Characteristic Constituents of Maocangzhu and Beicangzhu Revealed Using Electronic Nose, Electronic Tongue, HS-GC-IMS, and UPLC-Orbitrap Technologies

Abstract

Atractylodis Rhizoma is an important traditional Chinese medicinal material derived from two botanical origins, Maocangzhu (MCZ) and Beicangzhu (BCZ), which are difficult to distinguish by conventional morphological identification because of their similar appearance. However, differences in botanical origin may lead to variations in odor, taste, volatile constituents, and non-volatile metabolites, thereby affecting quality evaluation and clinical application. This study aimed to systematically characterize the sensory and chemical differences between MCZ and BCZ and to identify potential markers for their discrimination. A multi-dimensional analytical strategy combining electronic nose, electronic tongue, headspace gas chromatography–ion mobility spectrometry (HS-GC-IMS), and ultra-high-performance liquid chromatography–Orbitrap high-resolution mass spectrometry (UPLC-Orbitrap MS) was established. Electronic nose and electronic tongue were used to digitize odor and taste characteristics, HS-GC-IMS was employed to profile volatile organic compounds, and UPLC-Orbitrap MS was applied to characterize non-volatile metabolites. Principal component analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA), variable importance in projection (VIP) screening, permutation tests, and correlation analysis were further used to evaluate discrimination performance and screen characteristic markers. The electronic nose results showed that MCZ and BCZ exhibited distinct odor profiles, with W5S, W1W, and W1S identified as the main differential sensors, suggesting that nitrogen oxides, terpenoids, inorganic sulfides, and short-chain alkanes contributed to the odor differences between the two origins. Electronic tongue analysis further demonstrated clear taste discrimination, with sourness and richness identified as the key taste indicators. HS-GC-IMS detected 108 volatile organic compounds, and 24 volatile markers with VIP > 1.2 were screened as important contributors to the differentiation of MCZ and BCZ. Among them, propionic acid and 5-methyl-2-furancarboxaldehyde were mainly distributed in MCZ, whereas (E)-caryophyllene was present only or at higher levels in BCZ, indicating its potential as a characteristic volatile marker of BCZ. UPLC-Orbitrap MS detected 78 non-volatile constituents, and OPLS-DA screened 17 key non-volatile differential metabolites with VIP > 1.2. These results indicated that MCZ and BCZ could be clearly separated not only by sensory signals but also by volatile and non-volatile chemical profiles. This study revealed that the differences between MCZ and BCZ are mainly reflected in odor-active volatile compounds, key taste indicators, and non-volatile differential metabolites. The integration of electronic nose, electronic tongue, HS-GC-IMS, and UPLC-Orbitrap MS provides a comprehensive and reliable strategy for distinguishing the two botanical origins of Atractylodis Rhizoma. These findings provide valuable insights into the material basis underlying the sensory and chemical differences between MCZ and BCZ and offer scientific support for accurate authentication, quality evaluation, and rational clinical application of Atractylodis Rhizoma.

1. Introduction

Atractylodis Rhizoma is an important traditional Chinese medicinal material derived from the dried rhizomes of two Asteraceae species, Atractylodes lancea (Thunb.) DC. and Atractylodes chinensis (DC.) Koidz. In China, A. lancea is commonly known as Maocangzhu and is abbreviated as MCZ, whereas A. chinensis is commonly known as Beicangzhu and is abbreviated as BCZ. MCZ, also known as Nancangzhu, is mainly distributed in Jiangsu, Hubei, and Henan provinces of China, with the Maoshan region in Jiangsu Province traditionally recognized as an authentic production area. BCZ, also known as Shancangzhu, is primarily produced in Jilin, Liaoning, Inner Mongolia and Hebei provinces of China [1]. According to traditional Chinese medicine theory, Atractylodis Rhizoma has the effects of drying dampness, strengthening the spleen, dispelling wind, relieving cold, and improving vision. Modern pharmacological studies have reported that Atractylodis Rhizoma possesses multiple bioactivities, including antitumor [2], antibacterial [3], anti-inflammatory [4], and immunomodulatory effects [5].

Although MCZ and BCZ are both botanical origins of Atractylodis Rhizoma, their chemical compositions may differ because of genetic background, geographical distribution, and growth environment [6]. However, the two herbal materials are highly similar in external morphology, and conventional identification methods based mainly on appearance, odor, and empirical experience are often insufficient for accurate discrimination. This may lead to misidentification or mixed use in herbal markets and clinical practice, thereby affecting quality consistency and clinical efficacy. Therefore, it is necessary to establish reliable analytical methods for distinguishing MCZ from BCZ and to clarify the characteristic constituents responsible for their differences.

Atractylodis Rhizoma has a complex chemical profile, containing volatile constituents such as essential oils and terpenoids, as well as non-volatile constituents such as sesquiterpenes, polysaccharides, phenolic acids, and other secondary metabolites. Conventional analytical methods, such as gas chromatography–mass spectrometry (GC-MS), can effectively characterize essential oils [7]. However, these methods usually require relatively complex sample preparation and longer analysis time, which may limit their suitability for high-throughput rapid screening. Headspace gas chromatography–ion mobility spectrometry (HS-GC-IMS) combines chromatographic separation with ion mobility detection, enabling relatively rapid analysis of volatile organic compounds (VOCs) with simplified sample pretreatment. It is particularly suitable for detecting small-molecule VOCs and has been increasingly applied in the analysis of traditional Chinese medicines [8].

In addition to chemical profiling, sensory characteristics such as odor and taste are important features traditionally used in the identification and quality evaluation of Chinese medicinal materials. Electronic nose and electronic tongue technologies provide objective, reproducible, and digitalized evaluation of odor and taste, helping reduce the subjectivity associated with manual sensory assessment. These biomimetic techniques have shown considerable potential in quality control and authenticity evaluation of traditional Chinese medicines [9]. Meanwhile, Orbitrap mass spectrometry, characterized by high resolution, broad spectral coverage, and excellent mass accuracy, enables reliable multistage mass spectrometric analysis of complex samples and has been widely used for chemical component characterization [10,11].

Given the complexity of Atractylodis Rhizoma, a single analytical technique is insufficient to comprehensively reveal the differences between MCZ and BCZ. Therefore, the aim of this study was to systematically compare the two botanical origins by integrating electronic nose, electronic tongue, HS-GC-IMS, and UPLC-Orbitrap MS from sensory, volatile, and non-volatile perspectives, and to identify the characteristic constituents distinguishing MCZ from BCZ, thereby providing a scientific basis for the accurate authentication, quality evaluation, and rational utilization of Atractylodis Rhizoma.

2. Results

2.1. HS-GC-IMS Analysis of Volatile Metabolites

2.1.1. Identification of Major Volatile Metabolites in Atractylodes Samples Based on HS-GC-IMS

Figure 1 presents the three-dimensional (3D) surface characteristics of volatile metabolites identified in Atractylodis Rhizoma samples from different sources. The X, Y, and Z axes of the 3D plot correspond to migration time, retention time, and peak intensity, respectively. The 3D plots reveal pronounced differences in the types and concentrations of volatile organic compounds between BCZ and MCZ. For clearer comparison among samples, a differential comparison mode was used: each point represents a specific volatile organic compound, and the color intensity reflects concentration, with darker shades indicating higher concentrations. The spectrum of BCZ was set as the reference, with the spectrum of MCZ subtracted from it. Red signifies a higher concentration than the reference, while blue signifies a lower concentration. The results from the differential comparison spectrum (Figure 2) confirm differences in both component types and concentrations between MCZ and BCZ, consistent with the observations from the 3D spectrum.

Figure 1.

3D gas-phase ion mobility spectrum comparison.

Figure 2.

Differential comparison spectra.

Following background subtraction, chromatographic peak extraction, and peak alignment of the GC-IMS data from MCZ and BCZ samples, a total of 108 active components were identified and qualitatively analyzed (Table 1). These components primarily include esters, aldehydes, ketones, alcohols, and acids. Specifically, there were 13 aldehydes (12.04%), 14 ketones (12.96%), 21 alcohols (19.44%), 23 esters (21.30%), 4 carboxylic acids (3.70%), 8 phenols (7.41%), 6 alkenes (5.56%), 7 nitrogen-containing compounds (6.48%), 4 sulfur-containing compounds (3.70%), and 8 other compounds (7.41%). Some of these compounds are present as both monomers and dimers.

Table 1.

Identification of volatile metabolites in Atractylodes of different origins using gas-phase ion mobility spectrometry.

Class	Compound	CAS	Formula	MW	RI	Rt [sec]	Dt [a.u.]	Differential Components
Alcohol	Geraniol	106-24-1	C10H18O	154.3	1837.9	1136.69	1.21
	(Z)-4-Decen-1-ol	57074-37-0	C10H20O	156.3	1786.6	1061.73	1.44
	Propylene glycol	57-55-6	C3H8O2	76.1	1586.9	814.13	1.27
	Cedrol	77-53-2	C15H26O	222.4	1555.7	780.98	1.43
	Linalool oxide	1365-19-1	C10H18O2	170.3	1465.3	692.55	1.67
	2-Hexen-1-ol	2305-21-7	C6H12O	100.2	1395.0	628.38	1.19
	(E)-3-Hexen-1-ol	928-97-2	C6H12O	100.2	1350.9	542.58	1.25	C5
	α-Terpineol	98-55-5	C10H18O	154.3	1283.5	433.41	1.22	C8
	2-Butyl-1-octanol	3913-02-8	C12H26O	186.3	1277.2	424.23	1.67
	4-Methoxybenzyl alcohol	105-13-5	C8H10O2	138.2	1244.7	380.09	1.08	C11
	cis-3,7-Dimethyl-octa-2,6-dien-1-ol-M(FCC)	624-15-7	C10H18O	154.3	1219.4	349.03	1.22
	FCC dimer	624-15-7	C10H18O	154.3	1217.2	346.37	1.64	C1
	3,7-Dimethyl-1-octanol	106-21-8	C10H22O	158.3	1200.1	327.03	1.57
	(–)-4-Terpineol	20126-76-5	C10H18O	154.3	1178.5	304.86	1.22
	1-Penten-3-ol	616-25-1	C5H10O	86.1	1155.0	282.94	0.94
	2-Pentanol	6032-29-7	C5H12O	88.1	1153.9	281.93	1.44	C21
	β-Linalool	78-70-6	C10H18O	154.3	1098.7	236.61	1.21
	1,8-Cineole	470-82-6	C10H18O	154.3	1027.2	203.95	1.29
	2-Hexanol	626-93-7	C6H14O	102.2	966.5	179.99	1.28
	Isopropyl alcohol	67-63-0	C3H8O	60.1	942.9	171.44	1.09
	2-Butoxyethanol	111-76-2	C6H14O2	118.2	895.0	155.34	1.20
Aldehyde	(E)-2-Decenal	3913-81-3	C10H18O	154.3	1635.7	868.64	1.47
	(E,Z)-2,6-Nonadienal	557-48-2	C9H14O	138.2	1583.7	810.64	1.37
	5-Methylfurfural	620-02-0	C6H6O2	110.1	1555.7	780.97	1.13
	Undecanal	112-44-7	C11H22O	170.3	1316.3	483.49	1.594
	3,7-Dimethyl-2,6-octadienal	5392-40-5	C10H16O	152.2	1267.6	410.70	1.042
	Cuminaldehyde	122-03-2	C10H12O	148.2	1242.5	377.25	1.330
	(E)-2-Hexenal	6728-26-3	C6H10O	98.1	1207.2	334.86	1.173	C18
	4-Ethylbenzaldehyde	4748-78-1	C9H10O	134.2	1199.9	326.74	1.271
	(E)-2-Octenal	2548-87-0	C8H14O	126.2	1024.5	202.82	1.328
	3-Methylbutanal	590-86-3	C5H10O	86.1	921.3	164.00	1.194	C6
	(E,E)-2,4-Hexadienal	142-83-6	C6H8O	96.1	907.6	159.44	1.114
	(Z)-4-Heptenal	6728-31-0	C7H12O	112.2	881.9	151.20	1.145
	Furfural	98-01-1	C5H4O2	96.1	845.2	140.21	1.089
Ketone	2,4-Dimethylcyclopentan-1,3-dione	13494-06-9	C7H10O2	126.2	1900.2	1234.98	1.204
	2-Hydroxy-3-methyl-2-cyclopenten-1-one	80-71-7	C6H8O2	112.1	1834.6	1131.77	1.520	C22
	3-Methyl-2(5H)-furanone	22122-36-7	C5H6O2	98.1	1798.0	1078.04	1.381
	3-Methyl-2-cyclopenten-1-one	2758-18-1	C6H8O	96.1	1516.2	741.06	1.109
	2-Nonanone	821-55-6	C9H18O	142.2	1414.0	646.87	1.405
	6-Methyl-5-hepten-2-one	110-93-0	C8H14O	126.2	1335.2	514.89	1.184
	Hydroxyacetone	116-09-6	C3H6O2	74.1	1299.0	456.56	1.043
	Carvone (monomer)	2244-16-8	C10H14O	150.2	1261.2	401.89	1.303	C2
	Carvone (dimer)	2244-16-8	C10H14O	150.2	1260.1	400.37	1.831
	2-Hexanone	591-78-6	C6H12O	100.2	1106.3	242.39	1.191
	2,3-Pentanedione	1629-58-9	C5H8O	84.1	1005.0	194.83	1.075
	2-Octanone	111-13-7	C8H16O	128.2	982.5	186.01	1.335	C24
	2-Butanone	78-93-3	C4H8O	72.1	941.1	170.82	1.059	C13
	Cyclopentanone	120-92-3	C5H8O	84.1	810.1	130.42	1.102
Carboxylic acid	(E)-3-Hexenoic acid	1577-18-0	C6H10O2	114.1	1955.5	1329.31	1.229	C20
	Heptanoic acid	111-14-8	C7H14O2	130.2	1918.8	1265.87	1.369
	Propanoic acid	79-09-4	C3H6O2	74.1	1491.6	717.12	1.105
	Hexanoic acid	142-62-1	C6H12O2	116.2	1001.0	193.22	1.293
Ester	γ-Decalactone	706-14-9	C10H18O2	170.3	2119.5	1653.17	1.494
	Benzyl phenylacetate	140-11-4	C16H16O2	240.3	1713.1	962.89	1.326
	Citronellyl formate	105-85-1	C11H20O2	184.3	1615.7	845.86	1.335
	Furfuryl acetate	623-17-6	C7H8O3	140.1	1535.5	760.28	1.411	C9
	Propylene glycol diacetate	623-84-7	C7H12O4	160.2	1487.6	713.36	1.216
	Ethyl 4-methoxybenzoate	94-30-4	C10H12O3	180.2	1405.9	639.93	1.331
	Benzyl isobutyrate (dimer)	103-48-0	C12H16O2	192.3	1390.6	619.09	1.976
	Benzyl isobutyrate (monomer)	103-48-0	C12H16O2	192.3	1390.0	617.96	1.393
	Linalyl isobutyrate (monomer)	78-35-3	C14H24O2	224.3	1355.0	549.96	1.218	C3
	Linalyl isobutyrate (dimer)	78-35-3	C14H24O2	224.3	1351.9	544.34	1.696
	Ethyl 2-hydroxypropanoate	97-64-3	C5H10O3	118.1	1337.1	518.22	1.133
	Methyl anthranilate	134-20-3	C8H9NO2	151.2	1297.5	454.17	1.264	C7
	Methyl heptanoate	106-73-0	C8H16O2	144.2	1297.1	453.62	1.360
	Linalyl acetate	115-95-7	C12H20O2	196.3	1260.9	401.50	1.212
	β-Methyl-γ-octalactone	39212-23-2	C9H16O2	156.2	1201.2	328.21	1.371
	1-Phenylethyl acetate	93-92-5	C10H12O2	164.2	1192.4	318.60	1.049	C14
	Methyl phenylacetate	101-41-7	C9H10O2	150.2	1154.0	282.03	1.255
	2-Methylbutyl acetate	624-41-9	C7H14O2	130.2	1128.7	260.22	1.292
	Tetrahydrofurfuryl acetate	637-64-9	C7H12O3	144.2	1099.1	236.91	1.180
	3-Methylbutyl 2-methylbutanoate	27625-35-0	C10H20O2	172.3	1098.6	236.52	1.423
	γ-Hexalactone	695-06-7	C6H10O2	114.1	1064.6	220.29	1.185
	Prenyl acetate	1191-16-8	C7H12O2	128.2	932.5	167.82	0.944
	Butyl propionate	590-01-2	C7H14O2	130.2	893.6	154.88	1.282
Alkene	(E,E)-α-Farnesene	502-61-4	C15H24	204.4	1735.7	992.26	1.435	C10
	(E)-Caryophyllene	87-44-5	C15H24	204.4	1383.1	603.82	1.424	C19
	Limonene	138-86-3	C10H16	136.2	1233.6	366.14	1.218
	Naphthalene	91-20-3	C10H8	128.2	1155.6	283.44	1.113
	δ-3-Carene	13466-78-9	C10H16	136.2	1143.1	272.39	1.216	C16
	α-Phellandrene	99-83-2	C10H16	136.2	1014.0	198.49	1.221
Phenol	Methyl eugenol	93-15-2	C11H14O2	178.2	2040.9	1489.05	1.443	C15
	Butylated hydroxytoluene (monomer)	128-37-0	C15H24O	220.4	1517.8	742.64	1.329
	Butylated hydroxytoluene (dimer)	128-37-0	C15H24O	220.4	1496.9	722.25	1.331	C17
	Ethyl vanillin	121-32-4	C9H10O3	166.2	1428.7	659.58	1.278
	4-vinyl-2-methoxyphenol (monomer)	7786-61-0	C9H10O2	150.2	1319.6	488.95	1.218
	4-vinyl-2-methoxyphenol (dimer)	7786-61-0	C9H10O2	150.2	1319.2	488.22	1.732
	4-Vinylguaiacol	7786-61-0	C9H10O2	150.2	1290.7	443.97	1.218	C12
	Methyl salicylate	119-36-8	C8H8O3	152.1	1163.4	290.54	1.204
N-containing	2-Methoxy-3-sec-butylpyrazine	24168-70-5	C9H14N2O	166.2	1516.4	741.24	1.249	C23
	2-Ethyl-3-methylpyrazine	15707-23-0	C7H10N2	122.2	1434.0	664.25	1.161
	Pyrazine	290-37-9	C4H4N2	80.1	1211.5	339.85	1.044
	N-Ethyl-2-pyrrolidone	2687-91-4	C6H11NO	113.2	1152.9	281.06	1.154
	N,N-Dimethylaniline	121-69-7	C9H13N	135.2	1015.9	199.24	1.225
	2-Ethylpyridine	100-71-0	C7H9N	107.2	969.1	180.94	1.098
	1-Methyl-1H-pyrrole-2-carboxaldehyde	1192-58-1	C6H7NO	109.1	966.1	179.85	1.121
S-containing	5-Methyl-2-thiophenecarboxaldehyde	13679-70-4	C6H6OS	126.2	1129.1	260.60	1.179
	Ethyl thioacetate	625-60-5	C4H8OS	104.2	1117.4	251.12	1.131
	2,5-Dimethylthiophene	638-02-8	C6H8S	112.2	885.6	152.36	1.070
	Dimethyl disulfide	624-92-0	C2H6S2	94.2	737.7	112.35	1.140
Other	N,N-Dimethylacetamide	127-19-5	C4H9NO	87.1	1434.2	664.43	1.063
	N,N-Dimethylformamide	68-12-2	C3H7NO	73.1	1372.6	583.21	1.252	C4
	Dihydrocoumarin	119-84-6	C9H8O2	148.2	1352.3	545.04	1.287
	Anethole	4180-23-8	C10H12O	148.2	1288.2	440.29	1.761
	Triethyl phosphate	78-40-0	C6H15O4P	182.2	1138.1	268.17	1.299
	Octamethylcyclotetrasiloxane	556-67-2	C8H24O4Si4	296.6	1004.1	194.49	1.676
	2-Pentylfuran	3777-69-3	C9H14O	138.2	1001.7	193.51	1.246
	4-Methylanisole	104-93-8	C8H10O	122.2	993.5	190.29	1.116

2.1.2. Analysis of Major Volatile Metabolites in Atractylodes Samples Using HS-GC-IMS

2-Hexen-1-ol, geraniol, and α-Terpinolene are major alcohol-type volatile metabolites in Atractylodes (Figure 3b). 2-Hexen-1-ol is a nearly colorless liquid with a strong odor reminiscent of unripe fruit. Studies have shown that 2-hexen-1-ol exhibits inhibitory effects against Fusarium species infecting stored corn grains [12]. Geraniol, a monoterpenol with a rose-like fragrance, is widely used in cosmetics and high-end perfumes; it also demonstrates antibacterial [13], anticancer [14], neuroprotective [15,16], and antioxidant activities. α-Terpineol is a monocyclic monoterpene alcohol commonly found in the essential oils of various aromatic plants. It has been reported to possess a broad spectrum of biological activities, including anticancer, anti-inflammatory, and analgesic properties. α-Terpineol has been demonstrated to exert anticancer effects by suppressing NF-κB signaling and inducing apoptosis in tumor cells [17]. Moreover, its antinociceptive and anti-inflammatory effects have been validated in neuropathic pain models, where it significantly attenuated mechanical allodynia and hyperalgesia through the inhibition of spinal microglial activation and inflammatory cytokines [18].

Figure 3.

Composition of major volatile metabolites. (a) Pie chart of relative volatile metabolite content; (b) alcohols; (c) esters; (d) aldehydes.

Other bioactive compounds have also been identified among the volatile alcohol metabolites of Atractylodes, contributing to its complex and unique chemical profile and biological functions.

Linalyl acetate, Methyl anthranilate, ethyl 4-methoxybenzoate, and Linalyl isobutyrate together accounted for 50.6% of all ester compounds detected in this study (Figure 3c). Linalyl acetate is a monoterpene compound and the acetate ester of linalool, constituting a major component of bergamot and lavender essential oils. Its biological activity is closely related to that of linalool, and it may help alleviate psoriasiform inflammation by interfering with IL-12-driven inflammatory responses, mainly acting on Th-1 cell-specific cytokines [19]. Methyl anthranilate may inhibit the quorum sensing system in Pseudomonas aeruginosa by interfering with AHL biosynthesis and competitively binding to receptor proteins. It shows promise as a quorum sensing inhibitor and anti-biofilm agent, with potential for improving food safety [20]. Ethyl 4-methoxybenzoate is a colorless, oily liquid with a sweet, anise-like aroma. Occurring naturally in apple brandy, plums, and white wine, it is primarily used as a fragrance ingredient in cosmetics and perfumes. Safety evaluations indicate it poses no health risks, such as genotoxicity or skin sensitization, at current usage levels and is safe for the aquatic environment [21]. Linalyl isobutyrate is a colorless to pale yellow liquid with a sweet, fresh, rose-like scent, widely used as a fragrance ingredient in perfumes, cosmetics, and personal care products [22].

4-ethylbenzaldehyde and cuminaldehyde are primary aldehyde components. 4-ethylbenzaldehyde is a colorless to pale yellow liquid with a sweet, bitter almond scent. It occurs naturally in various foods such as prickly ash, dairy products, and peanuts, and is used as a fragrance ingredient in products like lip care items, perfumes, and air fresheners [23]. Cuminaldehyde, also a colorless to pale yellow liquid, has a strong, pungent, spicy, green herbal odor. As a bioactive component of essential oils, cuminaldehyde shows potential for analgesic, anti-inflammatory, and antibacterial applications [24,25,26]. Cuminaldehyde exerts analgesic and neuroprotective effects via opioid receptors, the L-arginine/NO/cGMP/KATP signaling pathway, and anti-inflammatory functions. It can modulate immune responses in inflammation and may be considered a lead compound for novel anti-inflammatory and analgesic drug development [24,25].

2.1.3. HS-GC-IMS-Based Volatile Organic Compound Fingerprint Analysis of Atractylodes Samples

To further examine the differences between BCZ and MCZ, the Gallery Plot plugin was employed for a visual comparison of volatile organic compounds among various samples. All peaks were included for fingerprint analysis; in the figure, each row corresponds to all signal peaks from a single sample, and each column represents the signal peaks of the same volatile organic compound across different samples. Fingerprint spectra were constructed for BCZ and MCZ, as shown in Figure 4. A deeper red indicates a higher concentration of that component. The samples display both unique characteristic signals and shared common signals.

Figure 4.

Component fingerprint profiles.

As illustrated in Figure 4, the components in regions A, C, and E are mainly concentrated in MCZ samples, including propionic acid and 5-methyl-2-furancarboxaldehyde, which may serve as volatile markers to differentiate BCZ from MCZ. The volatile profiles of batches 3 and 4 of BCZ are relatively similar in region A, while region E highlights the similarity between batches 1 and 2. Components in region B are present at higher levels in BCZ compared to MCZ. Region D includes 42 compounds common to both species. Region F comprises components found exclusively or at higher concentrations in MCZ; among these, (E)-Caryophyllene is unique to BCZ, consistent with previous literature [27], and can serve as a key marker for distinguishing between the two species.

2.1.4. Multivariate Statistical Analysis Based on HS-GC-IMS

The PCA scatter plot (see Figure 5) shows no clear separation between MCZ and BCZ samples, indicating that the data (variables and samples) used in this analysis were insufficient for PCA to clearly demonstrate the differences between these two categories. In contrast, the OPLS-DA scatter plot demonstrates a clear separation trend, suggesting that this model can reliably differentiate between MCZ and BCZ. Model assessment revealed that the cumulative explanatory power of independent variables (R²X) was 0.749, the goodness-of-fit for dependent variables (R²Y) was 0.994, and the predictive capability parameter (Q²) was 0.977. Both R² and Q² values are close to 1, reflecting excellent model fit and predictive reliability. The model’s robustness was further confirmed by a 200-permutation test, yielding a Q² intercept of −1.15 (less than 0), indicating the model is not overfitted and remains stable and valid.

Figure 5.

PCA and OPLS-DA based on HS-GC-IMS data. (a) PCA score plot; (b) OPLS-DA score plot; (c) permutation test; (d) VIP values.

Using a VIP > 1.2 criterion, 24 volatile biomarkers were identified as significant contributors to differentiation between the two sample types, including: FCC (dimer), carvone (dimer), linalyl isobutyrate (monomer), dimethylformamide (DMF), (E)-3-hexen-1-ol, 3-methylbutyraldehyde, methyl anthranilate, α-terpineol, acetyl furfuryl ester, (E,E)-α-Farnesene, 4-methoxybenzyl alcohol, 4-vinylguaiacol, 2-butanone, 1-phenylethyl acetate, eugenol methyl ether, 3-δ-carene, 2,6-di-tert-butyl-4-methylphenol (monomer), (E)-2-hexenal, (E)-Caryophyllene, (E)-3-hexenoic acid, 2-pentanol, 2-hydroxy-3-methyl-2-cyclopent-1-enone, 2-methoxy-3-sec-butylpyrazine, and 2-Octanone.

2.2. UPLC-Orbitrap Analysis of Non-Volatile Metabolites

2.2.1. Identification of Major Non-Volatile Metabolites in Atractylodes Samples Using UPLC-Orbitrap

High-resolution mass spectrometry (HRMS) analysis was conducted using ESI, alternating between positive and negative ion modes, with a full MS/ddMS2 scan mode (Figure 6). Compound identification was achieved by comparing characteristic ion fragmentation patterns against in-house databases (mzCloud and mzVault), the PubChem online database (https://pubchem.ncbi.nlm.nih.gov/), and literature data. A total of 78 chemical constituents were identified (Table 2), including terpenes (15), organic acids and their esters (38), coumarins (4), lignans (2), flavonoids (2), sugars and glycosides (8), amides (5), and other components (4).

Figure 6.

Total ion current chromatogram of non-volatile metabolites.

Table 2.

Chemical composition of non-volatile metabolites in Atractylodes.

Class	Name	Formula	DeltaMass [ppm]	MW	m/z	RT [min]	Fragment Ions
Flavonoids	Nobiletin	C21H22O8	−1.67	402.13	403.14	16.27	403.14, 388.11
	Daidzein	C15H10O4	−1.22	254.06	255.06	9.64	255.06, 227.07
Lignans	Matairesinol	C20H22O6	−1.46	358.14	359.15	9.64	323.12, 291.1
	(3R,4R)-4-[(3,4-dimethoxyphenyl)methyl]-3-[(4-hydroxy-3-methoxyphenyl)methyl]oxolan-2-one	C21H24O6	−1.49	372.16	373.16	11.39	306.12, 237.11
Others	2,4-di-tert-Butylphenol	C14H22O	−0.3	206.17	205.16	26.42	206.16, 205.16
	trans-3-Indoleacrylic acid	C11H9NO2	−1.81	187.06	188.07	3.1	188.07, 170.07
	1,2,3,4-Tetramethyl-1,3-cyclopentadiene	C9H14	−2	122.11	123.12	26.99	123.12, 95.09
	(±)-Cannabichromeorcin	C17H22O2	−0.84	258.16	259.17	14.8	259.17, 203.11
Sugars and their glycosides	Gluconic acid	C6H12O7	−0.79	196.06	195.05	1.07	195.05, 177.04
	α,α-Trehalose	C12H22O11	−0.79	342.12	341.11	1.13	341.11, 179.06
	Puerarin	C21H20O9	−0.84	416.11	417.12	4.17	381.1, 363.09
	3-Ethyl-4-hydroxy-4-methylpentyl 6-O-[(2S,3R,4R)-3,4-dihydroxy-4-(hydroxymethyl)tetrahydro-2-furanyl]-β-D-glucopyranoside	C19H36O11	−0.17	440.23	439.22	5.42	307.18, 161.05
	5′-S-Methyl-5′-thioadenosine	C11H15N5O3S	−1.44	297.09	298.1	4.32	136.06, 61.01
	D-(−)-Fructose	C6H12O6	−0.94	180.06	179.06	1.08	179.07, 161.05
	2′-O-Methyladenosine	C11H15N5O4	−1.71	281.11	282.12	3.23	283.17, 282.12
	8-Hydroxy-6-methoxy-2-oxo-2H-chromen-7-yl β-D-glucopyranoside	C16H18O10	−0.37	370.09	369.08	3.8	354.06, 207.03
Terpenoids	Nootkatone	C15H22O	−2.18	218.17	219.17	12.51	219.17, 163.11
	(−)-Caryophyllene oxide	C15H24O	−1.48	220.18	221.19	17.97	221.19, 203.18
	(3aS,5aR,6R,9aS,9bS)-6-hydroxy-5a-methyl-3,9-dimethylidene-dodecahydronaphtho[1,2-b]furan-2-one	C15H20O3	−1.45	248.14	249.15	10.71	249.15, 231.14
	(3aR,5aR,9bR)-3a-hydroxy-5a,9-dimethyl-3-methylidene-2H,3H,3aH,4H,5H,5aH,6H,7H,8H,9bH-naphtho[1,2-b]furan-2-one	C15H20O3	−1.06	248.14	249.15	14.52	249.15, 231.14
	(3aR,4aS,5R,7aS,8S,9aR)-5-Hydroxy-4a,8-dimethyl-3-methyleneoctahydroazuleno[6,5-b]furan-2,6(3H,4H)-dione	C15H20O4	−1.57	264.14	265.14	4.8	265.14, 229.12
	(3aR,7aS,8S,9aR)-5,8-dimethyl-3-methylidene-2H,3H,3aH,4H,6H,7H,7aH,8H,9H,9aH-azuleno[6,5-b]furan-2,6-dione	C15H18O3	−1.18	246.13	247.13	12.69	247.13, 229.12
	Ambrosic acid	C15H20O4	−1.54	264.14	265.14	7.37	265.14, 229.12
	6-hydroxy-3,5a,9-trimethyl-2H,3H,3aH,4H,5H,5aH,6H,7H,9aH,9bH-naphtho[1,2-b]furan-2-one	C15H22O3	−1.51	250.16	251.16	11.42	233.15, 187.15
	Zedoarondiol	C15H24O3	−1.38	252.17	253.18	9.47	235.17, 217.16
	D,L-Camphor	C10H16O	−1.69	152.12	153.13	8.29	153.13, 135.09
	(5E)-7-methylidene-10-oxo-4-(propan-2-yl)undec-5-enoic acid	C15H24O3	−1.41	252.17	253.18	8.34	235.17, 217.16
	Atractylone	C15H20O	−1.74	216.15	217.16	18.62	217.16, 175.15
	Atractylenolide IV	C17H22O5	−2.12	306.15	307.15	16.31	290.15, 215.11
	Atractylodinol	C13H10O2	−3.01	198.08	199.07	19.06	290.15, 215.11
	Muscone	C16H30O	−1.86	238.23	239.24	28.75	239.24, 151.15
Amides	Hexadecanamide	C16H33NO	−2.31	255.26	256.26	31.62	257.27, 116.11
	Oleamide	C18H35NO	−1.7	281.27	282.28	28.03	282.28, 149.13
	Stearamide	C18H37NO	−0.94	283.29	284.29	33.11	284.29, 172.17
	Erucamide	C22H43NO	−1.98	337.33	338.34	31.59	338.34, 303.31
	Linoleoyl ethanolamide	C20H37NO2	−2.06	323.28	324.29	30.17	306.28, 161.13
Coumarins	Scopoletin	C10H8O4	−1.48	192.04	191.03	5.16	193.05, 178.03
	Scoparone	C11H10O4	−2.15	206.06	207.06	6.43	207.06, 192.04
	4-Hydroxycoumarin	C9H6O3	−0.72	162.03	161.02	5.23	161.02, 162.03
	Fraxetin	C10H8O5	−1.44	208.04	209.04	4.52	209.04, 194.02
Organic acids and their esters	DL-Malic acid	C4H6O5	−0.55	134.02	133.01	1.44	133.01, 115
	DL-Arginine	C6H14N4O2	−1.8	174.11	175.12	1.66	175.12, 158.09
	D-(+)-Tryptophan	C11H12N2O2	−0.65	204.09	203.08	3.1	203.08, 159.09
	Proline	C5H9NO2	−2.36	115.06	116.07	1.09	116.07, 117.07
	Isoleucine	C6H13NO2	−1.74	131.09	132.1	1.53	132.1, 87.09
	Salicylic acid	C7H6O3	−0.46	138.03	137.02	3.51	138.03, 137.02
	4-Feruloylquinic acid	C17H20O9	−0.1	368.11	367.1	3.35	193.05, 173.05
	Benzoic acid	C7H6O2	−0.27	122.04	121.03	3.46	121.03, 122.02
	D-(−)-Quinic acid	C7H12O6	−0.37	192.06	191.06	2.37	191.06, 173.05
	α-Eleostearic acid	C18H30O2	−2.22	278.22	279.23	27.23	279.23, 173.13
	4,5-Dicaffeoylquinic acid	C25H24O12	−0.1	516.13	515.12	4.96	353.1, 191.06
	Picolinic acid	C6H5NO2	−0.91	123.03	122.02	3.16	122.02, 94.03
	Neochlorogenic acid	C16H18O9	−0.32	354.09	353.09	4.71	191.06, 179.03
	Chlorogenic acid	C16H18O9	−0.31	354.09	353.09	2.48	191.06, 179.04
	3-Furoic acid	C5H4O3	−0.81	112.02	111.01	1.23	111.01, 67.02
	Catechol	C6H6O2	−0.44	110.04	109.03	3.46	109.03, 108.02
	Pentadecanoic acid	C15H30O2	−0.64	242.22	241.22	31.68	241.22, 129.23
	16-Hydroxyhexadecanoic acid	C16H32O3	−0.44	272.24	271.23	27.1	271.23, 235.22
	(+/−)9,10-dihydroxy-12Z-octadecenoic acid	C18H34O4	0.1	314.25	313.24	24.35	313.24, 295.23
	Caffeic acid	C9H8O4	−0.33	180.04	179.03	2.88	179.03, 135.05
	Nervonic acid	C24H46O2	−0.74	366.35	365.34	32.28	365.34, 366.35
	9-Oxo-10(E),12(E)-octadecadienoic acid	C18H30O3	−0.86	294.22	295.23	26.65	277.22, 165.13
	Vanillic acid	C8H8O4	−0.52	168.04	167.03	2.71	167.03, 152.01
	Syringic acid	C9H10O5	−0.54	198.05	197.05	2.55	197.05, 182.02
	(±)13-HODE	C18H32O3	−0.91	296.23	295.23	28.09	295.23, 277.22
	8Z,11Z,14Z-Eicosatrienoic acid	C20H34O2	−0.38	306.26	305.25	32	305.25, 306.35
	Corchorifatty acid F	C18H32O5	0.38	328.23	327.22	14.56	327.22, 239.13
	(±)9-HpODE	C18H32O4	0.12	312.23	311.22	23.23	311.22, 293.21
	1-Linoleoyl glycerol	C21H38O4	−1.87	354.28	355.28	30.81	285.01, 263.24
	13(S)-HOTrE	C18H30O3	−0.81	294.22	295.23	23.18	277.22, 165.13
	Methyl salicylate	C8H8O3	−0.35	152.05	151.04	6.94	151.04, 136.02
	Citric acid	C6H8O7	−0.55	192.03	191.02	1.44	191.06, 129.02
	2-Methylbenzoic acid	C8H8O2	−0.57	136.05	135.05	6.6	136.05, 135.05
	3-Hydroxypicolinic acid	C6H5NO3	−1.44	139.03	140.03	4.26	183.08, 165.07
	(15Z)-9,12,13-Trihydroxy-15-octadecenoic acid	C18H34O5	−0.22	330.24	329.23	16.01	329.23, 311.22
	trans-10-Heptadecenoic acid	C17H32O2	−0.05	268.24	267.23	31.95	267.23, 249.22
	Methyl cinnamate	C10H10O2	−2.18	162.07	163.08	15.91	163.07, 131.05
	9-Oxo-ODE	C18H30O3	−1.03	294.22	295.23	28.67	277.22, 165.13

2.2.2. Analysis of Major Non-Volatile Metabolites in Atractylodes Samples Using UPLC-Orbitrap

Terpenoids are the signature metabolites of Atractylodes. Atractylone is a characteristic compound of Atractylodes from authentic production areas, with demonstrated antitumor and anti-inflammatory properties [28,29]. Zedoarondiol has demonstrated anti-inflammatory effects [30] and anti-atherosclerotic activity [31]. Caryophyllene oxide exhibits multiple pharmacological activities, including anti-inflammatory, antioxidant, and antitumor effects [32,33,34,35]. Nootkatone (NKT) displays a range of biological activities, such as anti-inflammatory, anticancer, antibacterial, and neuroprotective effects [36,37,38,39,40].

Organic acids and their esters: Quinic acid demonstrates a variety of pharmacological activities, including antioxidant, antibacterial, anti-inflammatory, antiviral, and antitumor effects [41,42]. Chlorogenic acid also exhibits significant biological activities, such as anti-inflammatory, immunomodulatory, glucose- and lipid-lowering, antibacterial, and antitumor effects [43,44,45,46,47], making it highly applicable in medicine, healthcare, food, cosmetics, and as an alternative to antibiotics in animal feed. Arginine displays immunomodulatory, antioxidant, and lipid-lowering properties, as well as anti-inflammatory effects [48,49,50].

Coumarins and flavonoids: The coumarin compounds scopoletin and scoparone possess anticoagulant and vasoprotective effects [51,52,53,54]. The flavonoids daidzein and puerarin are notable for their antioxidant and anti-inflammatory properties, garnering significant scientific interest [55,56].

The active metabolites of Atractylodes are mainly terpenoids (zedoarondiol, atractylone, and NKT), alongside phenolic acids (chlorogenic acid, caffeic acid), coumarins (scopoletin, scoparone), and flavonoid glycosides (puerarin). Together, these compounds form the pharmacological basis for the herb’s traditional effects of “strengthening the spleen, drying dampness, and providing anti-inflammatory and hepatoprotective actions”. These results are highly consistent with modern pharmacological interpretations of Atractylodes’ traditional uses [57], supporting its application in metabolic and digestive system disorders.

2.2.3. Multivariate Statistical Analysis Based on UPLC-Orbitrap

As shown in Figure 7a, although the PCA score plot can preliminarily distinguish between the two types of Atractylodes samples, its cumulative explanatory power is only 50.4%, indicating that PCA alone has limited ability to differentiate between varieties. In contrast, the OPLS-DA score plot in Figure 7b shows a strict separation boundary along the predicted component axis (t[1]): all BCZ samples (BCZ1–BCZ6) are distributed on the right, while all MCZ samples (MCZ1–MCZ6) are on the left, with no overlap, visually confirming the model’s high discriminatory capability. The OPLS-DA model (M3) explains 64.9% of the variance in the original data (R²X = 0.649), achieves a near-perfect fit for the classification labels (R²Y = 0.999), and its predictive power (Q² = 0.946) markedly surpasses that of PCA (Q² = 0.214) and PLS-DA (Q² = 0.870). After a 200-permutation test, the R² regression line intercept (0.969) was much lower than the original, and the Q² regression line intercept (−0.591) was below zero, confirming that the model is valid and not overfitted as Figure 7c.

Figure 7.

PCA and OPLS-DA based on UPLC-Orbitrap data. (a) PCA score plot; (b) OPLS-DA score plot; (c) Permutation test; (d) VIP values.

In this study, an OPLS-DA model was established to distinguish Atractylodes varieties based on non-volatile metabolites. The model’s performance was systematically validated through key statistical parameters and visualization. Furthermore, 17 non-volatile metabolites were identified as significant differentiators using UPLC-Orbitrap data with a VIP > 1.2 criterion as Figure 7d.

2.3. Electronic Nose Analysis Results

2.3.1. Odor Composition Analysis of Northern Atractylodes and Maocang Atractylodes

The average response values of electronic nose sensors for MCZ and BCZ samples were analyzed, and a radar chart was generated to visually illustrate the differences between the samples. As shown in Figure 8, all 10 sensors responded to the samples, with sensors W5S and W1W showing particularly strong responses. This suggests that both MCZ and BCZ contain higher levels of nitrogen oxides, terpenes, and inorganic sulfides, which are likely the main contributors to the odor differences between these species.

Figure 8.

Radar plots of Atractylodes sensor responses. (a) Radar plot of MCZ sensor responses; (b) radar plot of BCZ sensor responses.

2.3.2. Analysis of Contribution Rates to Sensor Discrimination via Loadings

Loadings analysis distinguishes the volatile components that play a primary role in each sample. As shown in Figure 9, the contribution rate of sensor response values for PC1 is 68.45% and for PC2 is 16.40%, with a cumulative contribution rate of 84.85%. Among these, W5S (Sensor No. 2) contributed most to discriminating the first principal component, while W1W (Sensor No. 7) contributed most to the second principal component. Based on the response characteristics, the first principal component is mainly sensitive to nitrogen oxides, and the second principal component to terpenes and inorganic sulfides. These findings are consistent with the earlier sensor response analysis.

Figure 9.

Analysis of sensor discrimination contribution rates.

2.3.3. Multivariate Statistical Analysis Based on the Electronic Nose

Multivariate statistical analysis was performed using SIMCA 14.1 software for OPLS-DA analysis, with VIP > 1 as the screening criterion. Additionally, 200 bootstrap tests were conducted on the OPLS-DA models. As shown in Figure 10, after bootstrapping, the R² and Q² values on the right were higher than the original points on the left, indicating that the models did not suffer from overfitting and could reliably discriminate between BCZ and MCZ. The results show that the two species cluster into distinct groups: MCZ is mainly distributed in the first and fourth quadrants, and BCZ in the first and third quadrants. Using VIP > 1, three differentially responsive sensors (W1W, W5S, W1S) were identified.

Figure 10.

OPLS-DA based on electronic nose data. (a) OPLS-DA score plot; (b) permutation test; (c) VIP values.

Analysis of the electronic nose data indicates significant differences in odor characteristics between BCZ and MCZ, mainly reflected in the responses of sensors W5S, W1W, and W1S. Nitrogen oxides, terpenes, inorganic sulfides, and short-chain alkanes such as methane are the key volatile components distinguishing the two species.

2.4. Electronic Tongue Analysis Results

2.4.1. Analysis of Taste Composition of BCZ and MCZ

Figure 11 presents the radar chart of electronic tongue sensor response intensities to BCZ and MCZ samples. The results indicate that neither sample exhibits sourness, but both show relatively high umami and bitterness values. This aligns with the description of Atractylodes as “bitter” in the 2025 edition of the Chinese Pharmacopoeia.

Figure 11.

Electronic tongue sensor response values.

2.4.2. Multivariate Statistical Analysis Based on the Electronic Tongue

The electronic tongue sensor response values for BCZ and MCZ were analyzed using SIMCA 14.1 software for PCA. The results (Figure 12) show that the two species can be clustered into separate groups, indicating that the electronic tongue can distinguish Atractylodes from different origins.

Figure 12.

PCA and OPLS-DA based on electronic tongue data. (a) PCA score plot; (b) OPLS-DA score plot; (c) permutation test; (d) VIP values.

To further identify distinguishing features between MCZ and BCZ, OPLS-DA analysis was performed on electronic tongue response values using SIMCA 14.1 software. With VIP > 1 as the criterion and 200 bootstrap tests, the results (Figure 12) show that the model is stable, reliable, and has strong predictive power without overfitting. The OPLS-DA plot shows that different Atractylodes species cluster into distinct groups, consistent with PCA results. Two discriminant features—sourness and richness—were identified. It should be noted that although the absolute sourness sensor response values were low for both BCZ and MCZ (Figure 11), the sourness sensor still generates a measurable electrical response, and the subtle difference in this response between MCZ and BCZ was statistically significant, which led to its identification as a discriminant feature by OPLS-DA (VIP > 1). In other words, electronic tongue sensors detect electrochemical signals rather than human taste perception; a statistically significant sensor difference does not necessarily indicate a perceptible taste difference at the human sensory threshold. Nevertheless, these subtle sensor-level differences, when combined with other taste attributes through multivariate modeling, contribute to the overall discrimination between the two species.

This study analyzed the flavor data of MCZ and BCZ and performed PCA using characteristic indicators. The results showed that the two species can be reliably distinguished by origin. OPLS-DA analysis further confirmed that flavor characteristics play a significant role in distinguishing Atractylodes origins. These findings provide a reference for origin identification and quality evaluation of Atractylodes.

2.5. Pearson Correlation Analysis of “Composition-Odor”

To further investigate the relationship between volatile metabolites and electronic nose sensor responses, a correlation analysis was conducted between major differentially expressed volatile metabolites and the sensors showing strong responses (W1W, W5S, and W1S). Twenty-four volatile metabolites with VIP > 1.2 were selected for correlation analysis with these three sensors.

The results (Figure 13) show that sensor W1W has a significant positive correlation with 3-methylbutyraldehyde (C6) (p < 0.05). Although W1W is sensitive to terpenes and inorganic sulfides, 3-methylbutyraldehyde—a short-chain aldehyde—may have produced a cross-response due to its molecular structure or sensor-surface interactions. No significant correlation was found between sensor W5S and any component (p > 0.05). Sensor W1S showed significant positive correlations with linalool (C1), carvone (C2), 4-vinylguaiacol (C12), 2-butanone (C13), 2,6-di-tert-butyl-p-cresol (C17), 2-pentanol (C21), and 2-methoxy-3-sec-butylpyrazine (C23), and significant negative correlations with farnesene (C10), δ-3-carene (C16), and 3-cyclopentene-1-one (C22).

Figure 13.

Correlation analysis plot (* p < 0.05).

Based on these findings, the main chemical basis for odor differences among the samples is attributed to short-chain oxygen-containing compounds (such as alcohols, ketones, and aldehydes) and terpenoids. Linalool, carvone, 4-vinylguaiacol, and 2-methoxy-3-sec-butylpyrazine produce strong positive responses on sensor W1S and are key contributors to the characteristic odors of the samples, while terpenes like farnesene and δ-3-carene may suppress or modify overall odor. Sensor W1S responds to the majority of differential components, making it the most critical sensor for distinguishing the odor characteristics of the samples.

2.6. Double Cross-Validation of OPLS-DA Models

Double cross-validation (2CV) was performed to rigorously evaluate the predictive ability of each OPLS-DA model. In Table 3, the 2CV Q² values confirmed the robust predictive ability of all four OPLS-DA models, with 2CV Q² of 0.624, 0.996, 0.880, and 0.878 for the E-nose, E-tongue, HS-GC-IMS, and UPLC-Orbitrap MS models, respectively. The classification accuracies ranged from 91.67% to 100%. All permutation test Q² intercepts were negative, confirming that none of the models suffered from overfitting.

Table 3.

Double cross-validation (2CV) results for OPLS-DA models.

Model	Q2 (2CV)	Permutation Q2 Intercept	Accuracy
Electronic nose	0.624	−1.680	91.67%
Electronic tongue	0.996	−1.621	100%
HS-GC-IMS	0.880	−2.281	100%
UPLC-Orbitrap MS	0.878	−1.269	100%

3. Discussion

This study integrated electronic nose, electronic tongue, HS-GC-IMS, and UPLC-Orbitrap MS to systematically analyze the sensory characteristics and chemical compositions of MCZ and BCZ. The results revealed significant differences between the two botanical origins in odor, taste, volatile compounds, and non-volatile metabolites.

HS-GC-IMS identified 108 volatile organic compounds, with 24 key volatile markers screened out, which contributed significantly to the differentiation of MCZ and BCZ. Notably, propionic acid and 5-methyl-2-furancarboxaldehyde were predominantly found in MCZ, while (E)-caryophyllene was enriched in BCZ, suggesting these compounds as potential volatile markers for distinguishing the two botanical origins. Ma et al. employed GC-MS coupled with chemometric analysis to differentiate four Atractylodes species, identifying 50 volatile components from 59 batches of samples and screening out hinesol, β-eudesmol, atractylon, atractylodin, and atractylenolide I as species characteristic markers [7]. However, traditional GC-MS analysis typically relies on hydrodistillation, which may cause loss or transformation of thermally labile compounds. In contrast, HS-GC-IMS requires no complex sample pretreatment and can more authentically reflect the original volatile profile. This technical advantage is particularly important for distinguishing botanical origins with subtle differences in volatile composition. Compared with conventional GC-MS, HS-GC-IMS offers advantages such as rapid analysis, minimal sample pretreatment, and high sensitivity for volatile organic compounds; however, its compound library is relatively limited compared to the well-established NIST mass spectral database used in GC-MS [58,59]. Therefore, combining both techniques provides a more comprehensive approach for Atractylodes quality evaluation.

Furthermore, UPLC-Orbitrap MS detected 78 non-volatile constituents, with 17 differential metabolites screened out by OPLS-DA with VIP > 1.2. Li et al. applied UHPLC-Orbitrap-HRMS for non-targeted metabolomics analysis of A. chinensis from eight origins, confirming significant metabolic variations among different regions [60]. These findings indicate that MCZ and BCZ can be effectively discriminated based on both sensory signals and chemical profiles, including volatile and non-volatile constituents.

Sensory technologies are also widely used in botanical origin identification. Wang et al. established a discrimination model for Codonopsis Radix from different origins by data fusion of electronic nose and electronic tongue, effectively addressing the issue of imbalanced sample categories [61], which confirmed the practical value of sensory technologies in botanical origin identification. This study found that the odor differences between MCZ and BCZ were mainly captured by the W5S, W1W, and W1S sensors, suggesting that nitrogen oxides, terpenoids, inorganic sulfides, and short-chain alkanes play key roles in the formation of distinct odor profiles of the two origins. Further electronic tongue analysis indicated that sourness and richness were the key taste indicators for distinguishing MCZ from BCZ. This finding underscores the value of electronic tongue-based digital taste profiling as a complementary tool to conventional chemical analysis, as it can detect subtle sensor-level differences that may not correspond to human sensory perception but nonetheless contribute to multivariate discrimination. Moreover, correlation analysis revealed that the response of the W1S sensor was strongly correlated with several key volatile compounds such as linalool and carvone. This analysis further confirmed that short-chain oxygenated compounds and terpenoids are jointly responsible for the distinctive odor differences between the two origins. Specifically, compounds such as linalool, carvone, and (E)-caryophyllene contribute significantly to the characteristic odor profiles of MCZ and BCZ.

This study has certain limitations. First, the sample size was relatively small (six batches per origin), and more samples from different geographical regions are needed for validation. Second, the relationship between these chemical differences and pharmacological effects requires further investigation. Third, the discriminatory efficiency of the established fingerprinting method needs validation with more authentic reference samples. Fourth, since MCZ and BCZ were collected from different geographic regions, the observed chemical differences, including the identified key volatile markers, may be confounded by geographic origin and cultivation conditions rather than solely reflecting species differences. Environmental factors such as temperature, light, soil properties, and water availability are known to significantly influence secondary metabolite accumulation in medicinal plants [62,63,64], and substantial geographic variation in essential oil composition has been documented across different Atractylodes cultivation regions [65,66]. Controlled cultivation experiments with both species grown under identical environmental conditions would be necessary to confirm whether these markers are truly species-specific. The 2CV results further validated the reliability of the OPLS-DA models, indicating that the discrimination between BCZ and MCZ was not an artifact of overfitting but reflected genuine metabolic and sensory differences between the two species.

Future studies should expand sample collection to include batches from different cultivation conditions and harvest times. In addition, metabolomics combined with pharmacological experiments may further elucidate the relationship between chemical constituents and therapeutic effects.

4. Materials and Methods

4.1. Reagents

Optima-grade methanol and acetonitrile were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Purified water was obtained from Guangzhou Watsons Food & Beverage Co., Ltd. (Guangzhou, China). All solvents were used without further purification.

4.2. Herbal Materials

A total of twelve batches of Atractylodis Rhizoma samples were used in this study, including six batches of MCZ and six batches of BCZ. The appearance characteristics of the herbal materials are shown in Figure 14, and the sample codes along with their geographical origins are provided in Table 4.

Figure 14.

(A–F) are MCZ1–MCZ6; (G–L) are BCZ1–BCZ6.

Table 4.

Herbal material information table.

No.	Origin	No.	Origin
MCZ-1	Yingshan County, Hubei Province	BCZ-1	Chifeng City, Inner Mongolia Autonomous Region
MCZ-2	Yingshan County, Hubei Province	BCZ-2	Chifeng City, Inner Mongolia Autonomous Region
MCZ-3	Yingshan County, Hubei Province	BCZ-3	Hulunbuir City, Inner Mongolia Autonomous Region
MCZ-4	Shiyan City, Hubei Province	BCZ-4	Hulunbuir City, Inner Mongolia Autonomous Region
MCZ-5	Shiyan City, Hubei Province	BCZ-5	Chengde City, Hebei Province
MCZ-6	Shiyan City, Hubei Province	BCZ-6	Chengde City, Hebei Province

The MCZ samples were derived from Atractylodes lancea (Thunb.) DC., while the BCZ samples were derived from Atractylodes chinensis (DC.) Koidz. All samples were authenticated by Professor Fang Zhang from Shandong University of Traditional Chinese Medicine.

4.3. HS-GC-IMS Analysis

Analysis was performed using the FlavourSpec^® GC-IMS (G.A.S., Dortmund, Germany). Precisely weigh 0.5 g each of powdered MCZ and BCZ, place the samples into 20 mL headspace vials with stoppers, and incubate at 80 °C for 15 min prior to injection. The system parameters were as follows: column temperature, 60 °C; carrier gas, N2; ion mobility detector temperature, 45 °C; total analysis time, 40 min. Injection volume was 200 μL, incubation time 15 min, incubation temperature 80 °C, injection needle temperature 80 °C, and incubation speed 500 rpm. E1 drift gas flow rate was 75 mL/min. E2 carrier gas flow rates were: 0–2 min, 2 mL/min; 2–10 min, 2–10 mL/min; 10–20 min, 10–100 mL/min; 20–25 min, 100–150 mL/min; 25–40 min, 150 mL/min. Qualitative analysis was conducted using the system’s built-in software, NIST database, and IMS migration time database. Component fingerprint profiles were generated using the Reporter and Gallery Plot plugins. Each sample was analyzed in triplicate.

4.4. UPLC-Orbitrap Analysis

4.4.1. Extraction of Non-Volatile Compounds

Pass the powders of MCZ and BCZ through a 60-mesh sieve. Accurately weigh 0.5 g of each sample, add exactly 20 mL of methanol, and weigh the mixture again to record its mass. Sonicate the mixture for 30 min, allow it to cool, and then reweigh. Add methanol to compensate for any weight loss, mix thoroughly, and transfer the solution to a 25 mL volumetric flask. Adjust to volume with methanol, filter the solution through a 0.22 µm microporous membrane, and store the resulting test solution at 4 °C in the dark until analysis. Each sample was prepared and analyzed in triplicate.

4.4.2. Instrumental Conditions

Analysis was performed using UPLC-Q-Exactive-Orbitrap-MS (Thermo Fisher Scientific, Waltham, MA, USA). Chromatographic conditions: Waters Acquity UPLC BEH C18 column (2.1 mm × 100 mm, 1.7 µm); mobile phase: A-water, B-acetonitrile, gradient elution; flow rate: 0.30 mL/min; column temperature: 30 °C; injection volume: 5 µL. The gradient elution program was as follows: 0–2 min, 5–30% B; 2–14 min, 30–60% B; 14–23 min, 60–70% B; 23–31 min, 70–95% B; 31–31.5 min, 95–5% B; 31.5–35 min, 5% B. MS conditions: Electrospray ionization (ESI) was performed in both positive and negative ion modes, scanning a mass range of m/z 50–1000. For negative ion mode, the optimized ESI parameters were as follows: capillary temperature, 350 °C; sheath gas (nitrogen) flow rate, 30 arb; auxiliary gas (nitrogen) flow rate, 10 arb; source voltage, 3.0 kV; capillary voltage, −35 V; tube lens voltage, −110 V. For positive ion mode: capillary voltage, 25 V; source voltage, 4.0 kV; tube lens voltage, 110 V; other parameters were the same as those for negative ion mode. The Orbitrap mass analyzer resolution was set to 30,000. The secondary mass spectrometer operated in data-dependent acquisition (DDA) mode with a normalized collision energy of 35%.

4.5. Electronic Nose Analysis

PEN3 electronic nose system (AIRSENSE Analytics GmbH, Schwerin, Germany) was used to analyze the odors of BCZ and MCZ. This system is equipped with 10 odor sensors (W1C, W5S, W3C, W6S, W5C, W1S, W1W, W2S, W2W, W3S). By evaluating each sensor’s sensitivity to various compound classes, the primary volatile odor characteristics of both Atractylodes species were identified. The response characteristics of the electronic nose sensors are listed in Table 5.

Table 5.

Electronic nose sensor information table.

Serial Number	Sensor Name	Representative Compound Categories
R (1)	W1C	Sensitive to aromatic compounds and benzene compounds
R (2)	W5S	High sensitivity, highly responsive to nitrogen oxides
R (3)	W3C	Sensitive to amines and aromatic compounds
R (4)	W6S	Primarily selective for hydrides
R (5)	W5C	Alkane and aromatic components
R (6)	W1S	Sensitive to methane and other short-chain alkanes
R (7)	W1W	Sensitive to terpenes and inorganic sulfides
R (8)	W2S	Sensitive to alcohols, aldehydes, and ketones
R (9)	W2W	Aromatic compounds, sensitive to organic sulfides
R (10)	W3S	Sensitive to alkanes, long-chain alkanes

Accurately weigh 2.0 g of each BCZ and MCZ powder and transfer to headspace vials. Seal the vials with double-layer sealing membranes and allow them to equilibrate at room temperature for 20 min before analysis. The headspace inhalation method was used, with stable response values serving as the evaluation criterion. Sampling was performed every 1 s, with a cleaning time of 120 s, zero adjustment for 5 s, pre-sampling for 5 s, and a sampling duration of 120 s. The injection flow rate was set at 600 mL/min. For each batch, three samples were weighed and analyzed in triplicate.

4.6. Electronic Tongue Analysis

An SA-402B electronic tongue (Insent Intelligent Sensor Technology Inc., Atsugi, Japan) was used to analyze samples of BCZ and MCZ. The samples were ground and passed through a No. 4 sieve (65 mesh). Accurately weigh 1.0 g of the powdered sample and place it in a 250 mL stoppered conical flask. Add 150 mL of deionized water, shake well, and sonicate at 100 W for 30 min. Transfer the entire extract to a centrifuge tube and centrifuge at 3000 rpm for 20 min. Collect 35 mL of the supernatant and place it in a dedicated electronic tongue sample cup for analysis. The raw data were converted into taste values using the instrument’s built-in data processing software and exported for further analysis. Detailed information on the electronic tongue sensors is provided in Table 6.

Table 6.

Information on electronic tongue sensors.

Sensor	Taste	Aftertaste
Umami (AAE)	Umami	Richness
Salty (CTO)	Salty	-
Acidity (CAO)	Acidity	-
Bitterness (COO)	Bitterness	Bitter aftertaste
Astringency (AE1)	Astringency	Astringent aftertaste

4.7. Software and Statistical Analysis

Qualitative analysis of volatile components and construction of fingerprint profiles were carried out using the built-in database and plugins of the FlavourSpec^® GC-IMS instrument. Principal Component Analysis (PCA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) were performed with SIMCA 14.1 software, and Variable Importance in Projection (VIP) values were calculated. Prism 10.1.2 was used to plot HS-GC-IMS component analysis charts; WinMuster software (AIRSENSE Analytics GmbH, Schwerin, Germany) was used to analyze electronic nose loadings plots; radar charts and stacked bar charts were created using Origin 2019b; and “component-odor” Pearson correlation analysis was conducted via the Microbioinformatics online platform. 2CV was performed using the pls package (version 2.9-0) in R (version 4.5.0). The outer loop employed leave-one-out cross-validation to assess prediction performance, while the inner loop utilized 7-fold cross-validation to optimize the number of orthogonal and predictive components. The 2CV Q² was calculated based on the predicted versus observed Y values across all 12 iterations. Permutation tests (100 random permutations) were conducted to confirm the absence of overfitting.

5. Conclusions

In this study, a multi-dimensional analytical strategy integrating HS-GC-IMS, UPLC-Orbitrap MS, electronic nose, and electronic tongue was established to systematically characterize the differences between MCZ and BCZ.

HS-GC-IMS detected 108 volatile organic compounds, from which 24 volatile markers with VIP > 1.2 were screened as key contributors to the differentiation of MCZ and BCZ. Among them, propionic acid and 5-methyl-2-furancarboxaldehyde were mainly distributed in MCZ, whereas (E)-caryophyllene was present only or at higher levels in BCZ, indicating its potential as a characteristic volatile marker of BCZ. UPLC-Orbitrap MS detected 78 non-volatile constituents, and OPLS-DA screened 17 key non-volatile differential metabolites with VIP > 1.2, providing complementary chemical evidence for distinguishing the two origins. Electronic nose results revealed distinct odor profiles between MCZ and BCZ. Sensors W5S, W1W, and W1S were identified as the main differential sensors, suggesting that nitrogen oxides, terpenoids, inorganic sulfides, and short-chain alkanes are the primary contributors to the odor differences. Electronic tongue analysis demonstrated clear taste discrimination between the two origins, with sourness and richness identified as the key taste indicators.

Taken together, MCZ and BCZ can be clearly separated not only by sensory signals but also by both volatile and non-volatile chemical profiles. The integration of these four techniques provides a comprehensive and reliable strategy for distinguishing the two botanical origins of Atractylodis Rhizoma, offering scientific support for accurate authentication, quality evaluation, and rational clinical application. Future studies should expand sample collection to include batches from different geographic regions and cultivation conditions to further validate these markers.

Author Contributions

Conceptualization, H.Z. and Z.Q.; methodology, H.Z. and Y.H.; formal analysis, H.Z.; investigation, H.Z.; data curation, F.W.; writing—original draft preparation, H.Z.; writing—review and editing, Y.L.; funding acquisition, Y.L. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This research was funded by the Young Scientists Fund of the National Natural Science Foundation of China (No. 82505122); Shandong Key Laboratory of Digital Traditional Chinese Medicine (No. 222025260); the Renowned Physicians and Classical Prescriptions Transformational Research and Development Program of Shandong University of Traditional Chinese Medicine (No. XK2025MY05); and the Youth Bian Que Scholars High-quality Development Program of Shandong University of Traditional Chinese Medicine (No. LIYANAN).