Identification of botanical origin of Chinese unifloral honeys by free amino acid profiles and chemometric methods.

The amino acid contents of five floral sources Chinese honeys (jujube, rape, chaste, acacia, and lungan) were measured using reversed phase high-performance liquid chromatography (RP-HPLC). The results showed that proline was the main amino acid in most of the analyzed samples. Phenylalanine presents at the highest content in chaste honey samples, and the total amino acid contents of chaste honeys were also significantly higher than those of other honey samples. Based on the amino acid contents, honey samples were classified using chemometric methods (cluster analysis (CA), principal component analysis (PCA), and discriminant analysis (DA)). According to the CA results, chaste honeys could be separated from other honeys, while the remaining samples were correctly grouped together when the chaste honey data were excluded. By using DA, the overall correct classification rate reached 100%. The results revealed that amino acid contents could potentially be used as indicators to identify the botanical origin of unifloral honeys.

Introduction

Honeys that come from different botanical origins vary widely in composition, nutritional contribution, and market price [1]. Thus, finding accurate ways to determine the origin of honey has become an important issue. Kaškoniené and Venskutonis [2] reviewed the chemical markers in honey, and indicated that volatile compounds [3], phenolic compounds [4], [5], carbohyrates [6], nitrogen-containing compounds [7], pollen [8], and other minor components might distinguish certain types of honey from others. However, there is no unambiguous and unique indicator that accurately identifies the botanical origin of honey. Therefore, in order to establish methods for differentiating honey types, research on chemical markers requires further development.

As characteristic small compounds found in honey, amino acids can provide some information about botanical origin. The amino acids present in honey mainly arise from the pollen of the plant, so they are derived directly from the plant source. Honey contains about 1% amino acids, of which proline is the dominant amino acid accounting for 50%–85% of total amino acids [9]. This is because proline is the main amino acid in pollen [10]. Furthermore, proline is used as a quality parameter of honey according to the laws in many countries, such as the AOAC [11]. There are many other amino acids in honey, such as glycine, alanine, leucine, and phenylananine. It has been previously reported that amino acid content can serve as an useful indicator for estimating the origin of honey [12], [13]. However, the classification of Chinese honey using amino acids data has not been described so far.

As an important honey importer and exporter, China is very rich in bee resources. Jujube, acacia, rape, lungan and chaste have the largest production capacity; especially jujube, acacia and chaste honey belong to the four most famous honey types. Jujube trees are grown originally in China, and jujube honey is mainly produced in Shaanxi, Shanxi, and Henan provinces of China and it is a specially Chinese honey [14]. Acacia, chaste and rape honeys are light color honey; the colors of these honey are from transparent to light yellow to light amber. In contrast, the color of jujube and lungan honeys is amber, more or less dark. Lungan honey is a top-grade spirit, and acacia, chaste and jujube honeys also have better quality. These four types of honeys are easily adulterated for their higher price. The price of rape honey is significantly lower than that of other honeys, so it is often used to adulterate.

Chemometrics can extract useful information from numerous data and make complicated analysis simple. Its combination with high-performance liquid chromatography (HPLC) has been used in the determination of certain chemical fingerprints of natural products including honey [4], [15]. The cluster analysis (CA), principal component analysis (PCA) and discriminant analysis (DA) are common to handle data [16]. By calculating the data of different parameters, such as physicochemical indicators, phenolic acids and polyphenols, proteins, volatiles, as well as free amino acids, these methods could analyze, identify and classify honeys [17].

The aim of the present study was to assess whether it is a useful way to identify the botanical origin of Chinese unifloral honeys by analyzing amino acid profiles. Five different Chinese honey types were investigated by analyzing their amino acid profiles using reversed phase high-performance liquid chromatography (RP-HPLC). In addition, PCA was performed to reduce dimensionality, and CA and DA were employed to distinguish and classify the botanical origins of Chinese unifloral honeys.

Materials and methods

Samples

A total of 75 honey samples (jujube, 15; rape, 15; chaste, 15; acacia, 15; and lungan, 15) were analyzed. All samples were collected from local beekeepers in China in 2013 or 2014, centrifuged and unpasteurized. All samples were stored at 4 °C until analysis. The botanical origins and geographical regions of the samples are presented in Table 1.

Table 1

Information of honey samples.

Sample code	Common name	Floral origin	Geographical origin	Year
C1-C5	Chaste honey	Vitex negundo L.	Huixian, Henan	2013
C6-C10	Chaste honey	Vitex negundo L.	Zanhuang, Henan	2013
C11-C15	Chaste honey	Vitex negundo L.	Pingding, Shanxi	2013
Y1-Y7	Rape honey	Brassica campestris L.	Hanzhong, Shannxi	2014
Y8-Y11	Rape honey	Brassica campestris L.	Ankang, Shannxi	2014
Y11-Y13	Rape honey	Brassica campestris L.	Neijiang, Sichuan	2013
Y14-Y15	Rape honey	Brassica campestris L.	Mianyang, Sichuan	2013
J1-J4	Jujube honey	Ziziphus jujuba Mlii.	Yulin, Shannxi	2013
J5-J8	Jujube honey	Ziziphus jujuba Mlii.	Dali, Shannxi	2013
J9-J11	Jujube honey	Ziziphus jujuba Mlii.	Jiaxian, Shannxi	2013
J12–14	Jujube honey	Ziziphus jujuba Mlii.	Linbao, Henan	2014
J15	Jujube honey	Ziziphus jujuba Mlii.	Zanhuang, Henan	2014
L1-L8	Lungan honey	Dimocarpus longan Lour.	Maoming, Guangdong	2014
L9-L11	Lungan honey	Dimocarpus longan Lour.	Pingnan, Guangxi	2013
L12-L13	Lungan honey	Dimocarpus longan Lour.	Guiping, Guangxi	2013
L14	Lungan honey	Dimocarpus longan Lour.	Lingshan, Sichuan	2014
L15	Lungan honey	Dimocarpus longan Lour.	Fujian	2014
A1-A6	Acacia honey	Robinin pseudoacacia L.	Lushi, Henan	2014
A7-A10	Acacia honey	Robinin pseudoacacia L.	Heshui, Gansu	2014
A11–12	Acacia honey	Robinin pseudoacacia L.	Ningxia, Gansu	2014
A13-A15	Acacia honey	Robinin pseudoacacia L.	Yanchang, Shanxi	2014

Reagents and standards

Acetonitrile of HPLC grade and 18 amino acid standards, including glycine (Gly), alanine (Ala), serine (Ser), cysteine (Cys), threonine (Thr), valine (Val), leucine (Leu), isoleucine (Iso), methionine (Met), phenylananine (Phe), tryptophan (Try), tyrosine (Tyr), aspartic acid (Asp), glutamic acid (Glu), glutamine (Gln), lysine (Lys), arginine (Arg), and histidine (His), were purchased from Sigma-Aldrich (St. Louis, MO, USA). The derivatizing reagent (kit Waters AccQ·Fluor™) and eluent A were supplied by Waters (Massachusetts, USA). Waters AccQ·Fluor™ kit consists of Waters AccQ·Fluor borate buffer, Waters AccQ·Fluor reagent powder, and Waters AccQ·Fluor reagent diluent. HPLC grade water was obtained from a Milli-Q system (Millipore, Bedford, MA, USA). Cellulose acetate filters (0.45 µm) were supplied by Anpel (Shanghai, China).

Sample pretreatment

Three grams of honey was accurately weighed and diluted with 25 mL of ultrapure water. The solution was sonicated for 5 min, and filtered through a 0.45 µm cellulose acetate filter.

Free amino acid analysis

The method of derivatization was as follows: aliquot of 10 μL sample was placed in a derivative tube; 70 μL AccQ·Fluor borate buffer was added to monitor pH in the range of 8.2–10.0; then approximately 20 μL AccQ·Fluor reagent was added, and the mixture was heated in an oven at 55 °C for 10 min [18].

Amino acids were simultaneously separated in an HPLC system using Dionex UltiMate 3000 with a diode array detector and a Waters AccQ·Tag amino acid analysis column (3.9 mm× 150 mm, 4 µm). The column temperature was 37 °C, the injection volume was 10 μL, and the flow rate was 1 mL/min [17]. The mobile phase consisted of 10% aqueous eluent A (A), ultrapure water (B), and acetonitrile (C). The elution gradients were as follows: 100%–99% A and 0%–1% C from 0 to 0.5 min; 99%–95% A and 1%–5% C from 0.5 to 18 min; 95%–91% A and 5%–9% C from 18 to 19 min; 91%–83% A and 9%–17% C from 19 to 29 min; 83%–0% A, 17%–60% C, and 0%–40% B from 29 to 38 min; 40% B and 60% C for 30 min; 100% A for 30 min; finally 40% B and 60% C to wash the column.

Statistical analysis

Statistical analysis was performed using IBM SPSS Statistics 19 software. The chemometric analysis methods, CA, PCA, and DA, were used to discriminate Chinese honeys according to their botanical origin. CA is an unsupervised classification model. Combining unsupervised clustering with supervised information could ensure whether these data are appropriate to determine the botanical origin of honey. PCA is a technique for linearly compressing multidimensional data into lower dimensions with minimal loss of information [19]. DA is a multivariate probabilistic classification method based on a multivariate probability distribution with the hypothesis of a normal distribution with the same variance–covariance matrix in all the considered classes [20].

Results and discussion

Method validation

Eighteen amino acids were determined in the present work. Fig. 1 shows a typical chromatogram of standards, using the proposed methodology. The calibration curves were established in Table 2 based on the corresponding relationship between the peak area and concentration. All of 18 amino acids had good linearity (R2>0.997). The limit of detection (LOD) was less than 0.56 μg/mL and the limit of quantitation (LOQ) was less than 1.68 μg/mL.

Fig. 1

The RP-HPLC chromatogram of 18 amino acid standards.

Table 2

Regression equation, R2, linear range, LOD and LOQ of 18 amino acids.

Analyte	Regression equation	R2	Linear range (μg/mL)	LOD (μg/mL)	LOQ (μg/mL)
Asp	y=14.831x+0.133	0.9991	1.23–1.08×101	0.31	0.94
Ser	y=21.547x+0.293	0.9976	3.05–1.04×102	0.25	1.03
Glu	y=14.622x+0.218	0.9995	2.53–1.50×101	0.32	1.04
Gly	y=29.244x+0.167	0.9992	1.02–1.10×101	0.27	0.82
His	y=12.107x+0.042	0.9993	3.25–4.10×101	0.56	1.68
Arg	y=12.836x−0.051	0.9998	0.95–9.50	0.19	0.53
Thr	y=21.929x+0.030	0.9998	0.85–7.50×101	0.23	0.67
Ala	y=27.332x+0.2070	0.9994	0.45–1.05×102	0.11	0.32
Pro	y=19.77x+0.038	0.9999	4.05×101−3.00×102	0.46	1.38
Cys	y=9.189x+0.067	0.9999	0.55–1.05×102	0.12	0.34
Tyr	y=12.337x+0.035	0.9991	2.03–1.25×102	0.31	0.93
Val	y=20.83x+0.052	0.9999	1.03–2.01×101	0.27	0.82
Met	y=17.23x−0.004	0.9999	0.52–1.50×102	0.14	0.43
Lys	y=26.342x+0.229	0.9976	0.85–2.34×101	0.21	0.63
Ile	y=19.259x+0.066	0.9999	0.95–1.54×101	0.24	0.71
Leu	y=19.771x+0.018	0.9999	0.82–1.34×101	0.21	0.62
Phe	y=17.125x+0.023	0.9998	0.93–7.04×102	0.16	0.47
Trp	y=12.186x+0.044	0.9998	2.34–4.54×101	0.48	1.43

Table 3 shows the relative standard deviation (RSD) of peak area and retention time of precision, stability, and repeatability, and recovery rate. It can be found that the RSD values of precision, stability, and repeatability were less than 2.54%. And recoveries were measured by adding the standards into the jujube honey samples at different concentrations. The average recovery rate varied from 96.32% to 103.71%; meanwhile the RSD varied from 0.74% to 2.79%. These results indicated that the accuracy and the precision of this method were adequate for the determination of amino acids in honey samples.

Table 3

The peak area (PA) and retention time (RT) of precision, stability, and repeatability, and recovery rate.

Analyte	Precision (n=6, % RSD)		Stability (n=12, % RSD)		Repeatability (n=6)		Recovery rate (n=3)
	PA	RT	PA	RT	PA	RT	Mean (%)	RSD (%)
Asp	1.03	1.23	1.93	1.78	2.49	2.34	98.32	1.07
Ser	2.34	2.13	2.35	2.41	1.35	1.23	99.32	1.65
Glu	1.58	1.34	1.45	1.39	1.93	1.53	97.54	1.98
Gly	1.32	1.45	2.12	2.30	2.54	1.93	98.78	0.78
His	2.13	2.01	1.67	1.52	2.19	1.83	100.92	2.13
Arg	1.93	1.78	1.86	1.72	1.67	1.52	97.29	1.39
Thr	1.29	1.32	1.35	1.42	1.39	1.09	102.72	2.07
Ala	1.34	1.54	1.37	1.29	1.73	1.34	103.25	1.94
Pro	1.73	1.93	2.09	1.97	1.82	1.32	100.83	1.85
Cys	2.01	1.78	2.21	2.05	2.23	1.92	96.57	2.79
Tyr	2.29	2.13	1.02	1.15	2.52	2.04	98.27	1.47
Val	1.92	2.13	1.35	1.43	1.59	1.34	99.35	2.48
Met	1.28	1.09	1.37	1.47	1.52	1.28	101.45	0.74
Lys	1.45	1.56	2.05	2.13	1.25	1.09	103.71	1.95
Ile	2.39	2.54	1.56	1.47	0.92	1.32	96.32	2.17
Leu	1.59	1.62	1.78	1.39	0.73	1.19	97.34	1.48
Phe	1.94	2.14	1.94	1.68	1.83	1.43	100.54	0.74
Trp	2.01	1.94	1.52	1.43	1.25	1.43	97.69	1.58

Amino acid profile

We studied systematically the amino acid profiles of five types of Chinese honey. Amino acid data obtained from RP-HPLC are shown in Table 4. Most of the amino acids were detected in all analyzed samples; however, the sulfur-containing amino acids (Cys and Met) were not found. This is in agreement with the results reported by Rebane et al. [21] involving Cys and Met content of Estonian honeys. That does not, however, mean that no honey samples contain Cys and Met. These amino acids had been found in honeys from some areas, such as certain French and Spanish honeys [22], [23]. Therefore, sulfur-containing amino acids could be potential geographical markers. Proline was found to have the highest concentration in most of samples, approximately accounting for 50%. Some other amino acids, including Gly, His, Ser, Ala, Tyr, and Phe, displayed similar mean values in different types of honey and were the second main amino acids in all analyzed honey samples with percent contents of 0.08%–4.11%, 0.66%–9.02%, 0.46%–9.99%, 1.35%–6.18%, 3.21%–5.65%, and 0.86%–3.99% (69.61% in chaste honey samples), respectively. Compared with these samples, Pro, Phe, Ala, Thr, and Arg were the dominant amino acids in Spanish honey [22], and Pro, Phe, Tyr, and Lys accounted for the majority in Serbian honey [24].

Table 4

The mean value of concentration (MV) and percent content (PC) of amino acids in five types of honey samples.

	Rape		Jujube		Acacia		Lungan		Chaste
Analyte	MV (mg/100 g)	PC (%)	MV (mg/100 g)	PC (%)	MV (mg/100 g)	PC (%)	MV (mg/100 g)	PC (%)	MV (mg/100 g)	PC (%)
Asp	6.3±3.1ab	1.59	15.6±10.7c	2.17	8.9±4.1b	1.96	13.3±4.8c	3.27	3.2±0.9a	0.20
Ser	39.4±2.4c	9.99	20.0±16.8b	2.78	16.9±5.5b	3.71	14.1±3.9b	3.47	7.2±1.4a	0.46
Glu	9.5±2.7b	2.42	20.2±6.5c	2.81	11.7±4.7b	2.58	20.5±6.3c	5.04	5.2±1.2a	0.33
Gly	16.2±8.4c	4.11	12.2±8.6bc	1.70	6.6±3.6ab	1.44	9.5±3.2b	2.33	1.3±0.6a	0.08
His	35.6±19.1b	9.02	26.5±19.4b	3.69	15.3±12.2a	3.37	14.2±5.0a	3.48	10.3±2.5a	0.66
Arg	16.2±4.0c	4.11	14.4±9.3bc	2.00	10.5±6.5ab	2.30	9.3±4.4a	2.27	7.3±4.8a	0.46
Thr	6.1±1.6ab	1.54	7.5±2.6b	1.04	9.3±4.4b	2.05	6.2±2.0ab	1.53	3.1±0.9a	0.20
Ala	5.3±1.6a	1.35	31.8±12.8c	4.42	15.7±9.4b	3.44	5.6±1.4a	1.38	97.2±10.2d	6.18
Pro	175.0±29.9a	44.33	457.5±50.8c	63.62	291.1±25.5b	63.93	184.3±35.1a	45.28	264.4±222b	16.81
Met	nd		nd		nd		nd		nd
Tyr	13.3±5.7a	3.38	40.6±26.8b	5.65	18.5±7.4a	4.06	16.4±3.9a	4.04	50.6±5.6c	3.21
Val	13.4±3.8b	3.40	15.7±3.3b	2.18	14.3±2.7b	3.15	19.8±7.2c	4.85	6.6±0.6a	0.42
Cys	nd		nd		nd		nd		nd
Lys	34.9±4.0d	8.84	4.6±2.0ab	0.64	7.1±5.4c	1.56	6.6±1.8bc	1.62	3.6±1.0a	0.23
Ile	5.7±1.5a	1.44	6.0±1.7ab	0.83	9.6±6.2c	2.11	8.5±3.8bc	2.09	3.5±0.9a	0.23
Leu	4.8±1.5a	1.21	4.6±1.1a	0.64	4.5±1.7a	0.98	5.2±1.4a	1.28	4.4±0.3a	0.28
Phe	13.0±4.7a	3.29	17.5±9.0a	2.43	3.9±0.4a	0.86	16.3±7.3a	3.99	1094.9±120.1b	69.61
Trp	nd		24.6±1.2c	3.42	11.5±1.2b	2.51	57.3±6.7d	14.07	11.2±1.9b	0.71
total AA	394.7±74.4	100.0	719.1±69.8	100.00	455.4±44.6	100.0	407.0±60.0	100.0	1572.9±140.8	100.0

Values within the same row follow by the same small letter are not significantly different at the level of 5%.

nd: Not detected.

Chaste samples were substantially different from other samples. The total content of amino acids in chaste type was 1527.9 mg/100 g, which was considerably higher than that in other honey types. The mean value of Phe was 1094.9 mg/100 g, and Phe was the dominant amino acid in all analyzed chaste honey samples. The same observation for Phe was also found in lavender honey [22], in which Phe content was 1152.8 mg/100 g. Meanwhile, the mean values of other amino acids in chaste type were higher than those in other types of honey: Ala (97.2 mg/100 g), and Tyr (50.6 mg/100 g).

Jujube honey is a special honey originating in China, and it is popular and well recognized by consumers because of its attractive taste and aroma. The contents of amino acids in jujube honey display special characteristics. The total amount of amino acids in the jujube honey was 719.1 mg/100 g, higher than that in the other three honey types. The mean value of Pro was 457.5 mg/100 g, and Pro was the dominant amino acid in jujube honey. Meanwhile Ala, Tyr, and Glu were the next most abundant amino acids in jujube honey.

For the rest analyzed honey samples, the total amount of amino acids ranged from 397.4 mg/100 g (rape honey) to 455.4 mg/100 g (acacia honey). Pro was the predominant amino acid in these honey samples, and its percent contents ranged between 44.33% (rape honey) and 63.93% (acacia honey). The contents of Ile and Leu were lower but also important in these honey samples. Trp is an important amino acid, but it was not detected in rape samples. Meanwhile, as compared to other honey types, rape honey displayed significant differences in percentages of Ser, His, and Lys (9.99%, 9.02%, and 8.84%). Acacia honey samples contained all detected amino acids except Met and Cys. The contents of Asp, Gly, Lys, Leu and Phe were very low in acacia honey samples. Arg, Ala, and Lys were comprised the minority amino acids in lungan honey, whereas the contents of Glu, His, Val, Phe, and Trp were higher in the other tested honey samples.

CA

CA was performed to further characterize the role of amino acids in classifying honey types. CA has been used in many fields, especially in honey classification [25]. Using CA, it is possible to group samples according to their similarity. In our study, all samples were divided into two categories, chaste honey samples and other samples (Fig. 2A). The jujube, rape, lungan, and acacia samples were grouped in one cluster, which showed that these honeys were similar compared with chaste honey in their amino acid contents.

Fig. 2

Dendrogram of the cluster analysis of honey samples. (A) Included chaste honey samples; (B) excluded chaste honey samples.

When CA was applied to the amino acid data from all samples, there were no significant differences between the jujube, rape, lungan, and acacia samples, so CA was applied to these samples again. The CA results in Fig. 2B were based on data with the chaste honey samples data removed. We can see that samples of different botanical origins gathered together in Fig. 2B. There were no samples assigned to the wrong group, and an ideal result was obtained. This result demonstrated that the amino acid contents could serve as indicators to authenticate the botanical origin of Chinese honeys.

PCA

Honey is a kind of natural product with a complex matrix, so it is difficult to reach a conclusion by analyzing the data of all classes of compounds. PCA is a chemometric analytical method that reduces data dimensions to make it possible to use fewer data to represent most of the variables and achieve the intended goal [19]. Data on all amino acids were analyzed by PCA and three principal components were extracted whose eigenvalues exceeded 1. Table 5 shows eigenvalue, variance and cumulative variance of three principal components. The first principal component (PC1) was strongly associated with the contents of Ser, Arg, Ala, Pro, Tyr, Lys, Leu, and Phe, representing 50.334% of variance. The second principal component (PC2) represented 14.595% of variance, and was mainly associated with Gly, His, and Thr. The third principal component (PC3) was mainly associated with the contents of Asp, Glu, and Trp, and represented 12.305% of variance. The cumulative variance accounted for by these principal components was 77.234%. This ratio was higher than that found in the case of Estonian (75.35%) [21] and Spanish (64%) [22] honeys.

Table 5

Component loadings matrix for the first three factors and the variance explained by each amino acid.

Amino acid	Principal component
	PC1	PC2	PC3
Asp	0.295	0.059	0.576
Ser	0.959	0.165	0.087
Glu	0.341	−0.032	0.752
Gly	0.179	0.902	−0.095
His	0.592	0.733	0.066
Arg	0.840	0.210	0.156
Thr	0.056	0.842	0.112
Ala	0.864	0.110	0.055
Pro	0.913	0.087	0.175
Tyr	0.854	0.161	0.205
Val	0.671	0.103	0.470
Lys	0.862	0.101	−0.229
Ile	0.721	0.169	0.270
Leu	0.900	0.256	0.186
Phe	0.890	0.194	0.163
Trp	−0.462	0.017	0.728
Variance (%)	50.334	14.595	12.305
Cumulative variance (%)	50.334	64.929	77.234

PC: Principal component.

To identity the relationship between the amino acid contents and the type of honey, a diplot was performed (Fig. 3A). This diplot was obtained from PC1 and PC2. The loading of PC1 and PC2 and the representative samples for which all amino acid contents had the mean value in this group are marked in this figure. The other analyzed samples were omitted to improve the graphical representation. It can be observed that all quantified amino acids are strongly related to some honeys in Fig. 3A. For example, the rape and acacia honeys are mainly associated with Glu and Trp. The same conclusion was proposed by Cometto et al. [26]. The first quadrant mainly includes chaste honey; jujube and acacia honeys are mostly distributed in the third quadrant; and lungan and rape honeys are in the fourth quadrant. Fig. 3B is a scatterplot of all samples with PC1 as the vertical axis and PC2 as the horizontal axis. PC1 of the chaste samples had positive values; some other samples had negative values. Most of the rape and acacia honey samples had negative values of PC1, and the lungan and jujube honeys had positive values. It can be seen that not all samples were correctly clustered together, with many samples distributed among other types of honey. This was not the case for chaste honeys analyzed.

Fig. 3

Scatter diagrams of PCA. (A) Loading points of PC1 and PC2; (B) Sample distribution map.

DA

Based on the results of CA and PCA, we realized that amino acids in Chinese honey could provide information about botanical origin. However, there was no definitive way to determine the botanical origin of a honey sample only based on its amino acids data. So DA was performed to set up a model for honey classification. DA is a multivariate statistical method used to identify the respective type of a sample according to various values, under a defined classification system [20]. Amino acid data obtained from RP-HPLC were selected as dependent variables, and the botanical origins of the honey samples as independent variables. The Fisher stepwise DA was the selected classification method. Four discriminant functions were obtained by DA, and all functions were demonstrated to be significant for correct classification. The first function could account for 47.8% of the total variance, with Wilks’ Lambda=0.000, X2=442.746, df=64, and sig.=0.000. The second function could account for 43.1% of the total variance, with Wilks’ Lambda=0.002, X2=277.196, df=45, and sig.=0.000. The third and fourth functions accounted for 8.3% and 0.7% of the total variance. Together, all functions accounted for 100% of the total variance. The centers of each group of honey mass were as follows: rape honey (3.221, –9.490, 3.406, 0.643), jujube honey (–2.752, 0.427, –3.640, 1.076), acacia honey (–0.078, –5.375, –2.289, –1.223), lungan honey (–9.887, 4.489, 2.496, –0.239), and chaste honey (8.530, 6.959, 0.546, –0.118) in a four-dimensional coordinate plot.

Fig. 4 shows a sample distribution plot with the first function as the horizontal axis and the second function as the vertical one. In it, all samples are divided into five categories, and the samples of different botanical origin cluster together. The results of DA are superior, and the accuracy reached 100%. Other researchers have used DA processing of data to classify honeys according to botanical orgin using phenolics, physicochemical parameters, and so on, but not all samples were grouped appropriately or there were overlapping areas between the different types of honey [27].

Fig. 4

The DA scores plot of honey samples.

Conclusion

The amino acid profiles of five Chinese honey types demonstrated that the botanical origin of honey influences the amino acid contents. Although most of the amino acid contents were similar, some significant differences could be found. The content of Phe in chaste honey samples was apparently higher than that in other samples. And Trp, Pro, Asp, Glu, Lys, and Ser exhibited differences according to honey botanical orgin. Multivariate statistical analyses employing CA, PCA, and DA showed that samples could be classified correctly according to their botanical origin. Additionally, DA offered a more precise mode to determine the botanical origin of Chinese honey. In conclusion, our results support the hypothesis that amino acid profiles are adequate for the determination of Chinese honey botanical origin.

Conflicts of interest

The authors declare that there are no conflicts of interest.