In this work, an effort was to find out whether square-wave anodic stripping voltammetry at GCPE can represent a suitable method for rapid determination of vitamin A.

All-trans-retinol (≥95%) and ethanol (≥99.5%) were purchased from Sigma-Aldrich. Universal 0.1 mol.L^-1 Britton-Robinson buffer (BRB), prepared by mixing of appropriate amounts of boric acid, glacial acetic acid, 85% phosphoric acid, and sodium hydroxide all from the aforementioned company, was used in selection of suitable detection medium. BRB was prepared using deionized water (minimum electric resistivity 18.2 MΩ cm, maximum 3 μg L^-1 of total organic carbon) made in a Milli-Q® ultrapure water system from Merck Millipore (Burlington, USA).

Voltammetric detection of accumulated lipophilic vitamins into the pasting liquid was performed in a conventional three-electrode arrangement containing always GCPE (working), silver chloride electrode with 3 mol.L^-1 KCl salt-bridge (reference) and platinum sheet (auxiliary electrode). These electrodes were connected to the potentiostat Autolab PGSTAT101 controlled by software Nova (Version 1.11.0), both from Metrohm (Prague, Česká republika).

Glassy carbon powder of type Sigradur G (mixture 5 – 20 μm, HTW Maintingen, Germany) and one of randomly selected pasting liquids were mixed in a ceramic mortar for 15 min to create homogenous glassy carbon paste. The amount of tested pasting liquid differed from 5 to 30% (w/w). The resulting glassy carbon paste was packed into the cavity of the Teflon® piston-driven electrode holder with an end-hole of 3 mm in diameter. It is necessary to mention that the height of column in the cavity must be less than 2 cm due to difficult extrusion of glassy carbon paste. It is recommended that freshly prepared GCPEs should not be employed in any experiments due to their rather unstable electrochemical behaviour attributed to an incomplete homogenization. Consequently, freshly prepared GCPEs were left at the laboratory conditions for one day. After this selfhomogenization process, GCPEs can be used for following voltammetric measurements (Sýs et al., 2017).

Principle of medium-exchange extractive stripping voltammetry is illustarted in Figure 1. The extraction of lipophilic vitamins into the pasting liquid was carried out from 10 mL non-treated milk and cream samples (available in Czech stores) without need to apply a potential in the electrode cell (nonelectrolytic preconcentration), which is an approach known as “open circle procedure”. After 10 min, GCPE enriched with analytes was rinsed with a stream of deionized water and immersed together with others electrodes into 0.1 mol.L^-1 BRB. Final voltammetric detection was performed using square-wave voltammetry at potential range from 0 to +1.4 V, potential step (E_step) of 5 mV, potential amplitude (E_ampl) of 25 mV and frequency (f) of 50 Hz.

Generally, the repeability may be expressed by several indexes, nemaly coefficient of repeatability (CR), coefficient of variation (CV) and intra-class correlation coefficient (ICC). CR defined by formula (1) is a precision measure which represents the value below which the absolute difference between two repeated test results may be expected to lie with a probability of 95%. The standard deviation (σ) under repeatability conditions is part of precision and accuracy.

(1) C R = 1.96 2 σ 2

However, the repeatability is more often given by CV defined as the ratio of the standard deviation (σ) and to the mean (μ). If this ratio is expressed as a percentage (see Eq. 2) then it will be referred to as relative standard deviation (RSD).

(2) R S D = σ μ ∗ 100

Sufficient extraction repeatability constitutes the main criterion for development of voltammetric methods utilizing extractive accumulation to be able to use them for analytical purposes. ICC could not be used because units of two physical quantities (variables) were statistically tested only. Therefore, using RSD can be probably expected to be sufficient.

Several GCPEs differing in the type of pasting liquid and containing always 20% (w/w) portion were investigated in SWASV of cow's milk (3.5% fat) to choose the optimum one. Working conditions for this experiment were as follows: accumulation for 10 min, stirring at 400 rpm, electrochemical detection in 0.1 mol.L^-1 BRB (pH 4.5) at E_start = 0 V, E_end = +1.4 V, E_step = 5 mV, E_ampl = 25 mV and f = 10 Hz. Due to relatively high current response and required reproducibility (Table 1), silicone oil should be taken for optimum extraction of lipophilic vitamins.

Under the prediction, an amount of extracted lipophilic vitamins would increase with a higher content of paste liquid in GCPE. However, electrochemical properties of GCPE are affected by ratio between glassy carbon powder and paste liquid. In principle, it can be stated that carbon particles remain in intimate contact (electrically conductive) until the amount of paste liquid exceeds 30% (w/w) (Švancara and Schachl, 1999).

Surprisingly, it was found that the highest peak current response (unfortunately, background current (I_b) as well) was obtained at GCPE containing 5% (w/w) silicone oil, as demonstrated in Table 2. Despite high current yield, the content of 15% (w/w) silicone oil was chosen as optimum, thanks to the high reproducibility of accumulation. A non-specific adsorption of milk fat onto electrode surface, which acts as an electrical insulator causing a significant increase of background current (baseline signal), can be considered as possible explanation.

Principally, the optimum value of accumulation time is defined as a period required for reaching the equilibrium of lipophilic vitamins distribution between a nonopolar pasting liquid of GCPE and used milk sample. The cow's milk can be considered as a direct emulsion (so-called the first type emulsion) because a small amount of fat droplets (organic phase) are uniformly distributed throughout the milk volume (aqueous phase).

Resulting saturation curve describing dependence of current peak height on accumulation time showed a typical extraction equilibrium isotherm, as shown in Figure 2. The extraction equilibrium has been achieved after 600 s because using accumulation for longer period did not cause any significant increase in peak current response. Hence, accumulation time of 10 min was chosen as optimum.

Stirring speed affects the rate of fat droplets transport to the electrode surface where these droples containing lipophilic vitamis are then extracted into the pasting liquid of GCPE. Under this study, it was found that setting the rate of magnetic stir bar higher than 300 rpm did not have any significant effect on increase in final peak current response. Therefore, the above mentioned value can be considered as an optimum for subsequent experiments.

The fat dispersed in cow's milks and creams is formed by non-polar TCGs that are surrounded by phospholipids and membrane lipoproteins (Heid and Keenan, 2005). At the natural pH of the cow's milk, they carry a negative charge and thus prevent bonding of MFGs. It is worth considering whether whole MFGs are extracted into the pasting liquid or a further equilibrium distribution of present lipophilic vitamins between the TCGs and the pasting liquid cannot exist. It can be assumed that both processes take place simultaneously. TCGs extracted (adsorbed) block surface of GCPE and therefore cause a dramatic increasing the background current.

From the literature (Indyk and Woollard, 1997; Hulshof et al., 2006; Trenerry et al., 2011; Musara and Nyagura, 2017), lipophilic vitamins are found primarily in the milk fat. Unlike cholecalciferol (vitamin D3) and phylloquinone (vitamin K1) present in limit amounts (0.1 μg per 100 g), α-tocopherol (vitamin E), and retinol together with its provitamines (carotenoids) such as β-carotene, zeaxanthine and luteine (vitamin A) are the major representatives (40- 110 μg per 100 g). Moreover, an artificially added retinyl palmitate (Jensen et al., 1991) can be present as well.

Generally, most extracted lipophilic vitamins and their provitamins usually provide very broad sensitive oxidation/reduction peaks (up to 250 mV) due to slow kinetic of corresponding electrode reactions occurred at liquid-liquid interface (Sýs et al., 2019). As confirmation, a broad anodic peak beginning +0.705 V at and ending at +1.007 V was obtained for all investigated cow's milks and creams.

It is therefore impossible to distinguish and determine the individual forms of retinoids and carotenoids (Žabčíková et al., 2018). Nevertheless, a number of published scientific papers suggest that the all-trans-retinol occupies a dominant position (Jensen, 1994; Hulshof et al., 2006; Hodulová et al., 2015). Thus, it can be assumed that the peak obtained most likely corresponds to the anodic oxidation of all-trans-retinol a +0.852 V (compare with (overlapping peak at +0.886 V for cow's milk), as shown in Figure 3.

It seems that proposed extractive stripping voltammetry (ExSV) based on direct immersing of GCPE into continuously stirred cow's milk (3.5% fat) and subsequent electrochemical detection using SWV provides the desired sensitivity for detecting the sum of retinoids and carotenoids. A quantitative or at least semi-quantitative determination of vitamin A in cow's milk and cream samples was not the aim of this study. Voltammetric analysis of cow's milk enriched by differently defined amounts of all-trans-retinol could be probably considered as semi-quantitative analytical method.

Optimisation consisted in finding out proper working conditions for an anodic oxidation of lipophilic vitamins which are presented in MFGs accumulated into silicone oil. At first, the electrochemical detection has been subjected to pH study which was investigated for 0.1 mol.L^-1 RBRs of pH values from 2 to 7. A linear relationship between peak potential and pH values of used supporting electrolytes, statistically evaluated as E_p = -0.0558 pH + 1.0891 (R² = 0.9978), was observed. The peak potential was shifted to more negative values with increased pH of used BRBs. This phenomenon probably occurs due to lowering the energy barrier and easier deprotonation of present lipophilic vitamins. The value of slope 0.0558 indicates the transition of electrons together with protons in a 1:1 ratio. The most sensitive peak current response was achieved using BRBs of pH values 4 and 5. It seems that BRB would be replaced by an acetate buffer of pH 4.5, more simple in composition.

Parameters of SWV, potential amplitude and frequency, were optimised at constant potential step of 5 mV. Extracted lipophilic vitamins provided a broad anodic peak which height increased with increasing potential amplitude up to value of 25 mV. Therefore, potential amplitude of 25 mV was chosen for following analysis of cow's milks and creams. As shown in Figure 4, the height of anodic peak linearly increased with higher frequency. However, it was observed that background current (baseline) increased as well. Hence, a value of 50 Hz representing a compromise was chosen as optimum.

At first, it is necessary to mention that a sample of whipped cream (40% fat content) could not be analysed using designed protocol due to high viscosity (Van Vliet and Walstra, 1980). The whipped cream completely covered GCPE used and did not allow rinsing the surface with distilled water. At first glance, someone may think that peak current will be higher for creams ('more milk fat more vitamins') (Gaucheron, 2011). According to Figure 5, demonstrating a dependence of anodic peak current on milk fat content in selected samples, it seems that the assumption is misleading. An explanation could be summarized as follows: lipophilic vitamins present in samples are equally distributed between pasting liquid and milk fat during ('liquid-liquid') extraction and these vitamins are more detained in creams (12 - 31%) than in milks (0.5 - 3.5%) due to many times higher fat content ('like dissolves like').

Assuming that MFGs are homogeneously dispersed throughout the volume of cow's milks and creams and their diameter is not higher than 1 μm (Robin and Paquin, 1991; Michalski et al., 2004), an extraction repeability will be affected only by homogeneity of glassy carbon paste used (Sýs et al., 2017). If recovery of developed HPLC-based methods ranging approximately from 85 to 110% (Blanco et al., 2000; Gomis et al., 2000) is taken to account, satisfactory repeatability of extraction characterized by RSD lower than 9% was achieved. Even despite the relatively short error bars (see Figure 5), it is not possible to determine, on the basis of the peak heights received, whether it is cream or milk with a 3.5% fat content.