Ginsenosides from Panax ginseng: Their mechanisms, microbiome determinants, and translational strategies in therapeutic activity against cancer

Abstract

Ginsenosides from Panax ginseng Meyer and their microbiota-derived metabolites have been explored as adjuncts in oncology, with pleiotropic activities across tumor-intrinsic and microenvironmental pathways. This review focuses on representative metabolites (Rg3, Rh2 and compound K) and summarizes mechanistic evidence for apoptosis and cell-cycle control, metabolic reprogramming (Warburg-type glycolysis), and epigenetic regulation (miRNA/lncRNA-linked programs), together with immunomodulatory effects in the tumor microenvironment. A central translational constraint is pharmacokinetics and the “microbiota gatekeeping” effect: many parent ginsenosides show limited oral absorption and require microbiota-mediated deglycosylation, resulting in delayed and variable systemic exposure across individuals. To address this bottleneck, we discuss pragmatic translational directions including bioconversion/pre-transformation to enrich active metabolites, biomimetic delivery platforms such as ginseng-derived exosome-like nanoparticles to improve exposure and tissue uptake, and mechanism-guided combinations (e.g., with immune checkpoint blockade). Finally, we outline how biomarker-informed stratification and AI/ML-enabled, hypothesis-generating optimization of delivery/formulation may support more predictable clinical development.

Graphical abstract

1. Introduction

Panax ginseng has been widely evaluated in supportive-care and adjunct settings in oncology, where randomized trials support benefits in symptom and quality-of-life–related outcomes [[1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13]]. However, evidence for direct anticancer efficacy in humans remains less consistent, and epidemiological studies have reported heterogeneous associations with cancer risk and outcomes [2,[14], [15], [16], [17], [18], [19], [20]]. A key translational reason is pharmacokinetics and “microbiota gatekeeping”: Many major ginsenosides have low oral absorption and require gut microbiota–mediated conversion to generate systemically active metabolites, leading to substantial inter-individual variability in exposure and response [[21], [22], [23]], as in the cases of other traditional and alternative medicines [[24], [25], [26], [27], [28], [29], [30], [31], [32]]. This review addresses how these constraints shape realistic clinical positioning and how they can be mitigated (see Fig. 1).

Fig. 1.

The Systems Pharmacology of Ginseng in Cancer.

(A) Gut Microbiota Biotransformation: Orally administered ginsenosides are converted by gut microbiota into their primary bioactive metabolite, Compound K (CK). This conversion is the rate-limiting step for systemic efficacy. (B) Multi-Target Action in the TME: CK exerts multi-faceted anticancer effects by directly inducing cancer cell apoptosis and cell cycle arrest, while also modulating the TME. This includes reprogramming tumor-associated macrophages (TAMs) to an anti-tumor M1 phenotype and enhancing T-cell-mediated cytotoxicity. (C) Future Optimization Strategies: To overcome clinical limitations, key strategies include: (1) enhancing bioavailability via nanodelivery systems or pre-transformation; (2) optimizing drug properties through chemical structural modification; and (3) designing synergistic combination therapies, particularly with immune checkpoint inhibitors (ICIs). Abbreviations: CK, compound K; TME, tumor microenvironment; TAMs, tumor-associated macrophages; ICIs, immune checkpoint inhibitors; PI3K, phosphoinositide 3-kinase; Akt, protein kinase B; mTOR, mechanistic target of rapamycin. Red arrows indicate key mechanistic effects and proposed intervention points; black arrows indicate process flow.

Ginsenosides are a structurally diverse group of over 150 triterpenoid saponins predominantly categorized into protopanaxadiol (PPD) and protopanaxatriol (PPT) types [22] with a variety of pharmacological activities such as anti-oxidative, anti-aging, anti-atherosclerotic, anti-diabetic, anti-obesity, anti-steatosis, anti-allergic, anti-thrombotic, anti-hypertension, anti-depressant, anti-hypertrophic, anti-cancer, anti-inflammatory and anti-neurological diseases [19,[33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]]. Beyond classical pathways such as apoptosis induction and cell-cycle regulation, recent studies have extended anticancer mechanisms of ginsenosides to metabolic reprogramming, epigenetic modulation, and remodeling of the tumor immune microenvironment [17,18,35,66,[69], [70], [71], [72], [73], [74]]. Moreover, specific post-translational mechanisms—such as modulation of PD-L1 glycosylation by Rg3—highlight actionable nodes for combination strategies with immunotherapy [35,75,76].

To bridge mechanistic advances with clinical realities, we integrate molecular evidence with a focused assessment of pharmacokinetic barriers and microbiota-dependent activation, emphasizing how biotransformation capacity and host handling can constrain or enable efficacy [70,[77], [78], [79]]. We further summarize translational solutions including bioconversion approaches, delivery innovations, and biomarker-informed stratification aimed at reducing exposure variability and improving predictability.

To provide a clinically oriented and evidence-weighted discussion, we primarily focus on Rg3, Rh2, and compound K (CK), as these exemplars have the most mature mechanistic-to-translational evidence base in oncology (including formulation-level clinical experience for Rg3 and microbiota-derived activity for CK). Other ginsenosides are discussed when they directly inform these concepts (microbiota gatekeeping, exposure variability, and delivery-enabled translation).

2. Chemical profile and pharmacokinetics of ginsenosides

Ginsenosides (ginseng saponins) are the major bioactive constituents of ginseng and belong to the dammarane-type triterpenoid saponin family. They are commonly classified into protopanaxadiol (PPD) types (e.g., Rb1, Rd, Rg3) and protopanaxatriol (PPT) types (e.g., Re, Rg1) [21,22]. Thermal processing of fresh ginseng into red ginseng alters glycosylation patterns and generates rare ginsenosides such as Rg3, Rk1, and Rg5 [80]. Representative classifications and structural features of anticancer ginsenosides are summarized in Table 1.

Table 1.

Representative classifications and characteristics of anticancer ginsenosides.

Class	Representative Compounds	Aglycone Skeleton	Key Structural Features	Pharmacological Notes	Representative References
PPD-type	Rb1, Rc, Rd	Protopanaxadiol	C-3/C-20 sugar residues	Abundant in fresh ginseng; metabolized into CK by gut flora; poor oral bioavailability	[79,81,82]
PPD-type	Rg3 (20S/20R)	Protopanaxadiol	Fewer sugar moieties; heat-processed rare ginsenoside	Approved in China as adjuvant anticancer agent; pro-apoptotic, anti-angiogenic, metabolic reprogramming	[58,80,83]
PPD-type	Rh2	Protopanaxadiol	C-3 monosaccharide	Strong apoptosis induction and immune microenvironment remodeling	[51,84,85]
PPD-type	Compound K (CK)	Protopanaxadiol	Final deglycosylated metabolite	Generated by gut β-glucosidases; better oral exposure than precursors; interindividual variability	[79,86,87]
PPT-type	Re, Rg1	Protopanaxatriol	C-6/C-20 sugar residues	Predominantly immunomodulatory; direct anticancer evidence weaker	[21,23]
PPT-type	Rh1, F1	Protopanaxatriol	Metabolites of Rg1	Anti-inflammatory/immune regulatory; linked to hepatic enzyme modulation	[21,79]

Most ginsenosides exhibit poor oral absorption because of their high polarity, large molecular weight, and low membrane permeability. Consequently, their systemic activity after oral administration largely depends on biotransformation by the gut microbiota [78,81].

Intestinal bacteria hydrolyze ginsenosides via stepwise deglycosylation: PPD-type ginsenosides can be converted into metabolites such as CK and Rh2, whereas PPT-type ginsenosides yield metabolites such as Rh1 and F1. Evidence indicates that these metabolites, rather than the parent ginsenosides, mediate many of ginseng's pharmacological activities (including anticancer effects). Among the key contributors, Bifidobacterium and Bacteroides species provide beta-glucosidases that drive this conversion [79]. Oral prototypical ginsenosides (e.g., Rb1) show poor bioavailability (<5%) [88]. Compared with parent ginsenosides, CK shows higher permeability and measurable systemic exposure; in rats, its absolute oral bioavailability is dose-dependent (reported from a few percent at lower doses up to ∼35% at higher doses), and in humans CK typically exhibits delayed absorption (Tmax ∼12 h), consistent with microbiota-mediated formation [86].

3. Molecular mechanisms and targets of anticancer activity

Note on evidence context: the mechanistic findings summarized in this section derive from a mixture of in vitro studies, animal models, and limited clinical observations; where possible, we indicate the predominant validation setting and phrase conclusions accordingly.

Ginseng saponins can exert anticancer effects by regulating multiple signaling pathways, all of which participate in the basic biological behavior of tumor cells. In subsequent chapters, we will focus on several key molecular targets discovered in recent studies, with Table 2 providing a brief overview of some of these representative targets. However, it is important to note that many compelling findings originate from in vitro studies, where the purified ginsenoside concentrations used may be difficult to achieve in vivo following oral administration. This represents an important translational challenge that will be discussed in subsequent sections.

Table 2.

Key ginsenosides in different cancer models: molecular targets and actions.

Compound	Cancer type/model	Major targets/pathways	Regulation	Observed effects (summary)	Refs	Evidence context
Rg3	Breast cancer	PI3K/Akt/mTOR; JNK/Beclin-1	↓/modulation	Autophagy-associated tumor suppression	[89]	In vitro
Rh2	Cervical cancer (nutrient stress)	Autophagic flux (fusion/processing)	↓	Inhibits cytoprotective autophagy; increases stress sensitivity	[90,91]	In vitro
CK (Compound K)	Various (G1/S control)	p21 ↑ → CDK4/6–Cyclin D ↓ → pRb/E2F	↑/↓	G0/G1 arrest; proliferation restraint	[92]	Both
Rp1	Breast cancer	IGF-1R/Akt	↓	G2/M arrest; proliferation inhibition	[93]	Both
Rh2	Multiple tumor models	BCL-2 family (Bcl-2↓/Bax↑); caspase-9/3	balance shift	Intrinsic apoptosis induction	[[94], [95], [96]]	In vitro
Rg3	HCC/breast cancer (apoptosis sensitization settings)	DR5/Fas → caspase-8/3	↑	Extrinsic apoptosis; chemosensitization phenotype in selected systems	[97,98]	In vitro
Rh2	CRPC (in vivo/mechanistic)	PINK1/Parkin-mitophagy; SLC7A11/GPX4	↑; ↓	Mitophagy + ferroptosis-associated tumor suppression	[99]	Both
Rg3/Rh2	HCC/lung cancer	HK2/PKM2 (glycolysis); STAT3/c-Myc	↓	Metabolic reprogramming (↓ glycolysis/Warburg-like output)	[100,101]	In vitro
Rg3/Rd	Breast cancer/endothelial context	Akt/mTOR/p70S6K → HIF-1α/VEGF/VEGFR2	↓	Anti-angiogenesis signaling suppression	[[102], [103], [104], [105]]	In vitro
Rh2	EMT/invasion models	EMT markers (E-cad↑, N-cad↓); MMP-2/9	reversal/↓	Reduced migration/invasion phenotypes	[106]	Both
Rg3	CSC-like phenotypes (breast/CRC)	HIF-1α; Wnt/β-catenin (β-catenin/TCF)	↓	Reduced stemness/self-renewal readouts	[[107], [108], [109]]	In vitro
Rh2/Rg3	Epigenetic regulation (selected models)	STXBP5-AS1/miR-4425; ATXN8OS/miR-424-5p; DNMT3A/miR-145	modulation	ncRNA-linked proliferation/EMT suppression signals	[[110], [111], [112]]	Both
Rh2 (+anti-PD-L1)	MC38 CRC/NSCLC	Intratumoral chemokine program (e.g., CXCL10); CD8+ T-cell recruitment; TAM state	↑	Enhanced checkpoint blockade efficacy; TME remodeling	[113,114]	Both
Rg3	NSCLC	PD-L1 N-glycosylation (EGFR–GSK3β–linked)	interference	Reduced PD-L1 stability; attenuated immune evasion signals	[76]	In vitro

Note: “Evidence context” indicates the predominant validation setting for each entry (in vitro, in vivo, or both). ↑/↓ denotes the direction of pathway output/phenotype as reported in the cited study.

3.1. Apoptosis

Evasion of apoptosis is a hallmark of cancer, and reactivating programmed cell death remains a central therapeutic strategy [115]. Available evidence supports engagement of both intrinsic and extrinsic apoptotic pathways [84].

Microbiota-derived metabolites such as Rh2 and CK have been reported to promote mitochondrial apoptosis by shifting BCL-2 family balance and facilitating cytochrome c release and caspase-9/3 activation [[94], [95], [96]]. However, the effective concentrations frequently used in vitro (e.g., >10 μM for Rh2) are rarely achieved after oral administration, underscoring an important translational limitation discussed in Section 4 [116].

For extrinsic signaling, Rg3 has been reported to increase death receptor expression (including Fas and DR5) [97],and some ginsenosides may reduce apoptosis resistance by suppressing c-FLIP in specific models [97,98].

Beyond canonical apoptosis, Rh2 has been reported to suppress castration-resistant prostate cancer (CRPC) by activating PINK1/Parkin-mediated mitophagy while concurrently promoting ferroptosis via SLC7A11/GPX4 downregulation [99]. Furthermore, ginsenoside Ro has been reported to modulate cuproptosis-related signaling via downregulation of ZnT1 in the cited model [117].

3.2. Cell cycle arrest

3.2.1. G0/G1 phase block

CK has been reported to upregulate p21 and inhibit CDK4/6–cyclin D activity, thereby reducing pRb phosphorylation and limiting E2F-dependent S-phase entry [92]. In breast cancer and drug-resistant CRC models, Rh2 has been reported to induce G1 arrest through p15/p27 upregulation, p38 activation, and cyclin D1 downregulation [118]. The 20(S) isomer of Rg3 can inhibit breast cancer cell proliferation by inducing G0/G1 phase block [119]; further studies have shown that Rk1 can block the G0/G1 checkpoint of triple-negative breast cancer cells by inhibiting cell cycle-related proteins [120].

3.2.2. G2/M phase block

At the G2/M checkpoint, Rp1 has been reported to impede progression in breast cancer cell lines such as MDA-MB-231 and T-47D [93].

3.3. Autophagy modulation

3.3.1. Inducing autophagy cell death

Rg3 has been reported to enhance autophagy-associated responses in breast cancer models, with signaling changes involving the mTOR/PI3K/Akt and JNK/Beclin-1 axes [89]. Rg5 has been linked to coordinated autophagy marker changes and PI3K/Akt inhibition in breast cancer models [121,122]. Rh1 has been reported to interfere with hypoxia-associated signaling in the cited model [123]. In contrast, F2-induced autophagy in breast cancer stem cells may support survival rather than elimination, highlighting context dependence [124].

3.3.2. Inhibit survival-promoting autophagy

Under nutritional stress, Rh2 has been reported to inhibit protective autophagy in cervical cancer models, potentially sensitizing cells to damage [90,91]. Collectively, these data support bidirectional autophagy modulation by ginsenosides in preclinical settings [125].

3.4. Anti-angiogenesis

The hypoxia-inducible factor-1α (HIF-1α)/VEGF axis is a central driver of tumor angiogenesis [126]. Rd has been reported to inhibit upstream Akt/mTOR/p70S6K signaling and thereby attenuate HIF-1α/VEGF outputs [102]. Rg3 has been reported to downregulate VEGF expression [103], and its 20(S) isomer has been linked to reduced VEGF and matrix metalloproteinase activity in breast cancer models [104].

These effects may reduce downstream activation of VEGFR2 signaling in endothelial cells in the cited contexts [105].

3.5. Precision modulation of the PD-L1 glycosylation axis

Beyond direct cytotoxicity, post-translational modifications have emerged as potentially actionable targets. In NSCLC models, Rg3 has been reported to inhibit PD-L1 N-linked glycosylation via the EGFR–GSK3β axis, promoting PD-L1 degradation and reducing immune evasion in preclinical settings [76].

3.6. Inhibition of invasion and metastasis

Several ginsenosides and intestinal metabolites have been reported to interfere with invasion/metastasis-associated programs, including ECM degradation and EMT. Rh2 has been reported to reduce MMP-1/2/9/13 and shift EMT markers toward an epithelial phenotype in the cited models [106]. In breast cancer, Rg1 has been linked to inhibition of PMA-induced MMP-9 via NF-κB suppression [127]. CK has been reported to suppress MMP-9 transcription through IKK/NF-κB inhibition and attenuate EMT via PI3K/Akt modulation [128].

These activities are exposure-sensitive because CK and related metabolites are generated by microbiota-dependent deglycosylation and can vary substantially between individuals.

3.7. Targeting cancer metabolic reprogramming

The Warburg effect describes the preference of many cancer cells for aerobic glycolysis over mitochondrial respiration even under oxygen-replete conditions [115]. In liver cancer models, Rg3 has been reported to reduce glycolytic activity by inhibiting HK2 and PKM2 and to promote mitochondrial respiration [100]; in lung cancer models, Rh2 has been linked to suppression of STAT3/c-Myc signaling [101].

By limiting glycolysis and lactate output, these effects may also mitigate microenvironment acidification and be associated with reduced invasive/metastatic phenotypes in the cited contexts [129].

3.8. Modulation of the epigenetic landscape

Ginsenosides have been reported to modulate epigenetic outputs through non-coding RNA circuits and DNA methylation–linked regulation of gene expression. Rh2 has been reported to upregulate lncRNA STXBP5-AS1, which functions as a ceRNA to sequester miR-4425 and derepress downstream tumor-suppressive programs in a breast cancer model [110]. In another model, Rg3 suppressed the oncogenic lncRNA ATXN8OS via promoter hypermethylation, restoring miR-424-5p activity and repressing downstream targets in the cited study [111]. In addition, 20(S)-Rg3 has been reported to increase miR-145 via DNMT3A downregulation and reduced methylation at the miR-145 precursor locus, linking miRNA regulation with chromatin modification [112].

These findings provide a mechanistic rationale for further in vivo validation in clinically representative models.

3.9. Targeting cancer stem cells

Ginsenoside Rg3 has been reported to attenuate cancer stem cell (CSC)–like phenotypes by targeting self-renewal signaling and microenvironmental support. In breast cancer, standardized red ginseng extract and Rg3 reduced CSC-like features with decreased nuclear HIF-1α and downregulation of stemness regulators [107]. In colorectal cancer models, Rg3 inhibited Wnt/β-catenin signaling with reduced β-catenin/TCF transcriptional activity and lower expression of canonical targets such as c-MYC [108]. Rg3 also decreased MDSC accumulation and blunted MDSC-driven EMT/stemness programs in a murine mammary carcinoma setting [109]. Consistent with an adjunctive positioning, Rg3 enhanced paclitaxel responses in TNBC models [130], and Rg3-based liposomal formulations have been developed to address multidrug resistance via dual targeting of tumor cells and the tumor microenvironment [131].

3.10. Modulation of the tumor immune microenvironment

Rh2 has been reported to remodel the tumor immune microenvironment (TME) toward more immunostimulatory states in preclinical models [132]. In the cited studies, Rh2 increased intratumoral CD4⁺ and CD8⁺ T-cell infiltration and enhanced cytotoxic activity [133]. Rh2 has also been linked to TAM polarization shifts away from an M2-like phenotype in the cited contexts [134,85]. In the MC38 colorectal cancer model, oral Rh2 combined with anti-PD-L1 antibodies has been reported to enhance intratumoral CD8⁺ T-cell infiltration, proliferation, and activation compared with monotherapy [113,114].

4. Gut Microbiota–Dependent bioactivation of ginsenosides

Most native ginsenosides are highly glycosylated and therefore poorly permeable, so their systemic exposure after oral dosing is largely shaped by microbiota-dependent deglycosylation into more lipophilic metabolites such as CK and Rh2 [79]. Interindividual variability in CK formation has been linked to differences in gut community structure; taxa including Bacteroides and Bifidobacterium are enriched in high CK-forming phenotypes and have been associated with Rb1-to-CK conversion capacity [135].

Beyond acting as a metabolic “gatekeeper,” ginseng can also reshape this ecosystem. Ginseng polysaccharides behave as prebiotic-like substrates that modulate microbial metabolism (including β-glucosidase activity) and can increase exposure to Rb1-derived metabolites (e.g., CK/PPD) in vivo and in microbiota-based assays [136,137]. Across ginseng interventions, enrichment of potentially beneficial taxa (including Akkermansia) has also been reported, supporting a bidirectional ginseng–microbiome interaction [138].

A recent mechanistic example is the Rh4–bile acid axis in colorectal cancer: Rh4 increased Akkermansia muciniphila and shifted bile-acid metabolism toward higher ursodeoxycholic acid (UDCA); UDCA then engaged FXR and attenuated TLR4–NF-κB signaling. Importantly, microbiota depletion/antibiotic intervention diminished the additional benefit of Rh4, consistent with a microbiota-dependent antitumor effect [139]. Key enzymes and bile-acid mediators are summarized in Table 3.

Table 3.

Gut microbial/enzymatic biotransformation of ginsenosides and metabolites.

Prototype ginsenoside	Key microbiota/enzymes	Active metabolite(s)	Functional consequence	References
Rb1/Rc/Rd (PPD)	Bifidobacterium, Bacteroides β-glucosidases	Compound K (CK)	Improved absorption and systemic exposure; enhanced bioactivity	[79,135,82]
Rg1 (PPT)	β-glucosidase, rhamnosidase	Rh1, F1, 20(S)-PPT	Anti-inflammatory/immune regulation; potential CYP3A modulation	[21,79,81]
Saponins + polysaccharides	Akkermansia expansion; β-glucosidase ↑	CK and others	Prebiotic effect, enhanced CK exposure and immunomodulation	[70,[136], [137], [138]]
Rh4→bile acid axis	Microbiota-regulated bile acid composition (↑UDCA)	FXR/TLR4-NF-κB signaling	Mediates colorectal cancer suppression; microbiota-dependent effect	[139]

Rg3 + first-line platinum-based chemotherapy vs chemotherapy alone (NSCLC).

5. Precision medicine: biomarkers and patient stratification

Inter-individual variability in the systemic exposure to bioactive ginsenoside metabolites is a major source of heterogeneity in clinical outcomes after oral ginseng administration, largely because bioactivation depends on gut microbial metabolism and host clearance pathways. Accordingly, biomarker-driven stratification should prioritize (1) microbiome-derived functional capacity for deglycosylation and (2) host factors that modulate exposure and safety in combination regimens [140].

5.1. Microbiota features and “converter” phenotypes as predictive biomarkers

Rather than relying solely on taxonomic profiling, a practical approach is to define “high-” versus “low-converter” phenotypes using fecal/intestinal CK-forming activity (a functional readout), which has been shown to correlate with plasma exposure to CK in humans [87]. Consistent with this concept, comparative microbiome analyses have reported compositional differences between individuals with high versus low ginsenoside-metabolizing capacity, supporting the feasibility of microbiota-informed stratification [135].

Because microbial bioactivation is susceptible to disruption, recent antibiotic exposure is a plausible exclusion/stratification variable: antibiotic-driven depletion of gut microbiota substantially attenuates the intestinal formation and systemic appearance of CK in preclinical models, consistent with a strong dependence on intact microbial metabolism to achieve CK exposure after oral dosing [141,82]. Conversely, dietary modulation may enhance bioactivation; for example, a soluble prebiotic fiber (NUTRIOSE) increased intestinal glycosidase activity and elevated plasma CK exposure after oral ginseng administration in rats [142].

5.2. Pharmacogenomics and safety-oriented stratification in combination therapy

Host pharmacogenomics may be most actionable when ginseng preparations are considered alongside standard anticancer drugs with established genotype–toxicity relationships. A practically relevant example is irinotecan, where reduced-function UGT1A1 variants (e.g., ∗28 and ∗6) are associated with higher risk of severe toxicity and are reflected in dosing recommendations (including DPWG guidance) [143,144]. In parallel, herb–drug interaction potential should be considered at the pathway level: major ginsenoside metabolites (including CK and related sapogenins) can modulate UGT activities in vitro, suggesting a mechanistic rationale to monitor for interactions when co-administered with UGT-cleared agents—although clinical significance depends on achievable exposures and requires confirmation in well-designed PK studies [145].

6. Clinical perspective and future directions

Clinical evidence to date supports the positioning of ginsenosides primarily as adjunctive agents, rather than stand-alone anticancer therapeutics. Among available formulations, ginsenoside Rg3 (commonly administered as Shenyi capsule in Chinese studies) is the most frequently evaluated. Meta-analyses in advanced non-small cell lung cancer (NSCLC) suggest that adding Rg3 to platinum-based chemotherapy is associated with improved objective response rate (ORR) and reduced chemotherapy-related adverse events (e.g., gastrointestinal reactions and myelosuppression), although heterogeneity in trial design and variable methodological quality remain key limitations [83] (Table 4).

Table 4.

Summary of Adjuvant Clinical Outcomes for Ginsenoside Rg3 (Shenyi Capsule) Data sourced from [83].

Outcome Metric	Relative Risk (RR)/OR	95% CI	P-value	Clinical Significance
Objective Response Rate (ORR)	1.44 (RR)	[1.27, 1.63]	<0.00001	Improved chemotherapy-associated outcomes
Disease Control Rate (DCR)	1.24 (RR)	[1.12, 1.38]	<0.0001	Improved disease control/tumor stabilization
2-Year Survival Rate	6.22 (RR)	[1.68, 22.95]	0.006	Potential long-term survival benefit
Myelosuppression Incidence	0.43 (RR)	[0.30, 0.61]	<0.00001	Reduced toxicity incidence
Gastrointestinal Reactions Incidence	0.66 (RR)	[0.47, 0.93]	0.02	Reduced gastrointestinal toxicity incidence
Karnofsky Performance (QOL)	1.62 (RR)	[1.42, 1.84]	<0.00001	Improved functional status (KPS)

In hepatocellular carcinoma (HCC), an open-label randomized study reported that Rg3 plus transcatheter arterial chemoembolization (TACE) prolonged median overall survival compared with TACE alone (approximately 13.2 vs 10.1 months) [146]. While these data support feasibility for integration into interventional regimens, larger multicenter trials with rigorous allocation concealment/blinding and standardized endpoints would likely be needed before broad guideline adoption.

A major translational bottleneck is pharmacokinetics. Most native ginsenosides have poor intestinal absorption, and systemic activity depends on microbiota-mediated deglycosylation followed by host handling. Human pharmacokinetic profiling of Korean red ginseng extract has reported delayed appearance of CK, with a time to peak concentration on the order of ∼10–12 h, consistent with slow biotransformation and substantial inter-individual variability in exposure [147]. This variability may be clinically relevant because it can propagate heterogeneity in downstream pharmacodynamics and ultimately clinical response.

Accordingly, three pragmatic optimization directions are emerging (Table 5): (i) pre-transformation/bioconversion (e.g., fermentation or enzymatic conversion) to enrich active metabolites and shorten absorption delay; (ii) delivery innovations to improve exposure, stability, and tissue distribution; and (iii) mechanism-guided combinations that leverage immunomodulatory effects. Preclinically, ginsenoside Rh2 can potentiate anti-PD-L1 therapy in the MC38 model by enhancing intratumoral chemokine signaling and reinvigorating CD8⁺ T-cell infiltration/activation, supporting a rationale for combination regimens [113]. In addition, Rh2 has been reported to reprogram tumor-associated macrophages (TAMs) toward an M1-like phenotype in vivo, consistent with a broader strategy of remodeling the TME to enhance standard therapies [85].

Table 5.

Comparative strategies to overcome oral bioavailability bottlenecks.

Strategy	Representative approach/formulation	Advantages	Limitations	PK/PD key info	References
Direct oral prototype PPD ginsenosides	Rb1/Rg3/Rh2 etc.	Readily available, simple preparation	Require gut biotransformation; high variability	Absolute bioavailability <5%; slow conversion; CV>60%	[79,81,88]
Probiotics/prebiotics co-administration	With ginseng polysaccharides or defined strains	Enhances β-glucosidase activity and CK exposure; immune benefits	Strain-specific, interindividual variability; clinical standardization lacking	Correlated with improved efficacy; clinical quantification pending	[70,136,137,142]
Direct administration of CK	Pure CK or fermented red ginseng enriched in CK	Bypasses conversion; more consistent PK	Production cost; quality control required	Oral CK F∼18.9 ± 2.7%; faster Tmax vs precursors	[86,87,147]
Delivery systems/nanoformulations	Cyclodextrin/PLGA/liposomes with adjuvants	Improved solubility, tumor targeting, immune synergy	Complex process; scalability and safety assessment	Enhanced efficacy in multiple models; improved TME	[131,148,149]
Structural modifications	Rh2 sulfated/esterified derivatives	Better uptake and intracellular delivery	Need safety profiling	Increased uptake, stronger anticancer/anti-inflammatory effects	[19,116]

Moving forward, trials should shift toward biomarker-informed precision use. Inter-individual differences in microbiota-driven conversion capacity (“converter phenotype”) have been linked to distinct baseline gut community structures, including differential abundance of taxa implicated in Rb1-to-CK biotransformation [135]. In parallel, experimental evidence indicates that intestinal bacterial hydrolysis is necessary for systemic appearance of CK after oral Rb1 exposure, supporting the concept that recent antibiotic exposure or dysbiosis could plausibly attenuate metabolic activation and diminish response [82]. Therefore, integrating baseline microbiome profiling with on-treatment exposure readouts (e.g., plasma CK monitoring) could support responder enrichment and reduce outcome heterogeneity.

Finally, host pharmacogenomics should be considered when ginsenosides are deployed alongside cytotoxic backbones. Because UGT1A1 activity is a determinant of irinotecan active metabolite (SN-38) glucuronidation and toxicity risk, genotype-guided dosing recommendations (including variants prevalent in Asian populations, such as UGT1A1∗28 and UGT1A1∗6) provide a practical framework for minimizing severe adverse events in combination regimens [143]. Collectively, biomarker-driven trial designs, standardized manufacturing/exposure reporting, and transparent risk-of-bias control are important to help bridge the current gap between reproducible adjunctive benefit and more predictable, individualized outcomes.

7. Emerging bio-delivery strategies: G-exos

Oral ginsenosides generally show limited intestinal absorption and low systemic exposure (reported oral bioavailability often <5%), with substantial inter-individual variability driven by metabolism and gut microbiota dependence, motivating strategies that improve exposure and tissue uptake [148,149]. Ginseng-derived exosome-like nanoparticles (G-Exos) isolated from Panax ginseng have emerged as biomimetic carriers that deliver intrinsic ginseng cargo and modulate host responses in preclinical models [[150], [151], [152]]. Because nomenclature and physicochemical attributes of G-Exos can vary with starting materials and extraction workflows, current evidence should be framed as preclinical and method-dependent [153].

7.1. Isolation and characterization

G-Exos are commonly obtained from root-derived fluids or homogenates through sequential clarification centrifugation (e.g., ∼300–20,000×g) followed by high-speed ultracentrifugation and/or density-gradient purification (e.g., sucrose gradient) to enrich nanoscale vesicle fractions [71,154]. Preparations typically include an exosome-sized subpopulation (∼50–150 nm), although broader EV heterogeneity may be present depending on fractionation stringency [71,153,154]. Representative datasets report mildly negative zeta potentials (∼−20 to −30 mV), consistent with colloidal stability, and TEM/NTA are widely used to document bilayer morphology, size distribution, and particle concentration [71,154]. Cargo profiling supports that ginseng-derived vesicles can contain proteins, lipids, small RNAs, and detectable ginsenosides, indicating that they may function as bioactive multi-component nanoparticles rather than “empty” carriers [152,153].

7.2. Biocompatibility and stability

Available in vitro assays suggest generally favorable baseline biocompatibility in tumor-relevant and immune cell systems. For example, a melanoma-oriented study reported no detectable cytotoxicity up to 30 μg/mL (72 h) and no major hematologic/biochemical abnormalities within the tested in vivo dosing window [150]. However, systematic head-to-head evaluations of long-term storage conditions, lyophilization workflows, and batch-to-batch reproducibility (including EV purity metrics and cargo drift) remain limited and should be treated as current technical gaps [153].

7.3. Functional implications for oncology

Mechanistically, G-Exos can remodel the TME through innate immune reprogramming. In melanoma models, G-Exos were reported to shift tumor-associated macrophages toward an M1-like phenotype via TLR4/MyD88-dependent signaling, accompanied by tumor growth inhibition in vivo [150]. In “cold” tumor settings, G-Exos increased TAM-derived chemokine programs (notably CCL5 and CXCL9) linked to CD8⁺ T-cell recruitment and enhanced the efficacy of anti-PD-1 blockade across multiple murine tumor models [151]. In lung cancer models, G-Exos have been reported to attenuate immune evasion via PD-L1–related mechanisms and potentiate anti-PD-L1 responses, supporting a combination-oriented positioning [155]. Beyond peripheral tumors, G-Exos have also shown anti-glioma activity in vivo, consistent with barrier-crossing potential and microenvironment modulation [71].

7.4. Clinical translation considerations

Evidence supporting G-Exo platforms in oncology remains preclinical. Key translational needs include harmonized isolation/characterization criteria (purity, size distribution, EV markers/cargo), scalable GMP-compatible manufacturing, rigorous toxicology, and quantitative PK/PD frameworks linking vesicle dosing to intratumoral cargo delivery and immune remodeling [153]. Addressing these standardization and PK/PD questions will likely be important prerequisites before credible first-in-human evaluation, particularly if G-Exos are positioned as delivery adjuvants for ginsenoside-based combination regimens [153].

8. Future directions and conclusion

Preclinical studies indicate that ginsenosides can modulate epigenetic programs and antitumor immunity, whereas current clinical data most consistently support an adjunctive role, with Rg3 (Shenyi capsule) being the best-studied formulation [83]. In NSCLC, meta-analyses suggest improved objective response and reduced chemotherapy-related toxicities when Rg3 is added to platinum-based regimens, and in advanced HCC, an open-label randomized study reported longer median overall survival with Rg3 plus TACE versus TACE alone, although trial heterogeneity and external validation remain limitations [83,146]. Translation is constrained by pharmacokinetics and microbiota “gatekeeping”: many parent ginsenosides have low oral absorption and delayed/variable generation of active metabolites (e.g., CK with Tmax ≈ 12 h), consistent with large inter-individual differences in conversion capacity [135,147]. Next-generation studies should move from empirical “tonic” use to biomarker-informed “precision supplementation”, integrating baseline microbiome features and exposure readouts (e.g., plasma CK monitoring, as a standardized exposure endpoint to operationalize the converter phenotype) with standardized rare-ginsenoside/delivery platforms and mechanism-guided combination regimens [135,83,147,156,157]. AI/ML-enabled nanomedicine optimization and multi-omics integration are potentially enabling approaches, but should be positioned as hypothesis-generating tools until prospectively validated [158].

Authorship contribution statement

Canglang Mou: Writing—original draft; Formal analysis; Conceptualization; Investigation. Yuhao Wang: Writing—review and editing; Supervision. Mi-Yeon Kim: Conceptualization; Writing—review and editing; Supervision. Jae Youl Cho: Conceptualization; Writing—review and editing; Supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.