Integrated Serum Pharmacochemistry and Investigation of the Anti-Depression Effect of Chaihu–Huangqin Drug Pair

1. Introduction

Depression is one of the most prevalent psychiatric disorders worldwide and remains a leading cause of disability. The Global Burden of Disease Study 2021 estimated 332 million prevalent cases of depressive disorders globally and ranked depressive disorders as the second leading cause of years lived with disability [1]. Although selective serotonin reuptake inhibitors (SSRIs) and other antidepressants are widely prescribed, their clinical utility is constrained by tolerability and adherence: in a retrospective cohort of 20,941 antidepressant users, discontinuation rates reached 51.6–57.3% for commonly prescribed agents, and adverse effects were a principal driver of withdrawal [2]. These limitations have prompted growing interest in multitarget natural products capable of simultaneously modulating inflammatory and monoaminergic systems [3]. The pharmacological actions of antidepressants are not uniform throughout the brain. Several brain regions, such as the prefrontal cortex (PFC), the ventral tegmental area (VTA) and nucleus accumbens (NAc) of the reward circuitry, the dorsal raphe nucleus (DRN) and lateral habenula (LHb) of the serotonergic circuit, and the hippocampal CA area, have become focal points of research on the pathogenesis of depression and on antidepressant development, because each of them makes a distinct contribution to stress-related pathology: the PFC and hippocampal CA area are mainly implicated in cognitive and emotional regulation, the VTA–NAc axis in reward and motivation, and the DRN–LHb circuit in serotonergic regulation of negative affect [4,6,8].

Accumulating evidence indicates that chronic stress-induced neuroinflammation, characterized by microglial activation and the release of pro-inflammatory cytokines, is a central mechanism in the pathogenesis of depression [4,5]. In both clinical and preclinical studies, elevated levels of tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β) and interleukin-18 (IL-18) have been reported [6]. Among the pathways driving this inflammatory response, the nuclear factor kappa-B (NF-κB)/NOD-like receptor family pyrin domain containing 3 (NLRP3) inflammasome axis plays a pivotal role: NF-κB activation promotes the transcription of NLRP3 and pro-inflammatory cytokines, whereas NLRP3 inflammasome assembly activates Caspase-1, which cleaves pro-IL-1β and pro-IL-18 into their mature forms [4,7]. Under chronic stress, sustained activation of this axis contributes to neuronal dysfunction and depressive-like behaviors. Notably, NLRP3 in the dorsal raphe nucleus (DRN) has recently been shown to be essential for depressive-like phenotypes: NLRP3 was upregulated in DRN microglia and neurons of chronic unpredictable mild stress (CUMS) mice, and cell-type-specific knockdown of NLRP3 in either microglia or neurons rescued the depressive phenotype [6]. The DRN, the principal source of serotonergic projections to the forebrain and limbic system, is a key hub in the regulation of mood and stress responses [8], and thus represents a critical site at which neuroinflammatory signaling and serotonergic dysfunction may converge.

In this context, traditional Chinese medicine (TCM) offers a promising source of multitarget therapeutic agents. Unlike single-target synthetic drugs, herbal formulas and drug pairs contain multiple bioactive constituents that may simultaneously modulate interconnected pathological processes—a characteristic particularly relevant to depression, which is shaped by the interplay of inflammatory, monoaminergic and neuroendocrine disturbances rather than a single pathogenic event.

Chaihu–Huangqin (CHHQ) (Bupleuri Radix–Scutellariae Radix), the core drug pair of the traditional formula Xiaochaihu Decoction, has long been used in TCM to soothe liver qi, clear heat and relieve emotional disorders. Chaihu (Bupleuri Radix) is traditionally valued for dispersing stagnated liver qi and alleviating depression-like symptoms, and its major bioactive constituents include saikosaponins, volatile oils and flavonoids. Huangqin (Scutellariae Radix) is traditionally used to clear heat and dry dampness, with baicalin, baicalein and wogonin as its principal bioactive constituents. In combination, the two herbs are considered to act in a complementary manner, with Chaihu addressing qi stagnation and Huangqin clearing the heat generated by prolonged stagnation. Experimental studies have reported that the Bupleurum chinense–Scutellaria baicalensis (BS) drug pair exerts antidepressant effects in CUMS mice, with LC-MS characterization identifying 29 constituents and network analysis and animal validation implicating the IL-6/JAK2/STAT3 and PI3K/AKT pathways [9]. Individual constituents of the pair have likewise been shown to counteract chronic stress-induced brain pathology: saikosaponin A improved depressive-like behaviors by activating TET1/DLL3/NOTCH1 signalling and promoting hippocampal neurogenesis in mice [10], and baicalin exerted neuroprotective effects by inhibiting GSK3β/NF-κB/NLRP3 pathway activation in rats [7].

Despite these advances, the antidepressant material basis of the CHHQ drug pair remains undefined, and three questions at successively deeper levels have not been answered in sequence. First, at the screening-of-absorbed-bioactives level: although serum pharmacochemistry has successfully revealed the absorbed constituents of other antidepressant herb pairs, such as Bupleurum chinense–Paeonia lactiflora [11], which constituents of the Chaihu–Huangqin pair enter the systemic circulation after oral administration, in what prototype and metabolite forms, and through which phase I/II biotransformation pathways, have not been systematically characterized. Second, at the brain-penetration and regional-distribution level: for constituents in the systemic circulation, crossing the blood–brain barrier is the prerequisite for entering the brain tissue and reaching the relevant brain regions. As noted above, depression involves region-specific pathological changes across multiple brain regions; previous studies on antidepressant TCM preparations, however, have mostly examined the whole brain without distinguishing individual brain regions, and whether the absorbed constituents penetrate the blood–brain barrier, and how their distribution differs among these regions, have not been systematically compared. Third, at the target and efficacy level: constituents that have entered the brain exert pharmacological effects only after binding to their targets; however, network pharmacology and molecular docking are usually initiated from the full constituent library of the herbs [12] rather than from brain-penetrant constituents identified in vivo, and the predictions are rarely verified experimentally. The evidence chain linking chemical exposure, brain penetration, target engagement and pharmacodynamic efficacy therefore remains incomplete. Answering these questions in sequence requires the screening of absorbed bioactives, the screening of region-specific bioactives, and in vivo validation to be connected within a single study design.

Accordingly, the present study established an integrated strategy of “screening of bioactives absorbed into blood — screening of region-specific bioactives — in vivo validation in animals” in a CUMS rat model to answer the above questions in sequence. In the first step, UPLC-MS/MS-based serum pharmacochemistry was used to characterize the chemical constituents of the herbal solution and to screen the bioactive-related candidates absorbed into serum, distinguishing candidate prototype constituents from candidate metabolites and delineating their potential phase I/II biotransformation features. In the second step, and in view of the blood–brain barrier, the candidates entering the brain tissue were profiled in six depression-related brain regions (hippocampal CA area, dorsal raphe nucleus (DRN), lateral habenula, nucleus accumbens, prefrontal cortex and ventral tegmental area); their numbers and composition were compared across regions to characterize the regional exposure pattern, and the candidates selectively detected in a single region were screened, among which the DRN showed the most region-specific candidates and was therefore selected as the focus of subsequent analysis. In the third step, the DRN-specific candidates were connected with in vivo efficacy: considering the DRN as the principal source of serotonergic innervation to the forebrain and limbic system and the essential involvement of NLRP3 activation within DRN microglia and neurons in depressive-like phenotypes, network pharmacology and molecular docking were initiated from the DRN-specific candidates to prioritize candidate constituent–target pairs among chemicals with proven regional exposure rather than among the full herbal constituent library; behavioral pharmacology (sucrose preference, forced swimming and open field tests), body weight and serum hepato-renal safety indicators, DRN histopathology and IBA-1/CD68 immunofluorescence, inflammatory cytokines, and Western blotting of NF-κB, NLRP3, Caspase-1 and SERT were then used to examine whether the drug pair ameliorates depressive-like behaviors and normalizes DRN microglial activation together with local inflammatory and serotonergic signaling in CUMS rats.

The distinctive features of this design are threefold. First, it establishes a brain-region-resolved chain of evidence for the CHHQ drug pair, extending the material-basis study from “which bioactive candidates are absorbed into blood” to “which of them penetrate into the brain and how their distribution differs across brain regions”, and further to “how the region-specific candidates of a given brain region act on their targets”; the selection of the DRN was driven by the measured exposure data rather than by assumption. Second, the prediction step is initiated from experimentally measured DRN-specific bioactive candidates rather than from the full herbal constituent library, which substantially improves the credibility of network pharmacology in TCM research. Third, to our knowledge, this is the first in vivo study to link a classical TCM drug pair with DRN microglial activation, the NF-κB/NLRP3/Caspase-1 axis and SERT within the same model, thereby connecting the chemical basis of the pair with a recently identified essential node of depressive-like behavior. This work is expected to clarify the constituent–target relationship underlying the antidepressant effect of the CHHQ drug pair and to provide a transferable, brain-region-resolved paradigm for dissecting the material basis of other TCM drug pairs.

2. Results

2.1. Chemical Characterization of the CHHQ

To define the original chemical composition of CHHQ and to provide a preliminary chemical basis for the subsequent analysis of in vivo prototype components and metabolites, the CHHQ decoction was analyzed by UPLC-Q-Exactive HF-MS. Raw LC-MS data were preprocessed using XCMS (v4.5.1) for peak detection, peak integration, retention time correction, peak alignment, and normalization, and candidate structures were annotated based on accurate mass-to-charge ratios, isotope patterns, and available MS/MS fragment information in combination with the Traditional Chinese Medicine (TCM) component database. According to the preset screening criteria, a total of 1390 candidate chemical components were annotated after integration and deduplication of the positive- and negative-mode results.

By chemical class, 221 flavonoids, 173 terpenoids, 164 phenylpropanoids, 125 sugars and glycosides, 96 fatty acyls, 92 organic heterocyclic compounds, 82 amino acids and peptides, 49 lipids and lipid-like molecules, 18 nucleotides and derivatives, and 14 organic acids and derivatives were identified, together with other classes such as phenols, alkaloids, and steroids, indicating that CHHQ has a relatively complex and diverse chemical composition. Detailed information on all candidate components is provided in Supplementary Table S1. The base peak chromatograms (BPCs) of the CHHQ decoction (TCM), blank serum (Control), and drug-containing serum (Treatment) in the positive and negative ion modes are shown in Figure 2.

To further characterize the mass spectrometric features of representative candidates from the major chemical classes, representative components were selected from the flavonoids, terpenoids, phenylpropanoids, sugars and glycosides, fatty acyls, amino acids and peptides, organic heterocyclic compounds, lipids and lipid-like molecules, and nucleotides and derivatives, taking into account the representativeness of the medicinal materials, the coverage of chemical classes, and the completeness of MS/MS information. Their experimental MS/MS spectra, reference spectra, and possible fragmentation behaviors were further analyzed (Figure 3).

2.1.1. Flavonoids

A total of 221 flavonoid candidates were annotated. Baicalin and wogonin 7-glucuronide, representative flavonoids of Scutellariae Radix, were selected for mass spectrometric characterization.

Baicalin was detected in the negative ion mode as the [M−H]⁻ ion at m/z 445.0779. Its MS/MS spectrum was dominated by the product ion at m/z 269.0456, with a mass difference of approximately 176.0323 Da, consistent with the neutral loss associated with C₆H₈O₆ and yielding a baicalein-related aglycone ion; product ions at m/z 175.0249 and 113.0246 were also observed. The major characteristic ions in the experimental MS/MS spectrum were in good agreement with those in the reference spectrum, supporting the candidate structural annotation of baicalin. Its representative fragmentation behavior is shown in Figure 3A.

Wogonin 7-glucuronide was detected in the positive ion mode as the [M+H]⁺ ion at m/z 461.1066. Its MS/MS spectrum was dominated by the product ion at m/z 285.0749, approximately 176.0317 Da lower than the precursor ion, consistent with the neutral loss associated with C₆H₈O₆ and yielding a wogonin-related aglycone ion. The experimental MS/MS spectrum agreed well with the reference spectrum, supporting the candidate structural annotation of wogonin 7-glucuronide. Its representative fragmentation behavior is shown in Figure 3B.

2.1.2. Terpenoids

A total of 173 terpenoid candidates were annotated. Saikosaponin A, a representative triterpenoid saponin of Bupleuri Radix, was selected for mass spectrometric characterization. Saikosaponin A was detected in the negative ion mode as the [M−H]⁻ ion at m/z 779.4615. Its MS/MS spectrum was dominated by the product ion at m/z 617.4108, with a mass difference of approximately 162.0507 Da, consistent with the neutral loss associated with C₆H₁₀O₅. The major characteristic ions in the experimental MS/MS spectrum were in good agreement with those in the reference spectrum, supporting the candidate structural annotation of saikosaponin A. Its representative fragmentation behavior is shown in Figure 3C.

2.1.3. Phenylpropanoids

A total of 164 phenylpropanoid candidates were annotated. Chlorogenic acid was selected as a representative compound for mass spectrometric characterization. Chlorogenic acid was detected in the negative ion mode as the [M−H]⁻ ion at m/z 353.0882. Its MS/MS spectrum was dominated by the product ion at m/z 191.0564, with a mass difference of approximately 162.0318 Da, consistent with the neutral loss associated with C₉H₆O₃. The ion at m/z 191.0564 is a characteristic quinic acid-related ion. The experimental MS/MS spectrum was in good agreement with the reference spectrum, supporting the candidate structural annotation of chlorogenic acid. Its representative fragmentation behavior is shown in Figure 3D.

2.1.4. Sugars and Glycosides

A total of 125 sugar and glycoside candidates were annotated. Sucrose was selected as a representative compound. In the positive ion mode, sucrose was mainly detected as the [M+NH₄]⁺ ion at m/z 360.1492, and product ions at m/z 198.0973, 180.0861, 163.0603, 145.0494, 127.0389, and 85.0289 were observed in its MS/MS spectrum. The ion at m/z 198.0973 can be formed from m/z 360.1492 through the neutral loss associated with C₆H₁₀O₅, corresponding to an ammoniated hexose-related ion, and can further yield product ions such as m/z 145.0494 through the neutral losses associated with NH₃ and 2H₂O. Multiple characteristic ions in the experimental MS/MS spectrum were in good agreement with those in the reference spectrum, supporting the candidate structural annotation of sucrose. Its representative fragmentation behavior is shown in Figure 3E.

2.1.5. Fatty Acyls

A total of 96 fatty acyl candidates were annotated. 2-Hydroxycaprylic acid was selected as a representative compound. It was detected in the negative ion mode as the [M−H]⁻ ion at m/z 159.1027. Its MS/MS spectrum was dominated by the product ion at m/z 113.0973, with a mass difference of approximately 46.0054 Da, consistent with the neutral loss of HCOOH. The corresponding product ions in the experimental and reference spectra were in good agreement, supporting the candidate structural annotation of 2-hydroxycaprylic acid. Its representative fragmentation behavior is shown in Figure 3F.

2.1.6. Amino Acids and Peptides

A total of 82 amino acid and peptide candidates were annotated. L-Phenylalanine was selected as a representative compound. It was detected in the positive ion mode as the [M+H]⁺ ion at m/z 166.0864. Its MS/MS spectrum was dominated by the product ion at m/z 120.0813, with a mass difference of approximately 46.0051 Da, consistent with the neutral loss of HCOOH. The ion at m/z 120.0813 in the experimental spectrum was in good agreement with the corresponding characteristic ion in the reference spectrum, supporting the candidate structural annotation of L-phenylalanine. Its representative fragmentation behavior is shown in Figure 3G.

2.1.7. Organic Heterocyclic Compounds

A total of 92 organic heterocyclic compounds were annotated. Owing to the considerable differences in the heterocyclic nuclei and substituent structures within this class, the compounds can exhibit distinct structure-related fragmentation behaviors during collision-induced dissociation. Taking 2-hydroxycinchoninic acid as an example, it was detected in the negative ion mode as the [M−H]⁻ ion at m/z 190.0503. Its MS/MS spectrum was dominated by the product ion at m/z 162.0554, with a mass difference of approximately 27.9949 Da, consistent with the neutral loss of CO. The major characteristic ions in the experimental and reference spectra were in good agreement, supporting the candidate structural annotation of 2-hydroxycinchoninic acid. Its representative fragmentation behavior is shown in Figure 3H.

2.1.8. Lipids and Lipid-like Molecules

A total of 49 lipids and lipid-like molecules were annotated. Reynosin was selected as a representative candidate. It was detected in the positive ion mode as the [M+H]⁺ ion at m/z 249.1481. Its MS/MS spectrum first showed the product ion at m/z 231.1379, approximately 18.0102 Da lower than the precursor ion, consistent with the neutral loss of H₂O; the ion at m/z 231.1379 further yielded the product ion at m/z 185.1323, with a mass difference of approximately 46.0056 Da, consistent with the neutral loss associated with HCOOH. The major characteristic ions in the experimental and reference spectra were in good agreement, supporting the candidate structural annotation of reynosin. Its representative fragmentation behavior is shown in Figure 3I.

2.1.9. Nucleotides and Their Derivatives

A total of 18 nucleotides and their derivatives were annotated. During collision-induced dissociation, these compounds commonly undergo cleavage related to the glycosidic bond, yielding product ions associated with the corresponding nitrogenous bases. Taking 2′-O-Methyladenosine as an example, it was detected in the positive ion mode as the [M+H]⁺ ion at m/z 282.1195. Its MS/MS spectrum was dominated by the adenine-related product ion at m/z 136.0617, with a mass difference of approximately 146.0578 Da, consistent with the neutral loss associated with C₆H₁₀O₄. The major characteristic ions in the experimental and reference spectra were in good agreement, supporting the candidate structural annotation of 2′-O-methyladenosine. Its representative fragmentation behavior is shown in Figure 3J.

In summary, multiple classes of candidate chemical components—including flavonoids, terpenoids, phenylpropanoids, sugars and glycosides, fatty acyls, amino acids and peptides, organic heterocyclic compounds, lipids and lipid-like molecules, and nucleotides and their derivatives—were detected in CHHQ, demonstrating a relatively complex chemical composition. The experimental MS/MS spectra of the representative candidates were generally in good agreement with the reference spectra, providing further mass spectrometric evidence for database-based candidate structural annotation.

These results established a candidate chemical profile of the original CHHQ formula. However, the chemical components detected in the original formula are not equivalent to the drug-related substances that actually enter systemic circulation after administration. Therefore, on the basis of the chemical characterization of the original formula, the prototype components and candidate metabolites in drug-containing serum were further analyzed.

2.2. Blood-Entry Drug-Related Components of the CHHQ and Their Potential in Vivo Biotransformation

Building on the chemical characterization of CHHQ, the drug-related components in drug-containing serum were further analyzed to screen the prototype components and metabolites that enter the blood circulation after administration and to examine their chemical class composition and potential in vivo metabolic characteristics.

2.2.1. Screening of Blood-Entry Related Components

By comparing the chromatographic peaks and mass spectrometric information of blank serum, drug-containing serum, and the CHHQ decoction samples, together with retention time (RT), accurate m/z, and available MS/MS fragment information, the blood-entry related components were screened by automated filtering combined with manual verification. According to the preset screening criteria, a total of 97 blood-entry drug-related candidate components were obtained, of which 50 were classified as prototype components and 47 as candidate metabolites. Detailed information is provided in Supplementary Table S2.

2.2.2. Chemical Classes and Relative Peak Area Distribution of Blood-Entry Prototype Components

To analyze the chemical composition characteristics of the blood-entry prototype components, the 50 screened prototype components were assigned to their respective chemical classes. In terms of the number of components, flavonoids were the most abundant, with 25 components accounting for 50.00%; phenylpropanoids numbered 10 (20.00%) and unclassified components 6 (12.00%); terpenoids, organic heterocyclic compounds, and lipids and lipid-like molecules each accounted for 2 components (4.00% each); and fatty acyls, amino acids and peptides, and nucleotides and their derivatives each accounted for 1 component (2.00% each). Flavonoids and phenylpropanoids accounted for relatively high proportions of the blood-entry prototype components, with flavonoids having the highest numerical proportion (Figure 4A).

Total peak area normalization was further performed based on the average peak areas of the 50 blood-entry prototype components in the CHHQ decoction samples. Taking the sum of the average peak areas of the 50 prototype components as 100%, the normalized peak area proportion of each component was calculated and summarized by chemical class. The results showed that flavonoids had the highest normalized peak area proportion, at 87.96%; unclassified components and phenylpropanoids accounted for 6.63% and 5.03%, respectively; and lipids and lipid-like molecules, organic heterocyclic compounds, terpenoids, fatty acyls, nucleotides and their derivatives, and amino acids and peptides accounted for 0.13%, 0.09%, 0.06%, 0.04%, 0.04%, and 0.01%, respectively (Figure 4B).

Combining the number of candidate components and the normalized peak area composition in the CHHQ decoction samples, flavonoids showed the highest proportions among the 50 blood-entry prototype components screened in this study, in terms of both number and normalized peak area in the decoction.

2.2.3. Blood-Entry Metabolites and Their Potential in Vivo Biotransformation Characteristics

Among the 97 blood-entry related components, in addition to the 50 prototype components, a total of 47 metabolites were screened. Based on the structural differences between the candidate parent compounds and the corresponding metabolites and the available mass spectrometric information, the possible types of in vivo biotransformation involved were analyzed. The results showed that the related components may undergo multiple phase I and phase II biotransformations. Phase I biotransformations mainly included hydroxylation, oxidation, reduction, hydrolysis, dealkylation, deglycosylation, demethylation, and carboxylation, whereas phase II biotransformations mainly included glucuronidation, sulfation, and methylation. Some metabolites may involve two or more consecutive or combined transformation processes.

Representative biotransformation relationships are shown in the figure: Figure 4C shows consecutive hydroxylation followed by further glucuronidation; Figure 4D,E show glucuronidation and sulfation, respectively; Figure 4F shows hydroxylation followed by further reduction; Figure 4G,H show dealkylation and deglycosylation, respectively; Figure 4I,J show carboxylation and oxidation, respectively; Figure 4K involves consecutive transformations such as hydrolysis followed by further hydroxylation; and Figure 4L involves consecutive transformation processes including demethylation, methylation, and sulfation. These results indicate that the blood-entry related components of CHHQ can present multiple phase I and phase II biotransformation forms and can yield products of multistep consecutive transformations. The above metabolic relationships were inferred mainly from the structural differences between the candidate parent compounds and the corresponding metabolites and from the mass spectrometric information, and the exact sequence of metabolic reactions and the modification sites remain to be further verified. In summary, the analysis of drug-containing serum screened 50 prototype components and 47 metabolites and revealed multiple phase I and phase II biotransformation characteristics. On this basis, the drug-related LC-MS features in tissue samples from six brain regions—CA, PFC, NAc, VTA, LHb, and DRN—were further compared to examine their cross-brain-region detection patterns.

2.3. Brain-Targeting Components in Reward and Anti-Reward Circuit-Related Brain Regions of the CHHQ

The CA, PFC, NAc, VTA, LHb, and DRN are all brain regions that have received attention in depression-related research, and some of the neural circuit relationships among these regions are shown in Figure 5A. Based on this, this study further compared the LC-MS candidate features in the tissues of the six brain regions to examine their cross-brain-region detection patterns and to screen candidate features with relatively pronounced regional restriction.

2.3.1. Integration and Detection Patterns of LC-MS Candidate Features Across the Six Brain Regions

The candidate features obtained from LC-MS analysis of the six brain regions (CA, PFC, NAc, VTA, LHb, and DRN) were integrated and deduplicated, and their cross-brain-region detection patterns were analyzed according to the detection of each feature in the different brain regions. The results showed that the detection range of the candidate features across the six brain regions varied: some features could be detected in multiple brain regions, whereas others were detected in only a single brain region. The complete integrated results of the six-brain-region candidate features are provided in Supplementary Table S3. As features detected in only a single brain region have more pronounced regional restriction, and to further utilize the cross-brain-region detection information to screen noteworthy candidate features, the single-region detection features of each brain region were subsequently compared.

2.3.2. Comparison of Relatively Region-Specific Detection Features Across Brain Regions

Based on the integrated six-brain-region LC-MS candidate features, the candidate features detected in only a single brain region under the current analytical conditions were further counted, hereafter referred to as “relatively region-specific detection features”. The results showed that three and four relatively region-specific detection features were present in the CA and DRN, respectively, whereas no features meeting the above criteria were found in the PFC, NAc, VTA, or LHb (Figure 5B). These results indicate that, among the six brain regions examined in this study, both the CA and DRN contained candidate features with relatively pronounced region-restricted detection patterns. Based on the purpose of exploratory candidate screening and combined with the important role of the DRN in serotonergic regulation and depression-related neural circuits, subsequent analysis further focused on the four relatively specific candidate features detected in the DRN.

2.3.3. Metabolite Annotation of the DRN Relatively Specific Candidate Features

The four candidate features detected only in the DRN were further annotated by combining the original metabolite analysis results with high-resolution mass spectrometric information. The original analysis results annotated them as 10-methoxy-20-O-acetylcamptothecin_M1, triptonoditerpenic acid_M1, L-phenylalanine_M1, and przewalskinic acid A_M1, and their composite scores all met the candidate component screening criteria preset in this study. As the above structures are all candidate annotations obtained based on the current mass spectrometric data, this paper uniformly treats these four features as “DRN relatively specific candidate metabolites” for subsequent analysis rather than regarding them as compounds with fully confirmed structures. The main LC-MS features, candidate parent compounds, and related annotation information of the four candidate metabolites are shown in Table 4.

Table 1. Main UPLC-MS/MS information, candidate parent compounds, and potential biotransformation annotations of the DRN relatively specific candidate metabolites.

Candidate metabolite	RT (min)	Measured m/z	Molecular formula	Adduct ion	Mass error (ppm)	MS/MS evidence	Parent compound	Annotated transformation(s)
10-methoxy-20-O-acetylcamptothecin_M1	5.42	454.1643	C23H20N2O7	[M+NH4]+	−7.49	Available	10-methoxy-20-O-acetylcamptothecin	Hydroxylation
Triptonoditerpenic acid_M1	5.95	551.2183	C27H36O12	[M−H]−	−8.89	Available	Triptonoditerpenic acid	Oxidation; Isopropenyl oxidation; Glucuronidation
L-Phenylalanine_M1	1.46	418.0972	C15H19NO10	[M+FA−H]−	4.54	Not obtained	L-Phenylalanine	Hydroxylation; Hydroxylation; Glucuronidation
Przewalskinic acid A_M1	4.04	575.0961	C24H24O15	[M+Na]+	8.00	Not obtained	Przewalskinic acid A	Reduction; Hydroxylation; Glucuronidation

Note: Parent compound and annotated transformation(s) were derived from the corresponding annotation fields in the original metabolite analysis results. All four candidate parent compounds could be detected among the candidate chemical components of the CHHQ decoction. “Available” indicates that a valid MS/MS spectrum was obtained under the current data acquisition conditions; “Not obtained” indicates that no valid MS/MS spectrum was acquired. The above candidate parent compounds and biotransformation relationships are annotation results and do not represent confirmed in vivo parent–metabolite transformation processes.

Table 1. Main UPLC-MS/MS information, candidate parent compounds, and potential biotransformation annotations of the DRN relatively specific candidate metabolites.

Candidate metabolite	RT (min)	Measured m/z	Molecular formula	Adduct ion	Mass error (ppm)	MS/MS evidence	Parent compound	Annotated transformation(s)
10-methoxy-20-O-acetylcamptothecin_M1	5.42	454.1643	C23H20N2O7	[M+NH4]+	−7.49	Available	10-methoxy-20-O-acetylcamptothecin	Hydroxylation
Triptonoditerpenic acid_M1	5.95	551.2183	C27H36O12	[M−H]−	−8.89	Available	Triptonoditerpenic acid	Oxidation; Isopropenyl oxidation; Glucuronidation
L-Phenylalanine_M1	1.46	418.0972	C15H19NO10	[M+FA−H]−	4.54	Not obtained	L-Phenylalanine	Hydroxylation; Hydroxylation; Glucuronidation
Przewalskinic acid A_M1	4.04	575.0961	C24H24O15	[M+Na]+	8.00	Not obtained	Przewalskinic acid A	Reduction; Hydroxylation; Glucuronidation

Note: Parent compound and annotated transformation(s) were derived from the corresponding annotation fields in the original metabolite analysis results. All four candidate parent compounds could be detected among the candidate chemical components of the CHHQ decoction. “Available” indicates that a valid MS/MS spectrum was obtained under the current data acquisition conditions; “Not obtained” indicates that no valid MS/MS spectrum was acquired. The above candidate parent compounds and biotransformation relationships are annotation results and do not represent confirmed in vivo parent–metabolite transformation processes.

2.3.4. Chromatographic and Mass Spectrometric Annotation Evidence for the DRN Relatively Specific Candidate Metabolites

To further evaluate the structural annotation evidence for the four DRN relatively specific candidate metabolites, their extracted ion chromatographic information and available MS/MS spectra were analyzed in combination (Figure 6).Among them, valid MS/MS spectra were obtained for both 10-methoxy-20-O-acetylcamptothecin_M1 and triptonoditerpenic acid_M1, in which corresponding MS/MS fragment information could be observed, providing further support for their current candidate structural annotations (Figure 6A,B). In contrast, no valid MS/MS spectra were obtained for L-phenylalanine_M1 or przewalskinic acid A_M1 under the current data acquisition conditions, and therefore no further MS/MS fragment analysis was performed (Figure 6C,D). Accordingly, the strength of structural annotation evidence differed among the four candidate metabolites: the first two were supported by MS/MS information, whereas the latter two were currently assigned mainly based on accurate mass, molecular formula, and the original metabolite annotation results. Subsequent analyses were all based on their current candidate structures.

2.3.5. Candidate Parent Compounds and Potential Biotransformation Annotations of the DRN Candidate Metabolites

To further analyze the possible parent sources and biotransformation relationships of the four DRN relatively specific candidate metabolites, this study combined the Parent Compound and Transformations annotation information from the original metabolite analysis results and further compared the molecular formula changes and theoretical neutral mass differences between the candidate parent compounds and the candidate metabolites to evaluate the consistency of the relevant annotations in terms of elemental composition.

The original data assigned 10-methoxy-20-O-acetylcamptothecin, triptonoditerpenic acid, L-phenylalanine, and przewalskinic acid A as the candidate parent compounds corresponding to the four candidate metabolites, and the above candidate parent compounds could all be detected among the candidate chemical components of the CHHQ decoction. Specifically, the potential biotransformation type corresponding to 10-methoxy-20-O-acetylcamptothecin_M1 was hydroxylation, with a net molecular formula change of +O relative to the candidate parent and a theoretical neutral mass increase of 15.9949 Da; triptonoditerpenic acid_M1 corresponded to oxidation, isopropenyl oxidation, and glucuronidation, with a net molecular formula change of +C₆H₈O₈ and a theoretical neutral mass increase of 208.0219 Da; L-phenylalanine_M1 corresponded to two-step hydroxylation and glucuronidation, with the same net molecular formula change of +C₆H₈O₈ and a theoretical neutral mass increase of 208.0219 Da; and przewalskinic acid A_M1 corresponded to reduction, hydroxylation, and glucuronidation, with a net molecular formula change of +C₆H₁₀O₇ and a theoretical neutral mass increase of 194.0427 Da (Figure 5). The above molecular formula changes and theoretical mass differences were consistent with the biotransformation annotations in the original data at the level of elemental composition.

It should be noted that the above candidate parent compounds and biotransformation types were all derived from the annotation information in the original metabolite analysis results, and the comparison of molecular formula changes and theoretical neutral mass differences in this study was used only to evaluate the consistency of the relevant annotations in terms of elemental composition. As direct evidence such as authentic standard comparison, metabolic enzyme experiments, or isotope tracing is currently lacking, the above results represent only the potential annotated biotransformation relationships between the candidate parent compounds and the DRN candidate metabolites; the exact reaction sequence, modification sites, and actual in vivo metabolic processes still require further experimental verification.

2.4. Network Pharmacology Analysis of Potential Targets and Biological Processes of the DRN Relatively Specific Candidate Metabolites

To further explore the biological processes potentially involving the four DRN relatively specific candidate metabolites, target prediction was carried out based on their current candidate structures, and network pharmacology analysis was conducted in combination with depression-related targets. Given that the structural assignments of the above metabolites still belong to candidate annotations and that network pharmacology analysis inherently relies on databases and computational prediction, the results of this section are mainly used to propose candidate targets and biological directions for subsequent experimental verification rather than as evidence of a direct mechanism of action.

2.4.1. Intersection of Candidate Metabolite-Related Targets and Depression-Related Targets, and PPI Network Analysis

Potential target prediction was performed based on the current putative structures of the four DRN relatively specific candidate metabolites, and a total of 404 candidate targets were obtained after merging and removing duplicates. Meanwhile, depression-related targets were retrieved through the GeneCards, OMIM, and DisGeNET databases using “depression” and “major depressive disorder” as keywords, and validated functional targets regulating depression in the DRN and lateral habenula (LHb) were supplemented through literature research; after merging, standardization, and deduplication, 18305 disease-related targets were obtained. Intersection analysis of the candidate metabolite-related targets and the depression-related targets yielded a total of 357 common targets, while another 47 candidate metabolite-related targets were not included in the currently constructed depression-related target set (Figure 7A).

It should be noted that the depression-related target set obtained in this study was large in scale; therefore, the 357 intersecting targets mainly reflect the overlap between the predicted targets and the broad set of depression-related genes. This result was used to define the candidate range for subsequent protein interaction and functional enrichment analyses, and the relatively high intersection proportion should not be interpreted as the candidate metabolites having specific effects on depression-related targets.

The 357 common targets were further imported into the STRING database to construct a protein–protein interaction (PPI) network under the condition of a combined score ≥ 0.400. The connectivity characteristics of network nodes were evaluated according to their degree values, and nodes such as TNF, ALB, STAT3, and BCL2 showed relatively high network connectivity (Figure 7B). These high-connectivity nodes suggest that they have numerous protein interaction relationships within the current predicted target network, but their network topological positions are not equivalent to the candidate metabolites having direct pharmacological effects on the corresponding proteins.

2.4.2. GO Functional and KEGG Pathway Enrichment Analyses

To further describe the functional features involving the 357 common targets, Gene Ontology (GO) functional enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on them.

GO enrichment analysis showed that, at the biological process (BP) level, the common targets were mainly involved in processes such as the response to oxidative stress, regulation of the inflammatory response, regulation of membrane potential, glutamatergic synaptic transmission, neurotransmitter uptake/reuptake, and regulation of neuronal apoptosis. At the cellular component (CC) level, the relevant targets were mainly enriched in cellular structures such as the synaptic membrane, presynaptic membrane, and postsynaptic membrane; at the molecular function (MF) level, they mainly involved functions related to neurotransmitter receptor activity and neurotransmitter transport (Figure 7D). Overall, the GO enrichment results showed that the predicted target set involves both synaptic/neurotransmitter regulation and biological processes related to inflammatory and stress responses.

KEGG pathway enrichment analysis further showed that the common targets were distributed across multiple pathways related to neurotransmitter transmission, synaptic function, and inflammatory signaling, including neuroactive ligand–receptor interaction, glutamatergic synapse, serotonergic synapse, the TNF signaling pathway, and the NF-kappa B signaling pathway, while also involving signaling pathways such as PI3K–Akt, cAMP, and MAPK (Figure 7C). These enrichment results suggest that the predicted targets of the four DRN relatively specific candidate metabolites may involve multiple interconnected neurotransmitter regulatory and inflammation-related processes rather than being concentrated in a single signaling pathway.

Combining the GO and KEGG analysis results, neurotransmitter regulation, synaptic function, and the inflammatory response constitute the relatively prominent functional directions in the current predicted target set. Among them, SLC6A4 and NOS1 were both located in the candidate metabolite–depression common target set and are associated with neurotransmitter-related functions; TNF had high node connectivity in the PPI network, and TNF- and NF-κB-related signaling was also reflected in the functional enrichment analysis. Therefore, the above results provide candidate targets and pathway clues for further analysis focusing on local neurotransmitter regulation and the inflammatory response in the DRN, but these predicted associations still need to be confirmed in combination with direct experimental verification.

To further focus on the anti-inflammatory direction, six anti-inflammatory candidate metabolites—undulatone_M1; (10S)-15-methoxypinusolidic acid_M1; przewalskinic acid A_M1; L-phenylalanine_M1; 10-methoxy-20-O-acetylcamptothecin_M1; and 1-O-vanillic acid-6-O-(3″,5″-dimethoxygalloyl)-β-D-glycoside_M1—were selected to construct an anti-inflammatory candidate metabolite–core target–KEGG pathway network presented as a Sankey–bubble diagram (Figure 7E). The linked core targets included proteins such as TNF, STAT3, RELA, NFKB1, IKBKB, JUN, AKT1, MAPK3, and PIK3CA, and the corresponding pathways were concentrated in inflammation-related signaling. The six metabolites in this panel form an anti-inflammatory-focused subset and differ in scope from the four DRN relatively specific candidate metabolites used above; the criteria used to select these six metabolites are to be specified.

Within the enriched inflammatory pathways, the core targets identified in this study were mapped onto the TNF signaling pathway (hsa04668) and the NOD-like receptor signaling pathway (hsa04621) (Figure 7F), illustrating their positions within these innate immune and inflammatory cascades. The meaning of the red and green highlighting is to be specified in a legend.

2.5. Molecular Docking Analysis of the DRN Relatively Specific Candidate Metabolites with Candidate Target Proteins

To further explore the potential molecular interactions between the DRN relatively specific candidate metabolites and candidate target proteins related to neurotransmitter regulation and inflammation, SERT, nNOS, TNF, NLRP3, and caspase-1 were selected as candidate target proteins based on the aforementioned network pharmacology analysis results and the direction of subsequent DRN pharmacodynamic evaluation, and molecular docking analysis was performed with the four DRN relatively specific candidate metabolites.

The molecular docking results showed that the docking scores of the four candidate metabolites with the five candidate target proteins ranged from −6.01 to −10.50 (GlideScore; more negative values indicate stronger predicted binding), and the different candidate metabolite–target protein combinations presented distinct predicted binding characteristics (Figure 8A; Supplementary Table S4). To further observe the predicted binding modes of representative complexes, based on comprehensive consideration of the docking scores, the classes of candidate target proteins, and the subsequent evaluation indicators of this study, przewalskinic acid A_M1–SERT, przewalskinic acid A_M1–nNOS, 10-methoxy-20-O-acetylcamptothecin_M1–TNF, 10-methoxy-20-O-acetylcamptothecin_M1–NLRP3, and przewalskinic acid A_M1–Caspase-1 were selected for visualization analysis (Figure 8B–F).

Among the candidate target proteins related to neurotransmitter regulation, the docking score of przewalskinic acid A_M1 with SERT was −8.90 (GlideScore). In the predicted binding pocket, it formed conventional hydrogen bonds with Asp98 and Phe335, carbon–hydrogen bonds with Gly100 and Gly442, π–π stacking with Phe341, and hydrophobic interactions with Ile172 and Val501 (Figure 8B). The docking score of przewalskinic acid A_M1 with nNOS was −9.60 (GlideScore); it formed a conventional hydrogen bond and a π-donor hydrogen bond with Trp592, a carbon–hydrogen bond with Ser462, and π–π stacking with Phe589 and Trp414 (Figure 8C).

Among the inflammation-related candidate target proteins, the docking score of 10-methoxy-20-O-acetylcamptothecin_M1 with TNF was −8.78 (GlideScore). It formed a hydrogen bond with Gly121 and hydrophobic interactions with residues such as Leu55, Leu57, Leu120, Tyr59, and Tyr119, and its fused-ring structure was located within the hydrophobic region formed by leucine and tyrosine residues (Figure 8D). The docking score of this candidate metabolite with NLRP3 was −9.57 (GlideScore); it formed hydrogen bonds with Arg38, Glu177, Gly100, Leu284, and Tyr252, hydrophobic interactions with Ile105, Ile392, Leu284, and Trp287, and π–π stacking with Trp287 (Figure 8E). In addition, the docking score of przewalskinic acid A_M1 with caspase-1 was −7.77 (GlideScore); it formed hydrogen bonds with Asp157, Gln142, His404, Ile155, Lys158, Met156, and Tyr153, hydrophobic contacts with Ala141, Ile152, Leu138, and Pro154, and π–π stacking with Phe401, Trp145, and Tyr153 (Figure 8F).

Overall, the molecular docking results showed that some of the DRN relatively specific candidate metabolites were able to form hydrogen bonds, hydrophobic interactions, and π-related interactions in the predicted binding pockets of SERT and nNOS, as well as TNF, NLRP3, and caspase-1, presenting distinct predicted molecular recognition features and providing further computational support for the neurotransmitter regulation and inflammation-related candidate directions suggested by the aforementioned network pharmacology.

2.6. Effects of CHHQ on Body Weight and Serum Liver/Kidney Function-Related Indicators in CUMS Rats

The experimental grouping, CUMS modeling, and drug intervention workflow is shown in Figure 9A. During the experiment, the body weight of rats in each group was monitored weekly, and serum ALT, AST, BUN, and creatinine (Cr) levels were measured at the end of the experiment to evaluate changes in body weight and serum biochemical indicators related to liver and kidney function during CUMS and CHHQ intervention.

2.6.1. Effects on Body Weight in CUMS Rats

The body weight changes of rats in each group during the experiment are shown in Figure 9B. At the beginning of the experiment, the body weights of rats in all groups were generally comparable. As the experiment progressed, the body weight of rats in the Control group increased continuously, whereas body weight gain was attenuated in the Model group. The CHHQ-treated groups and the Fluoxetine group also showed gradual increases in body weight during the experimental period.

At week 6, the body weight of the Model group was significantly lower than that of the Control group (p < 0.001; Figure 9C). Compared with the Model group, the body weights of the CHHQM and CHHQH groups were significantly increased (both p < 0.05), whereas no statistically significant difference was indicated for the CHHQL group in the figure. These results indicate that CUMS suppressed body weight gain and that medium- and high-dose CHHQ treatment partially ameliorated this change.

2.6.2. Effects on Serum Liver/Kidney Function-Related Indicators in CUMS Rats

At the end of the experiment, serum ALT, AST, BUN, and Cr levels were measured in each group (Figure 9D–G). The ALT and AST results are shown in Figure 9D and E, respectively. Although individual variation was observed among the groups, the low-, medium-, and high-dose CHHQ groups did not show a consistent dose-dependent increase in either indicator.

Similarly, the BUN and Cr results are shown in Figure 9F and G, respectively. No statistically significant differences in BUN or Cr were indicated between the Model group and the CHHQ-treated groups in the figure, and no consistent dose-dependent increase was observed across the three CHHQ dose groups.

Taken together, under the dose and 2-week intervention conditions used in this study, CHHQ treatment partially ameliorated the CUMS-associated reduction in body weight, particularly in the medium- and high-dose groups, while serum ALT, AST, BUN, and Cr did not show a consistent dose-dependent elevation.

2.7. CHHQ Ameliorates Depression-like Behaviors and HPA Axis Hyperactivity in CUMS Rats

To evaluate the effects of CHHQ on CUMS-induced depression-like behaviors, the sucrose preference test (SPT), open field test (OFT), and forced swimming test (FST) were performed. In addition, serum levels of corticotropin-releasing hormone (CRH) and corticosterone (CORT) were measured to assess hypothalamic–pituitary–adrenal (HPA) axis activity.

2.7.1. The Herb Pair Ameliorates the CUMS-Induced Decrease in Sucrose Preference

The SPT results are shown in Figure 10A. Before treatment, the sucrose preference index (SPI) was significantly lower in the Model, CHHQL, CHHQM, CHHQH, and Fluoxetine groups than in the Control group (all p < 0.001), indicating a marked reduction in sucrose preference following CUMS exposure.

After treatment, the SPI remained significantly lower in the Model group than in the Control group (p < 0.001). Compared with the Model group, the SPI was significantly increased in the CHHQL (p < 0.01), CHHQM (p < 0.05), CHHQH (p < 0.01), and Fluoxetine (p < 0.001) groups. These results indicate that CHHQ treatment ameliorated the CUMS-induced reduction in sucrose preference.

2.7.2. The Herb Pair Reduces the Immobility Time Ratio in the FST

As shown in Figure 10C, the immobility time ratio in the FST was significantly increased in the Model group compared with the Control group (p < 0.001). Compared with the Model group, the immobility time ratio was significantly decreased in the CHHQL, CHHQM, CHHQH, and Fluoxetine groups (all p < 0.01). These results indicate that CHHQ treatment attenuated the CUMS-induced increase in the immobility time ratio.

2.7.3. The Herb Pair Ameliorates the CUMS-Induced Reduction in Locomotor Activity

As shown in Figure 10B, the total distance traveled in the OFT was significantly lower in the Model group than in the Control group (p < 0.05). Compared with the Model group, the total distance traveled was significantly increased in the CHHQL, CHHQM, and CHHQH groups (all p < 0.05). No statistically significant difference between the Fluoxetine and Model groups was indicated in the figure. These results indicate that CHHQ treatment ameliorated the CUMS-induced reduction in locomotor activity.

2.7.4. The Herb Pair Attenuates CUMS-Induced HPA Axis Hyperactivity

As shown in Figure 10D, the serum CRH level was significantly increased in the Model group compared with the Control group (p < 0.0001). Compared with the Model group, serum CRH levels were significantly decreased in the CHHQL, CHHQM, and CHHQH groups (all p < 0.001), as well as in the Fluoxetine group (p < 0.01).

Similarly, the serum CORT level was significantly increased in the Model group compared with the Control group (p < 0.001; Figure 10E). Compared with the Model group, serum CORT levels were significantly decreased in the CHHQL (p < 0.01), CHHQM (p < 0.05), CHHQH (p < 0.05), and Fluoxetine (p < 0.05) groups. These results indicate that CHHQ treatment attenuated the CUMS-induced increases in serum CRH and CORT levels.

Collectively, CUMS induced depression-like behavioral changes, as evidenced by decreased sucrose preference, an increased immobility time ratio in the FST, and reduced total distance traveled in the OFT, accompanied by elevated serum CRH and CORT levels. CHHQ treatment ameliorated these behavioral changes and reduced the CUMS-induced elevations in serum CRH and CORT levels.

2.8. Effects of CHHQ on Microglial Activation-Related Phenotypes in the DRN of CUMS Rats

To evaluate microglial activation-related changes in the dorsal raphe nucleus (DRN) following CUMS exposure and CHHQ treatment, double-label immunofluorescence staining for IBA-1 and CD68 was performed. IBA-1 was used as a marker of microglia, whereas CD68 was used as a microglial activation-related marker.

As shown in Figure 11, the IBA-1-positive and CD68-positive fluorescence signals in the DRN were visibly increased in the Model group compared with the Control group. Following CHHQ treatment, the fluorescence signals were reduced to varying degrees in the CHHQL, CHHQM, and CHHQH groups. A similar reduction was observed in the Fluoxetine group.

Quantitative analysis of mean fluorescence intensity (MFI) further supported these observations (Figure 12A,B). IBA-1 MFI was significantly increased in the Model group compared with the Control group (p < 0.001; Figure 12A). Compared with the Model group, IBA-1 MFI was significantly decreased in the CHHQL, CHHQM, CHHQH, and Fluoxetine groups (all p < 0.001).

Similarly, CD68 MFI was significantly increased in the Model group compared with the Control group (p < 0.05; Figure 12B). Compared with the Model group, CD68 MFI was significantly decreased in the CHHQH and Fluoxetine groups (both p < 0.05), whereas no statistically significant differences were indicated for the CHHQL and CHHQM groups.

Taken together, these results indicate that CUMS enhanced microglial activation-related fluorescence signals in the DRN, whereas CHHQ treatment attenuated these changes to varying degrees.

2.9. Effects of CHHQ on Inflammatory Cytokine Levels in the DRN of CUMS Rats

To evaluate the effects of CHHQ on the local inflammatory response in the DRN of CUMS rats, the levels of TNF-α, IL-18, and IL-1β in DRN tissues were measured.

As shown in Figure 12C, the TNF-α level in the DRN was significantly increased in the Model group compared with the Control group (p < 0.0001). Compared with the Model group, TNF-α levels were significantly decreased in the CHHQM and CHHQH groups (both p < 0.001), whereas no statistically significant difference was indicated for the CHHQL group. The Fluoxetine group also showed a significant decrease in TNF-α level (p < 0.0001).

The IL-18 level was significantly increased in the Model group compared with the Control group (p < 0.01; Figure 12D). Compared with the Model group, IL-18 was significantly decreased in the CHHQH and Fluoxetine groups (both p < 0.05), whereas no statistically significant differences were indicated for the CHHQL and CHHQM groups.

As shown in Figure 12E, the IL-1β level was significantly increased in the Model group compared with the Control group (p < 0.0001). Compared with the Model group, IL-1β levels were significantly decreased in the CHHQL (p < 0.05), CHHQM (p < 0.01), CHHQH (p < 0.0001), and Fluoxetine (p < 0.0001) groups.

Collectively, these results indicate that CUMS increased the levels of the pro-inflammatory cytokines TNF-α, IL-18, and IL-1β in the DRN, whereas CHHQ treatment attenuated these increases to varying degrees.

2.10. Effects of CHHQ on NF-κB1, NLRP3, Caspase-1, and SERT Protein Expression in the DRN of CUMS Rats

To further evaluate the effects of CHHQ on inflammation-related and serotonin transport-related proteins in the DRN of CUMS rats, Western blotting was performed to determine the protein expression levels of NF-κB1, NLRP3, caspase-1, and SERT. Representative Western blot bands are shown in Figure 12F, and the corresponding quantitative analyses are shown in Figure 12G–J.

Caspase-1 protein expression was significantly increased in the Model group compared with the Control group (p < 0.0001; Figure 12G). Compared with the Model group, caspase-1 expression was significantly decreased in the CHHQL, CHHQM, CHHQH, and Fluoxetine groups (all p < 0.01).

As shown in Figure 12H, NLRP3 protein expression was significantly increased in the Model group compared with the Control group (p < 0.01). Compared with the Model group, NLRP3 expression was significantly decreased in the CHHQH and Fluoxetine groups (both p < 0.05), whereas no statistically significant differences were indicated for the CHHQL and CHHQM groups.

SERT protein expression was significantly increased in the Model group compared with the Control group (p < 0.0001; Figure 12I). Compared with the Model group, SERT expression was significantly decreased in the CHHQL and CHHQM groups (both p < 0.001), CHHQH group (p < 0.0001), and Fluoxetine group (p < 0.0001).

As shown in Figure 12J, NF-κB1 protein expression was significantly increased in the Model group compared with the Control group (p < 0.01). Compared with the Model group, NF-κB1 expression was significantly decreased in the CHHQM (p < 0.01) and CHHQH (p < 0.0001) groups, whereas no statistically significant difference was indicated for the CHHQL group. NF-κB1 expression was also significantly decreased in the Fluoxetine group (p < 0.0001).

Taken together, these results indicate that CUMS was associated with increased expression of NF-κB1, NLRP3, caspase-1, and SERT in the DRN, whereas CHHQ treatment attenuated these alterations to varying degrees.

3. Discussion

Depression is increasingly recognized as a disorder in which neuroinflammatory and serotonergic disturbances interact rather than act in isolation [4,6,8]. In the present study, an integrated strategy comprising UPLC-MS/MS-based serum pharmacochemistry, brain-region-resolved distribution profiling, network pharmacology, molecular docking and in vivo validation in a CUMS rat model was applied to characterize the antidepressant material basis of the Chaihu–Huangqin (CHHQ) drug pair. The principal findings were threefold. First, serum pharmacochemistry systematically characterized 97 constituents absorbed into the systemic circulation, comprising prototype constituents and metabolites generated through phase Ⅰ/Ⅱ biotransformation. Second, distribution profiling across six depression-related brain regions (hippocampal CA area, dorsal raphe nucleus (DRN), lateral habenula, nucleus accumbens, prefrontal cortex and ventral tegmental area) revealed distinct region-specific exposure patterns, among which the DRN exhibited the largest number of region-specific candidates, including 4 relatively DRN-specific candidate features, and was therefore selected as the focus brain region on the basis of the constituents detected in it. Third, in CUMS rats, the CHHQ drug pair attenuated depressive-like behaviors, ameliorated DRN microglial activation and local inflammatory responses, and normalized the NF-κB/NLRP3/Caspase-1 axis and SERT expression. Taken together with the chemical profiling, network pharmacology and pathway validation results, the findings support a multi-component pharmacological profile of the CHHQ drug pair acting through coordinated modulation of neuroinflammatory and serotonergic signaling in the DRN rather than through a single isolated target.

The constituents absorbed into the systemic circulation provide a potential chemical basis for the observed pharmacological effects. In the present study, the absorbed constituents were systematically characterized by UPLC-MS/MS-based serum pharmacochemistry, distinguishing candidate prototype constituents from candidate metabolites and delineating their phase Ⅰ/Ⅱ biotransformation features. This approach is consistent with the current direction of herb-pair pharmacology, in which chemical exposure is characterized in vivo before mechanisms are proposed [24], and with comparable exposure- and distribution-oriented designs reported for other herbal medicines and herb pairs [28]. Several absorbed constituents have documented activities relevant to the observed pharmacology: saikosaponin A, a representative saponin of Chaihu, relieved CUMS-induced depressive-like behavior in rats [13], saikosaponin D alleviated depressive-like behavior in CUMS mice by promoting NLRP3 ubiquitination and inhibiting inflammasome activation [17], and the BS drug pair as a whole has been shown to implicate inflammatory and survival-related pathways in CUMS mice [9]. Moreover, the metabolic pattern observed here is in agreement with the reported oral pharmacokinetics of Huangqin flavonoids: when the flavonoid glucuronides baicalin, wogonoside and oroxylin A 7-O-β-D-glucuronide were administered as a Scutellariae Radix extract, the glucuronides and their aglycones were simultaneously detectable in rat plasma throughout the experimental period (48 h), with systemic exposure shared between prototype and conjugated forms [14]. This is consistent with the co-detection of prototype constituents and phase Ⅱ metabolites in the present serum profile, and suggests sustained systemic availability of the flavonoid fraction after oral administration. Although the contribution of each individual constituent remains to be clarified, the coexistence of multiple constituents with potentially relevant activities suggests that the pair may contribute to the observed pharmacological effects through coordinated modulation of several pathological processes. This interpretation is consistent with the multicomponent pharmacological characteristics of herbal drug pairs rather than a single-compound mechanism [9,24].

For constituents in the systemic circulation, crossing the blood–brain barrier is the prerequisite for reaching brain tissue, and the present study further resolved their distribution across individual brain regions. The absorbed constituents were differentially detected in the six depression-related brain regions examined, with region-specific differences in both the number and the composition of detected candidates. This brain-region-resolved characterization extends previous serum pharmacochemistry studies of antidepressant herbal pairs, which successfully identified absorbed constituents but did not trace their regional distribution within the brain [11]. Although the pharmacokinetics and oral bioavailability of individual constituents of other herbs have been characterized in some detail [26], to the best of our knowledge, the present work is the first to report the distribution of the absorbed constituents of the CHHQ drug pair in the six depression-related brain regions of CUMS rats. The present finding is also consistent with the growing recognition that neuroinflammatory responses are regionally heterogeneous within the brain and that their regional specificity contributes to the associated behavioral comorbidities [21]. The region-specific exposure observed here suggests that systemic absorption does not translate into uniform brain exposure, and that brain-region-resolved profiling is an essential intermediate step between serum pharmacochemistry and target-level investigations.

Among the six regions examined, the DRN showed the largest number of region-specific candidates, and it was therefore selected as the focus brain region. The DRN is the principal source of serotonergic projections to the forebrain and limbic system and constitutes a key hub in mood and stress regulation [8], and NLRP3 activation within DRN microglia and neurons has recently been shown to be essential for depressive-like phenotypes in CUMS mice [6]. Microglia in the DRN are increasingly regarded as a therapeutic target for stress-related mood disorders [20]. The convergence of region-specific chemical exposure with this mechanistic background makes the DRN a plausible site at which the absorbed constituents of the pair may engage depressive-like circuitry. In addition, constituents of the pair have documented antidepressant-related activities: saikosaponin A relieved CUMS-induced depressive-like behavior and up-regulated hippocampal dopamine via PRRT2 [13], and other constituents of the pair have similarly been reported to act on neuroinflammatory targets in the central nervous system [15,16,31], supporting the biological plausibility of the candidate constituents prioritized here. Notably, the present study is, to our knowledge, the first to link the region-specific brain exposure of a classical herb pair to the DRN-centered neuroinflammatory circuitry.

Network pharmacology initiated from the DRN-specific candidates prioritized inflammatory and serotonergic signaling, including the nuclear factor-κB (NF-κB)/NOD-like receptor family pyrin domain-containing 3 (NLRP3)/Caspase-1 axis and SERT, as candidate pathways linking the region-specific constituents to the DRN-centered pharmacology. This design differs from conventional network pharmacology studies of herbal formulas, in which candidate targets are predicted from the full constituent library of the herbs [12]; initiating the prediction from chemicals with proven regional exposure restricts the candidate space to constituents that demonstrably reach the brain region of interest and thereby improves the biological credibility of the prioritized pairs. Similar exposure-to-target strategies have recently been advocated for herb-pair research [24], and comparable designs have been applied to other herbal medicines and herb pairs [28], which supports the rationale of the present design. It should be emphasized, however, that network pharmacology is predictive by nature. Its role in this study was not to establish a mechanism, but to prioritize candidate constituent–target pairs and guide the integration of the complementary experimental findings.

To further explore the potential molecular interactions between the DRN-specific candidate metabolites and the neurotransmitter- and inflammation-related candidate target proteins, and on the basis of the above network pharmacology results and the subsequent DRN pharmacodynamic evaluation, SERT, nNOS, TNF, NLRP3 and Caspase-1 were selected as candidate target proteins for molecular docking with the four DRN-specific candidate metabolites. Molecular docking performed with the Schrödinger Glide module showed that the GlideScore values of the four candidate metabolites against the five candidate target proteins ranged from −6.01 to −10.50, and that different metabolite–target combinations exhibited distinct predicted binding features. These docking results provided hypothesis-generating support for possible constituent–target interactions rather than evidence of experimentally confirmed binding. The prioritized pathways are biologically coherent: the BS drug pair has previously been shown by network analysis and animal validation to implicate IL-6/JAK2/STAT3 and PI3K/AKT signaling in CUMS mice [9], and the inflammatory and serotonergic targets prioritized in the present study further support the involvement of the DRN-specific constituents identified in vivo in these pathways. In addition, network- and docking-based predictions of flavonoid–target interactions have been reported for other central nervous system indications [18,22], indicating that the strategy applied here is consistent with current practice. Together, these observations suggest that the DRN-specific candidates and the inflammatory and serotonergic pathways may be functionally related in the depressive state.

In vivo validation in the CUMS model confirmed the antidepressant-like efficacy of the CHHQ drug pair. Chronic unpredictable mild stress reliably induced depressive-like phenotypes, as reflected by reduced sucrose preference and increased immobility in the forced swimming test, together with reduced body weight, consistent with the established validity of the model [9,10]. Behavioral assessment in the present study comprised the sucrose preference test (SPT), open field test (OFT), and forced swimming test (FST), which were performed after 6 weeks of CUMS and repeated after 2 weeks of drug intervention (n = 6 per group); in the FST, the immobility time of each rat during the last 4 min was recorded, and in the OFT, the total distance traveled over a 5 min session was recorded. The behavioral improvement observed here parallels previous reports in which saikosaponin D and baicalin improved sucrose preference and reduced immobility in CUMS rodents [15,17], and is consistent with the general finding that herbal preparations relieve depressive-like behaviors in chronic stress models [27]. Treatment with the drug pair significantly reversed these behavioral deficits, and no obvious toxic signs were observed during the treatment period. These findings are consistent with previous CUMS studies of the CHHQ drug pair and its individual constituents, in which saikosaponin A, baicalin and the pair as a whole improved sucrose preference and forced swimming outcomes [7,9,10,13].

The molecular findings provide pathway-level evidence for the involvement of NLRP3-related neuroinflammation and serotonergic signaling in the DRN in the antidepressant-like effect of the pair. IBA-1/CD68 immunofluorescence indicated microglial activation in the DRN of CUMS rats, which was attenuated following drug pair treatment; this observation agrees with the demonstrated involvement of microglial NLRP3 inflammasome activation in depressive-like phenotypes, from the pioneering characterization of microglial NLRP3 activation in the prefrontal cortex of depressive rats [4] to the recent cell-type-specific evidence that NLRP3 in DRN microglia and neurons is essential for the depressive phenotype [6]. Consistent with these relationships, treatment normalized the expression of NF-κB, NLRP3 and Caspase-1 in the DRN. Baicalin has been shown to inhibit NLRP3 inflammasome activation [7] and to suppress the GSK3β/NF-κB/NLRP3 pathway [15,18]; saikosaponin D alleviated depressive-like behavior through NLRP3 ubiquitination and inflammasome inhibition [17]; and the Huangqin flavonoids wogonoside and wogonin likewise attenuated microglia-mediated neuroinflammation [22,25,31], and single-cell transcriptomic profiling of inflammation-associated depression has begun to resolve the microglial subtypes involved [19]. Collectively, these studies provide constituent-level plausibility for the pathway changes observed here. In addition, the normalization of SERT expression observed in the present study is notable: chronic stress up-regulates SERT in the DRN and its projection regions [29], chronic unpredictable mild stress selectively increased the SERT–nNOS complex in the DRN of mice, and augmentation of SERT signaling in the DRN is sufficient to drive depressive-like behavior [6]. The reversal of DRN SERT changes may therefore represent a serotonergic correlate of the behavioral improvement.

The most distinctive finding of the present study concerns the DRN itself. Serotonergic plasticity in the DRN has been linked to stress and antidepressant responses at the epigenetic level, as chronic stress dysregulated histone serotonylation in the DRN and manipulation of this mark attenuated stress-related gene expression and behavior [13]. Transcriptomic profiling of DRN serotonergic neurons has further identified molecular signatures of stress resilience and of antidepressant action within this nucleus; spatial transcriptomic analysis has shown that SSRI treatment reorganizes the transcriptional landscape of DRN serotonergic neurons [23]. Whereas those studies addressed the DRN at the level of epigenetic and transcriptional regulation, the present study linked the DRN to the chemical exposure of a herb pair, showing that region-specific candidate constituents of the CHHQ drug pair were detected preferentially in the DRN and that their predicted targets converged on NF-κB/NLRP3/Caspase-1 and SERT. To our knowledge, this is the first report associating the region-specific brain distribution of a classical herb pair with the DRN-centered neuroinflammatory and serotonergic machinery, which places the material basis of the pair within the recently identified DRN-centered mechanism of depression.

Taken together, these coordinated changes support an association between DRN-targeted chemical exposure and improvement of neuroinflammatory and serotonergic disturbances in the CUMS state. The evidence chain established here—blood-absorbed constituents, brain-region-specific constituents, DRN-focused candidate prioritization, and in vivo behavioral and molecular validation—extends the material-basis characterization of the CHHQ drug pair from “which constituents are absorbed into blood” to “which brain-region-specific constituents may act on which targets”. This stepwise design is consistent with the recent trend in herb-pair research toward studying absorbed constituents and their tissue distribution [24,28], and with the recognition that the brain must be treated as a regionally heterogeneous target organ [21]. The DRN-centered account also provides a mechanistic rationale for the complementary actions of the two herbs observed at the constituent level, in which saikosaponin- and flavonoid-derived constituents jointly converge on inflammatory and serotonergic signaling.

Microglial activation is now recognized as a convergent node through which diverse antidepressant interventions, including herbal preparations, act on depressive-like pathology [30], and the present findings extend this picture by localizing such effects to a defined brain region. Several considerations differentiate the present findings from previous work on the same herbs. First, previous network pharmacology studies of the CHHQ drug pair predicted targets from the full constituent library of the two herbs [9], whereas the present study initiated target prediction from constituents with measured brain-region exposure, thereby restricting the candidate space to constituents that demonstrably reach the DRN. Second, previous studies of the pair examined hippocampal and cortical readouts of neuroinflammation [7,9,18], whereas the present findings localize the inflammatory signature to the DRN and analyze it jointly with serotonergic changes within the same animals. Third, although baicalin and saikosaponins have repeatedly been reported to relieve depressive-like behavior in chronic stress models [15,16,17,18], these studies administered isolated compounds, whereas the present work characterized the constituents that are actually absorbed and distributed after oral administration of the drug pair. Fourth, pharmacokinetic and tissue distribution studies of Scutellariae Radix flavonoids [14] and of Bupleuri Radix saponins [26] have mostly been performed in healthy or model animals, and comparative pharmacokinetic and tissue distribution designs have likewise been applied to other herb pairs [28], whereas the present design linked distribution data directly to in vivo antidepressant efficacy and pathway readouts. Finally, the multi-omics and systems pharmacology analyses reported for related herb pairs, such as Bupleurum–Paeonia [27], have provided comprehensive pathway maps but did not distinguish individual brain regions; the region-resolved strategy adopted here complements those approaches by basing pathway prediction on the constituents actually detected in each brain region.

Several limitations should be considered when interpreting the present findings. First, although the distribution, network pharmacology and molecular findings converged on DRN-centered inflammatory and serotonergic signaling, the directionality and causal relationship between the region-specific candidates and the observed molecular changes remain unresolved; pathway-specific intervention experiments, including NF-κB/NLRP3 inhibition, activation or gene silencing, were not performed and will be required to determine whether this signaling is necessary for the pharmacological effects of the pair. Second, the constituents detected in serum and brain tissues were not isolated as individual compounds. The molecular and pathway-level changes were therefore examined experimentally by Western blotting of NF-κB, NLRP3, Caspase-1 and SERT, which provides evidence of treatment-associated changes in these pathways rather than direct evidence of constituent-specific target engagement or of pathway activation state; the interaction between candidate constituents and their putative targets could only be assessed by virtual analysis using network pharmacology and molecular docking, because the metabolites could not be obtained as isolated, purified monomers and a direct experimental binding assay between individual constituents and their targets could therefore not be conducted. Isolation of individual constituents and experimental validation of constituent–target binding, together with bioactivity-guided fractionation, are needed to identify the constituents responsible for the observed effects. Third, pharmacokinetic, bioavailability and quantitative blood–brain barrier permeability studies of representative constituents were not performed, limiting conclusions regarding their in vivo contribution to the observed effects. Fourth, the present evidence indicates only that the region-specific candidates and the prioritized pathways are associated, and does not establish a direct causal relationship between them. These limitations notwithstanding, the present findings define a clear set of questions for future work. Further studies should combine region-specific intervention and gene-silencing experiments with the isolation and experimental validation of individual constituents, and should be extended to pharmacokinetic and quantitative blood–brain barrier permeability studies of the prioritized constituents. Such work, together with the evaluation of additional herb pairs, will help to determine whether the DRN-centered mechanism identified here is specific to the CHHQ drug pair or represents a more general principle underlying the antidepressant effects of multi-component herbal preparations.

Figure 13. Schematic diagram of the proposed antidepressant mechanism of the Chaihu–Huangqin drug pair (CHHQ) in the dorsal raphe nucleus (DRN). Under chronic stress (Depression, left), microglial activation promotes the NF-κB/NLRP3 inflammasome/Caspase-1 pathway, increasing the maturation and release of the pro-inflammatory cytokines IL-1β and IL-18 (together with TNF-α), while serotonergic neurotransmission is disrupted, as shown by reduced synaptic 5-HT, enhanced SERT-mediated reuptake, altered postsynaptic 5-HT1A signaling, and nNOS activation. After CHHQ intervention (Treatment, right), the absorbed active constituents enter DRN neurons, suppress microglial NF-κB/NLRP3 inflammasome activation and inflammatory cytokine production, and restore synaptic 5-HT by reducing SERT-mediated reuptake, thereby ameliorating depression-like behaviors.

Figure 13. Schematic diagram of the proposed antidepressant mechanism of the Chaihu–Huangqin drug pair (CHHQ) in the dorsal raphe nucleus (DRN). Under chronic stress (Depression, left), microglial activation promotes the NF-κB/NLRP3 inflammasome/Caspase-1 pathway, increasing the maturation and release of the pro-inflammatory cytokines IL-1β and IL-18 (together with TNF-α), while serotonergic neurotransmission is disrupted, as shown by reduced synaptic 5-HT, enhanced SERT-mediated reuptake, altered postsynaptic 5-HT1A signaling, and nNOS activation. After CHHQ intervention (Treatment, right), the absorbed active constituents enter DRN neurons, suppress microglial NF-κB/NLRP3 inflammasome activation and inflammatory cytokine production, and restore synaptic 5-HT by reducing SERT-mediated reuptake, thereby ameliorating depression-like behaviors.

4. Materials and Methods

4.1. Experimental Animals

Thirty male healthy specific-pathogen-free (SPF) Sprague–Dawley (SD) rats aged 5–7 weeks, with an initial body weight of 220–250 g, were purchased from the Hubei Provincial Center for Disease Control and Prevention (animal production license No. SCXK (E) 2020-0018). The animals were housed in the Laboratory Animal Center of Hubei University of Chinese Medicine at an ambient temperature of 25 ± 2 °C and a relative humidity of 50 ± 5%, with a 12 h light/12 h dark cycle and free access to food and water. All rats were acclimated for 1 week before the formal experiment. The animal experiment was approved by the Ethics Committee of Hubei University of Chinese Medicine (ethics approval No. HUCMS00310800).

4.2. Herbal Materials, Drugs, and Main Reagents

4.2.1. CHHQ

Bupleuri Radix was purchased from Huangjiahu Hospital of Hubei University of Chinese Medicine (batch No. 21003571), and Scutellariae Radix was purchased from the same hospital (batch No. 21002051). The herbal materials were authenticated by Dr. Yu Cai from the School of Pharmaceutical Sciences, Hubei University of Chinese Medicine, as the dried roots of Bupleurum chinense DC. (Apiaceae) and Scutellaria baicalensis Georgi (Lamiaceae), respectively. Bupleuri Radix and Scutellariae Radix were combined at a weight ratio of 2:1 and decocted three times with 10 volumes of water for 2 h each time. The resulting decoctions were combined and concentrated to a drug concentration of 100%, aliquoted, and stored at 4 °C until use.

4.2.2. Main Reagents

Fluoxetine hydrochloride dispersible tablets (20 mg per tablet; PATHEON FRANCE); methanol (A452-4), acetonitrile (A998-4), and formic acid (A117-50) (Thermo Fisher Scientific, Waltham, MA, USA); rat TNF-α, IL-1β, and IL-18 ELISA kits (E-EL-R0019, E-EL-R0012, and E-EL-R0567), PBS (PB180327), and the BCA protein quantification kit (E-BC-K318-M) (Elabscience, Wuhan, China). The catalog numbers of the NF-κB1, SERT/SLC6A4, GAPDH, and NLRP3 primary antibodies used for Western blotting were E-AB-32233, E-AB-64947, E-AB-40337, and E-AB-93112, respectively; the catalog numbers of the IBA-1 and CD68 primary antibodies used for immunofluorescence were AN300998L and E-AB-22013, respectively. All antibodies were purchased from Elabscience (Wuhan, China). Ultrapure water was used throughout the experiments.

4.3. Main Instruments

Ultra-performance liquid chromatograph (UPLC, ACQUITY I-Class; Waters, Milford, MA, USA); high-resolution liquid chromatography–mass spectrometer (Orbitrap Q Exactive HF; Thermo Fisher Scientific, Waltham, MA, USA); chromatographic column (ACQUITY UPLC HSS T3, 100 mm × 2.1 mm, 1.8 μm; Waters); benchtop high-speed refrigerated centrifuge (TGL-16MS; Shanghai Luxiangyi Centrifuge Instrument Co., Ltd., Shanghai, China); vortex oscillator (TYXH-I; Shanghai Hannuo Instrument Co., Ltd., Shanghai, China); ultrasonic cleaner (F-060SD; Shenzhen Fuyang Technology Group Co., Ltd., Shenzhen, China); open-field detection apparatus (ENV-520; Yihong Technology); swimming tank (HS-80UW-1D; Yihong Technology); and microplate reader (Spark 10M; TECAN, Männedorf, Switzerland).

4.4. Animal Grouping, CUMS Modeling, and Drug Administration

After 1 week of acclimation, the rats were randomly divided into six groups: the normal control group (Control), model group (Model), low-, medium-, and high-dose CHHQ groups (CHHQL, CHHQM, and CHHQH, respectively), and fluoxetine group (Fluoxetine), with six rats in each group. Except for the Control group, all other groups were subjected to chronic unpredictable mild stress (CUMS) to establish the depression model. One stressor was randomly applied each day, including 24 h food deprivation, 24 h water deprivation, ice-water swimming at 4 °C, damp bedding, 45° cage tilting, reversal of the light/dark cycle, 2 h restraint, 2 min tail pinch, and 10 min shaking. The CUMS procedure was continued for 6 weeks; the same stressor was not applied on consecutive days and was used no more than six times in total.

After 6 weeks of CUMS, behavioral assessments were performed, followed by a 2-week drug intervention. The CHHQL, CHHQM, and CHHQH groups were intragastrically administered the CHHQ decoction at crude drug doses of 1.05, 2.10, and 4.20 g/kg, respectively. The Fluoxetine group was intragastrically administered fluoxetine at 1.50 mg/kg, while the Control and Model groups received an equal volume of 0.9% normal saline. Administration was performed once daily for 2 consecutive weeks. Behavioral assessments were repeated after the drug intervention. After the final behavioral tests, serum and brain tissue samples were collected for subsequent LC–MS/MS chemical analysis and pharmacodynamic evaluation. The overall experimental design and analysis workflow is shown in Figure 1.

4.5. Sample Preparation for UPLC-Q-Exactive HF-MS

4.5.1. CHHQ Decoction Samples

An aliquot of 100 μL of the CHHQ decoction was placed in a 1.5 mL centrifuge tube, and 900 μL of 70% aqueous methanol containing 4 μg/mL mixed internal standards was added. The mixture was vortexed for 1 min and extracted by ultrasonication in an ice-water bath for 60 min. It was then centrifuged at 12,000 rpm at 4 °C for 10 min, and the supernatant was diluted twofold with 70% aqueous methanol containing 4 μg/mL mixed internal standards. An aliquot of 200 μL was transferred to an LC-MS injection vial for analysis.

4.5.2. Serum Samples

Serum samples were thawed at 4 °C and vortexed. An aliquot of 100 μL was placed in a 1.5 mL centrifuge tube, and 400 μL of precooled methanol containing 4 μg/mL mixed internal standards was added. After vortexing for 1 min, the mixture was allowed to stand at −20 °C for 30 min to precipitate proteins and was then centrifuged at 12,000 rpm at 4 °C for 10 min. An aliquot of 200 μL of the supernatant was placed in an LC-MS injection vial and stored at −80 °C until LC-MS analysis.

4.5.3. Brain Tissue Samples

Brain tissue (30 mg, accurately weighed) was placed in a 1.5 mL centrifuge tube, and two small steel beads and 180 μL of methanol–water (4:1, v/v) containing 4 μg/mL mixed internal standards were added. After precooling at −40 °C for 2 min, the tissue was ground at 60 Hz for 2 min, extracted by ultrasonication in an ice-water bath for 10 min, and allowed to stand overnight at −40 °C. The mixture was then centrifuged at 12,000 rpm at 4 °C for 10 min, and 120 μL of the supernatant was placed in an LC-MS injection vial and stored at −80 °C.

4.6. Chromatographic Conditions

An ACQUITY I-Class ultra-performance liquid chromatography (UPLC) system was used for chromatographic separation on an ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm) at a column temperature of 45 °C. Mobile phase A was 0.1% formic acid in water and mobile phase B was acetonitrile, with a flow rate of 0.35 mL/min and an injection volume of 5 μL. The gradient elution program was as follows: 0–2 min, 95% A; 2–4 min, 95–70% A; 4–8 min, 70–50% A; 8–10 min, 50–20% A; 10–14 min, 20–0% A; 14–15 min, 0% A; 15–15.1 min, 0–95% A; 15.1–16 min, 95% A.

4.7. Mass Spectrometric Conditions

An Orbitrap Q Exactive HF high-resolution mass spectrometer was used for mass spectrometric detection with a HESI electrospray ion source, and data were acquired in both positive and negative ion modes. Data were collected in data-dependent acquisition (DDA) mode with a Full MS/dd-MS² (Top 8) scan strategy. The spray voltages in the positive and negative modes were 3.8 kV and −3.2 kV, respectively; the capillary temperature was 320 °C and the auxiliary gas heater temperature was 350 °C; the sheath gas flow was 35 Arb and the auxiliary gas flow was 8 Arb; and the S-lens RF level was 50. The full-scan mass range was m/z 100–1500, with a full-scan resolution of 60,000 and an MS² resolution of 15,000. The normalized collision energies (NCEs) were set at 10, 20, and 40.

4.8. LC-MS Data Processing and Component Annotation

Raw LC-MS data were processed with XCMS v4.5.1 for peak detection, peak integration, retention time correction, peak alignment, and normalization. Candidate components were annotated based on a combination of accurate m/z, retention time, isotope distribution, and available MS/MS fragment information, together with the TCM database of traditional Chinese medicine components; candidates with a composite score ≥ 40 were retained for subsequent analysis. The results from the positive and negative ion modes were integrated and deduplicated before subsequent component analysis.

4.9. Rat Body Weight and Liver/Kidney Function Assessment

4.9.1. Body Weight Measurement

During the experiment, the body weight of each rat was measured once weekly between 10:00 and 12:00 on a fixed day, and the body weight changes of rats in each group were recorded.

4.9.2. Serum Liver and Kidney Function Indicators

At the end of the experiment, blood was collected and serum was isolated. The serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), and creatinine (Cr) were measured using Elabscience kits, and the procedures were performed according to the manufacturer’s instructions.

4.10. Behavioral Tests

Behavioral tests were performed after 6 weeks of CUMS and repeated after 2 weeks of drug intervention. The sucrose preference test (SPT), open field test (OFT), and forced swimming test (FST) were conducted as described below.

4.10.1. Sucrose Preference Test

Before the experiment, rats were trained to adapt to the sucrose solution. Each rat was provided with two identical drinking bottles containing 1% sucrose solution. After 24 h, one bottle was replaced with pure water, and the positions of the two bottles were switched every 12 h to reduce position preference. Before the formal test, the rats were deprived of water for 12 h. Each rat was then provided with one bottle of 1% sucrose solution and one bottle of pure water and allowed to drink freely for 1 h, after which the positions of the two bottles were switched and drinking continued for another 1 h. The bottles were weighed before and after the test, and the consumption of sucrose solution and pure water was calculated. Sucrose preference was calculated as follows:

Sucrose preference (%) = sucrose consumption / (sucrose consumption + pure water consumption) × 100%.

4.10.2. Open Field Test

A square open-field apparatus measuring 1 m × 1 m with a height of 40 cm was used. Each rat was placed in the center of the open field and allowed to move freely for 5 min. The total distance traveled during the 5-min test was recorded using a behavioral analysis system. After each test, urine and feces were removed, and the floor and inner walls of the apparatus were wiped with 75% ethanol to minimize olfactory interference.

4.10.3. Forced Swimming Test

A cylindrical transparent water tank with a diameter of 20 cm and a height of 60 cm was filled with water to approximately three-fifths of its height, and the water temperature was maintained at 25 °C. Each rat was placed in the water and forced to swim for 6 min. The immobility time during the last 4 min was observed and recorded and was expressed as the immobility time ratio (%). After the test, the rats were dried and returned to their cages.

4.11. Collection of Serum and Brain Tissue Samples

After the final behavioral tests, the rats were anesthetized by intraperitoneal injection of 1% sodium pentobarbital (65 mg/kg body weight), and blood was collected through the abdominal aorta. After standing, the blood was centrifuged at 12,000 rpm for 15 min, and the serum was isolated and stored at −80 °C until use. After blood collection, the hippocampus (CA), prefrontal cortex (PFC), nucleus accumbens (NAc), ventral tegmental area (VTA), lateral habenula (LHb), and dorsal raphe nucleus (DRN) were rapidly dissected and stored at −80 °C until use.

4.12. Determination of Serum CRH and CORT Levels

At the end of the experiment, blood was collected from the abdominal aorta and serum was isolated by centrifugation. The serum concentrations of corticotropin-releasing hormone (CRH) and corticosterone (CORT) were measured using corresponding rat ELISA kits (CRH, E-EL-R0270; CORT, E-EL-R0269; Elabscience, Wuhan, China) according to the manufacturer’s instructions. The absorbance was measured using a microplate reader, and the concentrations were calculated from the corresponding standard curves.

4.13. ELISA Detection of Inflammatory Cytokines in the DRN

DRN tissues were collected from rats in each group, and the levels of TNF-α, IL-1β, and IL-18 were detected using ELISA kits. The procedures were performed according to the kit manufacturer’s instructions.

4.14. Western Blot Detection of Protein Expression in the DRN

DRN tissues were mixed with RIPA protein lysis buffer and PMSF and homogenized on ice. The homogenates were centrifuged at 15,000 rpm at 4 °C for 15 min, and the supernatants were collected. Total protein concentrations were determined using the BCA method. Protein loading amounts were calculated according to the measured concentrations, followed by electrophoresis and membrane transfer at 400 mA for 30 min. The membranes were washed three times with TBST for 10 min each and blocked with 5% BSA at room temperature for 1 h. The membranes were then incubated with primary antibodies against NLRP3, caspase-1, NF-κB1, SERT, and GAPDH at 4 °C overnight. After three 10-min washes with TBST, the membranes were incubated with the corresponding secondary antibodies in the dark, followed by another three 10-min washes with TBST. Protein bands were visualized using ECL chemiluminescence. Band intensities were quantified using ImageJ software. The expression levels of NLRP3, caspase-1, NF-κB1, and SERT were normalized to GAPDH for quantitative analysis.

4.15. Immunofluorescence Detection of IBA-1 and CD68 in the DRN

Paraffin-embedded brain sections containing the DRN were prepared and blocked with blocking buffer at room temperature in the dark for 30 min, followed by washing with PBS. The sections were then incubated with primary antibodies against IBA-1 and CD68 at 4 °C overnight. After three 5-min washes with PBS, the sections were incubated with the corresponding secondary antibodies at room temperature in the dark for 50 min. The nuclei were counterstained with DAPI, and the sections were mounted and imaged using a fluorescence microscope. The fluorescence intensities of IBA-1 and CD68 in the DRN were quantified, and the mean fluorescence intensity (MFI) was used for quantitative analysis.

4.16. Network Pharmacology Analysis

4.16.1. Potential Target Prediction for the DRN Relatively Specific Candidate Metabolites

Four DRN relatively specific candidate metabolites, identified in our previous multi-brain-region brain-targeting analysis, were selected as the research objects. Their potential targets were predicted using SwissTargetPrediction (http://www.swisstargetprediction.ch/) and the PharmMapper database (http://lilab-ecust.cn/pharmmapper/), with the species restricted to Homo sapiens. Targets with a probability value ≥ 0.7 were selected and duplicates removed, and a target set of the four DRN relatively specific candidate metabolites was constructed.

4.16.2. Construction of the Depression-Related Target Set

Depression-related disease targets were retrieved from the GeneCards (https://www.genecards.org/), OMIM (https://www.omim.org/), and DisGeNET (https://www.disgenet.org/) databases using “depression” and “major depressive disorder” as keywords, and core targets were screened according to relevance scores. Meanwhile, validated functional targets that specifically regulate depression in the DRN and LHb were supplemented through literature research. The targets were merged to construct the depression disease target set.

4.16.3. Screening of Common Targets

An online Venn diagram tool was used to obtain the intersection of the candidate metabolite target set and the depression disease target set, yielding the potential common targets. The common targets were then imported into the STRING database (https://string-db.org/; species Homo sapiens) to construct a protein–protein interaction (PPI) network, with the minimum required interaction score set to medium confidence (combined score ≥ 0.4) and disconnected nodes hidden. The network was visualized, and the degree of each node was calculated to evaluate its connectivity; nodes with high degree values were defined as hub targets, and the top 10 hub targets were identified for subsequent analysis.

4.16.4. GO Functional Enrichment and KEGG Pathway Analysis

The common targets were imported into the clusterProfiler package in R for GO functional enrichment and KEGG pathway enrichment analyses. The Benjamini–Hochberg (BH) method was used for correction, and adjusted p < 0.05 was set as the enrichment significance threshold. Functional characteristics were analyzed from three dimensions: molecular function (MF), biological process (BP), and cellular component (CC). Core terms related to central nervous regulation and depressive mechanisms were screened for subsequent analysis.

4.16.5. Component–Target–Pathway Network Construction and Pathway Target Mapping

To further characterize the anti-inflammatory candidate metabolites, an anti-inflammatory candidate metabolite–core target–KEGG pathway network was constructed and presented as a Sankey–bubble diagram: the candidate metabolites were linked to their predicted core target proteins, which were in turn linked to the enriched inflammation-related KEGG pathways. The software/package used for network visualization and the criteria used to select the six anti-inflammatory metabolites are to be specified. In addition, the identified core targets were mapped onto the KEGG pathway maps of the TNF signaling pathway (hsa04668) and the NOD-like receptor signaling pathway (hsa04621) to indicate their positions; the meaning of the red and green highlighting is to be specified in a legend.

4.17. Molecular Docking

Four DRN relatively specific candidate metabolites were docked against five target proteins selected on the basis of the network pharmacology results and the DRN pharmacodynamic readouts: the serotonin transporter (SERT/SLC6A4; PDB ID: 6VRH), neuronal nitric oxide synthase (nNOS/NOS1; PDB ID: 6CID), tumor necrosis factor (TNF; PDB ID: 2AZ5), NOD-like receptor family pyrin domain-containing 3 (NLRP3; PDB ID: 9GU4), and caspase-1 (CASP1; PDB ID: 1IBC). Molecular docking simulations were then performed using the Schrödinger software suite (release 2022-4; Schrödinger LLC, New York, NY, USA) to predict the binding modes and affinities of the candidate metabolites to these target proteins. The three-dimensional crystal structures of each target were retrieved from the RCSB Protein Data Bank and preprocessed with the Protein Preparation Wizard module: missing amino acid side chains and flexible loop regions were completed, hydrogen atoms were added, the protonation states of amino acid residues were optimized at physiological pH 7.4, and non-essential crystallographic water molecules were removed. Constrained energy optimization was then performed using the OPLS-AA force field, with a heavy-atom root-mean-square deviation (RMSD) convergence threshold of 0.3 Å, yielding the energetically optimal protein conformations for subsequent docking. The LigPrep module was used to construct and optimize the three-dimensional conformations of the candidate metabolites; atomic charges were calculated based on the MMFF94s molecular force field, and ionization states and tautomers at pH 7.4 were enumerated and optimized. The lowest-energy dominant conformation of each molecule was selected as the docking ligand. Docking calculations were completed with the Glide module. Taking the geometric center of the native ligand in the co-crystal structure of each target as the origin, a docking grid box measuring 20 Å × 20 Å × 20 Å was constructed to fully cover the active binding pocket. Standard Precision (SP) mode was used first for preliminary screening and Extra Precision (XP) mode for refined docking, and the Glide Score was used as the core evaluation indicator of binding affinity. Finally, the Ligand Interaction Diagram module was used to generate two-dimensional interaction maps of the ligand–protein complexes, systematically analyzing non-covalent interactions at the binding interface, including hydrogen bonds, hydrophobic interactions, π–π stacking, and salt bridges, to identify the key amino acid residues involved in binding and verify the rationality of the binding modes.

4.18. Statistical Analysis

Statistical analyses were performed using GraphPad Prism 8.0. Data were tested for normality and homogeneity of variance and are presented as the mean ± standard deviation (mean ± SD). For comparisons among multiple groups at a given time point, one-way analysis of variance (one-way ANOVA) followed by Dunnett’s multiple comparisons test was used. Comparisons between the Control and Model groups were used to evaluate the effects of CUMS, whereas comparisons between the Model group and the treatment groups were used to evaluate the effects of the interventions. For longitudinal or repeated measurements, group and time were taken into consideration in the statistical analysis, with multiple comparisons performed where appropriate. A value of p < 0.05 was considered statistically significant.

5. Conclusions

In conclusion, the CHHQ drug pair attenuated depressive-like behaviors in CUMS rats, with effects accompanied by attenuation of DRN microglial activation and normalization of the NF-κB/NLRP3/Caspase-1 axis and SERT expression. UPLC-MS/MS-based serum pharmacochemistry identified 97 absorbed constituents, whose brain-region-resolved distribution was most prominent in the DRN, and network pharmacology initiated from DRN-specific candidates prioritized inflammatory and serotonergic targets. These findings identify the CHHQ drug pair as a source of bioactive constituents with multi-target antidepressant activity and support further preclinical evaluation of its candidate constituents for the development of multi-component, multi-target strategies for depression.