Analysis of 27 β-Blockers and Metabolites in Milk Powder by High Performance Liquid Chromatography Coupled to Quadrupole Orbitrap High-Resolution Mass Spectrometry

This paper presents an application of high performance liquid chromatography coupled with quadrupole orbitrap high-resolution mass spectrometry (HPLC-Q-Orbitrap HRMS) for the analysis of 27 β-blockers and metabolites in milk powder. Homogenized milk power samples were extracted by acetonitrile and purified by using Oasis PRiME HLB solid-phase extraction cartridges. The Ascentis® C8 chromatographic column was used to separate the analytes. The quantification was achieved by using matrix-matched standard calibration curves with carazolol-d7 and propranolol-d7 as the internal standards. The results show an exceptional linear relationship with the concentrations of analytes over wide concentration ranges (0.5–500 μg kg−1) as all the fitting coefficients of determination r2 are > 0.995. All the limits of detection (LODs) and quantitation (LOQs) values were within the respective range of 0.2–1.5 μg kg−1 and 0.5–5.0 μg kg−1. Overall average recoveries were able to reach 66.1–100.4% with the intra- and inter-day variability under 10%. This method has been successfully applied to the screening of β-blockers and metabolites in commercial milk powders. At the same time, the corresponding characteristic fragmentation behavior of the 27 compounds was explored. The characteristic product ions were determined and applied to the actual samples screening.


Introduction
β-blockers (BBS) are structurally analogous to the catecholamines, which can act as non-specific β-adrenergic receptor blocking agents. They play an extremely important role in the treatment of cardiovascular diseases such as coronary heart disease, hypertension, arrhythmia and cardiac insufficiency. However, improper use of β-blockers can cause an increase of myocardial oxygen consumption, vascular resistance, oxygen free radicals and myocardial cell apoptosis, etc. [1]. β-blockers are usually used in animals to reduce morbidity and mortality during transportation (to the slaughterhouse or livestock farm), mating, childbirth and in other stressful situations. Such stress usually results in a poor quality of meat, or even in the premature death of the animal [2][3][4]. The illegal use of β-blockers gives rise to drug residues in edible animal tissue, which can be metabolized in the observed by the first three columns; however, the Ascentis ® C8 column provided better shape and retention for hydroxyatenolol, as shown in Figure 1. Therefore, the Ascentis ® C8 column was chosen. Several mobile phases were tested using MeCN or MeOH as an organic solvent and water as a polar solvent with FA addition (from 0% to 0.5%). As shown in Figure 2A, the mixture of MeOH and water showed higher responses and better separation for isomers of practolol and atenolol. The proton donor tendency of MeOH contributed to the formation of positive adducts. The better volatility and lower surface tension of MeOH can also improve desolvation of the droplets. Therefore, MeOH was chosen as the organic solvent. On the other hand, the addition of FA could improve the phenomenon of peak tailing and the response, which is probably because the excess silanols in the stationary phase combine with the acid rather than with their targets. The retention time of isomers of atenolol and practolol were greatly affected by pH, so a good chromatographic separation could not obtained until 0.1% FA was added (mobile phase pH = 2.58) ( Figure 2B). Therefore, 0.1% FA-H2O was chosen as the water phase. Several mobile phases were tested using MeCN or MeOH as an organic solvent and water as a polar solvent with FA addition (from 0% to 0.5%). As shown in Figure 2A, the mixture of MeOH and water showed higher responses and better separation for isomers of practolol and atenolol. The proton donor tendency of MeOH contributed to the formation of positive adducts. The better volatility and lower surface tension of MeOH can also improve desolvation of the droplets. Therefore, MeOH was chosen as the organic solvent. On the other hand, the addition of FA could improve the phenomenon of peak tailing and the response, which is probably because the excess silanols in the stationary phase combine with the acid rather than with their targets. The retention time of isomers of atenolol and practolol were greatly affected by pH, so a good chromatographic separation could not obtained until 0.1% FA was added (mobile phase pH = 2.58) ( Figure 2B). Therefore, 0.1% FA-H 2 O was chosen as the water phase.
In addition, injection volumes of 2 µL to 5 µL were evaluated using the aforementioned conditions. Taking the isomers of atenolol and practolol as examples, the experimental results showed that the separation factor decreased from 1.41 to 0.77 with increasing injection volume. So, 2 µL was chosen as the experimental injection volume.
Furthermore, several elution gradient profiles were also optimized to obtain better chromatographic separation and less analysis time (within 15 min). Other parameters (flow rate, column temperature) were also characterized to achieve better target separation and peak shapes. Under these conditions (see Section 3.2), the retention times (RT) of these 27 analytes were constant, and ranged ranging from 3.85 min to 9.01 min. In addition, injection volumes of 2 μL to 5 μL were evaluated using the aforementioned conditions. Taking the isomers of atenolol and practolol as examples, the experimental results showed that the separation factor decreased from 1.41 to 0.77 with increasing injection volume. So, 2 μL was chosen as the experimental injection volume.
Furthermore, several elution gradient profiles were also optimized to obtain better chromatographic separation and less analysis time (within 15 min). Other parameters (flow rate,

Optimization of the Mass Spectrometric Parameters
The optimization of MS parameters was performed by infusing a standard solution of 100 µg L −1 of each β-blockers in methanol-water (50:50, v/v) as the mobile phase under Full scan mode (Full MS). The precursor ions were selected in both positive and negative modes. Consistent with previous report [2], β-blockers and metabolites tend to form [M + H] + adduct ions in positive mode. The Full MS/ddMS 2 scan mode, which can achieve the non-target list qualitative and quantitative detection in a single run, was used for screening all samples. All the MS parameters were optimized to provide the best responses of the analytes. The optimized parameters values are summarized in Section 3.2.

The Proposed Fragmentation Pathways for 27 β-Blockers
In this study, secondary mass spectrometry data of all the substances were extracted, and the fragment ions occurring many times were selected to analyze the fragmentation pathway. β-blockers are mainly classified into three kinds of structures. Labetalol and sotalol have the structure of phenylethanolamine, and others have the structure of aryloxypropanolamine except for timolol and hydroxytimolol, which have special chemical structures. The principal structures of phenylethanolamine and aryloxypropanolamine compounds are shown in Figure 3. and ranged ranging from 3.85 min to 9.01 min.

Optimization of the Mass Spectrometric Parameters
The optimization of MS parameters was performed by infusing a standard solution of 100 μg L −1 of each β-blockers in methanol-water (50:50, v/v) as the mobile phase under Full scan mode (Full MS). The precursor ions were selected in both positive and negative modes. Consistent with previous report [2], β-blockers and metabolites tend to form [M + H] + adduct ions in positive mode. The Full MS/ddMS 2 scan mode, which can achieve the non-target list qualitative and quantitative detection in a single run, was used for screening all samples. All the MS parameters were optimized to provide the best responses of the analytes. The optimized parameters values are summarized in Section 3.2.

The Proposed Fragmentation Pathways for 27 β-Blockers
In this study, secondary mass spectrometry data of all the substances were extracted, and the fragment ions occurring many times were selected to analyze the fragmentation pathway. β-blockers are mainly classified into three kinds of structures. Labetalol and sotalol have the structure of phenylethanolamine, and others have

Phenylethanolamines Structure
For phenylethanolamine structure, there is a characteristic loss of one molecule of water at position 2 at first, and then a characteristic cleavage at positions 1 and 3, as shown in Figure 3A. This phenomenon was consistent with the fragmentation of β-agonists compounds with the phenylethanolamine structure that were studied in our laboratory [23].

Aryloxypropanolamines Structure
For the aryloxypropanolamine structure ( Figure 3B), the proposed fragmentation pathways can be divided into three types, depending on the different substitution groups.
Type I (R6 was H, R7 and R8 were methyl) For type I, the bonds between carbon and oxygen were preferentially broken at position 2. At first, the phenyl structure (C6H4-R5) was lost to form the fragment ion [C6H14NO2 + H] + at m/z 133.06412. Then, the potential loss of [C3NH9] + or [OH] + gave m/z 74.06063 (formula C3H6O2) or 116.10702 (formula C6H14NO) moieties. And the loss of one molecule of water and a series of fragmentations occurred. The suggested fragmentation pathway of type I is shown in Figure 4.

Phenylethanolamines Structure
For phenylethanolamine structure, there is a characteristic loss of one molecule of water at position 2 at first, and then a characteristic cleavage at positions 1 and 3, as shown in Figure 3A. This phenomenon was consistent with the fragmentation of β-agonists compounds with the phenylethanolamine structure that were studied in our laboratory [23].

Aryloxypropanolamines Structure
For the aryloxypropanolamine structure ( Figure 3B), the proposed fragmentation pathways can be divided into three types, depending on the different substitution groups.
Type I (R 6 was H, R 7 and R 8 were methyl) For type I, the bonds between carbon and oxygen were preferentially broken at position 2. At first, the phenyl structure (C 6 H 4 -R 5 ) was lost to form the fragment ion [C 6 H 14 NO 2 + H] + at m/z 133.06412. Then, the potential loss of [C 3 NH 9 ] + or [OH] + gave m/z 74.06063 (formula C 3 H 6 O 2 ) or 116.10702 (formula C 6 H 14 NO) moieties. And the loss of one molecule of water and a series of fragmentations occurred. The suggested fragmentation pathway of type I is shown in Figure 4.
It is worth mentioning that the Υ-H of the amine structure with 1-propene (C 6 H 12 N at m/z 98.09663) rearranged to an unsaturated group, accompanied with the cleavage of the β-bond in the amine structure to produce the McLafferty Rearrangement. Imine structure of the fragment C 3 H 7 N (at m/z 57.07101) was produced by this reaction. The mechanism of the McLafferty Rearrangement reaction is shown in Figure 5.
Type II (R 6 , R 7 and R 8 were methyl group) For type II, the R 6 position of the aryloxypropanolamine structure was substituted with a methyl group. The methyl group can easily get lost to produce a type I structure, and then have a similar fragmentation pathway to the type I mentioned above. Type III (R 6 or R 7 was an isophthalic ether structure, others were H) For type III, the R 6 or R 7 position of the aryloxypropanolamine structure was substituted with an isophthalic ether structure, which has strong electronegativity. The bond between carbon and nitrogen at position 8 was easily broken, and it was difficult to form the fragment at m/z 116.10702 (formula C 6 H 14 NO).  It is worth mentioning that the Υ-H of the amine structure with 1-propene (C6H12N at m/z 98.09663) rearranged to an unsaturated group, accompanied with the cleavage of the β-bond in the amine structure to produce the McLafferty Rearrangement. Imine structure of the fragment C3H7N (at m/z 57.07101) was produced by this reaction. The mechanism of the McLafferty Rearrangement reaction is shown in Figure 5. Type II (R6, R7 and R8 were methyl group) For type II, the R6 position of the aryloxypropanolamine structure was substituted with a methyl group. The methyl group can easily get lost to produce a type I structure, and then have a similar fragmentation pathway to the type I mentioned above.
m/z 57.07101  It is worth mentioning that the Υ-H of the amine structure with 1-propene (C6H12N at m/z 98.09663) rearranged to an unsaturated group, accompanied with the cleavage of the β-bond in the amine structure to produce the McLafferty Rearrangement. Imine structure of the fragment C3H7N (at m/z 57.07101) was produced by this reaction. The mechanism of the McLafferty Rearrangement reaction is shown in Figure 5. Type II (R6, R7 and R8 were methyl group) For type II, the R6 position of the aryloxypropanolamine structure was substituted with a methyl group. The methyl group can easily get lost to produce a type I structure, and then have a similar fragmentation pathway to the type I mentioned above.
Type III (R6 or R7 was an isophthalic ether structure, others were H) For type III, the R6 or R7 position of the aryloxypropanolamine structure was substituted with an isophthalic ether structure, which has strong electronegativity.

Special Structures
For timolol and hydroxytimolol (structures see Figure 6.), they can be considered as β-blockers for newly-emerging aryloxypropanolamine compounds, which differ in structure from the others. These two compounds have characteristic fragments at m/z 74.06063, m/z 57.07101 and m/z 56.05025.
The structure types of the 21 β-blockers and 6 metabolites are listed in Table 1. The possible structures of the corresponding characteristic fragments are described and summarized in Table 2. By exploring the exact mass of these identification markers, it is possible to find newly-emerging β-blockers residues in a complex food matrix. nitrogen at position 8 was easily broken, and it was difficult to form the fragment at m/z 116.10702 (formula C6H14NO).

Special Structures
For timolol and hydroxytimolol (structures see Figure 6.), they can be considered as β-blockers for newly-emerging aryloxypropanolamine compounds, which differ in structure from the others. The structure types of the 21 β-blockers and 6 metabolites are listed in Table 1. The possible structures of the corresponding characteristic fragments are described and summarized in Table 2. By exploring the exact mass of these identification markers, it is possible to find newly-emerging βblockers residues in a complex food matrix.

The Optimization of the Sample Preparation Procedure
A rapid enzymolysis method was chosen to ensure the processing flux and dissociate the possible bound residual drug. Neutral enzymatic environment (closing to pH 5.2) can significantly reduce matrix co-extraction. Na 2 EDTA was added to the buffer to reduce the chelation between metal ions and strongly polar targets. The alkalized aqueous phase and salting out after the enzymolysis facilitate the target extraction into the organic solvent. pH at 9.0, 9.5, 10.0, 10.5, 11.0 and 12.0 were evaluated, and the best extraction efficiency was obtained at pH 10.0, which was consistent with the literature [1].
For food samples, MeCN was commonly used as the extraction solvent due to its protein precipitation ability. Since the acetic-buffer could be used to increase the recoveries of pH-dependent compounds, pure MeCN and different contents of acids (0.1% HOAc, 1% HOAc, 0.1% FA, 1% FA, v/v) in MeCN were compared for extraction efficiency in this study. For extraction solvents containing HOAc or FA, the recoveries of some analytes (such as sotalol, hydroxymetoprolol, labetalol, epractolol and hydroxytimoloven) were lower than 60%, as can be seen in Figure 7A. Pure MeCN provided better extraction efficiency with all analyte recoveries being higher than 65%; therefore, pure MeCN was found to be the most suitable extract solvent. Then, the solvent volume was investigated for optimization of the recoveries of the targets. It can be observed in Figure 7B that the recoveries of analytes increased with the solvent volume. When the solvent volume reached 15 mL, recoveries began to be stable. In order to ensure the stability of the recoveries, 20 mL extract solvent was chosen to extract all analytes. Taking into account the characteristics of the β-blockers and matrix interferences in milk powder samples, low temperature high-speed centrifugation, low temperature high-speed centrifugation + Taking into account the characteristics of the β-blockers and matrix interferences in milk powder samples, low temperature high-speed centrifugation, low temperature high-speed centrifugation + SPE (PRiME HLB column), and low temperature high-speed centrifugation + QuEChERS (quick, easy, cheap, effective, rugged and safe) methods were tested as purification steps. For the low temperature, high-speed centrifugation + QuEChERS method, the water removal step (using anhydrous magnesium sulfate) might take some water-soluble targets (such as sotalol) away, resulting in a low recovery rate. Compared with low temperature, high-speed centrifugation, PRiME HLB column used in SPE step could adsorb nonpolar interferences (some fats and phospholipids) in milk powder samples, which results in a smaller matrix effect and better target recoveries, as shown in Table 3. Therefore, the solid phase extraction PRiME HLB cartridge was selected for purification step.

Linearity and Sensitivity
The linearity of the proposed method was evaluated using matrix-matched spiked samples over the range of 0.5-500 µg kg −1 . Calibration curves resulted from the ratios of the peak area of the target compounds to the peak area of the isotope-labeled internal standards. The results showed a good linearity relationship with correlation coefficients (r 2 ) higher than 0.995 ( Table 4). Limits of detection (LODs) and quantification (LOQs) are fundamental parameters used to evaluate the sensitivity of instructions and methods. The LODs were determined by the injection of a series of diluted standard solutions corresponding to a signal-to-noise (S/N) ratio of 3. The LOQs were determined by the injection of a series of spiked samples corresponding to a signal-to-noise (S/N) ratio of 10. Under the optimum condition, the LODs and LOQs were in the range of 0.2-1.5 µg kg −1 and 0.5-5.0 µg kg −1 , respectively, which allows the quantification of analytes presented at low content, indicating that good sensitivity was obtained.

Matrix effect
Suppression or enhancement of the target signal usually occurs in the HESI source, especially for complicated food matrices. With a matrix effect (ME) value was between 80% and 120%, signal suppression or enhancement effect can be considered tolerable. As shown in Table 3, many of the analytes did not significantly express the matrix effect, except metoprolol, acebutolol, celiprolol, bisoprolol, diacetolol and α-hydroxyatenolol, which showed a significant matrix effect (ME < 80%, or ME > 120%). In order to accurately quantify the compounds, the assay was quantified with matrix-matched internal standard calibration.

Trueness and Precision
Recovery experiments were performed to evaluate the trueness of the method due to the lack of certified reference materials (CRM). As shown in Table 3, recoveries at three spiking levels (LOQ, 2 × LOQ, 4 × LOQ) ranged from 66.1% to 100.4%. The precision was calculated in terms of intra-day repeatability and inter-day reproducibility, which were expressed as relative standard deviations (RSDs). The results of intra-day and inter-day analyses performed at three spiking levels are presented in Table 3. Repeatability and reproducibility were in the range of 1.6-8.7% and 1.8-8.9%, respectively. Consequently, these results indicated that the developed method in this study is quite reliable, accurate and reproducible for determining β-blockers and their metabolites in milk powder samples.

Real Samples Analysis
In order to estimate the reliability and practicability of the developed method, 30 samples of infant formula milk powder purchased at local markets were analyzed in this study. The samples were regarded as representative, since they were ranged from phase 1 to stage 4 produced by reputable manufactures. High accuracy parent ions and product ions were used for qualitative analysis simultaneously. Full MS data of this mode were used for quantitative analysis. None of the 27 targeted analytes were detected by the developed Q-Orbitrap high resolution mass spectrometry method. However, an unknown compound which has the same molecule mass (m/z 309.18005) but different retention time (7.15 min vs. 5.75 min) with diacetolol has been screened in one sample. As shown in Figure 8, under the MS 2 conditions, the detected unknown compound had high-accuracy product ions 98.09692 (m/z) and 72.08157 (m/z), which was similar to diacetolol, and also the characteristic fragment ion m/z 57.07080. So, it is reasonable to speculate that this unknown substance may be an isomer of diacetolol. Considering the possible structure of product ions, the detected diacetolol's isomer could have two possible chemical structures (C 16 H 25 O 4 N 2 ), as shown in Figure 9. For the first one, the acetyl group on the phenyl ring is in the presence of an acetyl amino group, which is more likely. The reason for this is that intermediate isomers are produced during the synthesis process, resulting in the isomerism of the target compounds. The second possibility may be diastereoisomers. Imino NH, phenolic hydroxyl oxygen and carbonyl oxygen through hydrogen bonding make the nitrogen atom form a relatively stable chiral center, as shown in Figure 9B. The specific structure needs to be further confirmed by NMR or other techniques.
All standard stock solutions were prepared in MeOH at 100 µg mL −1 . The mixed working standard solutions were prepared daily via proportional dilution of the stock solutions. All of the standard solutions were stored at −20 • C in a dark amber bottle.
Matrix-matched standard working solutions were prepared in blank sample extracts, which were obtained from a commercial product purchased from a local market and affirmed in advance not to contain any of the tested analytes. All of the standard solutions were stored at −20 • C in a dark amber bottle.
Extraction solution: 37.5 g Na 2 EDTA was dissolved in an ammonium acetate buffer produced by dissolving 15.4 g ammonium acetate in 1 L deionized water and then using acetic acid to adjust the pH to 5.2.

Instrument and Analytical Conditions
The UHPLC/HESI Q-Orbitrap system consisted of a Thermo UltiMate 3000 UHPLC + system coupled with a Q Exactive mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). The system was controlled by Exactive Tune 1.1 and Xcalibur 2.2 software (Thermo Fisher Scientific, San Jose, CA, USA).
The Q-Orbitrap HRMS was equipped with a heated electrospray ionization (HESI, Waltham, MA, USA) source and the analysis was operated in the Full MS/dd-MS 2 (data-dependent MS 2 ) scanning mode, which includes a Full scan followed by MS/MS scan of precursors in the inclusion list. All analytes were measured in positive mode and precursor ion selected was [M + H] + in all cases. To keep a balance between the selectivity and the sensitivity with Full MS, a mass resolution of 70,000 FWHM was selected, and this turned out to be optimal for the majority of the analytes. For the dd-MS 2 scan, 35,000 FWHM was used for time-saving and to ensure sufficient scan points of the Full MS. The stepped normalized collision energy (NCE) was set to 15%, 25% and 35%. The spray voltage, capillary temperature, aux gas heater temperature were set to 3.0 kV, 350 • C and 350 • C, respectively. The sheath gas, auxiliary gas, sweep gas and S-lens RF level were set to 40, 10, 0 (arbitrary units) and 50 V, respectively. Main MS acquisitions parameters are listed in Table 1. All the extracted mass traces were based on a 2 ppm mass window (accuracy).

Sample Preparation
Two grams of each sample were precisely weighed in polypropylene centrifuge tube (50 mL). Then, 100 µL mixed ISTD solution (1µg mL −1 ), 40 µL β-glucuronidase/arylsulfatase and 5 mL EDTA extract solution were added to the sample after being fully dissolved in 5 mL H 2 O. The mixture was placed in a water bath shaker at 50 • C for 60 min after vortex-mixing for 1 min. After cooling to room temperature, the pH of the extract was adjusted to 10.0 with 3 mol L −1 NaOH solution. MeCN (20 mL) and NaCl (2.5 g) were added to the mixture and then shaken for 30 min. After that, the extract was centrifuged at 10,000 rpm at 4 • C for 10 min. The supernatant was decanted to another polypropylene centrifuge tube. The above procedure was repeated and combined with the supernatant. Next, 2 mL of the supernatant directly passed through the Oasis PRiME HLB (500 mg, 6 cm 3 ) cartridge. After sample loading, the cartridge was washed with 2 mL H 2 O/MeOH (95:5, v/v) and 2 mL MeOH/MeCN (1:9, v/v), and all of the effluent was collected. The mixture was evaporated with a gentle N 2 stream at 40 • C, and redissolved in 1 mL of H 2 O/MeOH (1:1). The sample extract was vortexed for 0.5 min and filtered through a 0.22 µm nylon membrane, and was ready for Q-Orbitrap HRMS analysis.

Method Validation
Linearity, precision and recovery were carried out to validate the method. An internal standard method which using carazolol-d 7 and propranolol-d 7 as ISTD was utilized for quantification. A matrix-matched calibration cure was constructed by linear regression of the ratios of chromatographic peak areas of the standards and the ISTD. The linearity was discussed by the coefficient of determination (r 2 ).
Blank milk sample powders spiked at three concentration levels (LOQ, 2 × LOQ and 4 × LOQ) which were tested for the recovery experiments. Each level was analyzed in five replicates. Intra-day precision was performed by spiking blank milk at three concentration levels (LOQ, 2 × LOQ and 4 × LOQ) with five replicates in one day. To evaluate inter-day precision, the same concentration levels were performed during over consecutive days.
The matrix effect (ME) was calculated by comparing the response of analytes prepared in solvent and in extracted blank matrix at the same concentration, respectively. The value of matrix effect can be calculated as (Equation (1)): ME (%) = B/A × 100 (1) A refers to the peak areas obtained from neat solution standards, while B refers to the corresponding peak areas of standards spiked after extraction from matrix [24,25].

Conclusions
In this study, a rapid HPLC-Q-Orbitrap HRMS method for simultaneous analyses of 27 compounds (21 β-blockers and 6 metabolites) in milk powder has been developed. Simultaneous qualitative and quantitative analysis of analytes were achieved using Full MS/dd-MS 2 acquisition mode of the Q-Orbitrap mass analyzer and the preparation procedure comprised a simple acetonitrile step, followed by a cleanup using cartridges. The method has been well validated, and is particularly effective and valuable for the routine screening of β-blockers and metabolites in infant formula milk powder. At the same time, the corresponding characteristic fragmentation behavior of the 27 compounds were explored, the characteristic product ions were determined and applied to the actual sample screening.