Using the online databases Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform(TCMSP, https://tcmspw.com/tcmsp.php) (Ru et al., 2014), HERB (http://herb.ac.cn/) (Fang et al., 2021), and Encyclopedia of Traditional Chinese Medicine (ETCM, http://www.tcmip.cn/ETCM/) (Xu et al., 2019) via a literature search, the active constituents of each herb in FXDH were evaluated. The molecular structures of the active constituents were obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) (Dashti et al., 2019); these were then saved in 3Dsdf format and then imported into the SwissADME platform (http://www.swissadme.ch/) (Daina et al., 2017) for screening. The screening criteria for potential active compounds were as follows: gastrointestinal absorption was “high” and drug-likeness (Lipinski, Ghose, Veber, Egan, and Muegge) had at least two “Yes” values. The compounds that did not meet these criteria were reviewed in the scientific literature and those with noticeable pharmacological effects in BCCD were retained. Using the Swiss Target Prediction database (http://www.swisstargetprediction.ch/) (Daina et al., 2019), target proteins for the included compounds were identified. A probability of >0 was set as the condition for screening.

Possible targets for breast cancer and depression were obtained from GeneCards (https://www.genecards.org/) (Safran et al., 2010), DisGeNet (https://www.disgenet.org/home/) (Pinero et al., 2017), OMIM (https://www.omim.org/) (Amberger and Hamosh, 2017), and DrugBank (https://www.drugbank.ca/) (Wishart et al., 2018). The DrugBank database is based on evidence from clinical trials, whereas the other databases are based on literature. The four databases complement each other to allow for increased precision and search. For example, the terms “breast cancer,” “breast carcinoma,” and “mammary cancer” were used to search breast cancer-related genes, whereas “depression” and “depressive disorder” were used to identify depression-related genes. After combining the genes from the four databases and eliminating duplicates, breast cancer- and depression-related target datasets were generated. Genes were uniformly named using gene symbols from the UniProt protein database (https://www.uniprot.org/) (Soudy et al., 2020).

Overlapping genes (OGEs) among FXDH, breast cancer, and depression targets were visualized using the E Venn tool (https://www.ehbio.com/test/venn/#/). The obtained OGEs were then inputted into the STRING database (https://string-db.org/) (Szklarczyk et al., 2019) with a threshold of confidence scores ≥0.9 and species restricted to Homo sapiens to generate the PPI network. Topological network analysis was conducted using Cytoscape 3.9.1 (Otasek et al., 2019) to visualize the PPI network. Hub targets were identified using a minimum criterion of having a node degree value of at least twice the median value.

The ClusterProfiler package (Yu et al., 2012) from the Bioconductor R3.6.2 project was used for GO and KEGG analyses. The genes identified using GO analysis were classified according to their role in either biological process (BP), molecular function (MF), or cellular component (CC). The results of GO and KEGG analyses were screened based on p-value, with a screening threshold of p < 0.05. The top thirty genes were then selected for visualization using bubble diagrams. Using the Cytoscape software, the mechanism and targets of FXDH in BCCD were clarified, along with the component–target–pathway network.

The 3D conformations of the hub target proteins were acquired from the RCSB PDB database (https://www.rcsb.org/) (Goodsell et al., 2020), whereas structures of the active compounds were obtained from PubChem. The Surflex-dock module of SYBYL-X 2.1.1 (Jain, 2007) was used for molecular docking with an FXDH core compound ligand and a hub target protein receptor. This matching method is based on the similarity of structure and shape and is highly accurate, has a high positive rate, and has a rapid matching speed. In general, the higher the Totalscore value, the lower the free binding energy between the receptor and ligand, indicating greater stability. A Total score of ≥5.0 indicates good binding ability between the ligand and the target protein, whereas a Totalscore of ≥7.0 indicates robust docking activity.

FXDH decoction was provided by the Pharmacy Department of the Xiamen Hospital of Traditional Chinese Medicine (Xiamen, China). Briefly, 420 g of decoction was weighed according to the ratio mentioned in Table 1. This was then mixed with distilled water at a ratio of 1:12 and then extracted at 90°C for 1.5 h. The filtrate obtained after three extractions was mixed and concentrated at 60°C in an oven, followed by vacuum drying. One gram of the dry powder was obtained from approximately 2.44 g of decoction.

The main chemical constituents of FXDH extract were assessed using a Vanquish UHPLC system (Thermo Scientific, Waltham, MA) equipped with an Accuity UPLC HSS-T3 column (100 × 2.1 mm; particle size: 1.8 μm size; Waters, Shanghai, China) at a temperature of 35°C. Mobile phase A was water + 0.1% formic acid, and mobile phase B was acetonitrile + 0.1% formic acid (LC-MS grade solvents, Fisher chemical). The samples were separated at a flow rate of 0.3 mL/min using the following gradient: 5% B for 1 min, 98% B for 16 min, 5% B for 0.2 min, and 5% B for 2.8 min. The UHPLC system was coupled with the Q-Exactive HFX mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Mass spectra were obtained using data-dependent acquisition mode under both positive and negative electrospray ionization settings, with a mass range of 90–1,300 m/z. The capillary and probe heater were set at 320°C and 370°C, respectively. This procedure was conducted for quality assurance, at the Shanghai Applied Protein Technology Co., Ltd. (Shanghai, China).

Female Balb/c mice (6–8 weeks old) were purchased from Beijing HFK Bioscience Co., Ltd. (Beijing, China; license: SCXK (Beijing) 2020-0004) and raised under specific pathogen-free conditions at the Animal Laboratory Center of Xiamen University (license: SYXK (MIN) 2018-0010). All experimental procedures were conducted in strict adherence to the guidelines provided by the Welfare and Ethics Committee of Laboratory Animals of Xiamen University (XMULAC20200055).

After 1 week of adaptive feeding, Balb/c mice were randomly divided into six groups: the blank group (BG, 0.9% saline), the 4T1 breast cancer transplantation group (4T1, 0.9% saline), the corticosterone group (CORT, 0.9% saline), the 4T1 tumor transplantation + corticosterone group (4T1+CORT, 0.9% saline), the FXDH group (FXDH, 13.65 g/(kg · d), the dosage of crude drug is same as the clinical equivalent dosages), and the capecitabine chemotherapy group (CAP, 390 mg/(kg · d), same as the clinical equivalent dosages). As depicted in Fig. 2, mice were injected subcutaneously with 4T1 cells on day 0 and received daily corticosterone injections for 28 days (Zhao et al., 2008). Then, on day 7, FXDH or CAP was administered intragastrically until 21 days. Every week, the body weight of mice was recorded. Every 3 days, tumor size was measured using a vernier caliper, and tumor volume was calculated using the following equation:

Volume (mm³) = 0.5 × [length (mm)] × [width (mm)]²

After 21 days of drug treatment, animals were sacrificed. Subsequently, the tumor tissues were resected, weighed, and photographed; in addition, spleen tissues were collected to detect epinephrine levels.

Depression-like behavior in mice was assessed using the sucrose preference test (SPT) and the open field test (OFT). For SPT, the animals were housed in solitary confinement and were provided free access to water and food. During the 48 h of drinking training, two bottles each of 2% sucrose water and pure water were provided. The amount of water consumed from the two bottles over 8 h was measured (the bottles’ positions were switched after 24 h). Sucrose preference was determined using the following equation: Sucrose preference (%) = Sucrose intake (g)/ [ sucrose intake (g) + water intake (g) ]×100% For OFT, mice were individually placed in a 40 × 40 × 40 cm box without a ceiling for 5 min. During this time, the mice were monitored to record the total moving distance, displacement speed, central area moving distance, and the number of central area entries using the TopScanLite 2.00 software (Clever Sys, USA).

Fresh tumor and hippocampal samples were fixed for 24 h using 4% paraformaldehyde. After dehydration with gradient concentration of alcohol, the tissues were fixed in paraffin wax and then sliced into 4-mm thick sections. The sections were then dewaxed with xylene and ethanol and were subsequently stained with hematoxylin and eosin (H&E), according to the instructions for H&E staining kit (Biyuntian, China). Finally, the sections were sealed with neutral resin. Light microscopy was used to visualize tissue histology.

Total RNA was extracted from tumor tissues using TRIzol reagent (Invitrogen, USA) and reverse transcribed to cDNA using the PrimeScript^™ RT reagent Kit with gDNA Eraser (TaKaRa, Japan), per the manufacturer’s instructions. Subsequently, RT-qPCR was performed using TB Green^™ Premix Ex Taq^™ (TaKaRa, Japan) and the Real-time PCR Detection System (Roche-LightCycler 480, USA). The relative abundance of mRNA was determined utilizing GAPDH as the internal control, and the expression was determined using the 2^−ΔΔCt method. The specific primers used (Sangon Biotech, Shanghai, China) are listed in Table 2.

The expression of epinephrine in the spleen was determined using an ELISA kit (Senbeijia Biotechnology, Nanjing, China), as per the instructions provided by the manufacturer. Then, using a tissuelyser (MB-48L-1, Zhejiang, China), 50 mg of spleen tissue was accurately weighed and homogenized with 450 μL phosphate buffer (0.01 mol/L). The homogenized mixture was then centrifuged at 3,500 rpm for 15 min at 4°C to obtain supernatant. The supernatant was removed and stored in 2 mL centrifuge tubes at −80°C until further analysis.

Using the bicinchoninic acid method, 100 mg tumor tissues were lysed with a potent RIPA buffer containing a complete protease inhibitor cocktail (Siding Biotechnology, Shanghai, China). 100 mg of lysed tissues were electrophoresed on sodium dodecyl sulfate-polyacrylamide gel to separate proteins. Equal amounts of proteins (approximately 30 µg) from each sample were separated on sodium dodecyl sulfate-polyacrylamide gel and transferred to the PVDF membrane (Merk Millipore, United States). The PVDF membranes were incubated overnight at 4°C with primary antibodies including anti-JAK2 (1:1000, ab108596, Abcam, USA), anti-PIK3 (1:1000, 20584-1-AP, Proteintech, USA), and anti-AKT (1:10000, ab179463, Abcam, USA). Subsequently, they were incubated with the secondary antibody against horseradish peroxidase-conjugated rabbit IgG (1:20000, FDR007, Fude Biological Technology, Hangzhou, China) at room temperature for 1 h. Using enhanced chemiluminescence reagents (Siding Biotechnology, Shanghai, China), immunoreactive proteins were detected and analyzed with Image J.

All statistical analyses were conducted using SPSS 26.0, and GraphPad Prism 8.0 was utilized to generate the graphs. Two-way or one-way analysis of variance was used to determine p-values, followed by a post hoc analysis with the least significant difference in multiple comparison tests. The statistical significance threshold was set at p < 0.05.

The TCMSP and ETCM databases and related literature searches yielded 174 potential active components of FXDH after eliminating redundancy. Details of these compounds are provided in Suppl. Table S1. Then, the Canonical SMILES format was imported into Swiss Target Prediction, and after removing duplicate genes, 676 prediction targets were obtained (Suppl. Table S2). Using the UniProt database for deduplication and correction, 4299 BC targets and 1649 depression targets were identified from the DisGeNET, GeneCards, OMIM, and DrugBank databases. Next, FXDH-related pharmacological targets, BC-related disease targets, and depression-related disease targets were mapped and intersected to yield 163 common targets (Suppl. Fig. S1 and Suppl. Table S2). The herb–compound-common target network was constructed using 362 nodes and 1,707 edges (Fig. 3). In descending order of “degrees,” the top six active compounds were identified as quercetin, methyl ferulate, luteolin, ferulaldehyde, wogonin, and diincarvilone A. The degrees represent the number of connected edges of nodes, and the greater the degree, the greater the regulatory function of nodes within the network.

The 163 common targets were uploaded to the String11.5 database, and a PPI network comprising 156 nodes and 1269 edges was generated (Fig. 4A). The median node degree was 16. In total, 19 hub targets were screened out according to the rule in method (Suppl. Fig. S2): AKT1, STAT3, EGFR, TNF, IL-6, TP53, VEGFR, IL-1B, ALB, MAPK1, EGF, HRAS, EP300, GAPDH, FN1, PIK3R1, ESR1, PIK3CA, and STAT1. A PPI network of these 19 hub targets was additionally constructed (Fig. 4B).

GO and KEGG pathway enrichment analyses were performed to classify the potential functions of 163 initial common targets of disease–compounds. In total, 2651 GO bio-function entries are associated with the interaction between FXDH anti-BC and depression. Among them, 2391 BPs are primarily associated with protein kinase activity, metabolic process, and signal response. In addition, 77 CCs are primarily associated with lumen, membrane, cytoplasm, caveola, alpha granules, and neuronal cell body; 183 MFs are associated with signaling receptor activator activity, receptor–ligand activity, binding functions, cytokine activity, and oxidoreductase activity (Suppl. Table S3). Fig. 5A depicts the top 10 terms enriched in each category. The top 10 BPs were: regulation of protein serine/threonine kinase activity, response to extracellular stimulus, response to nutrient levels, reactive oxygen species metabolism, response to the drug, response to lipopolysaccharide, activation of protein kinase activity, response to molecules of bacterial origin, regulation of reactive oxygen species metabolism, and positive regulation of reactive oxygen species metabolism. The top 10 CCs were: membrane raft, membrane microdomain, neuronal cell body, external side of the plasma membrane, vesicle lumen, endoplasmic reticulum lumen, plasma membrane raft, caveola, platelet alpha granule, and dendrite cytoplasm. The top 10 MFs identified were: signaling receptor activator activity, receptor–ligand activity, receptor binding, cytokine activity, oxidoreductase activity acting on paired donors with incorporation or reduction of molecular oxygen, heme binding, tetrapyrrole binding, protease binding, growth factor receptor binding, and insulin receptor substrate binding.

A total of 171 signaling pathways were identified via KEGG pathway enrichment analysis, and the top 30 pathways are shown according to their p-values in Fig. 5B and Suppl. Table S2. A literature search revealed that phosphoinositide-3-kinase–protein kinase B (PI3K/AKT), HIF-1, FoxO, JAK/STAT, chemical carcinogenesis receptor activation, endocrine resistance, and PD1/PDL1-related pathways may play an important role in BCCD. In addition, the target gene–signaling pathway network (Fig. 6) validated that the intervention effect of FXDH on BCCD might be mediated by the correlation between multiple targets and pathways. In particular, the PI3K-AKT signaling pathway had 36 intersection targets and may play a major role in depression-related breast cancer progression.

The compounds selected by network pharmacology were validated using qualitative UPLC-HRMS of the FXDH extract. Figs. 7A and 7B illustrate the positive- and negative-ion base peak chromatograms of FXDH. The abundant peaks observed in the base peak chromatograms were analyzed based on their peak shape and were confirmed by secondary spectra. As a result, 32 active components of FXDH were identified, the majority of which included aporphines, benzopyran, morphinans, proberberine alkaloids, flavonoids, phenols, cinnamic acids, glycerophospholipids, saccharolipids, carboxylic acids, and fatty acyls (Suppl. Table S4). Based on the network pharmacology results, quercetin, methyl ferulate, luteolin, ferulaldehyde, wogonin, and diincarvilone A were considered for molecular docking. The chemical structures of these six core components are depicted in Fig. 7C.

To understand the interactions between compounds and target genes, the six major components of FXDH identified by UPLC-HRMS were used as ligands for molecular docking. According to the literature and degree value, six core target proteins (IL-6, AKT1, EGFR, ESR1, PIK3, and STAT3) were identified from the PPI network as receptors for molecular docking. Fig. 8A displays the total score for every docking activity. In total, 22 of the 36 results exhibited exceptional docking activity (Totalscore ≥ 5.0), with 5 exhibiting strong matching activity (Totalscore ≥ 7.0). As shown in Fig. 8A, luteolin had the strongest binding affinity for the core targets, whereas AKT and STAT3 had the highest average binding levels with the six components. Fig. 8B depicts the 3D structures of the seven receptor–ligand interactions with the highest scores.

The plasma levels of corticosterone, a typical glucocorticoid in rodents, are significantly elevated following exposure to stress (Wu et al., 2022). In addition, multiple corticosterone injections have been shown to induce depression-like behaviors in mice (Ding et al., 2018). Accordingly, we exposed treated orthotopic 4T1 tumor-bearing mice to corticosterone for 28 consecutive days to establish a mice model of breast cancer complicated with comorbid depression. Compared with the blank group, 4T1, CORT, and 4T1+CORT groups had lower weight, total and central area distance, and sugar preference. As depicted in Fig. 9, the 4T1+CORT group had lower weight, movement ability, exploration behavior, and surface water preference, indicating a higher level of depression, which was mitigated by FXDH intervention. It is well-known that sympathetic nerve fibers innervate various peripheral tissues, particularly the spleen. Therefore, several studies have assessed the levels of catecholamine to evaluate the sympatho-adrenal-medullary activity in stressed mice (Cao et al., 2002; Thaker et al., 2006). Accordingly, the present study found that the CORT group had a higher splenic EPI level than the other groups. Moreover, FXDH could effectively alleviate this effect.

Sustained corticosterone stress may rapidly promote tumor progression, as evidenced by a significant increase in tumor volume and weight. This effect was mitigated by both FXDH and CAP (Figs. 10A–10C). H&E staining demonstrated that treatment with FXDH or CAP inhibited the proliferative effects of corticosterone, including tight arrangements of tumor cells, a high nuclear/cytoplasmic ratio, and increased karyokinesis as well as the development of large patchy necrosis in tumor tissue (Fig. 10D). Based on the KEGG pathway enrichment analysis and published literature, the regulatory effect of FXDH on BCCD in mice might be related to the PIK3CA-AKT and STAT-JAK signaling pathways. RT-qPCR revealed that in tumor tissues, the expression of PIK3, AKT, STAT3, and JAK2 mRNAs was upregulated after chronic stress. Moreover, FXDH treatment could downregulate the expression of these genes (Fig. 10E). In addition, the protein-level expression of PIK3, AKT, and JAK2 exhibited the same pattern, confirming the crucial role of hub targets in treating BCCD by FXDH (Fig. 10F).