Determination of E 472b emulsifiers in foamed food formulations by high-performance thin-layer chromatography‒fluorescence detection

E 472b emulsifiers, defined as lactic acid esters of mono- and diacylglycerides, are food emulsifiers widely used in foamed food formulations. So far, only qualitative methods for analyzing E 472b emulsifiers have been published. Thus, a new method was developed for determining E 472b in different foamed food formulations by high-performance thin-layer chromatography with fluorescence detection (HPTLC‒FLD). The proposed method allows simple and fast E 472b emulsifier extraction from the food matrix and points out an analytical approach for quantifying these emulsifiers using a commercial E 472b emulsifier as the reference standard. Limits of decision with 56‒59 ng of E 472b emulsifier/zone and limits of quantification with 172‒179 ng of E 472b emulsifier/zone in three foamed model food formulations, respectively, as well as satisfactory repeatability (n = 6) and reproducibility (n = 6) exposed by low relative standard deviation < 8% proved the method suitable for the sensitive and reliable determination of E 472b emulsifiers. Recoveries between 96 and 109.3% were obtained for all investigated model systems. In commercial foamed food formulations from the German market, the E 472b emulsifier content ranged between 0.1 and 0.6 g/100 g.


Introduction
Emulsifiers of the type E 472 comprise a group of mixtures of mono-and diacylglycerides (MAG and DAG) of fatty acids additionally esterified with fruit acids. Classification is based on the fruit acid moiety in E 472a-f. Emulsifiers of type E 472b are defined as lactic acid esters of MAG and DAG [1,2].
Emulsifiers of type E 472b are produced by direct esterification of glycerol with free fatty acids and lactic acid or by transesterification of triacylglycerides (TAG) from hydrogenated edible oils with glycerol followed by lactylation. The products are complex mixtures of various lactic acid esters of MAG and DAG, including positional isomers, byproducts like polymerized lactic acid, and educts. The degree of esterification, and thus, the composition, is also affected by the reaction conditions [3]. Emulsifiers of type E 472b are approved food emulsifiers in the European Union (EU) according to Commission Regulation EC No 1333/2008 and are generally considered safe. Thus, a numerical acceptable daily intake (ADI) has not been established [2,4]. They are applied quantum satis in various foods, e.g., dairy products, bread, and ice cream, to adjust technofunctional characteristics [2,5]. In many cases of product reclamation, it is unclear whether instabilities of the products originate from deviations of the emulsifiers or raw material, also because of the lack of analytical methods that allow rapid analysis of the emulsifiers. Prior research showed that variations in the emulsifier composition generally lead to deviating product properties, especially viscosity characteristics [6,7]. Therefore, a constant composition of the emulsifiers is mandatory to guarantee a long shelf life and high and consistent product quality. Thus, the qualitative and quantitative analysis of these substances should be one focus to gain knowledge on the influence of E 472b emulsifiers on product stability, which is the main focus of this research project.
There have only been a few publications dealing with analyzing E 472b emulsifiers, mainly aiming for qualitative analysis. Traditionally, identification of the emulsifier includes extraction and hydrolysis of the esters, and thus, the release of the fruit acid moiety, followed by analysis via gas chromatography with flame ionization detection (GC-FID) [8] or thin-layer chromatography (TLC) [9][10][11][12]. Those methods were applied to dairy and non-dairy coffee cream powder [8] and coffee cream [9]. Aside from being only suitable for qualitative analysis, other drawbacks of the available methods are (a) the application of large quantities of halogenated solvents that can have adverse effects on human health for extraction of the emulsifiers from the food matrix [9,12], (b) the necessity of an additional derivatization step to obtain sufficient volatility for GC-FID analysis [8], (c) the application of a time-and solvent-consuming hydrolysis step [8,9], and (d) the loss of further information on the emulsifier composition in detail [8,9,12]. Those methods were not applied on E 472b emulsifiers but only on E 472a (acetic acid esters of MAG and DAG) and E 472d-f (tartaric acid esters, diacetyl tartaric acid esters, and mixed acetic and tartaric acid esters of MAG and DAG). In addition, a few methods for the qualitative analysis of E 472b emulsifiers by liquid chromatography-mass spectrometry (LC-MS) are described in the literature [13,14].
Recently, a method for the characterization of E 472 emulsifiers by high-performance thin-layer chromatography-fluorescence detection (HPTLC-FLD) and HPTLC coupled to MS (HPTLC-MS) has been published by Oellig et al. [15]. This method is based on a twofold development on HPTLC silica gel plates before derivatization with primuline, allowing a direct visual comparison between and within E 472 emulsifier classes. Identification was achieved by HPTLC-MS analysis. In addition, Oellig et al. [16] introduced a rapid-to-perform method for determining E 471 emulsifiers by HPTLC-FLD in aerosol whipping cream. Quantification was based on the strategy according to Oellig et al. [17], where individual lipid classes are collectively detected and quantified.
Thus, this work aimed to develop a cost-effective, fast, robust, reliable, and selective quantitative method for determining E 472b emulsifiers in foamed food formulations by HPTLC-FLD, based on the recently published methods [15][16][17]. With this, the existing qualitative method for the characterization of E 472b emulsifiers [15] should be extended to include the aspect of quantification. In contrast to existing methods, extraction should be carried out without the need for halogenated solvents and derivatization, the latter allowing the possibility for simultaneous characterization of the emulsifiers. As a first step, the suitability of the former methods applied to E 471 emulsifiers, should be tested and optimized if needed for their application to E 472b emulsifiers in aerosol whipping cream and, in a second step, extended to other foamed food formulations using model matrices. Furthermore, a simple quantification strategy should be developed for E 472b emulsifiers in the tested food formulations. To account for differences in commercially available E 472b emulsifiers, response factors for available E 472b emulsifiers should be calculated to evaluate diversity in response. Finally, different samples of aerosol whipping cream, fruit foam, and foamed yogurt formulation from the German market should be analyzed for their E 472b content.
Model aerosol whipping cream, fruit foam, and foamed acidified milk gel samples, the latter being a model system for foamed yogurt formulation, were produced by the Department of Soft Matter Science and Dairy Technology, University of Hohenheim (Stuttgart, Germany) according to the "Model aerosol whipping cream", "Model fruit foam", and "Foamed acidified milk gel" sections. The manufacturers either provided commercially available aerosol whipping creams, foamed dairy formulations, and fruit foam samples, or the samples were bought in local supermarkets and stored in the refrigerator before analysis.

Model aerosol whipping cream
To prepare model aerosol whipping cream, raw bovine milk provided by the research station Meiereihof (University of Hohenheim) was separated at 60 °C using a separator (SA 10-T, Frautech SRL, Schio, Italy). Cream (lipid content > 30 g/100 g) was heated to 90 °C with a batch pasteurizer (Pasteurizer C600/45, Kaelte Rudi, Keltern, Germany), and skim milk (lipid content ≤ 0.1 g/100 g) was pasteurized (72 °C for 28 s) employing a plate heat exchanger (KS8FS1514, ATS-Suedmo, Feldkirch, Germany). The cream was standardized to a lipid content of 30 g/100 g and heated to 80 °C in a metal beaker set in a water bath (Julabo Labortechnik, Seelbach, Germany). Different amounts of E 472b emulsifier (0.1-2.0 g/100 g) were added under constant stirring at a temperature ≥ 60 °C. After temperature equilibration (80 °C, 5 min), a pre-emulsion was prepared by dispersing the samples at 10,000 min −1 for 3 min with a high-shear-blender (IKA, Staufen, Germany). Dispersing of the samples was conducted in a water bath (WarmMaster Deluxe, Merten&Storck, Drensteinfurt, Germany) set to 80 °C to prevent temperature loss during stirring. After sample homogenization with a two-stage homogenizer (APV-Gaulin, Lübeck, Germany) at a pressure setting of 6/1 MPa, samples were pasteurized for 10 min at 80 °C. Samples were collected in 1-L laboratory bottles with high-temperature screw caps (Schott, Mitterteich, Germany) and immediately cooled with ice water. Unhomogenized standardized model aerosol whipping cream (lipid content 30 g/100 g) without the addition of E 472b emulsifier was used as a reference sample. To ensure fat crystallization, samples were stored at 5 °C for at least 24 h before HPTLC analysis.

Model fruit foam
Sucrose (Bäko, Duisburg, Germany) was dissolved in ultrapure water under constant stirring in a Thermomix (Varona, Vorwerk, Wuppertal, Germany). Then, E 472b emulsifier was added, and the pH was adjusted with 6 M HCl to pH 3. Samples were preheated to 55 °C under constant stirring in a Thermomix (Varona, Vorwerk) before adding swollen waxy corn starch (company secret). After further heating to 95 °C, swollen gelatin (Congel 130, Condio, Werder (Havel), Germany) was added to the mixture. Holding at 95 °C for 10 min, gelatinized the starch, dissolved the gelatin, and lead to the inactivation of microorganisms. Samples were filled in metal cups and stored at 10 °C in a water bath overnight (≥ 12 h). After the cooling period, samples were whipped with a hand mixer (MFQ 4020, Bosch, Gerlingen, Germany) for 3 min at maximum speed. The model fruit foam samples were filled in 150 mL plastic cups, covered with parafilm, and stored at 5 °C for ≥ 12 h before HPTLC analysis. The exact composition is subject to company secrets and is therefore not listed.

Foamed acidified milk gel (model system for foamed yogurt formulation)
Raw bovine milk provided by the research station Meiereihof (University of Hohenheim) was separated at 60 °C using a separator (SA 10-T, Frautech SRL, Schio, Italy). The cream was heated to 90 °C for 5 min with a batchpasteurizer (Pasteurizer C600/45), and the skim milk was heated to 72 °C for 28 s using a plate heat exchanger (KS8FS1514). For standardization of the protein content to 4.0 ± 0.2 g/100 g, milk protein powder (TMP 80, Milei, Leutkirch im Allgaeu, Germany) was added to the skimmed milk (lipid content < 0.1 g/100 g) using a rotor-stator system at 10,000 min −1 for 2 min (Robomix). Samples were stored overnight at 4 °C to guarantee hydration of the milk proteins. Then, gelatin (1.0 g/100 g, Congel 130, Condio, Werder (Havel), Germany) and different amounts of E 472b emulsifier (0.1-1.0 g/100 g) were added. For the addition of the emulsifier, a small part of the milk was preheated. Subsequently, samples were heated for 5 min at 90 °C in a batch-pasteurizer. After cooling to 25 °C, glucono-δlactone (1.5 g/100 g, Acros Organics B.V.B.A., Fair Lawn, New Jersey, USA) was added to initiate acidification. To conduct acidification, samples were tempered at 25 °C for 16 h in a water bath, and the pH decreased from pH 6.6 ± 0.2 to 4.4 ± 0.2. After acidification, 1.5 g/100 g aqueous potassium sorbate solution (10%, w/w) was added using a rotor-stator system (Robomix at 10,000 min −1 for 30 s) to ensure microbiological stability, and the samples were stored at 4 °C overnight. Whipping of the samples was conducted at 1000 min −1 and 10 °C with a hand mixer (MFQ 4020, Bosch). Whipped samples were stored at 5 °C before HPTLC analysis.

Standard solutions
Stock solutions of E 472b emulsifiers (emulsifiers of technical grade) that were used in the model systems (E 472b emulsifier 4, 8, 15, and 16, Supplementary Information Table S1) were prepared in TBME (33.33 mg of E 472b/mL) to develop and optimize the extraction method, for method validation and sample analysis. These stock solutions were freshly prepared each working day. Single stock solutions of 21 E 472b emulsifiers of technical grade for calculating the response factor and determining the chromatographic fingerprint were prepared at a concentration of 0.5 mg/mL in TBME and stored at room temperature.

Internal standard preparation
The internal standard (ISTD) preparation was based on the procedure described by Oellig et al. [16] with some modifications. An amount of 0.95 g (5 mmol) 2-NCl was weighed in a 20 mL glass centrifuge tube equipped with a screw cap and was dissolved in 1 mL of anhydrous dichloromethane and 1.2 mL (15 mmol) of pyridine by briefly mixing on a vortex mixer (Vortex Genie 2, Scientific Industries, Bohemia, USA). After adding 0.3 g (4.8 mmol) of ethane-1,2diol, the tube was heated for 30 min at 60 °C in a thermoblock (Liebisch Labortechnik, Bielefeld, Deutschland) and subsequently cooled down in a fume hood to room temperature. The reaction mixture was transferred to a 40 mL glass centrifuge tube, and 5 mL of iced ultrapure water and 10 mL of diethyl ether were added. The mixture was vigorously shaken at 2,200 min −1 for 5 min, and the upper organic phase was transferred to a further 20 mL glass centrifuge tube. It was washed two times with an aqueous solution of HCl (1 M) and two times with an aqueous solution containing 10% bicarbonate. Next, the organic phase was evaporated to dryness under a stream of nitrogen, and the residue was dissolved in 1.5 mL of TBME. After centrifugation, the clear supernatant was collected and stored at room temperature until use.

Aerosol whipping cream and foamed yogurt formulation (standard LLE method)
Liquid-liquid extraction (LLE) was performed according to the method of Oellig et al. [16]. Briefly, 1 g of homogenized sample was weighed into a 20 mL glass centrifuge tube equipped with a screw cap. After adding 20 µL of ISTD and 3 mL of ethanol, the tube was vortexed before gentle shaking at 250 min −1 for 30 min (KS 125, IKA, Staufen, Deutschland). Subsequently, 7 mL of ultrapure water and 2 mL of TBME were added. After brief vortexing, the tube was stored at room temperature for 20 min. Then, the tube was vigorously shaken for 30 min at 2,200 min −1 (VXR Basic, IKA), and a centrifugation step at 6600g for 10 min at 18 °C was carried out. An aliquot of the organic phase was diluted 1:100 (v/v) with TBME (sample concentration of 5 mg/mL) and subjected to HPTLC analysis.

Fruit foam (shortened LLE method)
One g of homogenized sample was weighed into a 20 mL glass centrifuge tube equipped with a screw cap. After adding 4 µL of ISTD, 7 mL of ultrapure water, and 2 mL of TBME, the tube was vortexed, and LLE was performed for 15 min on a small shaker at 2200 min −1 (VXR Basic, IKA). After centrifugation at 12,200g for 10 min at room temperature (Biofuge Pico, Kendro laboratory products, Asheville, USA), an aliquot of the organic phase was diluted 1:20 (v/v) with TBME (sample concentration of 40 mg/mL) and subjected to HPTLC analysis.

High-performance thin-layer chromatographyfluorescence detection
An Automatic TLC Sampler 4 (ATS4, CAMAG, Muttenz, Switzerland) was used for application. Samples and standards were applied on 20 cm × 10 cm plates with a distance from the lower edge of 8 mm and the left and right edge of 10 mm. As the rinsing solvent with one rinsing cycle and one filling cycle, TBME was used. After a drying step for 10 min in a fume hood, a twofold development occurred in an Automatic Development Chamber (ADC2, CAMAG) equipped with a 20 cm × 10 cm twin-trough chamber (CAMAG). Before both developments, a saturated magnesium chloride solution controlled the plate activity to 33% relative humidity. For the first development, a mixture of chloroform/methanol/water/formic acid (67:6:1.2:0.2, v/v/v/v) was used up to a migration distance of 50 mm, followed by a drying step for 10 min. As the mixture used was a two-phase system, it was well shaken until it was filled in the development chamber, which took place briefly before the end of the plate activity control. The second development was carried out with a mixture of n-heptane/diethyl ether/formic acid (55:45:1, v/v/v) up to a migration distance of 80 mm, including 5 min of drying. Plate images were captured with the TLC Visualizer (CAMAG) under UV 254 nm and UV 366 nm illumination after both developments. Following the second development, the plates were scanned with the TLC Scanner 4 (CAMAG) in absorption mode at UV 254 nm for the detection of the ISTD, applying a scanning speed of 20 mm/s, a data resolution of 100 µm/step, and slit dimension of 4 mm × 0.3 mm. For derivatization, the plates were dipped into a solution of primuline (0.05% in acetone/water, 4:1 (v/v)) with the TLC Chromatogram Immersion Device III (CAMAG, immersion speed 1, immersion time 2) and dried in a stream of cold air for 30 s. Plates were stored in a desiccator at a constant relative humidity of 47%, adjusted by a saturated solution of potassium carbonate for 1 h. After that, the plates were scanned in fluorescence mode at UV 366/ > 400 nm with the TLC Scanner 4 (CAMAG) to detect the E 472b emulsifiers. Control of the HPTLC instruments and data evaluation was performed with the software winCATS, version 1.4.6.2002 (CAMAG).

Determination of response factors for E 472b emulsifiers
Twenty-one E 472b emulsifiers of technical grade, directly obtained from manufacturers or the German dairy product industry, were investigated. The sample set included different E 472b emulsifier products and lots of the same emulsifier product from 2015 to 2021 (Supplementary Information  Table S1). For analysis, 15 µL of ISTD were diluted with 1 mL TBME. 10 µL of the respective ISTD solution were added to an aliquot of each stock solution. For HPTLC-FLD analysis according to the "High-performance thin-layer chromatography-fluorescence detection" section, 10 µL of the stock solutions of the E 472b emulsifiers ("Standard solutions" section) were applied (5 µg E 472b emulsifier/ zone). Response factors for each emulsifier were calculated by comparing the sum of the peak areas of MAG and lactic acid esters of MAG to the sum of the same peak areas of a respective reference emulsifier (Supplementary Information Table S1) after normalization to the ISTD. Furthermore, response factors were calculated for selected E 472b emulsifiers at amounts of 0.3-0.7 µg E 472b emulsifiers/zone as this corresponds to emulsifier amounts that are relevant in the dairy matrix.

Sensitivity
Limits of decision and limits of quantification (LOQ) were determined in triplicate for three emulsifiers (E 472b emulsifier 8, 15, and 16, Supplementary Information Table S1) in solvent and the three model matrices, according to DIN 32645 [18] (equal to ISO 11843-2 [19]). These three emulsifiers were selected because they were used to prepare the model matrices. Therefore, aliquots of 1 mL of ultrapure water or 1 g of E 472b emulsifier-free model matrix were spiked with different volumes of E 472b stock solution ("Standard solutions" section). For the standard LLE method (for solvent, aerosol whipping cream, and foamed acidified milk gel), 40-280 µL, and for the shortened LLE method (for solvent and fruit foam), 5-40 µL stock solution were used and further analyzed according to the "Final sample preparation" section. HPTLC-FLD analysis was performed according to the "High-performance thin-layer chromatography-fluorescence detection" section. To determine the limit of decision and LOQ, the sum of the peak areas of MAG and lactic acid esters of MAG was used after normalization to the ISTD.

Repeatability and recovery
Samples of the three model matrices with defined E 472b emulsifier content (E 472b emulsifier 4, 8, 15, and 16, Supplementary Information Table S1, were used) were manufactured by the Department of Soft Matter Science and Dairy Technology (University of Hohenheim) according to the "Model aerosol whipping cream", "Model fruit foam", and "Foamed acidified milk gel" sections. The sample preparation was performed according to the "Final sample preparation" section. Quantification was done by six-point calibration with similarly prepared solvent standards. Therefore, aliquots of 1 mL of ultrapure water were spiked with different volumes of E 472b stock solution (E 472b emulsifier 4, 8, 15, and 16, Supplementary Information Table S1). For aerosol whipping cream and foamed acidified milk gel, 30-600 µL, and for fruit foam, 15-150 µL stock solution were used. Sample preparation was performed according to the "Sample preparation" section, and HPTLC-FLD analysis was done according to the "High-performance thin-layer chromatography-fluorescence detection" section. For evaluation, the sum of the peak areas of MAG and lactic acid esters of MAG was used after normalization to the ISTD.

E 472b emulsifiers in foamed food formulations from the German market
Four aerosol whipping cream samples, five whipped dairy formulations, and two foamed yogurt formulations from the German market, all containing E 472b emulsifier, as well as eighteen fruit foam samples, commercial products, and also two lots of foam ground matrices from a manufacturer, all containing E 472b emulsifiers, were analyzed. In addition, one aerosol whipping cream containing both E 472b and E 471 was investigated. The ground matrix, bottled by the manufacturer, and the fruit foam from the corresponding commercial product were analyzed for all six flavored fruit foams. Samples were stored at 4 °C in a refrigerator and were analyzed within the best-before date. Before the analysis of the commercial product, the fruit foam, which was on top of the foamed yogurt formulation, was scraped off and homogenized with a spatula. Generally, all samples were homogenized by stirring before analysis. The analysis was performed in triplicate for each sample according to the "Final sample preparation" and "High-performance thinlayer chromatography-fluorescence detection" sections. A six-point calibration with solvent standards similarly prepared as the samples was used for quantification. Therefore, aliquots of 1 mL of ultrapure water were spiked with different volumes of the stock solution (15-600 µL) and further analyzed according to the "Sample preparation" section. For quantification, the sum of the peak areas of MAG and lactic acid esters of MAG was used after normalization to the ISTD.

Results and discussion
The aim of the present study was the development of a simple and reliable sample preparation that provides selective and quantitative extraction of lactic acid esters of MAG and DAG from different foamed food formulations, namely aerosol whipping creams, foamed yogurt formulations, and fruit foams, and additionally, the chromatographic separation of interfering matrix components such as cholesterol. Furthermore, a simple quantification strategy should be developed for E 472b emulsifiers. As analytical standards are not available, a commercially available E 472b emulsifier should be chosen as a reference emulsifier for quantification. To account for differences in commercially available E 472b emulsifiers, response factors for available E 472b emulsifiers should be calculated to evaluate diversity in response. This is important for analyzing samples with unknown E 472b emulsifier composition.

Sample preparation
Oellig et al. [16] optimized the extraction method for E 471 emulsifiers in aerosol whipping cream regarding extraction times, storing times, and the addition of salts and organic solvents. They showed that denaturation of the protein by an organic solvent, e.g., ethanol, was necessary for the complete release of the DAG. Subsequent extraction with TBME delivered recoveries between 86.0 and 105.0% for MAG and DAG. To correct errors during sample preparation, they also added 1,2-bis-dinaphthoylethanediol as ISTD.
Because of the related chemical structure and behavior of E 471 and E 472b emulsifiers, it was assumed that the presented sample preparation [16] could also be suitable for analyzing E 472b emulsifiers in several foamed dairy formulations. Recoveries between 96.0 and 105.0% E 472b emulsifier (content 0.1-2.0 g/100 g, n = 3, RSD ≤ 10%) in model aerosol whipping cream confirmed the method's suitability for application to E 472b emulsifiers for this matrix without the need for optimization. Furthermore, it was also demonstrated that the extraction method was applicable for E 472b emulsifiers in fruit foam, which was shown with model fruit foam (recoveries 96.4-103.0%, E 472b content 0.1-0.5 g/100 g, n = 6, RSD ≤ 10%). Blank aerosol whipping creams and fruit foams were additionally analyzed to account for matrix interferences, but no interferences were observed. The analysis of E 472b emulsifiers in foamed acidified milk gel initially led to overestimation. By analyzing and comparing the densitograms of emulsifier-free foamed acidified milk gel and emulsifier-containing foamed acidified milk gel, fluorescence quenching, interfering with the lactic acid esters of MAG, was observed to be the reason for the overestimation (Fig. 1). Taking an emulsifierfree foamed acidified milk gel into account during integration, satisfactory recoveries in foamed acidified milk gel with 96.6-109.3% (E 472b content 0.1-1.0 g/100 g, n = 3, RSD ≤ 6%) were obtained, too. Fluorescence quenching was not detected in other acidified formulations like commercial foamed yogurt formulations.

Extraction of E 472b emulsifiers from fruit foam
Based on the negligible fat content in fruit foam, indicated by only slight light blue fluorescent zones at hR F 86-92 representing TAG and the distinct zones at hR F 59-69 showing Fig. 1 Densitograms of the fluorescence scan at UV 366/ > 400 nm of E 472b-free (---) and E 472b-containing (---) foamed acidified milk gel after separation on HPTLC silica gel with a twofold development applying a mixture of chloroform/methanol/water/formic acid (67:6:1.2:0.2, v/v/v/v) to a migration distance of 50 mm and a mixture of n-heptane/diethyl ether/formic acid (55:45:1, v/v/v) to a migration distance of 80 mm and performing a derivatization step with primuline. The amount of foamed acidified milk gel applied on the plate was 50 µg/zone for the blank sample. In addition, for the spiked sample, the amount of E 472b emulsifier applied on the plate was 0.4 µg/ zone. Densitograms were normalized to the internal standard for better comparability, resulting in the indication of relative signal intensity DAG and lactic acid esters of DAG (Fig. 2), the extraction procedure was shortened. Differing from the original method [16], adding ethanol to the sample could be omitted as a simple LLE with water and TBME completely extracted the emulsifier. Furthermore, the shaking time of the LLE procedure was reduced. Preliminary results for strawberry fruit foam indicated that an extraction time of 5 min was sufficient for complete emulsifier extraction; however, further investigations with different flavored commercial fruit foams showed that the RSD within peach-passion fruit foam was above 20% (n = 3). Thus, the extraction time was finally set to 15 min, where all RSD were below 9% (n = 3). Additionally, centrifugation parameters were optimized to obtain adequate phase separation, with 10 min at 12,300g proving to be the most suitable. Finally, the optimized method offered a quantitative and robust extraction procedure for E 472b emulsifiers from fruit foam ("Fruit foam" section).

Extraction of E 472b emulsifiers from acidified milk gel and foamed yogurt formulation
To test the shortened LLE method for other foamed dairy formulations, one skimmed milk foamed acidified milk gel, manufactured as described in the "Foamed acidified milk gel" section (lipid content < 0.1 g/100 g), containing 1 g/100 g of E 472b emulsifier, was analyzed in triplicate. The mean recovery of the low-fat foamed acidified milk gel was 109.3%, and RSD < 3% showed good repeatability, thus, verifying the use of the shortened LLE method for low-fat food formulations. In addition, one commercial foamed yogurt formulation (fat content of 10 g/100 g according to the product declaration) was analyzed with the shortened LLE method in triplicate to account for high-fat samples. A commercial product was used for evaluation to consider the emulsifiers incorporation in the matrix during the manufacturing process. For the high-fat foamed yogurt formulation, a recovery of only 42.3% (RSD > 15%) compared to the content found with the original LLE method was found. Furthermore, the shortened LLE method was tested on aerosol whipping cream. For this purpose, two different model aerosol whipping creams containing 0.8 g/100 g E 472b emulsifier were analyzed in triplicate, showing low recoveries with 51.5-85.1% and comparably high RSD (> 25%). Thus, treatment with ethanol before LLE was essential for high-fat food formulations, and the shortened LLE method was only suitable for low-fat food formulations.

Selection of signals
Several considerations and attempts were taken into account to develop a simple and reliable quantification strategy for E 472b emulsifiers in different foamed food formulations by HPTLC-FLD. For method development and validation, signals in the HPTLC-FLD densitograms suitable for quantification were evaluated. For this, three emulsifiers (E 472b emulsifiers 8, 15, and 16, Supplementary Information Table S1) were used, each applied during the manufacturing of the model matrices by the Department of Soft Matter and Dairy Science for the model matrices aerosol whipping cream, foamed acidified milk gel, and fruit foam. Primuline, which is typically used for the detection of lipophilic components, was used to stain the lactic acid esters. As it binds non-covalently to different lipid structures, it is a non-specific detection method for lactic acid esters. Thus, MAG, DAG, and TAG, also parts of E 472b emulsifiers, and phospholipids, are being visualized [20]. Nevertheless, the variation in the HPTLC fingerprints for different E 472b emulsifiers is expressed by different signal intensities of the same signals (Fig. 3), resulting in a specific E 472b fingerprint that can be clearly distinguished from other E 472 and E 471 species. Furthermore, for confirmation of E 472b, TLC-MS can be applied [15]. Moreover, as blank samples of the three model systems showed, the signal hump in the range of hR F 25-45, consisting of MAG and their lactic acid esters, was found to be suitable for reproducible and exact  (2) with the addition of E 472b emulsifier. A aerosol whipping cream, B foamed yogurt formulation, and C fruit foam with an amount of model matrix applied on the plate of 50 µg/zone corresponding to 0.4 µg E 472b emulsifier/zone in the spiked sample, and D E 472b solvent standard with the same E 472b emulsifier amount/ zone. Separation was performed on HPTLC silica gel under UV 366 nm illumination after a twofold development applying a mixture of chloroform/methanol/water/formic acid (67:6:1.2:0.2, v/v/v/v) to a migration distance of 50 mm and a mixture of n-heptane/diethyl ether/formic acid (55:45:1, v/v/v) to a migration distance of 80 mm and performing a derivatization step with primuline quantification, owing to no matrix interference. Due to coextracted DAG of the dairy matrix, co-eluting with the lactic acid esters of DAG (Fig. 2), the signal hump in the range of hR F 59-69 was unsuitable for quantification.

Determination of response factors for E 472b emulsifiers
To guarantee correct and robust quantification, response factors of 21 emulsifiers of technical grade were calculated in relation to a selected reference emulsifier. This is quite important for the analysis of commercial samples, because these samples contain E 472b emulsifiers with unknown compositions in contrast to the samples used for method development and validation. To determine the response factors, 21 E 472b emulsifiers were dissolved in TBME spiked with ISTD and directly subjected to HPTLC-FLD analysis according to the "High-performance thin-layer chromatography-fluorescence detection" section and the "Materials and methods" section "Determination of response factors for E 472b emulsifiers". Analysis was carried out three times on different days at an E 472b amount of 5 µg/zone. At first, response factors were calculated as the ratio of the sum signal at hR F 25-45 to emulsifier 1, an E 472b emulsifier from the German market with a typical composition ( Supplementary Information Table S1). Response factors ranged from 0.7 to 1.3, with RSD < 6%, showing good reproducibility, with a deviation of ≤ 30% (Table 1, response factor A). Further investigations focused on categorizing the investigated emulsifiers to minimize the deviation. Previous studies [15] indicated that visual fingerprints of different E 472b emulsifier products showed remarkable differences. The characterization and identification by TLC-MS showed that with increasing hR F value, higher polymerized lactic acid esters were present when esterification with up to four lactic acid molecules was found [15].
Different emulsifier fingerprints were also obtained for the 21 investigated E 472b emulsifiers in this study, which is exemplarily and impressively shown by the densitograms of four emulsifiers (Fig. 3). Considering all fingerprints and the calculated response factors A, it was noted that E 472b emulsifiers with similar fingerprints additionally showed similar response factors. Thus, based on the chromatographic fingerprints and the calculated response factors, categorizing the E 472b emulsifiers into four groups was possible (Table 1; Fig. 3). All emulsifiers of group 1 (three emulsifiers) showed response factors of 1.0. Group 2 comprised the vast majority of emulsifiers (thirteen emulsifiers) that showed response factors of 1.1-1.3 and revealed a shift in substance signals towards higher hR F values, indicating the presence of higher polymerized lactic acid esters. Emulsifiers of group 3 (four emulsifiers) rendered response factors of 0.9-1.0 and showed a shift in substance signals towards unesterified MAG. One emulsifier (emulsifier 21) showed a completely different and unique fingerprint with high-intensity signals in the hR F range of DAG and TAG and a very low response factor of 0.7. This emulsifier was categorized into a fourth group comprising only this. Product specifications were available for twelve of the investigated E 472b emulsifiers. Interestingly, emulsifier 21, showing a unique fingerprint and the lowest response factor, was based on edible, refined palmitic acid with a comparatively low lactic acid content of 13-16% according to the product specification, whereas the other emulsifiers were based on edible, refined palm or rapeseed oil with a lactic acid content of 20-30%. Response factors for the E 472b emulsifiers were also calculated for amounts of 0.3-0.7 µg of E 472b emulsifier/zone since these amounts correspond to E 472b emulsifier amounts that are relevant in foamed dairy formulations. Deviations to the response factors calculated for 5 µg of E 472b emulsifier/zone were not found (data not shown). Categorizing the investigated emulsifiers is a valuable result not only for exact quantification, but it might also be meaningful to correlate differences in technofunctional properties with the composition of the emulsifiers. These investigations are part of current research in cooperation with the Department of Soft Matter Science and Dairy Technology.
Response factors A are group-independent factors since the defined E 472b emulsifier groups were not considered. A different approach with group-dependent response factors was also evaluated to assess the grouping and thus lead to a more precise quantification with lower deviation. For this, the calculation of the response factors for the members of each group was performed with a specific emulsifier of the respective group as a reference emulsifier, namely emulsifier 1 for group 1, emulsifier 4 for group 2, and emulsifier 17 for group 3 (Table 1, response factor B). The calculated group-dependent response factors ranged between 0.9 and 1.1, leading to an even more precise quantification with a deviation ≤ 10%. This approach was finally used to analyze commercial E 472b containing samples. With recoveries of 96.0-109.3% for the three model systems, it was shown that using the signal hump of the MAG and their lactic acid esters for quantification, it was possible to determine the amount of E 472b emulsifier correctly.

Application of ISTD
To compensate for analyte losses during LLE and volume errors due to dilution and application of the emulsifier on the plate, 1,2-bis-dinaphthoylethanediol was added to the samples before the LLE as ISTD. Due to the high number of signals originating from the E 472b emulsifiers after primuline detection, leading to difficulties in finding an ISTD with similar chromatographic properties, this ISTD was chosen, although it is only detectable under UV 254 nm. The selfsynthesized ISTD contained both mono-and 1,2-bis-naphthoylated ethanediol, revealing two signals at UV 254 nm that can be considered. Whereas the more intense signal of the ISTD at hR F 51 was suitable for the matrices fruit foam and aerosol whipping cream, the signal at hR F 70 was used for the matrix foamed acidified milk gel due to matrix interferences at hR F 51. It has to be noted that a remarkable decrease in signal intensity of the ISTD was observed when the absorption scan was performed after derivatization with primuline; therefore, the plate was scanned before the derivatization for detection of the ISTD. As the ISTD is not detected by primuline, this unconventional approach was considered, for the reasons described above. As already realized in earlier studies [16], the signal intensities of the components of E 472b emulsifiers were distinctively influenced by the storage period at a relative humidity of 47% that followed the post-chromatographic derivatization with primuline. To evaluate the effect of the storage period on the signals of the E 472b emulsifiers, the derivatized plate was scanned after storage times of 0-360 min in fluorescence mode at UV 366/ > 400 nm. When performing the fluorescence scan directly after derivatization, including a short drying period with cold air and no storage time, a high plate background, poor signal intensities, and peak shapes were obtained. Peak performance remarkably improved and enhanced after a storage period of 60 min at 47% relative humidity. A further increase in the storage period did not show improvement both in signal intensities and peak shapes. Reducing the storage period to 30 min showed nearly equal results to the storage period of 60 min. However, a slightly enhanced background was still detected, leading to difficulties quantifying samples with low E 472b contents. Thus, a minimum storage period of the plate of 30 min at 47% relative humidity before the scan in fluorescence mode at UV 366/ > 400 nm was found to be most suited for reliable and sensitive quantification of E 472b.

Solvent-matched calibration
Furthermore, it was evaluated whether linear or polynomic regression was best suited for quantification. Signals of the MAG hump corrected with the ISTD were used for the calculation of linear and polynomic regressions for all model matrices and the solvent-matched standard in the range of 0.25-1.0 µg of E 472b emulsifier/zone, corresponding to 0.05-2.0 g/100 g of E 472b emulsifier/matrix. Polynomic regression was suitable, showing high accuracy for high and low emulsifier contents. The three model matrices were additionally used to verify whether a solvent-matched calibration could be applied to quantify the E 472b emulsifier content. Therefore, matrix-matched and solvent-matched calibrations were analyzed according to the respective LLE method in relevant emulsifier content ranges (0.1-2.0 g/100 g for aerosol whipping cream (n = 2), 0.1-1.0 g/100 g for foamed acidified milk gel (n = 2), and 0.05-0.3 g/100 g for model fruit foam (n = 2)). Considering the signal of the ISTD, calibration curves showed no difference and were congruent for both the solvent-matched and the matrix-matched calibrations (Fig. 4). Thus, solvent-matched calibration proved

Sensitivity
To evaluate the sensitivity of the developed methods for the determination of E 472b emulsifiers in foamed food formulations, limits of decision and LOQ were determined in triplicate according to DIN 32645 [18] (which is equal to ISO 11843-2 [19]) in pure solvent and three model matrices (aerosol whipping cream, foamed acidified milk gel, and fruit foam) for three reference emulsifiers (E 472b emulsifier 8, 15, and 16, Supplementary Information Table S1). For this method, at least five calibration standards near the suspected limit of decision were used, providing a linear correlation between the amount of analyte and the signal. In addition, variance homogeneity is required between the calibration standard with the highest amount and the determined limit of decision. Limits of decision and LOQ are calculated based on the calibration equation, the quality, and the applied calibration range. Calibration was performed according to the "Material and methods" section "Sensitivity" in the range of 70-400 ng of E 472b emulsifier/zone, corresponding to 0.07-0.8 g/100 g of E 472b emulsifier/sample for aerosol whipping cream and foamed acidified milk gel, and 0.01-0.1 g/100 g of E 472b emulsifier/sample for fruit foam, respectively, taking the final sample preparation (Final sample preparation" section) into account. Calibration resulted in graphs of good linearity with high correlation coefficients (R 2 > 0.994). Limits of decision were determined with 56-59 ng of E 472b emulsifier/zone (n = 3, RSD < 10%), and LOQ with 172-179 ng of E 472b emulsifier/zone (n = 3, RSD < 10%) in aerosol whipping cream, foamed acidified milk gel, and fruit foam ( Table 2). Taking the sample preparation into account, this corresponds to limits of decision and LOQ of 0.11 g/100 g and 0.34 g/100 g of E 472b emulsifier for aerosol whipping cream, of 0.11 g/100 g and 0.35 g/100 g of E 472b emulsifier for foamed acidified milk gel, and 0.01 g/100 g and 0.04 g/100 g of E 472b emulsifier for fruit foam. To achieve lower limits of decision and LOQ, higher sample application volumes could be applied.

Repeatability and recovery
To determine repeatability, samples of the different model matrices with known E 472b emulsifier content (0.8 g/100 g for aerosol whipping cream, 1.0 g/100 g for foamed acidified milk gel, and 0.1 g/100 g for fruit foam) were analyzed sixfold under the same conditions and at the same day by one operator. The same sample set was analyzed for the determination of reproducibility, but samples were analyzed in duplicate on three different days by one operator. The E 472b emulsifier content was quantified by six-point calibration with extracted solvent standards according to the "Material and methods" section "Repeatability and recovery". Repeatability and reproducibility expressed as RSD delivered satisfactory results with < 6 and < 8%, respectively, for all three matrices (Table 2), the latter resulting in a measurement uncertainty (1 − α two-sided = 0.95) of 10.5% for aerosol whipping cream, 14.9% for foamed acidified milk gel and 11.3% for fruit foam, respectively. To determine the trueness and precision of the sample preparation methods, recovery experiments were additionally carried out in triplicate for the three model matrices for at least two further E 472b contents in the range of 0.1-2.0 g/100 g (aerosol whipping cream), 0.1-1.0 g/100 g (foamed acidified milk gel), and 0.1-0.5 g/100 g (fruit foam). Recoveries were in the range of 91.0-107.8% for aerosol whipping cream with RSD < 8%, 96.2-109.3% for foamed acidified milk gel with RSD < 6%,  (Table 3). Thus, the developed methods proved suitable for the reliable and precise determination of E 472b emulsifiers in aerosol whipping cream, foamed acidified milk gel, and fruit foam in the tested calibration range.

E 472b emulsifiers in foamed food formulations from the German market
All selected samples generally contained E 472b, but no other emulsifier of the type E 472 or E 471. In addition, one aerosol whipping cream sample containing both E 472b and E 471 emulsifiers was purchased for testing the limits of the developed method. The samples were analyzed in triplicate according to the developed extraction method and the HPTLC-FLD method mentioned in the "Sample preparation" and "High-performance thin-layer chromatography-fluorescence detection" sections. In a first step, samples were analyzed by HPTLC-FLD to obtain information on the fingerprint of the E 472b emulsifier present in each sample, allowing categorization of the applied E 472b emulsifiers in the previously defined groups ("Results and discussion" section "Determination of response factors for E 472b emulsifiers"). Thereby, the reference emulsifier of the respective group (Supplementary Information  Table S1) was chosen for quantification. Quantification was done according to the "Materials and methods" section "E 472b emulsifiers in foamed food formulations from the German market" by a six-point calibration in the range of 50-600 ng of E 472b emulsifier/zone for aerosol whipping cream, whipped dairy formulation, and foamed yogurt formulation, and in the range of 200-2,000 ng of E 472b emulsifier/zone for fruit foam, corresponding to 0.1-1.2 g/100 g and 0.05-0.30 g/100 g of E 472b emulsifier/sample, respectively, depending on the sample preparation. Due to DAG being co-extracted from the dairy matrix, categorization of the E 472b emulsifiers in the sample matrix was solely based on the signal of MAG and lactic acid esters of MAG, which slightly differed from the procedure described in the "Results and discussion" section "Determination of response factors for E 472b emulsifiers". Thus, discrimination between groups 3 and 4 was not possible, leading to a possible deviation in the quantification of 25%. All analyzed samples, except for two, contained an E 472b emulsifier of group 2, showing a higher degree of esterification than emulsifiers of groups 1, 3, and 4. One of the two exceptions, an aerosol whipping cream, showed an emulsifier of group 3 or 4. The other exception, also an aerosol whipping cream, contained E 472b and E 471. Due to the large signal at hR F 24 resulting from the MAG of the E 471 emulsifier, quantifying the E 472b emulsifier was impossible in this sample. The E 472b emulsifier content in the analyzed samples ranged between 0.1 g/100 g and 0.59 g/100 g, with RSD < 8% (Table 4), with the lowest E 472b emulsifier contents found in fruit foams and the highest in foamed dairy formulations. The aerosol whipping creams and the whipped dairy formulations contained E 472b emulsifiers in the range of 0.27-0.59 g/100 g and 0.22-0.55 g/100 g, respectively. Preliminary, these results were a little surprising because experiments with aerosol whipping cream regarding the optimal E 472b emulsifier content using an emulsifier of group 2 by the Department of Soft Matter Science and Dairy Technology of the University of Hohenheim showed the optimal emulsifier content to be 0.8 g/100 g. However, according to the product declaration, all aerosol whipping creams and whipped dairy formulations contained one or more food additives like carrageenan, guar gum, or gelatin, which also contribute to product stabilization making a higher emulsifier content redundant. These additives were not used during the Department of Soft Matter Science and Dairy Technology experiments. Both the ground fruit matrix, bottled by the manufacturer, and the fruit foam from the directly related commercial product were analyzed for all six flavored fruit foams. E 472b emulsifier content in the commercial fruit foam products ranged between 0.1 and 0.2 g/100 g and revealed contents between 73.3 and 120.0% compared to the amount found in the corresponding ground matrix. Underestimation in the final product can occur due to losses during manufacturing and migration to other parts of the product.

Conclusion
A robust and reliable HPTLC-FLD method for quantifying E 472b emulsifiers in foamed food formulations, namely aerosol whipping cream, foamed yogurt formulation, and fruit foam, was developed. For the extraction of the E 472b emulsifiers from the food matrix, LLE methods for high-fat and low-fat foamed food formulations were developed for the first time. Sample extracts were analyzed by HPTLC-FLD on HPTLC silica gel plates with a twofold development whereby interfering matrix compounds like cholesterol and TAG were successfully separated. A commercial E 472b emulsifier with known composition was assigned as a reference emulsifier and used as the calibration standard. For quantification, MAG and lactic acid esters of MAG were detected as the sum after derivatization with primuline, both for the reference emulsifier and the sample. It was shown that using the signal hump of the MAG and their lactic acid esters, it was possible to conclude the amount of E 472b emulsifier initially used in the food formulation. Limits of decision and LOQ were determined with 56-59 ng and 172-179 ng of E 472b emulsifier/zone in aerosol whipping cream, foamed acidified milk gel, and fruit foam, respectively. Satisfactory repeatability and reproducibility exposed by RSD < 8% proved the method suitable for the sensitive, robust, and reliable detection of E 472b emulsifiers. Recoveries between 96.0 and 109.3% with RSD < 8% for E 472b showed complete extraction of E 472b emulsifier for the developed method. HPTLC fingerprints and calculated response factors enabled categorization of technical E 472b emulsifiers in four groups, each showing distinctive patterns and specific group-dependent response factors offering quantification with deviation ≤ 10%. In foamed food formulation samples from the German market, E 472b emulsifier content ranged between 0.1 and 0.6 g/100 g, with the lowest amounts found in fruit foam and the highest in foamed dairy formulations; however, there were substantial variations in contents between samples of one class.
Data availability Supplementary data is provided online in the Supplementary Information section. Further insight into data is available on request from the corresponding author.