Evaluation of the Antioxidant Activities and Phenolic Profile of Shennongjia Apis cerana Honey through a Comparison with Apis mellifera Honey in China

This study evaluates the phenolic profile as well as the antioxidant properties of Shennongjia Apis cerana honey through a comparison with Apis mellifera honey in China. The total phenolic content (TPC) ranges from 263 ± 2 to 681 ± 36 mg gallic acid/kg. The total flavonoids content (TFC) ranges from 35.9 ± 0.4 to 102.2 ± 0.8 mg epicatechin/kg. The correlations between TPC or TFC and the antioxidant results (FRAP, DPPH, and ABTS) were found to be statistically significant (p < 0.01). Furthermore, the phenolic compounds are quantified and qualified by high performance liquid chromatography-high resolution mass spectrometry (HPLC-HRMS), and a total of 83 phenolic compounds were tentatively identified in this study. A metabolomics analysis based on the 83 polyphenols was carried out and subjected to principal component analysis and orthogonal partial least squares-discriminant analysis. The results showed that it was possible to distinguish Apis cerana honey from Apis mellifera honey based on the phenolic profile.


Introduction
Honey can serve as a source of natural antioxidants. The antioxidant activity of honey is primarily provided by its polyphenols [1]. Thus, a considerable variation of antioxidant activity and polyphenols profile is found among different honey varieties around the world [2][3][4][5]. This variation is mainly due to different floral and geographical origins as well as the type of bees [6,7]. Therefore, the analysis of phenolic profile has been regarded as a very promising technique for studying the floral, geographical and honeybee origins of honeys.
In general, Apis cerana (A. cerana) honey is produced by Apis cerana grazing on various botanical sources. Traditionally, A. cerana honey is more nutritious than other honey varieties because of its long nectar cycle and the wide variety of nectar source [8]. There are recent research reports that have found various benefits for A. cerana honey, such as its' anti-inflammatory, anti-oxidant, and beneficial effects with regard to acute alcohol-induced liver damage [9][10][11]. These therapeutic activities have been attributed to the phenolic acid and flavonoids content of A. cerana honey [9,11]. Nonetheless, there is still a lack of understanding about the phenolic profile and antioxidant capacity of A. cerana honey. Until now, most of the studies have focused mainly on the phenolic profile and antioxidant activity in mono-floral honeys from Apis mellifera (A. mellifera) in China [12][13][14][15][16][17]. Therefore, it is necessary to evaluate the phenolic compounds in A. cerana honey here.
In ancient China, "Shen Nong's Herbal Classic" have recorded the use of A. cerana honey from Shennongjia district as the first use of medicine. The Shennongjia forestry

Quantification of Thirteen Polyphenols in Honeys Using Different Extraction Methods
Several common phenolic compounds and abscisic acid in honey that are reported in the literature were isolated using three different extraction methods and quantified in the present study. The LOD (Limit of Detection), LOQ (Limit of Quantitation), linear range, and MS characteristics of these compounds are listed in Supplementary Table S1. Table 2 shows the average amount of each compound isolated using different methods in the A. cerana and A. mellifera honeys. As seen, the average content of thirteen polyphenols in samples varied considerably depending on the extraction methods. EAC (ethyl acetate, liquid-liquid extraction) generated higher levels of kampferol (p < 0.0001), quercetin (p < 0.0001), vanillic acid (p < 0.0001), and trans-ferulic acid (p < 0.01), while, SPE (XA and PLS, solid-phase extraction) generated higher levels of rutin (p < 0.0001). Between the two SPE cartridges, the Strata XA cartridge showed lower recoveries of vanillic acid (p < 0.0001) and 4-hydroxybenzoic acid (p < 0.0001) compared to ProElut PLS SPE cartridges. The results suggested that different extraction methods have different extraction efficiency for phenolic acids and flavonoids. "*" represents values that differed significantly between A. cerana and A. mellifera honeys for the same compound using the uniform extraction method, * means p < 0.05, ** means p < 0.01, *** means p < 0.001, **** means p < 0.0001. " abc " letters represent values that differed significantly among different extraction methods for the same compound.
In general, most of the flavonoids showed a lower average content than phenolic acids. Among thirteen compounds, kaempferol and 4-hydroxybenzoic acid were the main flavonoid and phenolic acid found in the A. mellifera and A. cerana honeys in China. It was reported that kaempferol and 4-hydroxybenzoic acid were prevalent in A. mellifera honey of different geographic origins in previous studies [14,[25][26][27]. In addition, chrysin was present at the lowest levels in A. cerana honey, while rutin had the lowest content in A. mellifera honey. Furthermore, some flavonoids showed significant distinctions between A. cerana and A. mellifera honeys regardless of extraction methods. For example, the contents of quercetin (p < 0.05), rutin (p < 0.01) and p-coumaic acid (p < 0.05) were higher in A. cerana honeys than those in A. mellifera honeys. Three compounds includ-ing pinocembrin (p < 0.01), chrysin (p < 0.01), and galangin (p < 0.001) were considered as propolis-derived flavonoids [22,25], had lower contents in A. cerana honeys.

Identification of Individual Polyphenols
One hundred and eleven honey extracts were subjected to the identification of the flavonoids, phenolic acids and abscisic acid based on the optimization conditions of HPLC-QTOF-MS/MS. A total of 83 compounds were tentatively identified, and 13 of them were qualified by comparing the retention times (RT) and the MS spectra with available standards. In the absence of standards, the identification of a further 70 compounds was based on the search for the [M-H] − deprotonated molecule and its fragmentation referred to in the literature. Table 3 summarizes the data obtained for each of the identified compounds with their retention times, error in ppm (between the mass found and the accurate mass), as well as the MS/MS fragment ions.   Hydroxycinnamic acids such as caffeic acid and their derivatives were the main phenolic acids found in the study. Caffeic acid was present in all of the honey samples; in addition, ten caffeic acid derivatives were detected: caffeoylquinic acid isomers (compounds 6, 14 and 17), dicaffeoylquinic acid isomers (compounds 25, 27 and 30), and four ester derivatives of caffeic acid (compound 31, 32, 33 and 34). All of the caffeic acid derivatives showed negative product ions at 179 m/z due to the loss of the deprotonated molecule of caffeic acid. Caffeoylquinic acids and dicaffeoylquinic acids were reported in the European honeydew honey [33] and A. mellifera honey from different botanical and geographical origins [2,25,27,28,37]. Caffeic acid ester derivatives were detected in Chilean propolis [30] and Spanish A. mellifera honey [2]. As shown in Table 3, caffeic acid ester derivatives were commonly present in A. mellifera honey in this study, whereas they were relatively rare in A. cerana honey.
Furthermore, both isomers of abscisic acid previously described in other varieties of honey [2] were detected in all of the honey samples in the study. In addition, 4-ethoxy-3-methoxycinnamic acid (compound 24) was identified only in Manuka honeys in the study. By examining the empirical formula of this compound, it was concluded that it may be an ethylated derivative of ferulic acid. Concerning flavonoids, four subclasses of compounds were identified: flavonols, flavanonols, flavanones, and flavones, in addition to some flavanonol ester derivatives and flavonols glycosides. The flavanonol ester derivatives mainly came from pinobanksin (compounds 58, 59, 60 and 61), which showed a negative product ion at 271 m/z due to the loss of the deprotonated molecule of pinobanksin. Pinobanksin and its ester derivatives are characteristic flavonoids of propolis, and were found in Spanish A. mellifera honeys [2], sulla honey from the Sicilian native breed of black honeybee [36], as well as the Chilean propolis [30]. In this study, these compounds were present in almost all A. mellifera honeys, but very few were found in A. cerana honey. For example, pinobanksin-3-O-hexanoate (compound 61) was present in all A. mellifera honeys except for A.m_p7 honey, while it was undetectable in all A. cerana honey samples ( Table 3).
The flavonols' glycosides that were mainly from quercetin, kaempferol, methoxykaempferol, and isorhamnetin were previously described in different types of honey [2,33,34]. Numerous derivatives of flavonols' glycosides were identified in A. mellifera and A. cerana honey extracts in this study: rhamnosides (loss of 146 Da), hexosides (loss of 162 Da), neohesperidoside, rhamnosylhexoside (loss of 308 Da), and dihexosides (loss of 324 Da). For example, in MS2 spectra of compound 47 at 46.92 min and 431 m/z, base peak fragments at 285 m/z (loss of 146 Da) and additional two fragment ions resulting from the loss of 257 and 151 Da could be observed, and it was then concluded that it could be kaempferol-rhamnosides.
In conclusion, propolis-derived caffeic acid and pinobanksin ester derivatives were widely present in A. mellifera honeys in the study, but rarely in A. cerana honeys.

Metabolomics Analysis
A PCA was conducted to evaluate the effect of the honey species / extraction method on the 83 phenolic compounds from a descriptive point of view ( Figure 1). As shown in a PCA scores plot ( Figure 1A), all of the A. cerana honey extracts regardless of extraction method (n = 77) were designed in PC1 negative, and most of the A. mellifera honey extracts (n = 29) were designed in PC1 positive. These results suggested that different honey species, rather than extraction methods, could be distinguished based on the levels or the presence of phenolic compounds.
For A. cerana honeys distributed in PC1 negative, most of the honey extracts (n = 72) clustered tightly, except for five honey extracts which were far away from other honey extracts due to their high level of methoxy kaempferol ( Figure 1B). For A. mellifera honeys distributed in PC1 positive, the poly-floral A. mellifera honey extracts (n = 24) clustered tightly and were closest to the A. cerana honey group, followed by mono-floral A. mellifera (A.m_F) honey, and then by Manuka honey. The result indicated that botanical and geographical origins have an effect on the phenolic profile in A. mellifera honeys. Manuka honey was differentiated from other honeys for the high contents of 4-methoxyphenyllactic acid and p-hydroxy-hydrocinnamic acid. Fangxian A. mellifera honey was monofloral honey and characterized by a high content of pinobanksin ( Figure 2B). Wuhan A. mellifera honey was polyfloral honey, and thus it may be closer to Shennongjia A. cerana honey in its phenolic acid profile because of the diversity of plant sources.
Furthermore, an orthogonal partial least squares-discriminant analysis (OPLS-DA) was conducted to analyze the differences between A. mellifera and A. cerana honey. Figure 2A showed that A. cerana honey samples were located on the right side of the ellipse and were well separated from A. mellifera honey samples. This result indicated that there were significant differences in the two honey groups. In addition, seven-fold cross-validation and 200 permutations were conducted to further verify the predictability of the OPLS-DA model. As shown in Figure 2B, the intercept of Q 2 (−0.223) was negative on the vertical axis, and all blue Q 2 -values to the left were lower than the original points to the right, indicating that the established model was not overfitted for the experiment.
The variables responsible for discriminating A. cerana from A. mellifera honey were then identified using the OPLS-DA VIP ( Figure 2C, VIP > 1) and S-plot ( Figure 2D). The red variables ( Figure 2C, VIP > 1) were tested using a Student's t-test and the corresponding VIP and p values (p < 0.01) are listed in Supplementary Table S3. An S-plot ( Figure 2D) was used to visualize the covariance and correlation between A. mellifera and A. cerana honey. Here, eight variables (compound 1-8 in Supplementary Table S3, p < 0.01) were far from the origin and were located at the far left of the X-axis. This indicated that the contents of these compounds in A. mellifera honey were higher than those in A. cerana honey. Among these compounds, five propolis-derived flavonoids (pinobanksin, pinobanksin-5-methyl ether, galangin, chrysin and pinocembrin), were commonly present in all A. mellifera honeys in the present study (Table 3). These flavonoids have previously been identified in propolis, European honeydew honey, and mono-and polyfloral honey from A. mellifera [2,27,30,33,38]. For A. cerana honeys distributed in PC1 negative, most of the honey extracts (n = 72) clustered tightly, except for five honey extracts which were far away from other honey extracts due to their high level of methoxy kaempferol ( Figure 1B). For A. mellifera honeys distributed in PC1 positive, the poly-floral A. mellifera honey extracts (n = 24) clustered tightly and were closest to the A. cerana honey group, followed by mono-floral A. mellifera (A.m_F) honey, and then by Manuka honey. The result indicated that botanical and geographical origins have an effect on the phenolic profile in A. mellifera honeys. Manuka  Furthermore, an orthogonal partial least squares-discriminant analysis (OPLS-DA) was conducted to analyze the differences between A. mellifera and A. cerana honey. Figure  2A showed that A. cerana honey samples were located on the right side of the ellipse and were well separated from A. mellifera honey samples. This result indicated that there were significant differences in the two honey groups. In addition, seven-fold cross-validation and 200 permutations were conducted to further verify the predictability of the OPLS-DA model. As shown in Figure 2B, the intercept of Q 2 (−0.223) was negative on the vertical axis, and all blue Q 2 -values to the left were lower than the original points to the right, indicating that the established model was not overfitted for the experiment.
The variables responsible for discriminating A. cerana from A. mellifera honey were then identified using the OPLS-DA VIP ( Figure 2C, VIP > 1) and S-plot ( Figure 2D). The red variables ( Figure 2C, VIP > 1) were tested using a Student's t-test and the corresponding VIP and p values (p < 0.01) are listed in Supplementary Table S3. An S-plot ( Figure 2D) was used to visualize the covariance and correlation between A. mellifera and A. cerana honey. Here, eight variables (compound 1-8 in Supplementary Table S3, p < 0.01) were far from the origin and were located at the far left of the X-axis. This indicated that the contents of these compounds in A. mellifera honey were higher than those in A. cerana honey. Among these compounds, five propolis-derived flavonoids (pinobanksin, pinobanksin-5- As shown in Figure 2D, five variables (compound 9-13 in Supplementary Table S3, p < 0.01) were far from the origin and were located at the far right of the X-axis. The result indicated that the contents of these compounds in A. cerana honey were higher than those in A. mellifera honey. The five compounds have been previously reported in tilia, salvia officinalis L., and chestnut source honey samples [2,25,39]. In this study, they were commonly present in A. cerana and A. mellifera honey. The high content levels of these compounds in Shennongjia A. cerana honey may be due to the abundant sources of wild medicinal plants and nectar plants in this region.
Of course, whether these compounds can be used as appropriate markers to distinguish A. cerana honey from A. mellifera honey requires further study and confirmation by expanding the sample size and selecting A. cerana and A. mellifera honey from different geographical and plant sources in the future.

Extraction of Phenolic Compounds
The extraction of phenolic compounds was undertaken by solid-phase extraction. The SPE method was carried out according to the previous study [15] with minor modifications. A total of 10.0 g of honey samples were mixed with 50 mL of ultrapure water, and then the solution was adjusted to pH = 2 with HCl for the PLS cartridges or adjusted to pH = 7 with 5% ammonium (v/v) for the Strata X-A cartridges. After removing the impurity particles by centrifugation (8000 g, 10 min), the supernatants were loaded onto the previously conditioned cartridges (according to the manufacturer's instructions). After loading, these cartridges were washed with 4 mL of acidified ultrapure water (pH = 2) for the PLS SPE cartridges or washed with 4 mL of ultrapure water (pH = 7) for the Strata X-A SPE cartridges. The phenolic compounds retained on the cartridges were then eluted with 5 mL of formic acid: methanol (1:9, v/v). The extract was evaporated at 40 • C under a stream of nitrogen, and then reconstituted in 1 mL of methanol with 0.1% formic acid. The obtained extracts were filtered and stored at −20 • C until further analysis by high performance liquid chromatography-quadrupole time-of-flight mass spectrometry (HPLC-QTOF-MS).
The extraction of phenolic compounds was undertaken with liquid-liquid extraction. Briefly, 10.0 g of honey samples were mixed with 50 mL of ultrapure water, and the solution was then adjusted to pH = 2 with HCl. The honey solution was extracted three times with 20 mL of ethyl acetate. The extracts were evaporated to dryness on a rotary evaporator at 30-40 • C, and then dissolved in 1 mL of methanol with 0.1% formic acid. The obtained extracts were filtered and stored at −20 • C until further analysis by HPLC-QTOF-MS.

HPLC-QTOF-MS Conditions
HPLC analyses were performed using a Shimadzu LC-20A system (Shimadzu Corporation, Kyoto, Japan) coupled with a quadrupole time-of-flight mass spectrometer (AB Sciex Triple QTOF5600+, AB Sciex, Redwood, CA, USA). The chromatographic separation was carried out using an Eclipse XDB-C18 column (100 mm × 2.1 mm, 3.5 um) (Agilent, Wilmington, DE, USA). The mobile phase consisted of 0.1% formic acid in water (phase A) and 0.1% formic acid in methanol (phase B). The flow rate was 0.3 mL/min and the injection volume was 10 µL, while the temperature of the column oven was set at 35 • C. The gradient separation was performed as follows: 0-1 min, 0% (B); 1-6 min, 0-6% (B); 6-13 min TOF-MS and the data of ten TOF-MS/MS were collected in negative ion mode using the information-dependent acquisition (IDA) function. The parameters were as follows: dynamic background subtraction (DBS); charge monitoring to exclude multiply charged ions and isotopes; Ion Source Gas1: 55 psi; Ion Source Gas2: 60 psi; Curtain Gas: 30 psi; Temperature: 600 • C; IonSpray Voltage Floating: −4500 V; Declustering Potential: 100 V; Collision Energy: 25 V; Collision Energy Spread: 15 V. In order to ensure the stability of outcomes, the calibration reagent (sodium formate) was detected in every two sample intervals, and methanol was used as a blank control to avoid the misjudgment of characteristic markers. In parallel, quality control (QC) samples were prepared by mixing equal volumes (9 µL) from each sample. An aliquot of this pooled sample was analyzed every fourteen samples in order to provide the measure of the system's stability and performance. The system operation, data acquisition, and analysis were controlled and processed using Analyst 1.7.1,PeakView 2.2, and MultiQuant 3.0 softwares from AB Sciex Inc. (Vaughan, ON, Canada).

Determination of Total Phenolic Content (TPC) and Total Flavonoids Content (TFC)
TPC and TFC were measured on a UV-2550 Spectrophotometric Reader (Shimadzu Corporation, Kyoto, Japan). The absorbance was measured at 725 nm and 510 nm, respectively. All of the analyses were performed in triplicate. TPC and TFC analysis were performed using the photocolorimetric method, as described by Mohammed Moniruzzaman [41]. The TPC was expressed as milligrams of gallic acid equivalents per kilogram of honey (mg GAE/kg honey), and the standard curve was generated with gallic acid (10-160 µg. mL −1 ). The TFC were expressed as milligrams of epicatechin equivalents per kilogram of honey (mg EC/kg honey), and the standard curve was plotted using epicatechin (1-100 µg. mL −1 ).

Antioxidant Activity
Antioxidant activity assays including DPPH, ABTS and FRAP were studied as described by Habib et al. [42].
Radical scavenging activity assay (DPPH assay). The aqueous solution of honey (0.2 g. mL −1 ) was mixed with 3.8 mL of DPPH radical solution (0.25 mM). After incubating in the dark for 30 min, the absorbance of the solution was measured at 515 nm. The percentage of free radical scavenging activity that targeted DPPH was calculated using the following equation: DPPH radical savaging activity where A0 is the absorbance of the DPPH control, and A1 is the absorbance in the sample. ABTS cation radical scavenging. The cation radical ABTS+ was synthesized by the reaction of a 7 mM ABTS solution with a 2.4 mM potassium persulfate solution. The mixture was kept at room temperature in the dark for 14 h. Afterwards, the ABTS+ solution was diluted with methanol until an absorbance of 0.73 ± 0.01 units at 734 nm was achieved. 1.0 mL of the honey sample (20% w/v) was mixed with 1.0 mL of fresh diluted ABTS solution. After incubation at room temperature for 7 min, the absorbance of the solution was measured to be 734 nm. The percentage inhibition calculated as ABTS radical scavenging activity was according to Equation (1), as provided above.
Ferric reducing/antioxidant power assay. The FRAP reagent was prepared before the test by mixing 100 mL of acetate buffer (300 mM, pH 3.6) with 10 mL of TPTZ solution (10 mM in 40 mM HCl) and 10 mL of ferric chloride (FeCl 3 , 20 mM). A total of 100 µL of the honey solution (0.2 g·mL −1 ) was mixed with 900 µL of ultrapure water, followed by adding 2.0 mL of the FRAP reagent. The mixture was then vortexed and incubated at 37 • C for 30 min. The absorbance was then determined to be 593 nm using ferrous sulfate standards (0, 0.1, 0.2, 0.5, 1, 1.5, 2.0 mM). The units used for the FRAP values was µmol of ferrous equivalents/100 g of honey sample.