Development and validation of a rapid method for the determination of natamycin in wine by high-performance liquid chromatography coupled to high resolution mass spectrometry

Abstract

A selective, sensitive and rapid procedure based on liquid chromatography coupled to electrospray ionisation in positive mode high resolution mass spectrometry (LC-ESI-HR-MS) has been developed and single laboratory validated for the determination of natamycin, a naturally occurring fungicide, in white and red wine. This technique allowed accurate masses to be measured and interferences excluded from the extracted ion chromatograms. Natamycin measured at a high mass accuracy of m/z 666.31069 and a confirmatory ion at m/z 503.22672 facilitated the use of minimal sample preparation procedures. This allowed for the direct injection (20 µL) of wine, thereby minimising the photosensitive analyte degradation. Chromatographic separation was achieved on a C₁₈ column using a mobile phase gradient of 10% to 90% methanol : acetic acid (97 : 3, v/v) over 25 minutes with a total run time of 30 minutes. The limits of detection and quantification were 3 and 5 µg L⁻¹ respectively. The method was successfully validated to show the standard range linearity, sensitivity, accuracy and precision in the matrices tested. The method was applied to 190 commercial wines, with 50 (26%) positive results (>5 µg L⁻¹) over a period of three months between December 2009 and February 2010, with concentrations ranging from 7 to 2134 µg L⁻¹. Natamycin is not permitted for use in wine within the European Union and this method may be used to aid enforcement of regulations.

---

1. Introduction

Natamycin, also known as pimaricin (Fig. 1), is a naturally occurring product of Streptomyces species, specifically from actinomycete Streptomyces natalensis and also from Streptococcus species particularly Streptococcus lactis. It occurs as a white, virtually tasteless and odourless crystalline powder with low water solubility. Natamycin is an effective macrolide polyene agent, active against both yeast and moulds.

Fig. 1 Structure of natamycin ((1R,3S,5R,7R,8E,12R,14E,16E,18E,20E,22R,24S,25R,26S)-22-[(3-amino-3,6-dideoxy-D-mannopyranosyl)oxy]-1,3,26-trihydroxy-12-methyl-10-oxo-6,11,28-trioxatricyclo[22.3.1.0]octacosa-8,14,16,18,20-pentaene-25-carboxylic acid) showing the MS fragmentation site.

Natamycin (E235) is permitted in the European Union (EU) under Annex III Part C of European Directive 95/2/EC as a preservative for the surface treatment of semi-hard and semi-soft cheese and dry, cured sausage at a maximum level of 1 mg dm⁻² in the outer 5 mm of the surface, corresponding to 20 mg kg⁻¹.¹ It has the advantage over other preservatives that it does not affect the taste and appearance of the final products. Natamycin is not permitted for use in wines that are prepared or sold within the EU, however, the fraudulent addition of natamycin to wine may ensure that the final product does not degrade, thereby giving an unfair advantage to wines that have been adulterated over those that have been prepared using traditional techniques.

The present study was instigated in response to reports in the German press that natamycin had been detected in some Argentinean wines; hence UK wine distributors wished to extend their quality control tests to include natamycin.

All of the methods published to date to determine natamycin have been described with respect to their application to cheese, cheese rind, sausage and lactoserum. These methods use various analytical techniques, including microbiological assay,² immunochemistry,³ spectrophotometry and liquid chromatography^4–8 and LC tandem mass spectrometry.⁹ The LC methods were inter-laboratory validated and were accepted by the International Organization for Standardization (ISO) for the determination of natamycin in cheese and cheese rind.¹⁰ The stated quantification limit for the ISO method is 0.5 mg kg⁻¹ or 0.03 mg dm⁻² and therefore does not have adequate sensitivity for the determination of low levels of natamycin expected in adulterated wines. Lee et al.⁹ describe a method using LC-MS/MS for determination of the antifungal drug amphotericin using natamycin as an internal standard. This method demonstrates that natamycin is amenable to LC-MS analysis and, if optimised, could be improved with enhanced specificity and the required sensitivity for the determination of natamycin in wine.

Wine is a complex matrix comprising a mixture of water, acids, sugars, alcohols, polyphenols, etc. Unlike previous methods validated for natamycin in cheese using liquid chromatography^7,8 the method described does not incorporate any clean up or extraction steps, which are essential in cheese analysis in order to extract the analyte from lactoserum. Although wine is a complex mixture comprising of many components, the high resolution of the MS allows a simple direct injection approach without clean up (i.e. the ability to differentiate matrix components from target analyte by accurate mass). Also compared to previous methodology published for cheese, a further benefit of this rapid approach is the reduced sample handling time, decreasing the possibility of photo-degradation of natamycin.

We present here an in-house validated direct injection liquid chromatography high resolution mass spectrometric method for the determination of natamycin in wine.

2. Materials and methods

2.1. Chemicals and reagents

Pimaricin (i.e. natamycin, >95% from Streptomyces chattanoogensis, P9703) was purchased from Sigma-Aldrich (Gillingham, UK). Methanol and water (both HPLC grade) and acetic acid 100% (analytical reagent grade) were obtained from Fisher Scientific (Loughborough, UK).

The standard diluent solution comprised methanol : water : acetic acid (50 : 47 : 3, v/v/v).

Natamycin primary stock solution (1 g L⁻¹) was prepared by dissolving 10 mg natamycin in 10 mL standard diluent. The solution was stored in an amber bottle at 4 °C and was stable for 2 weeks.

Natamycin working solution I (10 mg L⁻¹) was prepared by diluting 100 µL of the primary stock solution to 10 mL in an amber volumetric flask with standard diluent. Natamycin working solution II (500 µg L⁻¹) was prepared by diluting 500 µL of working solution 1 to 10 mL in an amber volumetric flask with standard diluent.

Quantification was achieved by standard addition series to an unknown blank sample using a minimum of 4 calibration points.

2.2. Samples

Two different types of wine were used for the validation study. French Chardonnay and French Cabernet Sauvignon, both 2009 vintage, were selected to represent white and red wine respectively. The method was applied to over 190 wine samples from December 2009 to February 2010 obtained from various retail outlets in the UK and stored at ambient temperature prior to analysis.

2.3. Sample preparation

For the initial screening, 2 mL of the wine was pipetted into two separate 2 mL Eppendorf centrifuge tubes. One was labelled as blank and the other as a spike at 5 µg L⁻¹. To the spiked sample, 20 µL of a 500 µg L⁻¹ natamycin stock (working standard II) was added to the liquid. The spiked tube was then shaken for 1 minute mechanically. Both tubes were then centrifuged at 14 000 rpm for 10 minutes to separate out any particulate matter. An aliquot of the supernatant was then filtered through a 0.2 µm polytetrafluoroethylene (PTFE) syringe filter into a 2 mL amber glass vial prior to analysis by HR-LC-MS.

If the initial screen result was positive the analysis was repeated using the same protocol with an extended standard addition calibration series. The level of standard addition was varied depending upon the expected concentration of natamycin in the sample. Typically, 50, 100, 200 and 500 ng of natamycin was spiked into 2 mL of sample for accurate quantification to be achieved.

2.4. Instrumentation

High performance liquid chromatography was performed using an Accela™ autosampler injection and pump system (Thermo Fisher Scientific, San Jose, USA). The autosampler vial tray was maintained at 8 °C. The needle was washed with methanol–water (50 : 50, v/v). Separation was performed by injecting 20 µL onto a 150 × 2.1 mm, 3.5 µm Sunfire™ octadecylsilane (C₁₈) column (Waters UK, Elstree, UK) with the column oven maintained at 30 °C. Mobile phase A consisted water/acetic acid (97 : 3, v/v) and mobile phase B was methanol/acetic acid (97 : 3, v/v), delivered at 0.25 mL min⁻¹. A gradient was run from 90% to 10% mobile phase A over 25 minutes, this was held for 2 minutes before immediately returning to the start conditions. Equilibration was for 3 minutes, giving a total run time of 30 minutes.

A Thermo Scientific Exactive™ bench top Orbitrap technology high-resolution mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) was used for the detection of natamycin. The mass spectrometer was calibrated by infusion of a standard test mixture of caffeine, methionine–arginine–phenylalanine–alanine (MRFA) tetrapeptide and Ultramark 1621 fluorinated phosphazenes (Thermo Fisher Scientific, Waltham, USA). All subsequent mass analysis was performed following the initial mass calibration and no internal calibration was used. Electrospray ionisation was used as the interface and high purity nitrogen served as a drying gas at a flow rate of 60 L min⁻¹. Standard MS conditions compatible with the LC flow rate were used (capillary temperature 290 °C; auxiliary flow 5 L min⁻¹; spray voltage 3.5 V; auxiliary temperature 400 °C). Mass resolution was set at 50 000 at 2 Hz over the range m/z 480–670. Natamycin eluted at ca. 16.4 minutes with a main [M + H]⁺ ion of m/z 666.31069 and a confirmatory ion at m/z 503.22672.

3. Results and discussion

3.1. Method development

The method was developed in the first instance to be amenable to MS detection since peak identification was to be based on high resolution mass spectral data. Secondly, the method had to be rugged enough to allow direct injection of the sample thereby minimising losses of photosensitive natamycin.

In order to optimise the liquid chromatography conditions, both reversed and hydrophilic interaction chromatography (HILIC) phases were investigated to obtain good separation of natamycin from interfering compounds. A Sequant™ ZIC®-HILIC (100 × 2.1, 3.5 µm, Hichrom Ltd, Reading, UK) was found to provide good peak shape of a solvent standard solution, but when spiked into matrix, the sensitivity of the target analyte was reduced due to competition for ionisation with other compounds in the sample matrix. The extended gradient and equilibration time required for this column required an analysis time in excess of 45 minutes. Since direct injection of wine was a key objective, sufficient retention of the analyte on the column was necessary to allow most of the more hydrophilic endogenous substances to be eluted first. To achieve this, a short column with a sub 2 µm C₁₈ stationary phase was investigated, but this exhibited similar matrix interference problems to the HILIC column. A longer reversed phase Sunfire™ C₁₈ column with larger particle size (150 × 2.1 mm, 3.5 µm) was found to provide adequate separation of natamycin from interfering compounds, while maintaining excellent peak shape.

Many of the methods reported in the literature used a methanol : water gradient with a low pH to ensure that natamycin remained stable during the analysis. It was also reported to be beneficial to have a relatively high organic content in the mobile phase to allow formation of a fine spray of small droplets in the ESI interface. The acetic acid in the mobile phase at a concentration of 3% (v/v) provided an environment in the solution phase for the formation of protonated natamycin [M + H]⁺ ions. A gradient of 10–90% methanol over 25 minutes was employed in order to allow sufficient time for interfering compounds to elute before the natamycin at ca. 16.4 minutes. This delay allowed the LC flow to be diverted away from the mass spectrometer during the first 14 minutes. The column flow was then redirected into the MS to allow detection of the natamycin, before being diverted back to waste after 18 minutes. When flow diversion was not used, the MS ionization source became contaminated very rapidly accompanied by a significant reduction in sensitivity. A total ion chromatogram from a typical red wine sample is shown in Fig. 2.

Fig. 2 Total ion chromatograph of red wine spiked with natamycin at 200 µg L⁻¹.

The mass spectrometric conditions were optimised by infusion of a 100 µg L⁻¹ natamycin solvent standard at a flow rate of 10 µL min⁻¹ into a 100 µL min⁻¹ flow of methanol : water : acetic acid (50 : 47 : 3, v/v/v). Protonated molecular ions ([M + H]⁺) were produced in the ESI-MS positive ion mode. Electrospray voltage, capillary voltage, sheath gas flow and auxiliary gas flow were optimised for ionisation of natamycin in mass channel m/z 666.31069. During optimisation, a second confirmatory in-source fragment ion (m/z 503.22672) was observed that is attributable to loss of the substituted tetrahydropyran moiety (Fig. 1) and was used to confirm the presence of the target analyte. The ratio between the responses of the two mass channels was used as a quality control parameter during routine analysis. The accurate mass resolution provided by the instrumentation allowed natamycin to be resolved from isobaric interferences. The accurate mass capability also provided enhanced sensitivity due to the selectivity of the extracted ion chromatograms.

The quantitative analysis of target compounds in a matrix such as wine requires both selectivity and sensitivity. When using mass analysers such as time of flight and Orbitrap™, the sensitivity and selectivity are not mutually exclusive parameters. This is illustrated in Fig. 3 for the analysis of a red wine sample spiked at 5 µg L⁻¹. Data were collected at a resolution setting of 50 000. In Fig. 3, the chromatograms were generated by extracting the m/z 666.31069 signal using decreasing extraction window widths around the mass of interest. When a large mass extraction window was used (Fig. 3a) the selectivity was poor and accurate quantification was not possible. As the extraction window was narrowed from 500 mDa to 5 mDa (Fig. 3b–d), the selectivity was improved such that the target compound could be accurately quantified. It is important to note that mass accuracy must be maintained as the extraction window is decreased; otherwise the signal will be lost by falling outside the narrow window. A mass analyser that can maintain mass accuracy over time and over a wide dynamic range of the signal is therefore requisite for such a high-resolution application.

Fig. 3 Extracted ion chromatograms from the same data file of a red wine spiked at 5 µg L⁻¹, using decreasing extraction windows (a) 1000 mDa, (b) 500 mDa, (c) 100 mDa, and (d) 5 mDa. Mass windows were centred around m/z 666.31069.

3.2. Method validation

Due to the variation expected from different wine matrices and the potential effect on ionisation, it was decided that quantitation was to be achieved by standard addition.

3.3. Linearity and dynamic range

The linearity of the method was assessed in three different matrices: solvent, white wine and red wine. These were prepared by spiking natamycin at ten different concentrations 0, 1, 5, 10, 50, 100, 200, 1000, 2000, 2500 µg L⁻¹. Fig. 4 shows calibration graphs for the three matrices. All three showed excellent linearity over this range (r² > 0.99). Residuals were calculated for every point on all three calibration graphs. These results showed achieved linearity was over the range 1 to 2500 µg L⁻¹.

Fig. 4 Calibration graphs for natamycin in (a) solvent, (b) white wine, and (c) red wine.

Due to the different composition of red and white wines the gradient of the calibration curves differs for the three matrices. Over the calibration range, the overall response in the white wine matrix was approximately 75% of the response in solvent for both ions. For the red wine the overall response over the calibration range was 50% of the solvent response. This highlights a matrix suppression effect which can occur between different types of wine and this is the primary reason for the standard addition approach adopted. The matrix effect on calibration standards can be seen in Fig. 5.

Fig. 5 Overlaid calibration graphs for natamycin in solvent, white wine and red wine.

3.4. Limit of detection and quantitation

The limit of detection (LOD) and the limit of quantitation (LOQ) for natamycin were calculated from the response from the m/z signals at 503.22672 and 666.31069 at known concentrations. The LOD's and LOQ's corresponding to signal-to-noise ratios of 3 : 1 and 5 : 1, respectively were 3 and 5 µg L⁻¹ (lower responder confirmation ion 503.22672) in both red and white wines.

3.5. Precision and accuracy

The precision (repeatability) of the method was assessed by spiking solvent, white wine and red wine at two concentrations, 5 and 200 µg L⁻¹. Eight replicates of each matrix were analysed and quantified by standard addition. The results are shown in Table 1. The relative standard deviation expressed as a percentage (RSD%) was within acceptable limits ranging from 5.7–11.1% at 5 µg L⁻¹ and 2.7–4.2% at 200 µg L⁻¹. The accuracy was assessed by spiking a known amount of natamycin at two different concentrations into white and red wine. The analysis was then performed by a second analyst in triplicate without the knowledge of the spiked natamycin concentration (assigned value). The results are shown in Table 2. Recoveries of natamycin (spiked at 125 and 220 µg L⁻¹) were in the range 105–114% in white wine and 96–106% in red wine. The Z scores calculated from white wine show a very slight positive bias, which may indicate low level background contamination of the wine with natamycin.

Table 1 Precision (repeatability) of the analytical method in three matrices at two different concentrations

Matrix	Spike at 5 µg L^−1 range of results (n = 8)	Standard deviation/µg L^−1	RSD (%)	Spike at 200 µg L^−1 range of results (n = 8)	Standard deviation/µg L^−1	RSD (%)
Solvent	5.3–6.4	0.3	5.7	221–251	9.8	4.2
White wine	4.4–5.3	0.3	5.9	204–223	5.8	2.7
Red wine	3.8–5.5	0.5	11.1	171–198	7.5	4.1

---

Table 2 Accuracy of natamycin determination when spiked into white and red wine at two concentrations, 125 and 220 µg L⁻¹

	Assigned concentration/µg L^−1	Reported concentration/µg L^−1	Accuracy (%)	Z Score^a
a Calculations: Z scores = (reported concentration − spiked concentration)/target standard deviation, where: target standard deviation = 0.16 × spiked concentration, i.e. according to the Horwitz equation.^11
White wine A rep 1	125	135	108	0.50
White wine A rep 2	125	142	114	0.85
White wine A rep 3	125	138	110	0.65
White wine B rep 1	220	230	105	0.28
White wine B rep 2	220	230	105	0.28
White wine B rep 3	220	239	109	0.54
Red wine A rep 1	220	213	97	−0.20
Red wine A rep 2	220	234	106	0.40
Red wine A rep 3	220	223	101	0.09
Red wine B rep 1	125	129	103	0.20
Red wine B rep 2	125	129	103	0.20
Red wine B rep 3	125	120	96	−0.25

---

3.6. Natamycin in wine samples

The method was applied to the analysis of 190 wine samples, originating mainly from South Africa and South America, during the period from December 2009 to February 2010. Fifty samples (26%) gave positive results for natamycin (>5 µg L⁻¹) all of which originated from South America. The levels reported ranged from 7 to 2134 µg L⁻¹, with an average concentration of 390 µg L⁻¹.

4. Conclusions

A rapid, sensitive and highly selective method for the determination of natamycin in wine has been developed and single-laboratory validated. The combination of liquid chromatography and high resolution mass spectrometry has allowed for the direct injection of wine samples. The selectivity of the method has been enhanced by the presence of a confirmatory ion arising from in-source fragmentation, thereby providing a high degree of confidence in identifying positive samples. The method was successfully applied for the analysis of natamycin in wines at low µg L⁻¹ (ppb) levels in commercial samples of white and red wine and may therefore be considered for use as an enforcement tool.

References

1. EU, European Parliament and Council Directive No. 95/2/ECof 20 February 1995 on food additives other than colours and sweeteners, Off. J. Eur. Communities: Legis., 1995, 61, 23–24.
2. D. C. Grove and W. A. Randall, Assay Methods of Antibiotics, Medical Encyclopedia, Inc., New York, NY, 1955.
3. E. Martlbauer, H. Ali, R. Dietrich and G. Terplan, Arch. Lebensmittelhyg., 1990, 41, 112–115.
4. L. G. M. T. Tuinstra and W. A. Traag, J.– Assoc. Off. Anal. Chem., 1982, 65, 820–822.
5. W. G. De Ruig, J. J. Van Oostrom and K. Leenheer, J.– Assoc. Off. Anal. Chem., 1987, 70, 944–948.
6. D. J. Fletouris, N. A. Botsoglou and A. J. Mants, J. AOAC Int., 1995, 78, 1024–1029.
7. M. P. Lambros, S. A. Abbas and D. W. A. Bourne, J. Chromatogr., B: Biomed. Sci. Appl., 1996, 685, 135–140.
8. L. F. Capitan-Valley, R. Checa-Moreno and N. Navas, J. AOAC Int., 2000, 83, 802–808.
9. J. W. Lee, M. E. Petersen, P. Lin, D. Dressler and I. Bekersky, Ther. Drug Monit., 2001, 23, 268–276.
10. International Organization for Standardization, Cheese and Cheese Rinds: Determination of Natamycin Contents Method by Molecular Absorption Spectrometry and by High-Performance Liquid Chromatography, ISO Standard 9233-2, Geneva, Switzerland, 2007.
11. M. Thompson, Recent trends in inter-laboratory precision at ppb and sub-ppb concentrations in relation to fitness for purpose criteria in proficiency testing, Analyst, 2000, 125, 385–386.

---