A rapid and sensitive method for simultaneous quantification of seven furfural compounds in milk powder based on GC-MS/MS combined with QuEChERS method

Abstract

Abstract: A precise analysis method using gas chromatography coupled with triple quadrupole mass spectrometry (GC-MS/MS) was developed to detect seven furfural compounds in milk powder, including furfural, 2-acetylfuran, 5-methyl-2-furfural, 1-(5-methylfuran-2-yl)ethanone, 2(5H)-furanone, 5-hydroxymethyl-2-furfural and furfuryl alcohol. Applying QuEChERS (quick, easy, cheap, effective, rugged, and safe) extraction methodology, the detection condition for all compounds was optimized to obtain good linearity between 0.001–2 mg/kg with correlation coefficients (R²) above 0.9990. The recoveries for the seven furfurals from milk powder samples are 92.06–106.19 %, with intra-day relative standard deviation ranging from 0.01 % to 3.32 % and inter-day relative standard deviation from 0.09 % to 1.58 %. The limits of detection (LODs) were between 0.0003–0.001 mg/kg and the limits of quantification (LOQs) were in the range of 0.001–0.003 mg/kg. A total of 34 real samples were analyzed to estimate the reliability and practicality of the developed method. The results obtained in the samples studied showed that galacto-oligosaccharide and fructose evidently increased the content of 5-hydroxymethyl-2-furfural, and the levels of 2(5H)-furanone was positively connected to the proportion of carbohydrate, in milk powders.

Introduction

Milk powder is becoming more popular because it provides equivalent nutrients but with advantages of lower weight, simpler transport and longer shelf life (Aktağ et al., 2019). To ensure the safety and longer shelf life, compared with liquid milk, relatively high temperatures were applied for not only sterilization but also drying (Albalá-Hurtado et al., 1998). As a result, milk powder is more easily to trigger Maillard Reaction (MR). Though MR can reduce the nutritional value and change the flavor and color of milk powder, but some undesirable compounds produced during the MR at advanced states can be helpful indicators to evaluate the extent of damage in milk powder (Madani-Tonekaboni et al., 2015; Ferrer et al., 2005; Wagner et al., 2010).

To date, the most widely studied furfural compounds in processed foods are 5-hydroxymethyl-2-furfural (5-HMF), furfural (F), 2-acetylfuran (FMC) and 5-methyl-2-furfural (MF) (Abraham et al., 2011; Capuano et al., 2011), which result in cytotoxicity, carcinogenicity and genotoxicity once over a certain limit of the body absorption (Teixidó et al., 2006; Murkovic et al., 2006; Sun et al., 2017). Moreover, 2(5H)-furanone (2(5H)-F) (Totlani et al., 2007), furfuryl alcohol (FA) and 1-(5-methylfuran-2-yl)ethanone (MMF) (Li et al., 2016), which affect organoleptic quality of milk powder, are also produced in MR, but are rarely investigated (van Boekel, 2006; Nursten et al., 2005). Therefore, it is meaningful to establish methods to assess these seven furfural compounds and thus evaluate the quality of milk powder.

Until now, researchers have developed several methods to detect furfural compounds including spectrophotometric (de Andrade et al., 2017; Melo et al., 2016), gas chromatography (GC) (Senila et al., 2012), and high-performance liquid chromatography (HPLC) (Gheisari et al., 2018; da Silva et al., 2015) methods. The application of spectrophotometric method in dairy product is simple to handle, but the analysis results are variable due to its matrix effects and low selectivity (Espinosa-Mansilla et al., 1993). Therefore, HPLC and GC methods have become major strategies to analyze furfural compounds. For example, Demirhan et al. (2015) used HPLC to measure5-HMF and F in milks and infant formulas. Gaspar et al. (2009) developed a method using direct immersion solid-phase microextraction (SPME) that extracted for 2 h at 25 °C and desorbed for 5 min, coupled to GC-FID or GC–TOF-MS for the direct analysis of 5-HMF, F and MF in liquid and water soluble foods. Tsai et al. (2012) used SPME technique with on-fibre derivatisation to extract at 80 °C, following by GC-MS for determination of furfural in infant formulas, beers, and vinegars. Compared to GC and GC-MS, GC-MS/MS has advantages of better sensitivity and confirmation reliability, which become the effective tools for the qualitative and quantitative analysis of trace target compounds in several matrices (Rodríguez-Carrasco et al., 2014). In this study, we successfully applied GC-MS/MS to simultaneously analyze seven furfurals in milk powder.

Sample preparation is a critical step, which not only impacts the accuracy of experimental results but also determines the complexity of the analysis process. Madani-Tonekaboni et al. (2015) extracted furfural and hydroxymethyl furfural in baby-food by using dispersive liquid-liquid microextraction (DLLME) and the recoveries of F and 5-HMF were 92.0 and 98.1 %, respectively. Gökmen et al. (2006) used aqueous extraction of 5-HMF and solid-phase extraction (SPE) for cleanup with the recovery ranging between 91.8 and 94.7 %. Therefore, choosing suitable preparation approach to improve the selectivity and the effectiveness of the method is obviously important.

In the present study, a sensitive and efficient analytical method for F, 5-HMF, FMC, MF, MMF, FA and 2(5H)-F analysis in milk powder was developed. This method using QuEChERS followed by GC-MS/MS detects in the selective reaction monitoring (SRM) mode for fast and efficient qualitative and quantification. In addition, as one of the crucial effects in MR, kinds of sugars, such as galacto-oligosaccharide (GOS), were investigated by analyzing commercial milk powder.

Materials and Methods

Chemicals and samples    Furfural (purity 99.0 %), 5-hydroxymethyl-2-furfural (purity 97.0 %), 5-methyl-2-furfural (purity 98.0 %), 2-acetylfuran (purity 99.0 %), 2(5H)-Furanone (purity 98.0 %), furfuryl alcohol (purity 99.0 %) and 1-(5-methylfuran-2-yl) ethanone (purity 97.0 %) were obtained from Bioroyee Biotechnology Co., Ltd (Beijing, China). Acetonitrile (ACN, LC-MS grade) was purchased from Sigma-Aldrich (Santa Clara, USA). The analytical-grade reagents, sodium acetate anhydrous (NaAc), sodium chloride (NaCl) and magnesium sulfate anhydrous (MgSO₄) were obtained from XiLong Scientific Co., Ltd. Primary secondary amine (PSA) (40–63 µm, ultraclean bulk for QuEChERS) and octadecylsilyl -modified silica gel (C18) (50 µm, ultraclean bulk for QuEChERS) were supplied by Agela Technologies (Shanghai, China). Milk powder samples were purchased from retail stores and covered all ages containing 18 infant formulas, 7 whole milk powders, a skim milk powder, 3 maternal formulas, 2 special purpose infant formulas, a middle-aged and elderly milk powder, a high-fiber milk powder and a family nutrition milk powder.

Instruments    All measurements were performed on a triple quadrupole mass spectrometer, model number TSQ8000, and manufactured by Thermo Fisher. The desktop high-speed centrifuge (Sorvall ST 16R Centrifuge), manufactured by Thermo Fisher, was used in this study. The MX-S mixer used in the extraction and mixing step was manufactured by Shanghai Bajiu Industrial Co., Ltd. The ultrasonic cleaner (SB-120 DTN) was from Ningbo Scientz Biotechnology Co., Ltd.

Standard solutions    Standard stock solutions of each target compound were prepared by dissolving in methanol. In order to guarantee the simplicity and accuracy of standard curve, the mixed standard solutions (10 mg/L) were prepared by diluting the mixture of seven standard solutions with methanol and stored at −20°C. On this basis, calibration standard solutions were prepared by further diluting from the mixed standard solutions and the concentrations were between 0.002 and 2 mg/L.

Sample preparation    Milk powder sample (4.0 g) was dissolved by ultra-pure water (40°C, 20 mL) and vortexed for 30s, then a 5.0 g portion of such solution was treated with three different preparation methods.

Solid-phase extraction: 0.3 g oxalic acid (C₂H₂O₄) was added to the samples and heated at 100 °C for 30 minutes, then 2.5 mL of 40 % trichloroacetic acid (TCA) was added prior to be shaken and centrifuged for 3 min at 10 000 rpm, following collected the supernatant. The deposit was added with 10 mL of 4 % TCA and centrifuged at 10 000 rpm for 3 min, and the upper layer was combined with the first extraction. After that, the mixture was applied to SPE (OASIS PRiME HLB) tubes which were washed with 5 mL of ultrapure water and target compounds were eluted from the column with 5 mL of methanol. The solvent was evaporated to dryness by purging nitrogen and the remainder was dissolved with 1 mL of methanol and then filtered. Finally, the filtrate was injected into GC-MS/MS for analysis.

Liquid-liquid extraction: The proteins were precipitated according to the process of the first part of solid-phase extraction. The samples were added with 5 mL of diethyl ether and mixed on a vortex for 2 min. A total of 0.5 mL aliquot of the supernatant was evaporated to dryness under nitrogen flow and was added with 0.5 mL of methanol before filter and transfer to an autosampler vial for the chromatographic analysis.

QuEChERS: The samples obtained above was added with 10 mL ACN, 3.0 g of NaCl and 1.0 g of NaAc, then mixed thoroughly on a vortex for 2 min and sonicated for 20 min. Then, 3.0 g of MgSO₄ was added to the mixture and centrifuged for 5 minutes at 6000 r/min. A 1 mL aliquot of the supernatant was added into a 10-mL centrifuge tube with 40 mg of C18 adsorbent, 10 mg of PSA adsorbent and 100 mg of MgSO₄, vortexed for 1 min, and then centrifuged for 5 min at 4000 r/min. The supernatant was filtered and then analyzed by GC–MS/MS.

Validation of the Method    Matrix effect, limits of detection (LODs), limits of quantification (LOQs) and recovery of target compounds were evaluated by spiking blank samples. Any matrix effect was investigated by comparing the calibration curves of standard in solvent and matrix. Recoveries were calculated by analyzing spiked sample at the concentrations of single fold, triple, and tenfold of LOQs. And linearity was calculated by analyzing standard mixtures with different concentrations ranged from LOQs. And linearity was calculated by analyzing standard mixtures with different concentrations ranged from LOQs to 2 mg/kg.

GC-MS/MS conditions    The chemical components were analyzed by using a Thermo TSQ8000 Triple Quadrupole GC–MS/MS system equipped with Rtx-WAX column (30 m × 0.25 mm i.d., with 0.25 µm film thickness). The column was first programmed to heat to 60 °C for 2 min followed by heating to 120 °C at a rate of 20°C/min, and then increased to 220 °C at 10 °C/min and held for another 10 min. Helium was used as carrier gas at a flow rate of 35.7 cm/s and the injector was used in the split mode. The temperature of the transfer line was maintained at 280 °C, and argon was used as the collision gas. Data acquisition and processing were performed with the Trace Finder GC software from Thermo Fisher.

Results and Discussion

Optimization of sample preparation methods    In order to ensure the extractive efficiency of target substances and reducing the influence of impurities during the sample pretreatment, different sample preparation methods were applied in this study, including solid-phase extraction, liquidliquid extraction and QuEChERS method. A mixed standard solution of seven analytes with concentration of 1 mg/kg was added into the same milk sample to assess the feasibility of different methods and the results is shown in Fig. 1.

Fig. 1.

Average recoveries (%) in different sample preparation approach in milk powder (n = 3).

During solid-phase extraction process, the disturbing ingredients such as fats will increase the matrix effect. After removed those ingredients, most of the target compounds were retained by SPE tubes, then the results of recoveries were close to 50 %. At the same time, the steps of deproteinization and concentration were time-consuming and complex. Therefore, it was unfit for the study. For liquid-liquid extraction method, although volatile diethyl ether had relatively extracting effect for seven furfurals, the recoveries were still not satisfactory. In contrast to the above-mentioned extraction method, the QuEChERS method was simple and fast while guaranteeing the high recoveries of furfurals and repeatability. Therefore, QuEChERS was chosen as the sample treatment method in this study.

Optimization of GC-MS/MS conditions    In this study, gas chromatography with triple quadruple pole mass spectrometry was used for detecting F, FMC, MF, MMF, FA, 2(5H)-F and 5-HMF simultaneously. In order to ensure the accuracy of qualitative and quantitative analysis and improve the detection sensitivity, a series of parameters such as ion pair, collision energy and number of monitoring reaction ion pair must be optimized. To acquire the retention time of furfurals and the optimal precursor ion, individual injections of each standard solution in the mode of full scan were performed. In order to find the product ions with optimal selectivity and sensitivity, three kinds of precursor ions were conducted product ion scan for each compound. The collision energy of each ion pair was optimized in SRM mode. Finally, the ion pair with highest response value was applied for quantification and other two ion pair were qualitative.

As a result, the AUTO-SRM mode was adapted in the optimum condition of mass spectrum because this mode could be completed automatically and greatly save the time of parameter optimization. The final parameters for seven furfural compounds are shown in the Table 1.

Table 1. Mass spectrometric conditions for seven furfural compounds.

Name	Retention time (min)	Mass (m/z)	Product Mass (m/z)	Collision Energy (eV)
F	5.35	95	38.8	16
		96.1	37.8	30
		96.1	95	10
FMC	5.72	95	37.4	32
		95	38.8	18
		110	95	10
MF	6.39	109	53	10
		110	53	18
		110	109	10
MMF	6.84	109	53	10
		124.1	53	20
		124.1	109	10
FA	7.31	81	53	10
		98.1	41.9	10
		98.1	41.9	15
2(5H)-F	8.24	84	55	15
		84	55	10
		84	55	5
5-HMF	15.31	97	69	6
		126	69	10
		126	97	6

Matrix interference    Milk powder that in rich sugars and lipids may produce the matrix effects which have negative influence in accuracy of the method. Therefore, we need to calculate the matrix effect (ME) according to the formula (1):   

where S_m and S_s are the slopes of calibration plot with matrix-matched and solvent standard solution, respectively. A value between −20 % and 20 % indicates that the matrix enhancement or suppression effect is weak and relatively cceptable, meanwhile, a value higher than 20 % or below - 20 % indicates there is a strong matrix effect (Tian et al., 2016). The results of matrix effect in milk powder were listed in Table 2. Matrix effects of seven furfurals were calculated in experiment and the values were in the range of −20 % to 0 %, indicating the matrix had no significant effect on the accuracy.

Table 2. Matrix effect of seven furfurals in milk powder.

Analyte	Sm / Ss	ME (%)
F	0.92	−8.41
FMC	0.89	−10.76
MF	0.89	−11.04
MMF	0.88	−12.17
FA	0.88	−12.14
2(5H)-F	0.85	−14.90
5-HMF	0.82	−17.99

LODs, LOQs and Linearity    LODs and LOQs were calculated in sextuplicate using blank samples spiked from 0.1 to 5 µg/kg. The LODs of furfurals were between 0.0003 and 0.001 mg/kg and LOQs were 0.001 and 0.003 mg/kg (Table 3). These limits were close to or less than 20 % of those obtained by using the difference spectrophotometry with good detection capability (Habibi et al., 2017; Truzzi et al., 2012).

Table 3. Linear range, LODs and correlation coefficient of seven kinds of furfural compounds.

Furfurals	Regression Equation (y = Ax + B)	R2	Linear range	LOD	LOQ
			(mg/L)	(mg/kg)	(mg/kg)
F	y = 83705.56 x + 2348230.38	0.9992	0.002-2	0.0006	0.002
FMC	y = 94396.39 x + 2229051.02	0.9991	0.002-2	0.0006	0.002
MF	y = 86662.28 x + 2193179.21	0.9990	0.001-2	0.0003	0.001
MMF	y = 84407.55 x + 1872188.81	0.9994	0.001-2	0.0003	0.001
FA	y = 5210.25 x + 60358.16	0.9991	0.003-2	0.001	0.003
2(5H)-F	y = 8865.67 x + 276898.56	0.9991	0.003-2	0.001	0.003
5-HMF	y = 11050.13 x + 347459.87	0.9998	0.003-2	0.001	0.003

The analytical curves were constructed by the detection results of a series of mixed standard solutions of F, FMC, MF, MMF, FA, 5-HMF and 2(5H)-F and the GC-MS/MS chromatogram of 1 mg/L mixed standard solution is shown in Fig. 2. The linear ranges were between 0.001–2 mg/L and correlation coefficients (R2) were above 0.9990 for seven furfurals.

Fig. 2.

GC-MS/MS chromatogram of 1 mg/L standard furfural compounds.

Recovery    Mixed standard solutions with low, medium, and high concentrations of standard solution were added in milk powder samples to assess recovery efficiency. Each sample was measured six times in parallel a day for three consecutive days. As shown in Table 4, the average recoveries varied between 92.06 % and 106.19 %, and the reproducibility of the proposed method, expressed by intra-day and inter-day relative standard deviation (RSD), were in the range of 0.01 % to 3.32 % and 0.09 % to 1.58 %, respectively. Therefore, the good repeatability and reproducibility of the method was achieved and met the analysis requirements.

Table 4. Average recoveries (%) and repeatability (RSD, %) obtained with the modified QuEChERS method in milk powder (n=6).

Analyte	Spiking levels (mg/kg)	Recovery percentage (%)	RSD (%)
			Intra-day	Inter-day
F	0.020	105.56	0.76	0.50
	0.006	99.01	0.93	0.83
	0.002	100.55	0.02	0.56
FMC	0.020	101.20	1.64	0.09
	0.006	100.27	0.91	0.33
	0.002	100.35	3.32	0.96
MF	0.010	106.19	0.18	0.16
	0.003	104.76	0.74	0.94
	0.001	104.13	2.40	1.23
MMF	0.010	105.26	0.12	0.83
	0.003	105.98	1.07	0.24
	0.001	100.38	0.61	0.23
FA	0.030	97.06	0.27	0.41
	0.009	99.88	0.56	0.48
	0.003	93.83	1.65	0.91
2(5H)-F	0.030	97.10	0.28	0.10
	0.009	92.06	0.12	1.11
	0.003	97.10	0.14	0.97
5-HMF	0.030	103.31	0.40	1.58
	0.009	103.81	0.18	1.34
	0.003	101.43	0.01	0.82

Analysis of commercially available milk powder samples    The optimized QuEChERS method was applied to analyze 34 commercially available milk powder samples. The concentration data of detected compounds were listed in Table 5. The results showed that the content of 5-HMF and 2(5H)-F were higher than other compounds, while MMF was hardly detected in samples.

Table 5. Actual sample analysis results (n=6).

Number	Product Type	Contents (RSD (%)) (mg/kg)
		F	2(5H)-F	MF	5-HMF	MMF	FA	FMC
1	Infant formula (stage 3)	0.058	1.418 (2.72)	0.012	0.493	ND	0.105	0.007
		(6.64)		(4.10)	(4.35)		(0.97)	(8.71)
2	Infant formula (stage 3)	0.024	0.210 (1.45)	0.004	0.052	ND	0.010	0.004
		(0.11)		(5.75)	(2.17)		(1.53)	(5.17)
3	Infant formula (stage 3)	0.029	0.212 (1.08)	0.004	0.078	ND	0.016	0.004
		(0.44)		(2.13)	(3.14)		(7.76)	(0.88)
4	Infant formula (stage 3)	0.024	0.511 (0.21)	0.004	0.029	ND	0.022	0.004
		(2.16)		(2.70)	(5,59)		(0.16)	(6.34)
5	Infant formula (stage 3)	0.037	0.238 (0.94)	0.007	0.219	ND	0.023	0.005
		(3.84)		(1.13)	(0.01)		(3.54)	(0.31)
6	Infant formula (stage 3)	0.024	0.228 (0.22)	0.004	0.113	ND	0.012	0.004
		(2.75)		(3.69)	(0.64)		(1.32)	(8.99)
7	Infant formula (stage 3)	0.024	0.160 (0.69)	0.004	0.394	ND	0.029	0.004
		(1.85)		(4.61)	(1.76)		(3.27)	(2.10)
8	Infant formula (stage 3)	0.027	0.142 (0.31)	0.004	0.507	ND	0.028	0.004
		(0.14)		(3.28)	(0.01)		(2.44)	(9.50)
9	Infant formula (stage 3)	0.023	0.231 (0.59)	0.005	0.665	ND	0.014	0.004
		(4.09)		(0.65)	(0.51)		(6.37)	(1.44)
10	Infant formula (stage 3)	0.025	0.358 (0.02)	0.005	0.210	ND	0.034	0.004
		(0.66)		(0.79)	(1.41)		(2.41)	(0.65)
11	Infant formula (stage 3)	0.024	0.163 (0.05)	0.004	0.073	ND	0.012	0.004
		(0.78)		(0.22)	(1.71)		(8.04)	(4.13)
12	Infant formula (stage 3)	0.029	0.139 (0.92)	0.004	0.082	ND	0.010	0.004
		(1.10)		(2.05)	(2.36)		(1.10)	(0.07)
13	Infant formula (stage 3)	0.113	0.270 (0.49)	0.013	0.454	ND	0.026	0.010
		(3.89)		(2.15)	(3.63)		(7.73)	(6.77)
14	Infant formula (stage 3)	0.107	0.372 (3.26)	0.013	0.737	ND	0.075	0.008
		(3.36)		(7.05)	(1.43)		(5.59)	(1.41)
15	Infant formula (stage 3)	0.079	0.301 (2.65)	0.011	0.214	ND	0.049	0.008
		(0.78)		(2.03)	(0.51)		(4.15)	(0.19)
16	Infant formula (stage 3)	0.057	0.277 (1.45)	0.008	0.197	ND	0.018	0.006
		(1.59)		(2.14)	(0.50)		(6.92)	(1.86)
17	Infant formula (stage 3)	0.050	0.263 (0.19)	0.008	0.314	ND	0.081	0.007
		(2.06)		(3.27)	(0.50)		(0.41)	(1.74)
18	Infant formula (stage 3)	0.131	0.321 (2.13)	0.032	1.451	0.002	0.064	0.015
		(1.69)		(1.25)	(3.12)	(1.58)	(5.63)	(0.47)
19	Whole milk powder	0.031	0.115 (2.05)	0.005	0.082	ND	0.009	0.005
		(3.13)		(6.29)	(1.74)		(0.05)	(8.69)
20	Whole milk powder	0.035	0.147 (0.87)	0.005	0.113	ND	0.008	0.005
		(7.67)		(8.96)	(0.76)		(0.65)	(9.47)
21	Whole milk powder	0.034	0.195 (2.20)	0.005	0.409	ND	0.016	0.005
		(2.53)		(4.65)	(1.13)		(4.40)	(8.84)
22	Whole milk powder	0.086	0.215 (0.07)	0.014	0.172	ND	0.014	0.008
		(4.66)		(3.82)	(0.09)		(7.77)	(4.59)
23	Whole milk powder	0.043	0.163 (0.42)	0.007	0.082	ND	0.009	0.005
		(3.18)		(2.14)	(2.80)		(1.30)	(0.15)
24	Whole milk powder	0.038	0.137 (3.31)	0.006	0.101	ND	0.008	0.006
		(4.84)		(7.87)	(3.74)		(4.61)	(7.25)
25	Whole milk powder	0.046	0.182 (2.58)	0.011	0.079	ND	0.011	0.007
		(3.42)		(0.73)	(1.45)		(0.56)	(5.74)
26	Skim milk powder	0.035	0.141 (0.50)	0.006	0.079	ND	0.012	0.006
		(4.55)		(4.09)	(4.08)		(8.44)	(7.75)
27	Maternal formula samples	0.043	0.170 (0.30)	0.006	0.196	ND	0.026	0.006
		(2.89)		(3.13)	(2.49)		(4.88)	(2.75)
28	Maternal formula samples	0.039	0.327 (0.58)	0.006	0.269	ND	0.032	0.005
		(3.69)		(7.91)	(1.04)		(1.65)	(8.89)
29	Maternal formula samples	0.072	0.260 (0.46)	0.007	0.384	ND	0.033	0.010
		(0.85)		(1.50)	(1.17)		(4.20)	(1.41)
30	Special purpose infant formula	0.064	0.255 (0.67)	0.007	1.090	ND	0.058	0.007
	samples	(2.75)		(3.42)	(2.36)		(3.32)	(7.74)
31	Special purpose infant formula	0.051	0.210 (1.92)	0.009	0.232	ND	0.032	0.006
	samples	(2.51)		(8.04)	(3.10)		(7.27)	(5.30)
32	Middle-aged and aged	0.054	0.263 (0.64)	0.007	0.506	ND	0.031	0.006
	modulated milk powder	(0.07)		(9.43)	(2.05)		(8.25)	(0.46)
33	High-fiber milk powder	0.069	0.308 (0.63)	0.010	0.680	ND	0.029	0.007
		(3.40)		(2.99)	(0.07)		(3.23)	(3.29)
34	Family nutrition milk powder	0.038	0.137 (0.14)	0.005	0.093	ND	0.022	0.005
		(6.97)		(3.65)	(1.17)		(2.84)	(0.24)

ND: not detected.

In addition, we analyzed furfural levels in milk powders with various sugar additives. GOS is also a common additive in milk powder. However, the presence of GOS markedly increased the content of 5-HMF (i.e. number 18 and 30 in Table 5), especially in special purpose infant formula with only GOS added, probably because GOS was much easier to be degraded to C₆ α-dicarbonyls species, namely 3-deoxyglucosone and 3-deoxygalactosone, thereby resulting in higher concentrations of 5-HMF during MR (Zhang et al., 2018). Oligo-fructose (FOS) was added as a bifidobacterium actinogen in infant formula to solve the problem of getting inflamed during infant nutrition supplement. Our data showed that the ingredients of F and 5-HMF in infant milk powder containing FOS were lower than those without GOS (i.e. compare number 4 and number 1 in Table 5). This result may show FOS, compared with GOS, is harder to react with proteins and more stable at ambient temperature.

Sucrose and fructose are often added in adult and senior milk powders respectively to increase sweetness. The 5-HMF content in middle-aged milk powder was up to 0.506 mg/kg, significantly higher than that in the whole milk powder, possibly due to the formation of fructofuranosyl cation from fructose which was confirmed to produce extra 5-HMF during Maillard Reaction (Perez et al., 2008; Zhao et al., 2016).

In summary, the type of sugar was a crucial factor in the Maillard Reaction. Among all samples, the F levels ranged between 0.023 and 0.131 mg/kg, and the 2(5H)-F levels ranged between 0.115 and 1.418 mg/kg, and the MF levels ranged between 0.004 and 0.032 mg/kg, and the 5-HMF levels ranged between 0.029 and 1.451 mg/kg, and the contents of FA was between 0.008 and 0.105 mg/kg, while MMF in most samples is below the detection limit. According to these results, in comparison with milk powders with other sugars, milk powders with GOS have relatively high content of 5-HMF. Moreover, over 0.100 mg/kg content of 2(5H)-F levels were in all samples, which is relatively high and may be positively connected to the proportion of carbohydrate in milk powders.

Conclusions

In this study, a rapid and effective method for simultaneous detection of furfural, 2-acetylfuran, 5-methyl-2-furfural, 1-(5-methylfuran-2-yl)ethanone, 2(5H)-furanone, 5-hydroxymethyl-2-furfural and furfuryl alcohol in milk powder by QuEChERS and GC-MS/MS was developed. The validation data showed that this method has good repeatability, high sensitivity, and good selectivity. Thirty-four actual samples added with different sugars including galacto-oligosaccharide, oligo-fructose, sucrose and fructose, were analyzed. The results revealed that the contents of 5-hydroxymethyl-2-furfural were obviously directly connected with the type of sugar and the content of 2(5H)-furanone was influenced by the addition of sugar, in the milk powder. Features such as simple, rapid speed and low costs make the method useful and the consequence provide a guide for the process of adding sugars.

Acknowledgements    This study funded by the National Key R&D Program of China (No. 2017YFC1600404) is gratefully acknowledged.

Conflicts of interest    There are no conflicts to declare.

References

•  Abraham,  K.,  Gürtler,  R.,  Berg,  K.,  Heinemeyer,  G.,  Lampen,  A., and  Appel,  K.E. (2011). Toxicology and risk assessment of 5 - Hydroxymethylfurfural in food. Mol. Nutr. Food Res., 55, 667-678.
•  Aktağa,  I.G.,  Hamzalıoğlua,  A., and  Gökmen,  V. (2019). Lactose hydrolysis and protein fortification pose an increased risk for the formation of Maillard reaction products in UHT treated milk products. J. Food Compos. Anal., 84, 103308.
•  Albalá-Hurtado,  S.,  Veciana-Nogués,  M.T.,  Mariné-Font,  A., and  Vidal-Carou,  M.C. (1998). Changes in furfural compounds during storage of infant milks. J. Agric. Food Chem., 46, 2998-3003.
•  Capuano,  E., and  Fogliano,  V. (2011). Acrylamide and 5-hydroxymethylfurfural (HMF): A review on metabolism, toxicity, occurrence in food and mitigation strategies. LWT-food Sci. Technol., 44, 793-810.
•  da Silva,  J.L.,  Beluomini,  M.A., and  Stradiotto,  N.R. (2015). Determination of furanic aldehydes in sugarcane bagasse by high - performance liquid chromatography with pulsed amperometric detection using a modified electrode with nickel nanoparticles. J. Sep Sci., 38, 3176-3182.
•  de Andrade,  J.K.,  de Andrade,  C.K.,  Komatsu,  E.,  Perreault,  H.,  Torres,  Y.R.,  da Rosa,  M.R., and  Felsner,  M.L. (2017). A validated fast difference spectrophotometric method for 5-hydroxymethyl-2-furfural (HMF) determination in corn syrups. Food Chem., 228, 197-203.
•  Demirhan,  B.E.,  Demirhan,  B.,  Sönmez,  C.,  Torul,  H.,  Tamer,  U., and  Yentür,  G. (2015). Determination of potential 5-hydroxymethyl-2-furaldehyde and 2-furaldehyde compounds in follow-on milks and infant formulas using the high-performance liquid chromatography method. J. Dairy Sci., 98, 818-822.
•  Espinosa-Mansilla,  A.,  Peña,  A.M.D., and  Salinas,  F. (1993). Semiautomatic determination of furanic aldehydes in food and pharmaceutical samples by a stopped-flow injection analysis method. J. AOAC Int., 76, 1255-1261.
•  Ferrer,  E.,  Alegría,  A.,  Farré,  R.,  Abella´n,  P., and  Romero,  F. (2005). High-performance liquid chromatographic determination of furfural compounds in infant formulas during full shelf-life. Food Chem., 89, 639-645.
•  Gaspar,  E.M.S.M., and  Lopes,  J.F. (2009). Simple gas chromatographic method for furfural analysis. J. Chromatogr. A, 1216, 2762-2767.
•  Gheisari,  H.R.,  Mazkour,  S.,  Shekarforoush,  S.S., and  Keshavarzi,  Z. (2018). Finding a rapid, simple and precise method for determination of skim milk powder adulteration in non reconstituted milk. Int. Food Res. J., 25, 2261-2267.
•  Gökmen,  V., and  Şenyuva,  H.Z. (2006). Improved method for the determination of hydroxy-methylfurfural in baby foods using liquid chromatography-mass spectrometry. J. Agric. Food Chem., 54, 2845-2849.
•  Habibi,  H.,  Mohammadi,  A., and  Kamankesh,  M. (2017). Application and optimization of dispersive liquid-liquid microextraction coupled with high-performance liquid chromatography for sensitive determination of furfural and hydroxymethyl furfural in jarred and canned baby-foods. Nutrition. Food Sci. Res., 4, 25-32.
•  Li,  Y.N.,  Liu,  W.Y.,  Zhang,  S.L.,  Li,  J.P.,  Chen,  W.H.,  Qu,  C.,  Ai,  T., and  Cheng,  X.Y. (2016). Effect of reaction time on the flavor quality of maillard reaction products from beef meat and bone hydrolysates. Meat Research, 30, 1-5.
•  Locas,  C.P., and  Yaylayan,  V.A. (2008). Isotope labeling studies on the formation of 5-(hydroxymethyl)-2-furaldehyde (HMF) from sucrose by pyrolysis-GC/MS. J. Agric. Food Chem., 56, 6717-6723.
•  Madani-Tonekaboni,  M.,  Kamankesh,  M,, and  Mohammadi,  A. (2015). Determination of furfural and hydroxymethyl furfural from baby formula using dispersive liquid–liquid microextraction coupled with high performance liquid chromatography and method optimization by response surface methodology. J. Food Compos. Anal., 40, 1-7.
•  Melo,  E.M.G.,  Fernandes,  J.P.A.,  Cunha,  F.A.S.,  Lira,  J.O.B.,  Lima,  R.A.C.,  Lima,  K.M.G., and  Almeida,  L.F. (2016). A new method for the determination of total furanic aldehyde compounds in Brazilian cachaça samples using liquid–liquid extraction and UV detection. Anal. Methods, 8, 7047-7053.
•  Murkovic,  M., and  Pichler,  N. (2006). Analysis of 5 - hydroxymethylfurfual in coffee, dried fruits and urine. Mol. Nutr. Food Res., 50, 842-846.
•  Nursten,  H.E. (2005). The Maillard reaction: chemistry, biochemistry, and implications. Royal Society of Chemistry, London.
•  Rodríguez-Carrasco,  Y.,  Moltó,  J.C.,  Mañes,  J., and  Berrada,  H. (2014). Development of a GC–MS/MS strategy to determine 15 mycotoxins and metabolites in human urine. Talanta, 128, 125-131.
•  Senila,  L.,  Gog,  A.,  Senila,  M.,  Roman,  C., and  Silaghidumitrescu,  L. (2012). Development of a GC-MS method for 5-hydroxymethylfurfural determination in wood after steam-explosion pretreatment. Rev. Chim., 63, 559-563.
•  Sun,  Y.,  Guan,  Z.,  Cai,  H.,  Huang,  Y.Y.,  Lin,  Y.W., and  Hu,  X.S. (2017). Highly sensitive method for aldehydes detection: Application to furfurals analysis in raisin and bovine milk powder. Anal. Chimica Acta, 987, 47-55.
•  Teixidó,  E.,  Santos,  F.J.,  Puignou,  L., and  Galceran,  M.T. (2006). Analysis of 5-hydroxymethylfurfural in foods by gas chromatography–mass spectrometry. J. Chromatogr. A, 1135, 85-90. 11.
•  Tian,  H.,  Wang,  J.Q.,  Zhang,  Y.D.,  Li,  S.L.,  Jiang,  J.D.,  Tao,  D.L., and  Zheng,  N. (2016). Quantitative multiresidue analysis of antibiotics in milk and milk powder by ultra-performance liquid chromatography coupled to tandem quadrupole mass spectrometry. J. Chromatogr. B, 1033, 172-179.
•  Totlani,  V.M.,  Peterson,  D.G. (2007). Influence of epicatechin reactions on the mechanisms of Maillard product formation in low moisture model systems. J. Agric. Food Chem., 55, 414-420.
•  Truzzi,  C.,  Annibaldi,  A.,  Illuminati,  S.,  Finale,  C.,  Rossetti,  M., and  Scarponi,  G. (2012). Determination of Very Low Levels of 5-(Hydroxymethyl)-2-furaldehyde (HMF) in Natural Honey: Comparison between the HPLC Technique and the Spectrophotometric White Method. J. Food Sci., 77, 784-790.
•  Tsai,  S.W., and  Kao,  K.Y. (2012). Determination of furfural in beers, vinegars and infant formulas by solid-phase microextraction and gas chromatography/mass spectrometry. Intern. J. Environ. Anal. Chem., 92, 76-84.
•  van Boekel,  M.A.J.S. (2006). Formation of flavour compounds in the Maillard reaction. Biotechnol. Adv., 24, 230-233.
•  Wagner,  B., and  Beil-Seidler,  S. (2010). Agent and method for identifying furfurals. U.S. Patent, 7718441.
•  Zhang,  W.,  Ray,  C.,  Poojary,  M.M.,  Jansson,  T.,  Olsen,  K., and  Lund,  M.N. (2018). Inhibition of Maillard reactions by replacing galactose with galacto-oligosaccharides in casein model systems. J. Agric. Food Chem., 67, 875-886.
•  Zhao,  Z.,  Umemura,  K., and  Kanayama,  K. (2016). Effects of the addition of citric acid on tannin-sucrose adhesive and physical properties of the particleboard. BioResour., 11, 1319-1333.