQL).

Table 3. Contents of chlorogenic acid lactones and chlorogenic acids in coffee samples: means of duplicate determinations

CQL, Caffeoylquinic acid lactone; COA, Caffeoylquinic acid; n.d., not determined; -, not detectable; R.S.D., Relative standard deviation; d.m., dry matter; a ORV Organic roasting loss b Decaffeinated c Dark roast (espresso)

Sample

Degree of roasting a

3-CQL g/kg d.m.

4-CQL g/kg d.m.

Sum CQL g/kg d.m.

Sum CQA g/kg d.m.

Brand A Brand B Brand C b Brand D c Brand E c Columbia arabica Columbia arabica Columbia arabica Columbia arabica Columbia arabica Columbia arabica R.S.D.

n.d. n.d. n.d. n.d. n.d. 0.79 1.11 2.80 5.72 10.53 13.08

1.38 1.86 1.26 0.91 0.78 0.17 0.41 1.71 1.56 0.68 9.04

1.19 1.58 1.01 0.77 0.77 0.19 0.37 1.44 1.34 0.68 9.98

2.57 3.44 2.27 1.68 1.55 0.36 0.78 3.15 2.90 1.36

26.98 26.36 19.37 8.13 10.04 58.24 52.98 49.15 32.27 14.36 6.96 7.99

21

Quantification

References

A series of coffees of different degrees of roasting produced from green arabica coffee was analysed (see Table 3). Results indicate that the formation of lactones reaches a maxim u m in medium roasted coffee. In coffees of a higher degree of roasting the contents are lower. This means that a decomposition of chlorogenic acid lactones takes place. Figure 4 shows the amounts of the lactones calculated for green coffee as a function of the organic roasting loss. In addition, five different commercial roasted coffees were examined. The caffeoylquinides were also detectable in these samples. The average content was about 2.3 g/kg dry matter. The results are given in Table 3. The content of the lactones in an instant coffee sample was very low probably because of hydrolysis during the extraction process (data are not shown here).

1. Maier HG (1988) ASIC Colloq Sci Int Care 12:229-237 2. Clifford MN, Kellard B, Birch GG (1989) Food Chem 33:115-123 3. Stegen GHD van der, Duijn J van (1980) ASIC Colloq Sci Int Cafe 9:107-112 4. Morishita H0 Iwahashi H, Kido R (1986) Phytochemistry 25:2679-2680 5. Clifford MN, Jarvis T (1988) Food Chem 29:291-298 6. Scholz-B6ttcher BM, Maier HG (1991) ASIC Colloq Sci Int Cafe 14:220-229 7. SturmR (1983) Thesis, University of Hamburg 8. Wynne KN, Familari M, Boublik JH, Drummer OH, Rae D, Funder JW (1987) Blin Exp Pharmacol Physiol 14:785-790 9. Bundesgesundheitsamt (1992) Amtliche Sammlung von Untersuchungsverfahren nach w35 LMBG: L 46.00-2 10. Clifford MN, Ramirez-Martinez JR (1991) Food Chem 40:35-42 11. Morishita H, Iwahashi H, Kido R (1989) Bull Fac Edu Wakayama Univ Nat Sci 38:33-39 12. Clifford MN (1985) Chlorogenic acids. In: Clarke RJ, Macrae R (eds) Coffee, vol 1. Chemistry. Elsevier, London New York, pp 153-202 13. Ruveda EA, Deulofeu V, Galmarini OL (1964) Chem Ind 8:239-240

Acknowledgements. The

authors are grateful to the Deutsche Forschungsgemeinschaft for financial support of this project. We thank Dr. V. Wray (GBF, Braunschweig) for providing the NMR data and for his advice.