Determination of pyridine and furfuryl alcohol in breath after coffee consumption

Chemicals

Standards of Py (ReagentPlus, ≥99%, CAS: 110-86-1) and FFA (≥98 %, CAS 98-00-0) were purchased from Sigma-Aldrich. Mobile phase solvents methanol CHROMASOLV^TM (gradient grade, for HPLC ≥99.9%, Honeywell) were used when running the HPLC system, and deionized water was obtained using a deionizer system NANOpure Infinity (Barnstead/Thermolyne, USA). Acetic acid glacial (USP, BP, Ph. Eur.) pure, pharma grade (Applichem) was used to acidify the water mobile phase (0.1%).

Coffee drink preparation

Coffee beans (dark roasted, 100% Arabica) were purchased from a local store. Coffee drink samples were obtained using two different brewing methods, CafeRomatica (NIVONA) espresso coffee machine and Italian moka pot. Two different parameters could be adjusted on the CafeRomatica coffee machine when the coffee beverages were prepared. The first parameter was the quantity of coffee beans used, and the second was the quantity of water used to brew the coffee. Light, normal and strong coffee could be chosen and 3.50, 7.50 and 9.00 g of coffee beans were used, respectively. The most common volumes of brewing water used by our volunteers were analysed (i.e. 30, 100 and 240 ml of water). For Italian moka pot coffee drinks, 12 g of ground coffee and 200 ml of water were used.

HPLC analysis

Coffee drinks were prepared as described above. A sample of 2 ml was collected into a closed 2 ml plastic centrifuge vial to avoid evaporation of the VOCs and left to cool. The samples were then filtered using a 0.45 μm PTFE Captiva Econofilter (Agilent). An aliquot of 100 μl was spiked with Py and FFA standards and taken to a final volume of 1000 μl by adding deionized water with 0.1% acetic acid. An aliquot of 50 μl of each sample was analysed using an Agilent 1100 series HPLC system with a diode array detector and an Agilent Zorbax XDB-CN (3.0 × 150 mm, 3.5 μm) column.

Separation was achieved using a gradient mobile phase of 0.1% acetic acid in water (A) and methanol (B) at 25°C. The gradient started at 95% Phase A, increased to 20% Phase B after 2.5 min, reaching 100% phase B after 10 min, and was held until the end of the run (15 min). The flow rate used for the entire run was 0.5 ml/min. The absorption recorded for the detection of Py was 254 nm, and 217 nm for FFA. HPLC conditions were similar to the proposed [16] where furan derivates were analysed in coffee. As we wanted to analyse both Py and FFA in a single run using the same parameters, different columns and mobile phase gradients were studied until the desired separation of both compounds was obtained.

TD/GC-MS analysis

Once the coffee drink was prepared and an aliquot of 2 ml was collected for HPLC analysis, the rest of the sample was consumed by the volunteer. Five volunteers participated in the study, providing samples during several days. All the volunteers were non-smokers and could drink and eat any time before the study. Because of that, a breath sample was collected from the volunteer just before the coffee was consumed, using a self-modified 2-liter plastic bag [31]. The background signal obtained was subtracted from the breath sample after coffee consumption. This helped to ensure that the concentrations of MRPs obtained were coming only from the coffee ingested for the experiment. Immediately after finishing the drink, another breath sample was collected.

The breath sample was passed through a thermal desorption tube (Markes International, with Tenax TA sorbent) at 250 ml/min (air sampler Dupont Alpha-2). The thermal desorption (TD) tube with the sample was then inserted in the TD system (Perkin Elmer ATD400) and heated to 280°C for 10 min at a desorption flow of 60 ml/min. The compounds were released inside the system using helium gas (purity 5.0) and refocused inside the system’s cold trap (–15°C). Fast heating of the trap to 250°C (and hold for 3 min) at 8 ml/min desorption flow released the compounds into the gas chromatographic system where the analytes were separated using a temperature ramp. An outlet split of 6 ml/min was used to allow a flow of 2 ml/ml through the chromatographic system. The initial temperature of 40°C was held for 3 min and then increased to 220°C at a rate of 20°C/min.

An Agilent 6890N gas chromatography system with DB-5MS (30 m × 0.25 mm × 0.25 μm) column was used for the separation of the analytes, followed by a Waters AutoSpec Premier (Waters/MicroMass Technologies) mass spectrometer for detection and identification. The transfer line to the mass spectrometer was maintained at 250°C. Mass spectra were obtained by electronic ionisation at 70 eV, and magnetic scan was performed in the range m/z 50–100. For the quantification of the analytes, m/z 79 and m/z 98 were extracted for Py and FFA, respectively.

Standards preparation

For the HPLC analysis, each coffee sample was spiked with a standard solution of 1, 2 and 4 ppm of Py and 10, 20 and 40 ppm of FFA and injected into the HPLC system. The linearity obtained for this method was satisfactory (Table 1). Each coffee sample was prepared in triplicate to assess repeatability.

Table 1. r² values from the standard addition calibration of coffee samples spiked with Py and FFA for 3 different aliquots

[a] Si (Light coffee), 3.5 g; N (Normal coffee), 7.5 g; St (Strong coffee), 9 g; Mk (Moka coffee), 12 g and brewing water volume (ml).

[b] Not enough sample to perform a third run of samples.
 

To calibrate the breath analysis, two water solutions of 5 and 50 ppm of both analytes were prepared. From each solution, 5, 10 and 20 μl were selected to obtain a 6-point calibration curve. As the range of concentrations in breath was very wide, we wanted to obtain the highest possible range of concentration by using the same volume intake for both solutions to avoid errors. Therefore, the calibration points were not equidistant.

A self-made 2-liter plastic bag was partly filled with nitrogen gas and placed into the oven for a few minutes at 40°C. Then, a volume from the standard solution was collected in a gas-tight syringe and quickly injected into the conditioned bag through the plastic valve and closed. The bag was then completely filled with nitrogen gas and a sample was collected for analysis. Before a set of breath samples was analysed, a standard injection was performed to assess reproducibility. When standard levels differed more than 10%, a new calibration was made.

Amount of MRPs in breath does not depend on the amount of ingested MRPs

Regardless of the quantity of coffee beans used, levels of Py and FFA extracted per gram of coffee beans with the CafeRomantica coffee machine were similar (Table 2). Higher quantities of both MRPs were obtained when preparing the coffee using an Italian moka pot due to higher temperature reached by the steam and longer time to brew the coffee [32, 33]. The quantities of toxic MRPs present in the coffee drinks ranged from 0.2 to 3 mg/cup for Py, and from 7 to 30 mg/cup for FFA both depending on the amount of water used for brewing the coffee.

Table 2. Quantities of pyridine (Py) and furfuryl alcohol (FFA) determined in coffee drink

^[a] Si (Light coffee), N (Normal coffee), St (Strong coffee), Mk (Moka coffee).

^[b] Grams of coffee used.

^[c] Millilitres of water used for brewing.

^[d] Concentration of target compounds in the sample, in micrograms per millilitre.

^[e] Milligrams of target compounds in the full cup of coffee.

^[f] Milligrams of target compounds per gram of coffee bean.

Note: standard deviation <10%.

Similar levels of FFA in espresso coffee (1.6 mg/g) were found in other studies [34], but also lower levels were found in the literature, where maximum levels of 0.41 mg/g were detected (Okaru, 2017). Also, lower levels of Py (0.04 mg/g) were detected [35]. This difference in the concentrations of MRPs in coffee could be due to the different roasting procedures of the coffee beans [16], as well as the different bean growing locations, which might change the composition of the coffee [36].

Breath samples from all volunteers showed the concentrations of both toxic MRPs in a range from 1 to 1200 ng/l for pyridine and to 760 ng/l for FFA (Fig. 2). We could observe a big dispersion of the results for each sample replicate. According to the literature, the values of Py in breath from 0.66 to 141 μg/l were found in patients with end-stage renal disease [37], and up to 300 μg/l in active smokers [38]. Regarding FFA, we did not find other studies reporting FFA in breath.

The levels of toxic MRPs in coffee drink were compared with their levels found in breath of each volunteer after drinking the beverage (Fig. 3) using the Pearson test. No correlation between the levels of toxics ingested and their presence in breath was observed (Py = 0.00135 and FFA = –0.168). Furthermore, we observed a big dispersion of the results for each volunteer ingesting the same amount of MRPs. Thus, we studied different factors that could influence the MRP levels in breath when drinking coffee or collecting the sample.