The Evolving Landscape of Hepatocellular Carcinoma Therapy: From Conventional Modalities to TME-Responsive Prodrug Design Strategies Authors: Jiayao Chen, Hongrui Yan, Haoyu Wang, Huixin Li, Zhiyu Wang, Qian Ding, Yi Zhun Zhu Published: 2026 ABSTRACT AbstractHepatocellular carcinoma (HCC), the predominant form of primary liver cancer, remains a major cause of cancer-related mortality. Despite advances in resection, transplantation, locoregional therapy, and systemic treatment, recurrence, resistance, toxicity, tumor heterogeneity, and impaired hepatic reserve limit durable benefit. Features of the HCC tumor microenvironment (TME) may serve as conditional activation signals for prodrug design. A prodrug is a pharmacologically inactive or less active derivative that undergoes chemical or enzymatic conversion after administration to generate or release an active drug. This review links the limitations of HCC treatments to prodrug-design requirements and evaluates bona fide prodrugs activated by acidity, redox imbalance, hypoxia, disease-associated enzymes, or lactate-coupled processes. Responsive carriers containing unmodified active drugs, imaging probes, and nanodelivery systems without a prodrug component are considered separately. Unlike previous reviews centered on broad stimuli-responsive nanomedicine or individual TME cues, this review uses a clinically anchored framework to assess whether proposed activation signals discriminate HCC from adjacent inflamed, fibrotic, or cirrhotic liver. It further integrates interpatient and intratumoral heterogeneity, off-target activation, impaired hepatic function, and the maturity of the evidence into the assessment of selectivity and translational feasibility. Because most HCC-directed systems remain supported by cellular or animal studies, improved selectivity, reversal of resistance, reduced toxicity, and survival benefit cannot yet be assumed. Future development should prioritize quantitative biomarker-guided activation, multi-input designs, spatial pharmacokinetic/pharmacodynamic assessment in disease-relevant models, and mechanism-based combinations that remain manufacturable. These advances may enable more controlled intratumoral drug exposure, but clinical value remains to be established. Introduction According to the 2022 global cancer statistics from the International Agency for Research on Cancer (IARC), primary liver cancer is the sixth most common cancer worldwide and the third leading cause of cancer-related deaths.1 Its incidence and mortality rates have been increasing annually due to contemporary unhealthy dietary habits and lifestyles. From 2010 to 2021, the number of liver cancer cases increased by 26%, while the number of deaths rose by 25%.2 Primary liver cancer can be classified into two categories: hepatocellular carcinoma (HCC), which accounts for nearly 90% of the approximately 906,000 global liver cancer cases, and cholangiocarcinoma (CCA), the second most common type after HCC.3 The etiology of primary liver cancer is multifactorial, including, but not limited to, hepatitis B (HBV), hepatitis C (HCV), alcohol-related liver disease, and metabolic dysfunction-associated steatotic liver disease (MASLD or MAFLD).2 In recent years, the high incidence of HCC among liver cancers has led to an increased focus on research in this area. HCC is the dominant contributor to the elevated mortality rate of primary liver cancer and is designated as the fifth leading cause of cancer-related deaths on a global scale.4 The disease has a relative 5-year survival rate of approximately 18%, which indicates an extremely poor prognosis. Its incidence is comparable to its mortality rate, which is 830,000 deaths annually.5 HBV infection is the most significant risk factor for HCC development, accounting for approximately 50% of cases.6 However, since the 2000s, the incidence of viral hepatitis-related malignancies has declined globally due to the implementation of neonatal HBV vaccination programs and the widespread availability of highly effective antiviral therapies. According to a study on global epidemiological trends in liver cancer from 2010 to 2019, metabolic dysfunction-associated steatohepatitis (MASH) is the fastest-growing cause of liver cancer,7 with an upward trend in incidence consistent with that of HCC. MASLD is gradually becoming the major cause of high HCC prevalence. At present, approximately 38.2% of the global population is affected by MASLD,8 and the incidence may surge to as high as 56% over the next decade.7 However, only two drugs have currently received approval for clinical use. One is resmetirom, a selective thyroid hormone receptor β agonist, employed in the treatment of MASH with moderate-to-severe liver fibrosis. The other is semaglutide, a GLP-1 receptor agonist, utilized in the treatment of non-cirrhotic MASH with moderate-to-severe liver fibrosis. There is a paucity of pharmacotherapeutic agents available for the treatment of the full progression of MASLD or for the treatment of HCC driven by MASLD. Treatment options for early-stage HCC include liver resection, transplantation, and radiofrequency ablation as well. Transcatheter arterial chemoembolization (TACE) has been the chief treatment method and gold standard for intermediate-stage HCC for the past two decades.9 However, due to the asymptomatic nature of early-stage HCC and the lack of sensitive and specific diagnostic markers, most patients are not diagnosed until intermediate or advanced stages, missing the optimal window for early surgical intervention. Systemic therapy is the treatment for patients with advanced HCC and for those with intermediate-stage HCC who are ineligible for local treatment. For many years, chemotherapy has been the standard systemic treatment for advanced HCC. However, the emergence of sorafenib (the first targeted therapy for advanced HCC)9 and nivolumab (the first immunotherapy for HCC),10 has gradually shifted intermediate- and advanced-stage HCC treatment strategies toward targeted therapy and immunotherapy. Despite significant progress in treatment across all stages, key challenges remain, including insufficient early diagnosis, high recurrence rates, tumor heterogeneity, an immunosuppressive microenvironment, primary and secondary resistance to targeted therapy, limited liver function reserve, a lack of precise predictive biomarkers, toxicity and cost barriers of combination therapy and so on. There is an urgent clinical need for therapeutic approaches offering higher selectivity, lower toxicity, long-term applicability, and greater accessibility. The prodrug strategy is a potential solution to the aforementioned issues. Prodrugs are pharmacologically inactive derivatives that undergo enzymatic or chemical conversion in vivo to regenerate or release the active parent drug, thereby improving the physicochemical, pharmacokinetic, or biopharmaceutical properties of the therapeutic agent.11 In recent years, anticancer prodrugs have become a focal point in the development of cancer treatment strategies. In the presence of cancer-specific biomarkers, the transition from “masked to unmasked” can produce therapeutic effects at the target site.12 Several prodrug cases have been reported to exhibit high tumor specificity, extremely low off-target toxicity, higher maximum tolerated doses,11,13–15 improved water solubility, and cellular permeability.16 Zytiga (for hormone-resistant prostate cancer) and Ixazomib Citrate (for multiple myeloma) are FDA-approved prodrug-based cancer therapies.12 Applying prodrug strategies to HCC therapeutic agent development can improve drug delivery efficiency, enhance tumor targeting, mitigate drug resistance, and reduce systemic toxic side effects. The tumor microenvironment (TME) of HCC exhibits characteristics distinct from those of normal tissues and the TMEs of tumors in other organs, providing a potential switch for the activation of prodrugs used in HCC. A growing body of recent studies has systematically summarized prodrug design strategies and applications in response to different TME characteristics. However, these studies have either focused on broad TME features or concentrated on a single characteristic.12,17,18 This review first summarizes current clinical treatments for HCC and their major limitations (Figure 1). It then examines bona fide TME-responsive prodrugs, focusing on their activation mechanisms, current evidence, and key translational challenges, including tumor heterogeneity, overlap with adjacent diseased liver, off-target activation, and the limited therapeutic window associated with impaired liver function. Finally, future design priorities are discussed to link unmet clinical needs with the development of safer and more selective HCC prodrugs. Figure 1Clinical management algorithm for HCC based on the BCLC staging system. This schematic delineates the recommended therapeutic strategies stratified by Barcelona Clinic Liver Cancer (BCLC) stages (Early, Intermediate, and Advanced) indicated by green, orange, and red headings respectively. The upper part of the figure shows the clinical progression and causative factors of HCC. Dashed grey lines represent causative factors in the development and progression of HCC. Solid black arrows represent progressive deterioration. Solid black, orange and red lines (no arrows) indicate established frontline or standard-of-care clinical interventions. The green check-marked “Application” boxes summarize the clinical indications for each modality, while the red warning-boxed “Limitation” sections highlight the primary constraints, risks, or contraindications associated with each treatment approach.An infographic flowchart about clinical management algorithm for HCC based on the BCLC staging system. Top row shows risk factors and disease progression: Alcohol, HBV/HCV, MAFLD, Healthy Liver, Liver Fibrosis, Liver Cirrhosis, Hepatocellular Carcinoma. Bottom section is a three column layout. EARLY STAGE (BCLC 0/A): Resection, Transplantation, Local Ablation. Current Main Modalities of Local Ablation: Radiofrequency Ablation (RA), Microwave Ablation (MWA). Selected or Emerging Options of Local Ablation: Cryoablation (CA), Irreversible Electroporation (IRE). Application: Curative options for localized disease; Transplantation treats tumor and cirrhosis; Ablation for small tumors or non-surgical patients. Limitation: Resection: recurrence remains common; Transplantation: donor shortage and strict selection; Ablation: heat-sink effect and nearby-organ risk. INTERMEDIATE STAGE (BCLC B): Radiotherapy and Transarterial Therapies. Radiotherapy list: Stereotactic Body Radiation Therapy (SBRT), Proton therapy, Interstitial brachytherapy. Transarterial list: Transarterial embolization (TAE), Conventional TACE (cTACE), Drug-eluting bead TACE (DEB-TACE), Radio-embolization (TARE). Application: cTACE: standard of care for suitable BCLC B; DEB-TACE: selected delivery strategy; TARE or radiotherapy: selected patients. Limitation: Usually not curative; Reduced efficacy with atypical blood supply; Systemic therapy if locoregional therapy is unsuitable. ADVANCED STAGE (BCLC C): A numbered list of systemic therapy options, 1 to 5. 1, Preferred 1st line: Atezolizumab plus bevacizumab. 2, Other preferred 1st line: Tremelimumab plus durvalumab (STRIDE). 3, Other 1st-line options: Lenvatinib or sorafenib. 4, Later-line options: Regorafenib after sorafenib; cabozantinib; ramucirumab if AFP greater than or equal to 400 ng/ml. 5, Selected subsequent options: setting-dependent immunotherapy. Application: Preferred for advanced HCC; Also for BCLC B unsuitable for locoregional therapy; Sequence if liver function and fitness are preserved. Limitation: Limited long-term survival benefit; Drug resistance and tumor heterogeneity; Impaired liver function may restrict therapy.An infographic flowchart of HCC progression and BCLC stage treatments: Early, Intermediate, Advanced. Clinical management algorithm for HCC based on the BCLC staging system. This schematic delineates the recommended therapeutic strategies stratified by Barcelona Clinic Liver Cancer (BCLC) stages (Early, Intermediate, and Advanced) indicated by green, orange, and red headings respectively. The upper part of the figure shows the clinical progression and causative factors of HCC. Dashed grey lines represent causative factors in the development and progression of HCC. Solid black arrows represent progressive deterioration. Solid black, orange and red lines (no arrows) indicate established frontline or standard-of-care clinical interventions. The green check-marked “Application” boxes summarize the clinical indications for each modality, while the red warning-boxed “Limitation” sections highlight the primary constraints, risks, or contraindications associated with each treatment approach. Methodology Literature Search A literature search was conducted in PubMed and Scopus. Search terms combined “hepatocellular carcinoma” or “HCC” with terms related to resection, transplantation, local ablation, transarterial therapy, radiotherapy, systemic therapy, targeted therapy, or immunotherapy. For prodrug studies, HCC terms were combined with “prodrug” and acidity-, hypoxia-, GSH-, ROS-, enzyme-, or lactate-related terms. Google Scholar and ScienceDirect were used for supplementary citation tracking and full-text retrieval, and reference lists were screened manually. No lower date limit was applied, but recent evidence was prioritized. Inclusion Criteria Human HCC trials and observational or comparative studies reporting efficacy or safety were included in the clinical section; guidelines and reviews provided clinical context. Core prodrug examples had to investigate HCC or liver cancer, contain a bona fide prodrug, and use an HCC-relevant TME cue for activation. Preclinical and clinical prodrug studies were eligible. Studies comparing HCC with adjacent diseased liver were included to assess trigger specificity. Exclusion Criteria Duplicate records, conference abstracts, case reports, letters, and studies without accessible full text or adequate methodological detail were excluded. Non-HCC studies, carriers containing unmodified drugs, imaging probes, and nanodelivery systems without a prodrug component were excluded from the core examples. Non-HCC prodrug studies were retained only when no HCC-specific example was available and were identified as contextual evidence. Results The reference set comprised 130 records dated 1956–2026; 81 (62.3%) were from 2022–2026. The clinical-treatment section included 42 references dated 2002–2025, of which 20 (47.6%) were from 2022–2026. The prodrug-design section included 71 references dated 1956–2026, of which 51 (71.8%) were from 2022–2026. The 11 core HCC prodrug studies were published between 2013 and 2026, and 8 (72.7%) were published in 2022–2026. Current Landscape of HCC Treatment Liver Resection Liver resection is regarded as the optimal curative treatment for HCC and is recommended for patients with solitary tumors and adequate liver function. The objective of treatment is to achieve R0 resection while preserving as much normal liver parenchyma as possible to reduce the risk of postoperative liver decompensation.5 At present, minimally invasive hepatectomy has become the prevailing clinical procedure, as its efficacy is comparable to that of open surgery and it offers a superior short-term prognosis.19,20 Advances in minimally invasive techniques have enabled the safe execution of minimally invasive hepatectomy for some patients with concomitant portal hypertension.21 However, the presence of multifocal tumors, satellite lesions, and vascular invasion has been demonstrated to result in suboptimal postoperative outcomes.22 The 5-year overall survival rate following minimally invasive hepatectomy is approximately 70%, but caution is warranted due to an intrahepatic recurrence rate as high as 80%. At present, numerous clinical trials are investigating the potential of immunotherapy as an adjuvant treatment.5 Liver Transplantation Liver transplantation has been demonstrated to address both HCC and liver cirrhosis, with a 5-year survival rate ranging from 75% to 80%, accompanied by a recurrence rate of approximately 15%. This approach is particularly well-suited for patients presenting with multifocal lesions (one lesion < 5cm or three or fewer lesions, each < 3cm; no extrahepatic manifestations; no evidence of macrovascular invasion) or decompensated cirrhosis.5,23 Nevertheless, the main constraints associated with it pertain to a paucity of donors and the necessity for meticulous screening procedures. Typically, patients must wait 6–9 months for a donor, during which time they undergo bridging therapies, such as TACE, transarterial radioembolization (TARE), radiotherapy (RT) and ablation to prevent tumor progression. In this context, living donor liver transplantation (LDLT) has emerged as a preferable option for patients with HCC, as it can reduce the rate of patients being removed from the waiting list and enhance treatment outcomes.24 However, its implementation necessitates a cautious approach. Local Ablation Thermal ablation is an appropriate treatment for early-stage HCC (≤2 cm) and inoperable lesions measuring 2–4 cm. However, this treatment has limitations, including the thermal sink effect (a cooling effect resulting in a smaller ablation volume on the side adjacent to blood vessels due to blood flow) and toxic reactions in nearby organs.23 Radiofrequency ablation (RFA) is a frequently utilized ablation technique. The alternating radiofrequency waves utilized in this process induce alterations in the direction of ions, thereby inducing ionic oscillation and frictional heat. This elevated thermal energy leads to the destruction of tumor tissue, with temperatures reaching over 60°C.25 Local ablation is considered a potentially curative treatment for small HCCs measuring < 3 cm, with RFA recommended by several clinical practice guidelines as a first-line option for solitary tumors < 2cm.23 For solitary HCCs measuring 2–5 cm, no-touch multibipolar RFA yields survival rates comparable to surgical resection (SR), but is associated with a higher recurrence rate.26 Microwave ablation (MWA) has emerged as the prevailing ablation modality for early-stage HCC, owing to its ability to attain higher ablation temperatures, reduce procedural duration, potentially enhance efficacy against HCCs ≤ 5 cm in size, and moreover, overcome the thermal sink effect.27 The present study explores the potential of alternative liver ablation techniques, including cryoablation (CA) and irreversible electroporation, which are currently under investigation. Research has demonstrated that the advent of smaller argon-helium probes has led to a significant advancement in the field of cryoablation, offering a level of safety and efficacy comparable to that of RFA and MWA.28 This development has led to a shift in treatment paradigms, with CA emerging as the preferred choice for precise treatment of lesions near delicate structures. In a propensity-score-matched population analysis, CA was not associated with significantly inferior liver cancer-specific survival (LCSS) or OS compared with RFA; however, the retrospective design and the small matched cohort limit the strength of this finding.29 The combination of local ablation with other therapeutic modalities is also currently under investigation. Intra-Arterial Therapy HCC is a highly vascularized neoplasm. Embolization within the hepatic artery has been demonstrated to induce tumor hypoxia and necrosis; consequently, intra-arterial therapy is the cornerstone of treatment for intermediate-stage HCC. The classification system is divided into three categories: transarterial embolization (TAE), transarterial chemoembolization (conventional transarterial chemoembolization (cTACE) or drug-eluting bead (DEB)-TACE), and transarterial radioembolization (TARE). cTACE is the international standard of care, with an overall survival (OS) of 20–36 months.30 DEB-TACE technology facilitates prolonged and targeted drug release, along with a reduction in adverse reactions, but nevertheless, the efficacy of cTACE and DEB-TACE remains comparable to that of TAE.31,32 TARE delivers 90Y-loaded glass (20–30 μm) or resin (20–60 μm) microspheres through the hepatic artery, with limited embolic effects and preservation of hepatic arterial flow. Consequently, post-embolization syndrome is largely minimized, with fatigue being its most prominent manifestation, and the procedure can generally be performed on an outpatient basis.5,33 TARE offers high tumor response rates and prolonged progression-free survival; it has been primarily used for locally advanced HCC early and was included in the Barcelona Clinic Liver Cancer (BCLC) curative treatment category in 2021.5,34 Several early trials evaluating TACE combined with antiangiogenic molecularly targeted agents failed to achieve their expected progression-free survival (PFS) or OS endpoints. The TACTICS trial was an exception: although the addition of sorafenib to TACE did not significantly improve OS, it significantly prolonged TACE-specific PFS.35 More recently, the Phase III EMERALD-1 and LEAP-012 trials showed that TACE combined with immune checkpoint blockade and antiangiogenic therapy significantly improved PFS. However, additional overall survival data and longer follow-up are required to establish the clinical benefit of these combinations.36,37 However, intra-arterial therapy cannot achieve curative eradication of the tumor, and its efficacy is limited in tumors with atypical blood supply. When the tumor develops collateral circulation or exhibits progressive metastasis, treatment must be transitioned to systemic therapy. Clinicians must maintain vigilance regarding safety and the risk of complications throughout the treatment process. Radiotherapy Common radiotherapy modalities for HCC include stereotactic body radiation therapy (SBRT), proton therapy and interstitial brachytherapy. The high precision of these localized treatments enables high-dose irradiation while concomitantly reducing the risk of radiation-induced liver disease.5 SBRT achieves a one-year local control rate of 92.3%,38 and is suitable for patients with early-stage disease and those at high risk of portal vein invasion.39 It can be used as a palliative treatment or to prepare patients for liver transplantation.38,40 Systemic Therapy Systemic Therapy is the treatment of choice for advanced HCC, and it is also suitable for patients with intermediate-stage disease who are ineligible for local therapy. Since 2017, the treatment of advanced HCC has entered the era of sequential therapy.5 In first-line treatment, sorafenib, a tyrosine kinase inhibitor (TKI), was the first targeted therapy. The SHARP study demonstrated that the OS rate was 10.7 months, which was significantly longer than the 7.9 months observed in the placebo group.41 The adverse reactions that were primarily observed included hand-foot syndrome and diarrhea. Furthermore, the safety outcomes were suboptimal.41,42 Lenvatinib, another first-line treatment, demonstrated superior median progression-free survival and objective response rates compared to sorafenib; however, it was associated with more pronounced adverse effects, particularly hypertension and proteinuria.42 Nivolumab was the first immune checkpoint inhibitor to receive accelerated FDA approval for HCC after sorafenib in 2017, although this monotherapy indication was subsequently withdrawn in 2021.43 Nivolumab plus ipilimumab significantly improved OS compared with Lenvatinib or sorafenib in patients with previously untreated unresectable HCC. However, hepatic adverse events (AEs) remain an important safety concern and require close monitoring during treatment.44 In 2025, this combination received regular FDA approval as first-line treatment for unresectable or metastatic HCC, and the most common adverse reactions (>20%) were rash, pruritus, fatigue and diarrhea.45 IMbrave150 was the first positive phase III study based on an ICI regimen to demonstrate a significant survival benefit; the combination of the VEGF-A antibody, bevacizumab, and the PD-L1 antibody, atezolizumab, significantly improved OS, overall response rate (ORR), and PFS, outperforming sorafenib.46 The median OS (19.2 months) of IMbrave150 is the longest median OS of any first-line treatment for advanced HCC to date.47 The IMbrave150 trial signifies a paradigm shift toward ICI therapy for HCC and establishes a new standard for first-line treatment of advanced HCC.48 The risk of gastrointestinal bleeding associated with bevacizumab requires particular attention. In the IMbrave150 trial, patients underwent endoscopic evaluation within 6 months before enrolment, and those with incompletely treated varices or varices associated with a high risk of bleeding were excluded. Bleeding events of any grade, including fatal events, were more frequent with atezolizumab plus bevacizumab than with sorafenib.47 The HIMALAYA trial was the first to report outcomes of dual ICI therapy and the largest Phase III first-line study for advanced HCC. A comparative analysis was conducted to assess the efficacy of the STRIDE regimen (a single priming dose of tremelimumab, an anti-cytotoxic T-lymphocyte-associated antigen 4 antibody, plus regular interval durvalumab, an anti-programmed cell death ligand-1 antibody, administered once every 4 weeks) and durvalumab monotherapy in comparison to sorafenib.49,50 The STRIDE regimen demonstrated a sustained and clinically meaningful OS benefit, while durvalumab monotherapy exhibited non-inferiority to sorafenib. STRIDE and durvalumab monotherapy resulted in 5-year OS rates of 19.6% and 14.4%, respectively, compared with 9.4% for sorafenib.23 However, the addition of tremelimumab was associated with a higher incidence of immune-mediated AEs, including grade 3–4 events. Grade 3–4 immune-mediated AEs occurred in 12.6% of patients receiving STRIDE and 6.2% of those receiving durvalumab monotherapy, while high-dose corticosteroids were required in 20.1% and 9.5% of patients respectively.49 The CARES-310 trial demonstrated significant improvements in PFS at the primary analysis and OS at the interim analysis with camrelizumab plus rivoceranib compared with sorafenib. At the final analysis, the median follow-up was 22.1 months in the camrelizumab-rivoceranib group and 14.9 months in the sorafenib group. Median OS was 23.8 versus 15.2 months, while median PFS was 5.6 versus 3.7 months, respectively. These findings further confirmed the clinically meaningful survival benefit of camrelizumab plus rivoceranib and support this combination as a new first-line treatment option for unresectable HCC.51 Although its safety profile was reported to be manageable, increased aspartate aminotransferase, increased blood bilirubin, upper gastrointestinal hemorrhage, increased alanine aminotransferase, and the relatively high overall incidence of grade 3–4 toxicities remain important clinical limitations.52 Tislelizumab is a monoclonal antibody with high affinity and binding specificity for PD-1. In the RATIONALE-301 trial, tislelizumab demonstrated noninferior OS and produced a higher ORR with more durable responses than sorafenib. However, disease control rate and median PFS favored sorafenib. These findings support tislelizumab as a potential first-line treatment option for unresectable HCC. Although tislelizumab showed a more favorable overall safety profile than sorafenib, clinically relevant immune-mediated toxicity remained. Moreover, RATIONALE-301 included only patients with preserved liver function (Child-Pugh class A). The safety and efficacy of tislelizumab in patients with moderate-to-severe hepatic impairment remain insufficiently established.53 A number of immunotherapy agents, including nivolumab, have received FDA accelerated approval for second-line treatment.54 However, only pembrolizumab has demonstrated positive results in Asian populations.55 Standard second-line treatments include regorafenib, cabozantinib, and ramucirumab. Regorafenib is an oral multikinase inhibitor that blocks the activity of protein kinase involved in angiogenesis, oncogenesis, metastasis, and tumor immunity. In a randomized, double-blind, placebo-controlled phase III trial, regorafenib significantly improved OS in patients whose disease had progressed during sorafenib treatment. Significant improvements over placebo were also observed in secondary endpoints, including PFS, time to progression, disease control rate, and objective tumor response. The most common grade 3–4 AEs included hypertension, hand-foot syndrome, fatigue and diarrhea.56 Cabozantinib, a multi-targeted TKI including VEGF receptors 1,2, and 3, MET and AXL, inhibits tumor growth in murine models of HCC. In the phase III CELESTIAL trial, cabozantinib significantly prolonged OS and PFS compared with placebo in patients with previously treated advanced HCC. The most common grade 3–4 AEs associated with cabozantinib were palmar-plantar erythrodysesthesia, hypertension, increased aspartate aminotransferase level, fatigue, and diarrhea. The incidence of grade 3–4 AEs was nearly twice as high with cabozantinib as with placebo (68% vs 36%).57 Ramucirumab, a human IgG1 monoclonal antibody that inhibits ligand activation of VEGFR2, improved median OS (8.5 months vs 7.3 months) and PFS (2.8 months vs 1.6 months) compared with placebo in patients with HCC who had previously received sorafenib and had AFP levels ≥ 400 ng/mL in the phase III REACH-2 trial. Ramucirumab was well-tolerated. Hypertension and hyponatraemia were the only grade ≥3 treatment-related adverse events that occurred in at least 5% of patients and were more frequent with ramucirumab than with placebo. However, the generalizability of REACH-2 was limited because the trial excluded patients with Child–Pugh class B or C liver function, those with AFP levels <400 ng/mL, and those who had received first-line treatments other than sorafenib. The efficacy of ramucirumab as a second-line treatment requires further evaluation in larger and more representative patient cohorts.58 Immune-Checkpoint Mechanisms, Biomarkers, Resistance, Toxicity, and Combination Rationale PD‑1 signaling suppresses T‑cell activation by recruiting SHP2 upon ligand engagement, thereby attenuating both T‑cell receptor- and CD28‑dependent signaling. CTLA‑4 functions primarily during T‑cell priming, competing with CD28 for CD80/CD86 binding, and can further reduce antigen presentation through CD80 trans‑endocytosis.59 In HCC, these pathways are accompanied by additional inhibitory receptors, including TIM‑3, LAG‑3, and TIGIT, which may sustain T‑cell dysfunction and provide compensatory routes for immune evasion.59 Reliable biomarkers for patient selection in ICI therapy remain an unmet need. PD‑L1 expression alone has inconsistent predictive value in HCC, partly because the relative abundance of PD‑L1 and CD80 can modulate both PD‑1‑ and CTLA‑4‑mediated signaling.59 Candidate biomarkers—including the PD‑L1/CD80 expression ratio, tumor mutational burden, circulating tumor DNA, TME characteristics, and microbiome profiles—require prospective validation before incorporation into routine clinical practice.59 ICI resistance is multifactorial. Impaired antigen recognition or presentation, immunosuppressive cytokines and cells, loss of tumor antigens, tumor heterogeneity, and microbiome dysbiosis can all contribute to primary or acquired resistance.59 Compensatory upregulation of the PVR/TIGIT axis, as well as TIM‑3 and LAG‑3, may provide alternative inhibitory signals following PD‑1 blockade. Recruitment of myeloid‑derived suppressor cells, TGF‑β signaling, and WNT/β‑catenin activation can further suppress antitumor immunity by limiting dendritic‑cell recruitment and T‑cell infiltration.59 Because checkpoint blockade lacks tumor specificity, activation of self‑reactive immune responses can cause immune‑related adverse events affecting the skin, gastrointestinal tract, liver, lungs, endocrine organs, and other systems.59 Immune‑mediated hepatitis is particularly relevant in HCC, given that most patients have underlying chronic liver disease. Thus, treatment selection should account for hepatic reserve, tumor burden, prior therapy, and the feasibility of close toxicity monitoring.59 The rationale for ICI‑based combinations is mechanistically complementary. Dual‑checkpoint blockade targets non‑redundant inhibitory pathways, locoregional therapy can release tumor antigens and promote immune priming, and antiangiogenic agents may improve vascular function and immune‑cell infiltration.59 These combinations may nevertheless increase toxicity and do not eliminate the need for predictive biomarkers. From a prodrug‑design perspective, conditional intratumoral release of cytotoxic or immunomodulatory payloads could theoretically support immunogenic cell death or attenuate local immunosuppression while limiting systemic exposure. This concept, however, remains preclinical and will require tumor‑selective activation relative to adjacent diseased tissue to achieve meaningful therapeutic benefit. Although systemic therapy options for advanced HCC have expanded, durable disease control remains unattained for many patients. Treatment selection and efficacy are constrained by primary and acquired resistance, the absence of validated predictive biomarkers, marked molecular and immune heterogeneity, and the limited evidence available for patients with impaired liver function.9,23 Treatment-related hepatic, vascular and immune toxicities can further narrow the therapeutic window in patients with underlying chronic liver disease.23 Against this background, TME-responsive prodrugs provide a rational experimental framework for conditionally generating or releasing active agents within HCC while limiting systemic drug exposure.11,12 These strategies should nevertheless be regarded as potential adjuncts to, rather than replacements for, established clinical modalities. Their translational value will depend on achieving consistent activation across heterogeneous HCC lesions and adequate discrimination between tumor tissues and the surrounding diseased liver (Table 1). Table 1Clinical Limitations of Current HCC Therapies and Potential Implications for TME-Responsive Prodrug DesignClinical ModalityMajor Clinical LimitationRelevant TME or Treatment-Induced FeaturesPotential Prodrug Design DirectionsKey Translational ConsiderationsLiver resectionEligibility is constrained by tumor burden and location, portal hypertension, and hepatic reserve. Recurrence remains common after curative-intent resection.22,23Postoperative recurrence may arise from occult intrahepatic metastases or de novo tumors in the diseased remnant liver. Whether microscopic residual disease provides a sufficiently strong and reproducible hypoxic, acidic, redox, or enzymatic activation differential remains unestablished22,60If investigated in the postoperative setting, a prodrug should require an HCC-associated marker plus an independent TME input, rather than relying on acidity or hypoxia alone.12,61Fibrotic or cirrhotic remnant liver can share HCC-derived stromal programs with HCC. Postoperative prevention has not been clinically demonstrated, and evaluation requires disease-relevant models.60,62Liver transplantationClinical use is limited by donor availability, strict eligibility criteria, and waiting-list progression.23,24No transplant-specific endogenous activation signal has been established. Maintenance immunosuppression alters host immune competence and may affect the biology and treatment tolerability of recurrent HCC.23TME-responsive prodrugs could be investigated as antitumor therapy during bridging or downstaging, or for post-transplant recurrence; they cannot address donor scarcity or transplant eligibility.12,23Drug-immunosuppressant interactions, infection risk, hepatic safety, and the altered immune context require dedicated evaluation.23Local ablationEfficacy depends on tumor size and location, adequate ablation margins, and modality-specific technical constraints. Residual viable tumor can cause local progression.23Local ablation can remodel tumor signaling and immunity, but the effects differ by modality. Incomplete thermal ablation, particularly RFA, can expose residual HCC to sublethal heat, hypoxia/HIF signaling, inflammation, ROS, and increased MMP2/MMP9 expression. Cryoablation, IRE, and chemical ablation produce distinct injury patterns.63–65Current mechanistic evidence most directly supports evaluating hypoxia-, ROS-, or protease-responsive prodrugs in residual disease after incomplete thermal ablation. For cryoablation, IRE, or chemical ablation, the trigger should be selected only after a reproducible modality-specific signal has been demonstrated.18,65,66Injury-associated signals also arise in non-malignant liver. An additional HCC-associated input may therefore be required. Evidence remains preclinical and is not interchangeable among ablation modalities.61,63–65TACETACE can cause incomplete and spatially heterogeneous necrosis. Repeated procedures may worsen hepatic reserve.23Arterial embolization intensifies hypoxia and HIF signaling in residual tumor. The magnitude and distribution of hypoxia vary within and among lesions.67,68Hypoxia-activated prodrugs are directly relevant to embolization-induced hypoxia. Intra-arterial administration and sustained local retention may limit systemic exposure, while an additional HCC-associated input may improve selectivity.61,66,68Severe embolization may reduce subsequent drug delivery. Hypoxia is heterogeneous, and the interval between embolization and prodrug administration requires pharmacokinetic and pharmacodynamic optimization.60,68TARE and radiotherapyTreatment is constrained by hepatic reserve, radiation exposure to non-tumor liver, dosimetry, and the risk of radiation-induced liver injury.23,40Ionizing radiation can generate ROS that contribute to DNA damage and cell death. Pre-existing tumor hypoxia can reduce radiosensitivity and produce heterogeneous responses to radiotherapy.69,70Radiation-amplified ROS-responsive prodrugs or hypoxia-activated cytotoxins could be evaluated as radiosensitizing strategies.18,66Radiation-generated ROS are not tumor-exclusive. Dosimetry and the interval between irradiation and prodrug administration therefore require definition.40,60TKIs and antiangiogenic therapyClinical benefit is limited by primary or acquired resistance, chronic systemic exposure, treatment-related toxicity, and impaired hepatic function.23Antiangiogenic therapy dynamically remodels tumor vasculature. Depending on dose and timing, vessel pruning can reduce perfusion and drug delivery, whereas transient vascular normalization can improve both.71 Hypoxia-associated metabolic adaptation may contribute to treatment resistance in HCC.67The active drug could be chemically masked and released through a linker responsive to hypoxia, acidity, ROS, or an HCC-associated enzyme. Multi-input activation may further restrict active-drug exposure.12,61,66Heterogeneous perfusion may limit prodrug delivery, and hepatic dysfunction may alter exposure and safety. Improvement in resistance or tolerability remains unproven clinically.23,60Immune-checkpoint blockade and immunotherapy-based combinationsOnly a subset of patients achieves durable responses. Primary or acquired resistance, immune-related hepatic events, and limited evidence in patients with impaired liver function remain important constraints.72Excess lactate production and the associated extracellular acidification can contribute to an immunosuppressive microenvironment in HCC.73Prodrugs that couple TME-restricted cytotoxic activation with lactate modulation or ICD induction could be investigated with checkpoint blockade. Evidence in HCC is currently preclinical.74,75An antibody or active drug encapsulated in a responsive carrier is not, by itself, a bona fide prodrug. Overlap with inflammatory liver signals and immune-mediated hepatotoxicity may narrow the therapeutic window.12,62,72Impaired hepatic reserve and adjacent diseased liverTreatment eligibility and tolerability depend on underlying liver function as well as tumor burden.23,72HSCs generate activated myofibroblasts in liver fibrosis and contribute to CAF populations in HCC; HSC-associated stromal activation is therefore not tumor-exclusive.62Multi-input prodrugs requiring an HCC-associated marker plus an independent physicochemical or enzymatic input should be prioritized for investigation.61The relevant benchmark is the quantitative activation differential between HCC and adjacent diseased liver, not between HCC and healthy liver. Validation should use fibrotic or cirrhotic models and paired human tissues.60,62Interpatient and intratumoral heterogeneityHCC heterogeneity and the lack of validated predictive biomarkers impede patient stratification and durable treatment response.23,60Immune infiltration and stromal composition contribute to interpatient heterogeneity in HCC.60 Spatial single-cell profiling has further demonstrated heterogeneous TME architectures within HCC.76Biomarker-guided selection of single- or multi-input prodrugs should match treatment to a measurable patient-specific activation phenotype.60,61Clinically applicable activation thresholds, pharmacokinetics, safety margins, and sampling strategies remain to be established.12,60Notes: The proposed design directions are testable hypotheses, not established clinical benefits. A bona fide prodrug is defined here as a chemically or biochemically masked therapeutic entity that is converted to, or releases, its active form. Responsive carriers that merely contain unmodified active drugs, diagnostic probes, and non-cleavable delivery systems are excluded. Treatment-induced signals are distinguished from endogenous TME cues, and evidence for these signals is modality-specific.Abbreviations: CAF, cancer-associated fibroblast; EBRT, external-beam radiotherapy; HCC, hepatocellular carcinoma; HIF, hypoxia-inducible factor; HSC, hepatic stellate cell; ICD, immunogenic cell death; IRE, irreversible electroporation; MMP, matrix metalloproteinase; MWA, microwave ablation; RFA, radiofrequency ablation; ROS, reactive oxygen species; TACE, transarterial chemoembolization; TARE, transarterial radioembolization; TKI, tyrosine kinase inhibitor; TME, tumor microenvironment. Table 1Clinical Limitations of Current HCC Therapies and Potential Implications for TME-Responsive Prodrug DesignClinical ModalityMajor Clinical LimitationRelevant TME or Treatment-Induced FeaturesPotential Prodrug Design DirectionsKey Translational ConsiderationsLiver resectionEligibility is constrained by tumor burden and location, portal hypertension, and hepatic reserve. Recurrence remains common after curative-intent resection.22,23Postoperative recurrence may arise from occult intrahepatic metastases or de novo tumors in the diseased remnant liver. Whether microscopic residual disease provides a sufficiently strong and reproducible hypoxic, acidic, redox, or enzymatic activation differential remains unestablished22,60If investigated in the postoperative setting, a prodrug should require an HCC-associated marker plus an independent TME input, rather than relying on acidity or hypoxia alone.12,61Fibrotic or cirrhotic remnant liver can share HCC-derived stromal programs with HCC. Postoperative prevention has not been clinically demonstrated, and evaluation requires disease-relevant models.60,62Liver transplantationClinical use is limited by donor availability, strict eligibility criteria, and waiting-list progression.23,24No transplant-specific endogenous activation signal has been established. Maintenance immunosuppression alters host immune competence and may affect the biology and treatment tolerability of recurrent HCC.23TME-responsive prodrugs could be investigated as antitumor therapy during bridging or downstaging, or for post-transplant recurrence; they cannot address donor scarcity or transplant eligibility.12,23Drug-immunosuppressant interactions, infection risk, hepatic safety, and the altered immune context require dedicated evaluation.23Local ablationEfficacy depends on tumor size and location, adequate ablation margins, and modality-specific technical constraints. Residual viable tumor can cause local progression.23Local ablation can remodel tumor signaling and immunity, but the effects differ by modality. Incomplete thermal ablation, particularly RFA, can expose residual HCC to sublethal heat, hypoxia/HIF signaling, inflammation, ROS, and increased MMP2/MMP9 expression. Cryoablation, IRE, and chemical ablation produce distinct injury patterns.63–65Current mechanistic evidence most directly supports evaluating hypoxia-, ROS-, or protease-responsive prodrugs in residual disease after incomplete thermal ablation. For cryoablation, IRE, or chemical ablation, the trigger should be selected only after a reproducible modality-specific signal has been demonstrated.18,65,66Injury-associated signals also arise in non-malignant liver. An additional HCC-associated input may therefore be required. Evidence remains preclinical and is not interchangeable among ablation modalities.61,63–65TACETACE can cause incomplete and spatially heterogeneous necrosis. Repeated procedures may worsen hepatic reserve.23Arterial embolization intensifies hypoxia and HIF signaling in residual tumor. The magnitude and distribution of hypoxia vary within and among lesions.67,68Hypoxia-activated prodrugs are directly relevant to embolization-induced hypoxia. Intra-arterial administration and sustained local retention may limit systemic exposure, while an additional HCC-associated input may improve selectivity.61,66,68Severe embolization may reduce subsequent drug delivery. Hypoxia is heterogeneous, and the interval between embolization and prodrug administration requires pharmacokinetic and pharmacodynamic optimization.60,68TARE and radiotherapyTreatment is constrained by hepatic reserve, radiation exposure to non-tumor liver, dosimetry, and the risk of radiation-induced liver injury.23,40Ionizing radiation can generate ROS that contribute to DNA damage and cell death. Pre-existing tumor hypoxia can reduce radiosensitivity and produce heterogeneous responses to radiotherapy.69,70Radiation-amplified ROS-responsive prodrugs or hypoxia-activated cytotoxins could be evaluated as radiosensitizing strategies.18,66Radiation-generated ROS are not tumor-exclusive. Dosimetry and the interval between irradiation and prodrug administration therefore require definition.40,60TKIs and antiangiogenic therapyClinical benefit is limited by primary or acquired resistance, chronic systemic exposure, treatment-related toxicity, and impaired hepatic function.23Antiangiogenic therapy dynamically remodels tumor vasculature. Depending on dose and timing, vessel pruning can reduce perfusion and drug delivery, whereas transient vascular normalization can improve both.71 Hypoxia-associated metabolic adaptation may contribute to treatment resistance in HCC.67The active drug could be chemically masked and released through a linker responsive to hypoxia, acidity, ROS, or an HCC-associated enzyme. Multi-input activation may further restrict active-drug exposure.12,61,66Heterogeneous perfusion may limit prodrug delivery, and hepatic dysfunction may alter exposure and safety. Improvement in resistance or tolerability remains unproven clinically.23,60Immune-checkpoint blockade and immunotherapy-based combinationsOnly a subset of patients achieves durable responses. Primary or acquired resistance, immune-related hepatic events, and limited evidence in patients with impaired liver function remain important constraints.72Excess lactate production and the associated extracellular acidification can contribute to an immunosuppressive microenvironment in HCC.73Prodrugs that couple TME-restricted cytotoxic activation with lactate modulation or ICD induction could be investigated with checkpoint blockade. Evidence in HCC is currently preclinical.74,75An antibody or active drug encapsulated in a responsive carrier is not, by itself, a bona fide prodrug. Overlap with inflammatory liver signals and immune-mediated hepatotoxicity may narrow the therapeutic window.12,62,72Impaired hepatic reserve and adjacent diseased liverTreatment eligibility and tolerability depend on underlying liver function as well as tumor burden.23,72HSCs generate activated myofibroblasts in liver fibrosis and contribute to CAF populations in HCC; HSC-associated stromal activation is therefore not tumor-exclusive.62Multi-input prodrugs requiring an HCC-associated marker plus an independent physicochemical or enzymatic input should be prioritized for investigation.61The relevant benchmark is the quantitative activation differential between HCC and adjacent diseased liver, not between HCC and healthy liver. Validation should use fibrotic or cirrhotic models and paired human tissues.60,62Interpatient and intratumoral heterogeneityHCC heterogeneity and the lack of validated predictive biomarkers impede patient stratification and durable treatment response.23,60Immune infiltration and stromal composition contribute to interpatient heterogeneity in HCC.60 Spatial single-cell profiling has further demonstrated heterogeneous TME architectures within HCC.76Biomarker-guided selection of single- or multi-input prodrugs should match treatment to a measurable patient-specific activation phenotype.60,61Clinically applicable activation thresholds, pharmacokinetics, safety margins, and sampling strategies remain to be established.12,60Notes: The proposed design directions are testable hypotheses, not established clinical benefits. A bona fide prodrug is defined here as a chemically or biochemically masked therapeutic entity that is converted to, or releases, its active form. Responsive carriers that merely contain unmodified active drugs, diagnostic probes, and non-cleavable delivery systems are excluded. Treatment-induced signals are distinguished from endogenous TME cues, and evidence for these signals is modality-specific.Abbreviations: CAF, cancer-associated fibroblast; EBRT, external-beam radiotherapy; HCC, hepatocellular carcinoma; HIF, hypoxia-inducible factor; HSC, hepatic stellate cell; ICD, immunogenic cell death; IRE, irreversible electroporation; MMP, matrix metalloproteinase; MWA, microwave ablation; RFA, radiofrequency ablation; ROS, reactive oxygen species; TACE, transarterial chemoembolization; TARE, transarterial radioembolization; TKI, tyrosine kinase inhibitor; TME, tumor microenvironment. Clinical Limitations of Current HCC Therapies and Potential Implications for TME-Responsive Prodrug Design Notes: The proposed design directions are testable hypotheses, not established clinical benefits. A bona fide prodrug is defined here as a chemically or biochemically masked therapeutic entity that is converted to, or releases, its active form. Responsive carriers that merely contain unmodified active drugs, diagnostic probes, and non-cleavable delivery systems are excluded. Treatment-induced signals are distinguished from endogenous TME cues, and evidence for these signals is modality-specific. Abbreviations: CAF, cancer-associated fibroblast; EBRT, external-beam radiotherapy; HCC, hepatocellular carcinoma; HIF, hypoxia-inducible factor; HSC, hepatic stellate cell; ICD, immunogenic cell death; IRE, irreversible electroporation; MMP, matrix metalloproteinase; MWA, microwave ablation; RFA, radiofrequency ablation; ROS, reactive oxygen species; TACE, transarterial chemoembolization; TARE, transarterial radioembolization; TKI, tyrosine kinase inhibitor; TME, tumor microenvironment. Classification and Activation Mechanisms of Prodrugs Targeting the HCC Microenvironment Cancer development is a multistep process that is strongly influenced by the surrounding microenvironment. Compared with normal tissues, tumors exhibit distinct cellular, biochemical, and physicochemical features that may serve as activation triggers for prodrug design. The TME comprises immune cells, such as T lymphocytes, macrophages, neutrophils, and dendritic cells, and non-immune cells, such as fibroblasts, blood and lymphatic vessel endothelial cells, and the extracellular matrix.77 It is frequently characterized by hypoxia, weak acidity, altered redox homeostasis, including elevated reactive oxygen species and glutathione levels, dysregulated enzyme expression and metabolic reprogramming.15 These features can be exploited to achieve preferential drug activation within tumors, thereby limiting systemic exposure to the active drug. The microenvironment of HCC exhibits disease-specific and tissue-specific characteristics that distinguish it from other solid tumors. The tumor develops in the context of chronic liver injury, inflammation, fibrosis, and cirrhosis. Additionally, the liver’s immunoprivileged nature contributes to the formation of a distinctive immunosuppressive microenvironment. Furthermore, the stromal components, mainly activated hepatic stellate cells (HSCs), and abnormal angiogenesis associated with sinusoidal capillary formation collectively constitute features that distinguish HCC from other solid tumors.78 The liver, as the metabolic hub of the body, undergoes increasingly extreme and complex metabolic reprogramming as HCC progresses. It is important to note that the microenvironment surrounding the diseased liver tissue shares similar or overlapping characteristics with the HCC TME. This may lead to off-target activation and compromise the efficacy of single-response prodrugs. We will also provide detailed explanations in the corresponding sections regarding prodrug activation. A comprehensive analysis of the TME characteristics of HCC, in conjunction with their alignment and classification with prodrug activation mechanisms, constitutes the theoretical foundation for establishing an efficient and safe HCC prodrug system (Figure 2 and Table 2). Table 2Representative Bona Fide Tumor Microenvironment-Responsive Prodrug Systems Evaluated in HCC Models or PatientsActivation MechanismDrug or FormulationProdrug ClassificationPayloadDisease ModelPharmacokinetic and Distribution FeaturesHepatic and Systemic ToxicityControl Groups and Treatment DesignDevelopment StagepH-responsiveHis-CA@Pt/NPs79His–CA is a small-molecule prodrug containing a Schiff-base linkage.CA alone: drug-loading efficiency (DLE), approximately 95%; drug-loading content (DLC), approximately 50%. Dual loading: CA DLC, approximately 35%; CDDP DLC increased with feed ratio. The selected CDDP: His–CA mass ratio was 2.5:10.In vitro: HepG2 cells. In vivo: H22 tumor-bearing BALB/c mice.After intravenous administration at 5 mg/kg, the half-life was 0.45 h for the free drug and 2.55 h for the nanomedicine. The nanomedicine AUC was approximately twice that of free CDDP. Tumor concentrations were 4.92 μg/g at 4 h and 3.92 μg/g at 12 h. Pulmonary and renal accumulation was lower than with free CDDP.No overt hepatic or renal histopathological injury was observed under the tested conditions. ALT, AST, and MDA values were lower than those in the free-CDDP group. The hemolysis rate was 3.89% (<5%).Five parallel in vivo groups (n=6) received three tail-vein injections: saline, free CA, free CDDP, CA plus CDDP, or His-CA@Pt/NPs.Preclinical; in vivo efficacy and preliminary safety were evaluated; mouse model.PGCA@PA NPs80PGCA is a polymeric prodrug containing acid-cleavable acetal linkages.PA loading content was 7.0%; a specific loading-efficiency value was not reported in the main text. CA constituted 27.3% by mass of the polymer.In vitro: HepG2, BEL-7402, and B16 cells. In vivo: HepG2 tumor-bearing BALB/c nude mice for efficacy and B16 tumor-bearing C57BL/6 mice for immunotherapy.After intravenous injection, the half-life was 1.15 h for free PA and 16.06 h for the nanomedicine. Tumor fluorescence was higher with the nanoparticle formulation than with free PA.No marked body-weight loss or overt morphological injury in the heart, liver, spleen, lungs, or kidneys was observed. ROS levels were lower in liver tissue than in tumor tissue. These findings provide preliminary evidence of tolerability in the tested models.The in vivo efficacy study included five groups (n=5): saline, free PA, free CA, blank PGCA NPs, and PGCA@PA NPs. Three tail-vein injections were administered; irradiated groups received light 24 h after dosing. The immunotherapy study included four groups (n=5): saline, anti-PD-1, irradiated PGCA@PA NPs, and anti-PD-1 plus irradiated PGCA@PA NPs.Preclinical; mouse modelGSH-responsiveFU-SS-IND NPs74Small-molecule prodrug containing a disulfide linkage.5-FU and IND were conjugated at a 1:1 molar ratio. The loading percentage was not specified. The nanoparticles were approximately 186.9 nm in diameter, with a polydispersity index of 0.1.In vitro: Hepa1-6 cells and 5-FU-resistant Hepa1-6/FU cells. In vivo: Hepa1-6/FU tumor-bearing C57BL/6J mice.IR780 fluorescence tracking showed a clear tumor signal at 12 h after tail-vein injection, which increased through 48 h, whereas hepatic fluorescence declined from 6 to 48 h. Ex vivo imaging at 48 h showed 46.5% tumor-associated fluorescence for FU-SS-IND NPs versus approximately 5.5% for free IR780 (a 41.0-percentage-point difference). These fluorescence data track the probe rather than directly quantifying released 5-FU or IND.Mice receiving free 5-FU lost approximately 22.8% of body weight and showed severe hepatic and renal histological injury. Body weight remained stable and no overt major-organ injury was reported with FU-SS-IND NPs, indicating reduced systemic toxicity in this model.The in vivo efficacy study included five groups (n=5): saline, free 5-FU, free IND, 5-FU plus IND, and FU-SS-IND NPs; four tail-vein injections were administered. A separate combination study compared saline, anti-PD-L1 alone, FU-SS-IND NPs alone, and FU-SS-IND NPs plus anti-PD-L1.Preclinical; mouse modelLipase/ROS-responsive(light-amplified)Hybrid adipocyte-derived exosome platform encapsulating DSTG and PPLA81DSTG is a triglyceride-mimetic docetaxel prodrug containing a lipase-cleavable ester and a ROS-responsive thioether. PPLA is a PUFAylated protoporphyrin IX prodrug that also serves as the photosensitizer.DSTG and PPLA co-assembled at a 1:1 mass ratio to form lipid-core nanoparticles. The loading percentage was not reported. Particle diameters were 132.5 nm for NPs, 181.1 nm for NEMPs, and 183.8 nm for HEMPs.In vitro: Hepa1-6, HepG2, and 3T3-L1 adipocytes. In vivo: Hepa1-6 tumor-bearing C57BL/6 mice.After intravenous injection, the half-life of the free prodrug mixture was 0.18 h. The AUC0–24 h values of the prodrugs in NPs, NEMPs, and HEMPs were 18.5-, 35.7-, and 53.3-fold higher, respectively, than that of the free mixture. HEMP tumor fluorescence peaked at 8 h and remained elevated at 24 h, when it was 1.32-fold that of NPs. PPLA fluorescence tracked the labeled prodrug and did not directly quantify released docetaxel.Body weight did not differ markedly among treatment groups. CREA, BUN, ALT, and AST showed no significant between-group differences, providing no evidence of overt hepatic or renal injury under the tested conditions.The in vivo study included nine groups (n=5): saline; free mixture with or without irradiation; NPs with or without irradiation; NEMPs with or without irradiation; and HEMPs with or without irradiation. A single tail-vein dose was administered on day 0, and irradiated groups received 660-nm light 8 h later.Preclinical; mouse modelHypoxia-responsiveTirapazamine (TPZ)82Bioreductive small-molecule prodrug.TPZ was administered as a free-drug solution by intra-arterial injection; drug-loading metrics are not applicable.Clinical study in patients with unresectable early- or intermediate-stage HCC (BCLC A or B); some had progressed after surgery, radiofrequency ablation, or TACE.Systemic pharmacokinetic exposure was not considered an appropriate surrogate for efficacy or toxicity because the intended effect depended on first-pass exposure after intra-arterial TPZ administration. After embolization, devascularized HCC was no longer exposed to circulating TPZ.The most frequent treatment-related events were increased AST and ALT (82.4% each). Grade ≥3 AST/ALT elevations occurred in 41.2%–47.1% of patients and were consistent with post-embolization syndrome. No dose-limiting toxicity was observed. Transaminase elevations were transient and generally returned to baseline within 2–3 days.Phase I, open-label, single-arm 3+3 dose-escalation study (n=17) without a parallel control. Three cohorts received 5, 10, or 20 mg/m2 intra-arterial TPZ, and the expansion cohort received a fixed 35-mg dose. Response was assessed using mRECIST.Phase I clinical studyTH302-DS@FMT83TH-302 is a small-molecule hypoxia-activated prodrug.The TH-302 loading rate was 21.99 ± 1.34% before FMT modification and 68.61 ± 6.44% after modification. After five FMT-modification/loading cycles, the total TH-302 payload reached 208.79 ± 6.67 mg.In vitro: RH35 cells under hypoxia (1% O2) or normoxia (21% O2). In vivo: diethylnitrosamine-induced primary multifocal HCC in rats.The study did not report half-life, AUC, or other conventional pharmacokinetic parameters for TH302-DS@FMT.TH302-DS@FMT caused mild hemolysis. ALT and AST were lower than in controls but did not fully normalize. TBIL and urea remained within the normal range; CREA was slightly elevated in controls and normalized in treated groups. No overt major-organ abnormality was observed by H&E staining.TH302-DS@FMT produced minimal hemolysis. ALT and AST were lower than in controls but did not fully normalize. TBIL and urea remained within the normal range, and creatinine was within the normal range in all treatment groups. No overt major-organ abnormality was observed by H&E staining.Preclinical; diethylnitrosamine-induced primary multifocal HCC in rats.Enzyme-responsivePDOX (Ac-Phe-Lys-PABC-DOX)84Small-molecule prodrug containing a Cat B-cleavable Phe–Lys dipeptide and a self-immolative PABC spacer.Not applicable. PDOX is a discrete small-molecule prodrug rather than a drug-loaded carrier.In vitro: highly metastatic human HCCLM9 cells. In vivo: orthotopic HCCLM9 tumors established by intrahepatic implantation in BALB/c nu/nu mice, with spontaneous pulmonary and peritoneal metastasis.The study did not report half-life or AUC values. PDOX was stable in human and mouse blood.ALT, total bilirubin, and direct bilirubin did not differ significantly among the control, DOX, and PDOX groups. Compared with DOX, PDOX was associated with a higher white-blood-cell count, lower serum creatinine and creatine kinase, and no observable myocardial histopathological injury. Healthy-liver histopathology was not separately assessed.Three groups received intravenous treatment: saline control (n=10), DOX at 4 mg/kg (n=11), or PDOX at 21.6 mg/kg (n=12).Preclinical; orthotopic HCC xenograft model in nude mice.FA/PF-LP-CaAs85As(V) is an inorganic prodrug reduced to As(III) by GSH. MMP2 cleavage of the PLGLAG sequence in the PF peptide enhances tumor penetration but does not directly activate As(V).Arsenic-loading efficiency in calcium arsenate nanoparticles was 34.28 ± 2.15%. The administered As dose for FA/PF-LP-CaAs was 1.9 mg/kg intravenously.In vitro: HepG2 and HCCLM3 cells. In vivo: HCCLM3 subcutaneous xenografts in BALB/c nude mice.Half-life and AUC were not reported. DiR-labeled FA/PF-LP-CaAs became detectable at the tumor site at 8 h and increased through 24 h. ICP–MS measured a tumor arsenic concentration of 276.82 ± 7.79 μg/kg, higher than those of the comparator formulations. Renal arsenic measured by ICP–MS was higher than suggested by DiR fluorescence, indicating that label-based imaging did not fully represent arsenic disposition.No marked body-weight fluctuation or major-organ histological injury was observed. Hemolysis remained below 5%, supporting intravenous compatibility in the tested model.Seven groups (n=5) received saline, free As(V), CaAs NP, CaAs LP, FA-LP-CaAs, PF-LP-CaAs, or FA/PF-LP-CaAs by tail-vein injection at 1.9 mg As/kg every other day for 10 doses.Preclinical; mouse xenograft model.HA/Lipid@Pro-Dox86Pro-Dox is a small-molecule γ-GGT-activated prodrug in which a γ-glutamyl group masks Dox and is enzymatically cleaved to release the active drug.Pro-Dox loading in HA/Lipid@Pro-Dox was 3.64%. No drug leakage was detected in buffer over 96 h.In vitro: Hepa1-6 and AML12 cells. In vivo: C57BL/6 mice bearing proximal and distal Hepa1-6 tumors.Half-life and AUC were not reported. After direct injection of Pro-Dox into tumor tissue and contralateral normal tissue, Dox fluorescence was stronger in the tumor. After intravenous administration, HA/Lipid@Pro-Dox tumor fluorescence peaked at 12 h and exceeded that of Lipid@Pro-Dox.Free Dox caused marked cytotoxicity in both Hepa1-6 and AML12 cells and significant body-weight loss in mice. Pro-Dox and HA/Lipid@Pro-Dox were more cytotoxic to Hepa1-6 than to AML12 cells. Mice receiving HA/Lipid@Pro-Dox gained weight, had normal blood-biochemistry indices, and showed no overt major-organ histopathological injury under the tested conditions.In vitro cytotoxicity compared free Dox, Pro-Dox, and HA/Lipid@Pro-Dox in Hepa1-6 and AML12 cells. The bilateral-tumor study included five groups (n=5): PBS, free Dox, Pro-Dox, HA/Lipid@Pro-Dox, and HA/Lipid@Pro-Dox plus anti-PD-L1.Preclinical; mouse model.Lactate-Coupled Prodrug ActivationGD-A NPs75Pentavalent arsenic [As(V)] is an inorganic prodrug reduced by H2S to trivalent arsenic [As(III)].Loading contents were 4.35% for As, 55.20% for diclofenac, and 33.40% for GYY4137. Encapsulation efficiencies were 68.14 ± 2.46% for diclofenac and 60.22 ± 6.85% for GYY4137.In vitro: H22 and HepG2 cells. In vivo: H22 subcutaneous tumors in BALB/c mice.Conventional half-life and AUC values were not reported. ICG-GD-A NP tumor fluorescence at 4 h was higher than that of free ICG. Ex vivo imaging at 24 h showed sustained tumor retention, but substantial hepatic fluorescence was also observed because of nanoparticle sequestration. ICG fluorescence did not directly quantify arsenic disposition or As(V)-to-As(III) conversion.Toxicity was lower than with free As(III). ALT and AST increased in the free-As(III) group but not in the GD-A NP group. No overt major-organ histopathological injury or body-weight loss was observed with GD-A NPs. Approximately 70% of the excreted arsenic was recovered in urine, and nearly all arsenic was eliminated through urine and feces by day 5.The in vitro study compared PBS, As(III), As(V), GD NPs, and GD-A NPs. The in vivo efficacy study used the same five groups (n=5) and administered 2 mg As/kg by tail-vein injection every other day for seven doses.Preclinical; mouse model.Notes: Data are reported as presented in the cited studies. Direct cross-study comparisons should be made cautiously because the formulations, doses, administration routes, disease models, sampling methods, and endpoints differed. Fluorescence-based biodistribution generally tracks a fluorescent label or labeled formulation and does not necessarily quantify the released active drug. In FA/PF-LP-CaAs, MMP2 cleavage enhances tumor penetration, whereas As(V) is activated by GSH-mediated reduction. In GD-A NPs, H2S activates As(V), whereas lactate accumulation is a downstream pharmacological effect rather than the direct activation trigger. Except for TPZ, all entries are supported only by preclinical evidence and do not establish clinical efficacy or safety.Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUC, area under the concentration–time curve; BCLC, Barcelona Clinic Liver Cancer; CA, cinnamaldehyde; CDDP, cisplatin; Dox/DOX, doxorubicin; DTX, docetaxel; FMT, Fe3⁺/Mn2⁺/tannic-acid metal–phenolic network; GGT, γ-glutamyl transpeptidase; GSH, glutathione; HCC, hepatocellular carcinoma; MMP2, matrix metalloproteinase-2; mRECIST, modified Response Evaluation Criteria in Solid Tumors; NPs, nanoparticles; PABC, para-aminobenzyloxycarbonyl; ROS, reactive oxygen species; TACE, transarterial chemoembolization; TPZ, tirapazamine. Figure 2Schematic representation of TME-responsive prodrug activation mechanisms in HCC. This schematic illustrates the molecular rationale for leveraging differential features of the HCC TME compared with normal tissues, including weak acidity, GSH and ROS levels, hypoxia, enzyme overexpression, and lactate accumulation, to trigger prodrug activation and release the active parent drug into cancer cells. The nine color-coded text boxes surrounding the central mechanism diagram provide concise descriptions of these TME-specific features and the corresponding prodrug design strategies. The numbering shown in the figure (eg, 4.1) corresponds to the relevant section header numbering in the main text of the article. Dashed black lines extending from each module indicate their conceptual anchoring to the central mechanism. Within the central illustration, the magnified tumor region is demarcated by solid gray lines and irregular light-brown shaded areas, while solid black arrows with arrowheads depict the differential expression features within the TME. The vertical red upward arrows indicate increased abundance, activity, or expression, as applicable (eg, GSH, HIF-1α, reductases, Cat B, GGT, MMP2/9, and GUS). The prodrug is rationally engineered as an active drug conjugated to a masking moiety. Upon entry into the TME (demonstrated by orange arrows), the prodrug is subjected to activation triggered by these differentially expressed features (represented by solid blue arrows), leading to the release of the active drug (emphasized by bold black arrows), which then enters and exerts therapeutic effects within the tumor cells (indicated by orange arrows).A schematic representation of TME-responsive prodrug activation mechanisms in HCC. The central diagram illustrates the tumor microenvironment (TME) with features like weak acidity (pH 6.5-6.9), high GSH level, high ROS level, hypoxic microenvironment and lactate-coupled prodrug activation. Surrounding the central illustration are nine text boxes detailing specific features and strategies: 4.1 Weak Acidity, 4.2.1 High GSH Level, 4.2.2 High ROS Level, 4.3 Hypoxic Microenvironment, 4.4.1 Cathepsin B Overexpression, 4.4.2 Matrix Metalloproteinase 2/9 Overexpression, 4.4.3 gamma-Glutamyltransferase Overexpression, 4.4.4 beta-Glucuronidase Overexpression and 4.5 Lactate-Coupled Prodrug Activation. Each box describes the mechanism and strategy for prodrug activation, such as the use of pH-responsive units, GSH-mediated cleavage, ROS-sensitive linkers and enzyme overexpression. The diagram shows how these features trigger prodrug activation and release the active drug into cancer cells.Diagram of TME-responsive prodrug activation mechanisms in HCC with nine feature descriptions. Table 2Representative Bona Fide Tumor Microenvironment-Responsive Prodrug Systems Evaluated in HCC Models or PatientsActivation MechanismDrug or FormulationProdrug ClassificationPayloadDisease ModelPharmacokinetic and Distribution FeaturesHepatic and Systemic ToxicityControl Groups and Treatment DesignDevelopment StagepH-responsiveHis-CA@Pt/NPs79His–CA is a small-molecule prodrug containing a Schiff-base linkage.CA alone: drug-loading efficiency (DLE), approximately 95%; drug-loading content (DLC), approximately 50%. Dual loading: CA DLC, approximately 35%; CDDP DLC increased with feed ratio. The selected CDDP: His–CA mass ratio was 2.5:10.In vitro: HepG2 cells. In vivo: H22 tumor-bearing BALB/c mice.After intravenous administration at 5 mg/kg, the half-life was 0.45 h for the free drug and 2.55 h for the nanomedicine. The nanomedicine AUC was approximately twice that of free CDDP. Tumor concentrations were 4.92 μg/g at 4 h and 3.92 μg/g at 12 h. Pulmonary and renal accumulation was lower than with free CDDP.No overt hepatic or renal histopathological injury was observed under the tested conditions. ALT, AST, and MDA values were lower than those in the free-CDDP group. The hemolysis rate was 3.89% (<5%).Five parallel in vivo groups (n=6) received three tail-vein injections: saline, free CA, free CDDP, CA plus CDDP, or His-CA@Pt/NPs.Preclinical; in vivo efficacy and preliminary safety were evaluated; mouse model.PGCA@PA NPs80PGCA is a polymeric prodrug containing acid-cleavable acetal linkages.PA loading content was 7.0%; a specific loading-efficiency value was not reported in the main text. CA constituted 27.3% by mass of the polymer.In vitro: HepG2, BEL-7402, and B16 cells. In vivo: HepG2 tumor-bearing BALB/c nude mice for efficacy and B16 tumor-bearing C57BL/6 mice for immunotherapy.After intravenous injection, the half-life was 1.15 h for free PA and 16.06 h for the nanomedicine. Tumor fluorescence was higher with the nanoparticle formulation than with free PA.No marked body-weight loss or overt morphological injury in the heart, liver, spleen, lungs, or kidneys was observed. ROS levels were lower in liver tissue than in tumor tissue. These findings provide preliminary evidence of tolerability in the tested models.The in vivo efficacy study included five groups (n=5): saline, free PA, free CA, blank PGCA NPs, and PGCA@PA NPs. Three tail-vein injections were administered; irradiated groups received light 24 h after dosing. The immunotherapy study included four groups (n=5): saline, anti-PD-1, irradiated PGCA@PA NPs, and anti-PD-1 plus irradiated PGCA@PA NPs.Preclinical; mouse modelGSH-responsiveFU-SS-IND NPs74Small-molecule prodrug containing a disulfide linkage.5-FU and IND were conjugated at a 1:1 molar ratio. The loading percentage was not specified. The nanoparticles were approximately 186.9 nm in diameter, with a polydispersity index of 0.1.In vitro: Hepa1-6 cells and 5-FU-resistant Hepa1-6/FU cells. In vivo: Hepa1-6/FU tumor-bearing C57BL/6J mice.IR780 fluorescence tracking showed a clear tumor signal at 12 h after tail-vein injection, which increased through 48 h, whereas hepatic fluorescence declined from 6 to 48 h. Ex vivo imaging at 48 h showed 46.5% tumor-associated fluorescence for FU-SS-IND NPs versus approximately 5.5% for free IR780 (a 41.0-percentage-point difference). These fluorescence data track the probe rather than directly quantifying released 5-FU or IND.Mice receiving free 5-FU lost approximately 22.8% of body weight and showed severe hepatic and renal histological injury. Body weight remained stable and no overt major-organ injury was reported with FU-SS-IND NPs, indicating reduced systemic toxicity in this model.The in vivo efficacy study included five groups (n=5): saline, free 5-FU, free IND, 5-FU plus IND, and FU-SS-IND NPs; four tail-vein injections were administered. A separate combination study compared saline, anti-PD-L1 alone, FU-SS-IND NPs alone, and FU-SS-IND NPs plus anti-PD-L1.Preclinical; mouse modelLipase/ROS-responsive(light-amplified)Hybrid adipocyte-derived exosome platform encapsulating DSTG and PPLA81DSTG is a triglyceride-mimetic docetaxel prodrug containing a lipase-cleavable ester and a ROS-responsive thioether. PPLA is a PUFAylated protoporphyrin IX prodrug that also serves as the photosensitizer.DSTG and PPLA co-assembled at a 1:1 mass ratio to form lipid-core nanoparticles. The loading percentage was not reported. Particle diameters were 132.5 nm for NPs, 181.1 nm for NEMPs, and 183.8 nm for HEMPs.In vitro: Hepa1-6, HepG2, and 3T3-L1 adipocytes. In vivo: Hepa1-6 tumor-bearing C57BL/6 mice.After intravenous injection, the half-life of the free prodrug mixture was 0.18 h. The AUC0–24 h values of the prodrugs in NPs, NEMPs, and HEMPs were 18.5-, 35.7-, and 53.3-fold higher, respectively, than that of the free mixture. HEMP tumor fluorescence peaked at 8 h and remained elevated at 24 h, when it was 1.32-fold that of NPs. PPLA fluorescence tracked the labeled prodrug and did not directly quantify released docetaxel.Body weight did not differ markedly among treatment groups. CREA, BUN, ALT, and AST showed no significant between-group differences, providing no evidence of overt hepatic or renal injury under the tested conditions.The in vivo study included nine groups (n=5): saline; free mixture with or without irradiation; NPs with or without irradiation; NEMPs with or without irradiation; and HEMPs with or without irradiation. A single tail-vein dose was administered on day 0, and irradiated groups received 660-nm light 8 h later.Preclinical; mouse modelHypoxia-responsiveTirapazamine (TPZ)82Bioreductive small-molecule prodrug.TPZ was administered as a free-drug solution by intra-arterial injection; drug-loading metrics are not applicable.Clinical study in patients with unresectable early- or intermediate-stage HCC (BCLC A or B); some had progressed after surgery, radiofrequency ablation, or TACE.Systemic pharmacokinetic exposure was not considered an appropriate surrogate for efficacy or toxicity because the intended effect depended on first-pass exposure after intra-arterial TPZ administration. After embolization, devascularized HCC was no longer exposed to circulating TPZ.The most frequent treatment-related events were increased AST and ALT (82.4% each). Grade ≥3 AST/ALT elevations occurred in 41.2%–47.1% of patients and were consistent with post-embolization syndrome. No dose-limiting toxicity was observed. Transaminase elevations were transient and generally returned to baseline within 2–3 days.Phase I, open-label, single-arm 3+3 dose-escalation study (n=17) without a parallel control. Three cohorts received 5, 10, or 20 mg/m2 intra-arterial TPZ, and the expansion cohort received a fixed 35-mg dose. Response was assessed using mRECIST.Phase I clinical studyTH302-DS@FMT83TH-302 is a small-molecule hypoxia-activated prodrug.The TH-302 loading rate was 21.99 ± 1.34% before FMT modification and 68.61 ± 6.44% after modification. After five FMT-modification/loading cycles, the total TH-302 payload reached 208.79 ± 6.67 mg.In vitro: RH35 cells under hypoxia (1% O2) or normoxia (21% O2). In vivo: diethylnitrosamine-induced primary multifocal HCC in rats.The study did not report half-life, AUC, or other conventional pharmacokinetic parameters for TH302-DS@FMT.TH302-DS@FMT caused mild hemolysis. ALT and AST were lower than in controls but did not fully normalize. TBIL and urea remained within the normal range; CREA was slightly elevated in controls and normalized in treated groups. No overt major-organ abnormality was observed by H&E staining.TH302-DS@FMT produced minimal hemolysis. ALT and AST were lower than in controls but did not fully normalize. TBIL and urea remained within the normal range, and creatinine was within the normal range in all treatment groups. No overt major-organ abnormality was observed by H&E staining.Preclinical; diethylnitrosamine-induced primary multifocal HCC in rats.Enzyme-responsivePDOX (Ac-Phe-Lys-PABC-DOX)84Small-molecule prodrug containing a Cat B-cleavable Phe–Lys dipeptide and a self-immolative PABC spacer.Not applicable. PDOX is a discrete small-molecule prodrug rather than a drug-loaded carrier.In vitro: highly metastatic human HCCLM9 cells. In vivo: orthotopic HCCLM9 tumors established by intrahepatic implantation in BALB/c nu/nu mice, with spontaneous pulmonary and peritoneal metastasis.The study did not report half-life or AUC values. PDOX was stable in human and mouse blood.ALT, total bilirubin, and direct bilirubin did not differ significantly among the control, DOX, and PDOX groups. Compared with DOX, PDOX was associated with a higher white-blood-cell count, lower serum creatinine and creatine kinase, and no observable myocardial histopathological injury. Healthy-liver histopathology was not separately assessed.Three groups received intravenous treatment: saline control (n=10), DOX at 4 mg/kg (n=11), or PDOX at 21.6 mg/kg (n=12).Preclinical; orthotopic HCC xenograft model in nude mice.FA/PF-LP-CaAs85As(V) is an inorganic prodrug reduced to As(III) by GSH. MMP2 cleavage of the PLGLAG sequence in the PF peptide enhances tumor penetration but does not directly activate As(V).Arsenic-loading efficiency in calcium arsenate nanoparticles was 34.28 ± 2.15%. The administered As dose for FA/PF-LP-CaAs was 1.9 mg/kg intravenously.In vitro: HepG2 and HCCLM3 cells. In vivo: HCCLM3 subcutaneous xenografts in BALB/c nude mice.Half-life and AUC were not reported. DiR-labeled FA/PF-LP-CaAs became detectable at the tumor site at 8 h and increased through 24 h. ICP–MS measured a tumor arsenic concentration of 276.82 ± 7.79 μg/kg, higher than those of the comparator formulations. Renal arsenic measured by ICP–MS was higher than suggested by DiR fluorescence, indicating that label-based imaging did not fully represent arsenic disposition.No marked body-weight fluctuation or major-organ histological injury was observed. Hemolysis remained below 5%, supporting intravenous compatibility in the tested model.Seven groups (n=5) received saline, free As(V), CaAs NP, CaAs LP, FA-LP-CaAs, PF-LP-CaAs, or FA/PF-LP-CaAs by tail-vein injection at 1.9 mg As/kg every other day for 10 doses.Preclinical; mouse xenograft model.HA/Lipid@Pro-Dox86Pro-Dox is a small-molecule γ-GGT-activated prodrug in which a γ-glutamyl group masks Dox and is enzymatically cleaved to release the active drug.Pro-Dox loading in HA/Lipid@Pro-Dox was 3.64%. No drug leakage was detected in buffer over 96 h.In vitro: Hepa1-6 and AML12 cells. In vivo: C57BL/6 mice bearing proximal and distal Hepa1-6 tumors.Half-life and AUC were not reported. After direct injection of Pro-Dox into tumor tissue and contralateral normal tissue, Dox fluorescence was stronger in the tumor. After intravenous administration, HA/Lipid@Pro-Dox tumor fluorescence peaked at 12 h and exceeded that of Lipid@Pro-Dox.Free Dox caused marked cytotoxicity in both Hepa1-6 and AML12 cells and significant body-weight loss in mice. Pro-Dox and HA/Lipid@Pro-Dox were more cytotoxic to Hepa1-6 than to AML12 cells. Mice receiving HA/Lipid@Pro-Dox gained weight, had normal blood-biochemistry indices, and showed no overt major-organ histopathological injury under the tested conditions.In vitro cytotoxicity compared free Dox, Pro-Dox, and HA/Lipid@Pro-Dox in Hepa1-6 and AML12 cells. The bilateral-tumor study included five groups (n=5): PBS, free Dox, Pro-Dox, HA/Lipid@Pro-Dox, and HA/Lipid@Pro-Dox plus anti-PD-L1.Preclinical; mouse model.Lactate-Coupled Prodrug ActivationGD-A NPs75Pentavalent arsenic [As(V)] is an inorganic prodrug reduced by H2S to trivalent arsenic [As(III)].Loading contents were 4.35% for As, 55.20% for diclofenac, and 33.40% for GYY4137. Encapsulation efficiencies were 68.14 ± 2.46% for diclofenac and 60.22 ± 6.85% for GYY4137.In vitro: H22 and HepG2 cells. In vivo: H22 subcutaneous tumors in BALB/c mice.Conventional half-life and AUC values were not reported. ICG-GD-A NP tumor fluorescence at 4 h was higher than that of free ICG. Ex vivo imaging at 24 h showed sustained tumor retention, but substantial hepatic fluorescence was also observed because of nanoparticle sequestration. ICG fluorescence did not directly quantify arsenic disposition or As(V)-to-As(III) conversion.Toxicity was lower than with free As(III). ALT and AST increased in the free-As(III) group but not in the GD-A NP group. No overt major-organ histopathological injury or body-weight loss was observed with GD-A NPs. Approximately 70% of the excreted arsenic was recovered in urine, and nearly all arsenic was eliminated through urine and feces by day 5.The in vitro study compared PBS, As(III), As(V), GD NPs, and GD-A NPs. The in vivo efficacy study used the same five groups (n=5) and administered 2 mg As/kg by tail-vein injection every other day for seven doses.Preclinical; mouse model.Notes: Data are reported as presented in the cited studies. Direct cross-study comparisons should be made cautiously because the formulations, doses, administration routes, disease models, sampling methods, and endpoints differed. Fluorescence-based biodistribution generally tracks a fluorescent label or labeled formulation and does not necessarily quantify the released active drug. In FA/PF-LP-CaAs, MMP2 cleavage enhances tumor penetration, whereas As(V) is activated by GSH-mediated reduction. In GD-A NPs, H2S activates As(V), whereas lactate accumulation is a downstream pharmacological effect rather than the direct activation trigger. Except for TPZ, all entries are supported only by preclinical evidence and do not establish clinical efficacy or safety.Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUC, area under the concentration–time curve; BCLC, Barcelona Clinic Liver Cancer; CA, cinnamaldehyde; CDDP, cisplatin; Dox/DOX, doxorubicin; DTX, docetaxel; FMT, Fe3⁺/Mn2⁺/tannic-acid metal–phenolic network; GGT, γ-glutamyl transpeptidase; GSH, glutathione; HCC, hepatocellular carcinoma; MMP2, matrix metalloproteinase-2; mRECIST, modified Response Evaluation Criteria in Solid Tumors; NPs, nanoparticles; PABC, para-aminobenzyloxycarbonyl; ROS, reactive oxygen species; TACE, transarterial chemoembolization; TPZ, tirapazamine. Representative Bona Fide Tumor Microenvironment-Responsive Prodrug Systems Evaluated in HCC Models or Patients Notes: Data are reported as presented in the cited studies. Direct cross-study comparisons should be made cautiously because the formulations, doses, administration routes, disease models, sampling methods, and endpoints differed. Fluorescence-based biodistribution generally tracks a fluorescent label or labeled formulation and does not necessarily quantify the released active drug. In FA/PF-LP-CaAs, MMP2 cleavage enhances tumor penetration, whereas As(V) is activated by GSH-mediated reduction. In GD-A NPs, H2S activates As(V), whereas lactate accumulation is a downstream pharmacological effect rather than the direct activation trigger. Except for TPZ, all entries are supported only by preclinical evidence and do not establish clinical efficacy or safety. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUC, area under the concentration–time curve; BCLC, Barcelona Clinic Liver Cancer; CA, cinnamaldehyde; CDDP, cisplatin; Dox/DOX, doxorubicin; DTX, docetaxel; FMT, Fe3⁺/Mn2⁺/tannic-acid metal–phenolic network; GGT, γ-glutamyl transpeptidase; GSH, glutathione; HCC, hepatocellular carcinoma; MMP2, matrix metalloproteinase-2; mRECIST, modified Response Evaluation Criteria in Solid Tumors; NPs, nanoparticles; PABC, para-aminobenzyloxycarbonyl; ROS, reactive oxygen species; TACE, transarterial chemoembolization; TPZ, tirapazamine. Schematic representation of TME-responsive prodrug activation mechanisms in HCC. This schematic illustrates the molecular rationale for leveraging differential features of the HCC TME compared with normal tissues, including weak acidity, GSH and ROS levels, hypoxia, enzyme overexpression, and lactate accumulation, to trigger prodrug activation and release the active parent drug into cancer cells. The nine color-coded text boxes surrounding the central mechanism diagram provide concise descriptions of these TME-specific features and the corresponding prodrug design strategies. The numbering shown in the figure (eg, 4.1) corresponds to the relevant section header numbering in the main text of the article. Dashed black lines extending from each module indicate their conceptual anchoring to the central mechanism. Within the central illustration, the magnified tumor region is demarcated by solid gray lines and irregular light-brown shaded areas, while solid black arrows with arrowheads depict the differential expression features within the TME. The vertical red upward arrows indicate increased abundance, activity, or expression, as applicable (eg, GSH, HIF-1α, reductases, Cat B, GGT, MMP2/9, and GUS). The prodrug is rationally engineered as an active drug conjugated to a masking moiety. Upon entry into the TME (demonstrated by orange arrows), the prodrug is subjected to activation triggered by these differentially expressed features (represented by solid blue arrows), leading to the release of the active drug (emphasized by bold black arrows), which then enters and exerts therapeutic effects within the tumor cells (indicated by orange arrows). Weakly Acidic Microenvironment Warburg’s seminal work revealed that tumor cells exhibit a proclivity for converting glucose into lactate through glycolytic metabolism. This process precipitates a decline in the extracellular pH of tumor tissue to a range of 6.5–7.2, whereas the pH of normal tissue is approximately 7.4.12,87 In HCC, the activation of HSCs and the infiltration of inflammatory cells result in a substantial reduction of the extracellular pH of tumor cells to a range of 6.5–6.9. The weak acidity of HCC TME could serve as a potential target for the design of HCC prodrugs.78,88 Design strategies for acid-sensitive prodrugs principally entail the introduction of chemical bonds or functional groups into the drug molecule that can cleave or undergo conformational changes under acidic conditions, thereby enabling selective drug release within the TME or acidic intracellular organelles. A taxonomic classification of pH-responsive units, predicated on their chemical structures and cleavage mechanisms, can be made into three categories.17,89 First one is the acid-sensitive hydrolytic bonds. These include hydrazone, acetal/ketal bonds, orthoesters, cis-Aconityl, imine bonds, and ester bonds. The cleavage of hydrazone occurs via hydrolysis at a pH range of 4.5–6.5. The hydrolysis of acetals and ketals occurs in acidic environments, resulting in the formation of aldehydes, ketones, and alcohols. Orthoesters undergo rapid hydrolysis at a pH lower than 6.0. The cleavage of cis-Aconityl occurs at a pH lower than 6.0, and imine bonds (including Schiff bases) undergo hydrolysis under acidic conditions. The hydrolysis of ester bonds occurs slowly in weakly acidic environments, and these bonds are often utilized in conjunction with enzyme-responsive mechanisms. Secondly, there are phosphorus-containing sensitive bonds, including phosphoester bond, phosphonate ester bond, and phosphoramidate bond (P-O, P-N). These bonds undergo hydrolysis under acidic conditions and are frequently employed in the design of nucleotide prodrugs. Thirdly, protonatable functional groups, including imidazole, tertiary amines, and pyridine, have been identified as crucial elements in this process. Under acidic conditions, these groups undergo protonation, a process that induces alterations in conformation, charge reversal, and shifts in the hydrophilicity-hydrophobicity balance. These alterations, in turn, serve to trigger the release of the drug substance. Cinnamaldehyde (CN/CA) is a small-molecule natural aldehyde with antitumor activity, but its application is limited by poor stability, low bioavailability, nonspecific biodistribution, rapid metabolic clearance, and potential systemic toxicity.79 Wang et al conjugated histidine (His) to CA through a Schiff base to form an amphiphilic prodrug (His–CA), which co-assembled with cisplatin (CDDP) into stable His-CA@Pt/NPs through hydrophobic interactions and platinum-coordination crosslinking.79 In HepG2 cells and H22 tumor-bearing mice, His-CA@Pt/NPs prolonged circulation (half-life, 2.55 h), increased tumor accumulation, achieved 72.81% tumor inhibition, and reduced systemic toxicity. At extracellular pH 6.5, protonation of the His imidazole groups triggered charge reversal and partial dissociation of the platinum-coordination domains. At lysosomal pH 5.0, acid-catalyzed Schiff-base hydrolysis released CA. CA induced mitochondrial dysfunction and amplified oxidative stress, whereas His and CA depleted intracellular glutathione (GSH) and tetrahydrofolate (THF), respectively. This dual depletion enhanced CDDP-mediated DNA damage and chemosensitization through complementary mechanisms. Feng et al conjugated CA to a hydrophobic polymer block through acetal linkages to synthesize an acid-responsive poly(galactose-co-cinnamaldehyde) polyprodrug (PGCA).80 PGCA self-assembled into core–shell nanoparticles (PGCA@PA NPs) encapsulating the photosensitizer pheophorbide A (PA) for co-delivery. The galactose units of PGCA can facilitate positive targeting to tumor sites. After endocytosis, acid-triggered hydrolysis of the acetal linkages led to the synchronous release of CA and PA. Upon light irradiation, CA and PA synergistically enhanced the pro-apoptotic effect in cancer cells via a cascade of ROS generation, ultimately leading to apoptosis mediated by the intrinsic mitochondrial pathway. Furthermore, when combined with anti‑PD‑1 therapy, PGCA@PA NPs potentiated antitumor immunity in melanoma models, as evidenced by the induction of T‑cell responses and markedly increased T‑cell infiltration within the tumor microenvironment. By orchestrating redox modulation through both endogenous and exogenous pathways—a hallmark of cancer cell adaptation—this polyprodrug nanoplatform establishes a promising paradigm for achieving synergistic antitumor efficacy with enhanced selectivity. A non‑invasive CEST MRI study involving 15 patients with hepatic carcinoma revealed that the pH within tumor tissues was lower than that in adjacent normal tissues, thereby providing supportive evidence for the development of pH‑responsive prodrugs in HCC.90 However, the small sample size, the lack of strict histopathological confirmation of HCC, and the absence of stratification of adjacent tissues according to fibrosis or cirrhosis status represent notable limitations. Moreover, it should be noted that acidic microenvironments are also present in peritumoral tissues. Glycolytic metabolism, hypoxia, and poor perfusion collectively contribute to proton accumulation and microenvironmental acidification, and similar pH reductions can also occur in ischemic and inflamed tissues.91 In addition, acidic microenvironment in the perisinusoidal space has been shown to enhance bile salt uptake and activate HSCs, thereby promoting the progression of hepatic fibrosis.92 Although direct evidence remains lacking, it is plausible that in HCC patients with underlying hepatitis or fibrosis, the pH gradient between peritumoral tissues and the tumor itself may not be sufficiently pronounced, raising the concern of premature prodrug activation in non‑malignant regions. High Glutathione Levels and Oxidative Stress In order to counteract oxidative stress, HCC cells upregulate the expression of γ-glutamylcysteine synthetase (γ-GCS) and glucose-6-phosphate dehydrogenase (G6PD). This may result in intracellular GSH concentrations of 1–10 mM, which are significantly higher than extracellular levels (2–20 μM).93 Concurrently, elevated reactive oxygen species (ROS) levels in HCC cells are attributable to metabolic abnormalities, resulting in the establishment of an oxidative stress microenvironment.94 Reduction-sensitive prodrugs depend on GSH-mediated cleavage of disulfide or diselenium bonds to achieve drug release. A disulfide bond, for instance, undergoes a thiol-disulfide exchange reaction when exposed to the reducing action of GSH. This reaction results in the formation of thiol compounds and the release of the active drug.93 Conversely, oxidation-sensitive prodrugs leverage elevated ROS levels to facilitate targeted and controlled drug release. ROS can function as therapeutic agents for chemodynamic therapy (CDT) and as stimulatory signals for drug release. The reactive sites of ROS-sensitive prodrugs can currently be classified into two categories: cleavable linkers and non-cleavable linkers. Cleavable linkers mainly consist of boronic acids and boronates (such as phenylboronic acids and phenylboronates), thioketal, diselenides, peroxalate esters, vinyl disulfides, polysaccharides, aminoacylates, among others. Non-cleavable linkers comprise the chalcogen ether family, which includes thioethers, selenethers, and tellurethers. Xing’s team summarized seven key activation mechanisms for ROS-sensitive prodrugs, as follows: 1. Under acidic conditions, a phenylborate ester reacts with H2O2, undergoing a 1,2-shift and subsequently undergoing hydrolysis, which generates a phenol and a hydroxyborate ester. 2. Thioketal undergoes a series of chemical reactions, including oxidation, hydrolysis, and proton transfer, resulting in the formation of thiol and carbonyl compounds. 3. Aminoacrylate undergoes oxidation and cleavage via a peroxide-ring intermediate, forming carboxylate ester and imide. 4. The peroxalate ester undergoes an oxidation-induced rearrangement, which subsequently results in the formation of cyclic peroxide. This, in turn, leads to the release of carbon dioxide. 5. Diselenide linkers undergo a stepwise oxidation process initiated by H2O2, resulting in the formation of selenium-containing oxidation products. 6. Selenium ether undergoes a two-step oxidation process. Initially, a selenoxide is formed; subsequently, a selenone is formed. 7. The thioether undergoes a two-stage oxidation process. In the initial stage, a sulfoxide is generated, and in the subsequent stage, a sulfone is produced.18 5‑Fluorouracil (5‑FU), a first‑line chemotherapeutic agent for HCC, is hampered by rapid clearance, poor targeting efficiency, and systemic toxicity. Its clinical efficacy is further compromised by acquired drug resistance and the complex immunosuppressive tumor microenvironment characteristic of HCC. FU‑SS‑IND NPs are self‑assembled nanoparticles derived from a minimalist prodrug (FU‑SS‑IND), in which 5‑FU is conjugated to an indoleamine‑2,3‑dioxygenase (IDO) inhibitor (IND) via a disulfide bond.74 The disulfide linkage serves a dual role: it not only facilitates the self‑assembly of the amphiphilic prodrug into nanoparticles and enables GSH‑responsive dissociation for accelerated drug release, but also depletes GSH, thereby further amplifying ROS generation and inducing oxidative stress. Concurrently, the released IND inhibits GSH biosynthesis and enhances the effector function of T cells, effectively reversing the immunosuppressive tumor microenvironment. More importantly, FU‑SS‑IND NPs significantly upregulate PD‑L1 expression on the tumor cell surface, enabling synergistic chemo‑immunotherapy and extending the survival of 5‑FU‑resistant HCC‑bearing mice. A hybrid adipocyte‑derived exosome (ad‑exo) nanoplatform was engineered by substituting the natural core of ad‑exos with a self‑assembled lipid core composed of a lipid‑like docetaxel prodrug (DSTG), bearing a ROS‑cleavable linker, and a lipid‑conjugated photosensitizer (PPLA). Upon remote activation by near‑infrared (NIR) light, this platform enables synergistic chemo‑photodynamic therapy to induce cancer remission.81 DSTG was designed with a ROS‑responsive thioether bond and a lipase‑sensitive triglyceride backbone. Upon ester bond hydrolysis by lipase, the increased hydrophilicity and structural disruption facilitate subsequent ROS‑mediated cleavage. Under NIR irradiation, the photosensitizer generates ROS, which in turn promotes the self‑immolation of the adjacent thioether bond, thereby activating the cytotoxic agent docetaxel (DTX) and accelerating tumor cell death. The synergistic convergence of DTX‑initiated chemotherapy, PpIX‑mediated photodynamic therapy, and lipid peroxidation effects potentiates the overall antitumor efficacy. Huang et al compared GSH levels between HCC tumor tissues and normal liver tissues resected from non‑HCC patients, and found a twofold increase in GSH content within HCC tissues.95 However, a separate LC‑MS analysis comparing surgically resected HCC tumor tissues with paired non‑tumor tissues from the same patients revealed that GSH content was significantly lower in tumors than in adjacent non‑tumor tissues.96 These discrepant findings suggest that GSH levels in HCC may exhibit methodological and patient‑related heterogeneity, and therefore cannot be presumed to be uniformly elevated in prodrug design and development. Moreover, ROS elevation is not exclusive to tumors; oxidative stress also constitutes a key pathological mechanism in MAFLD.97 Hepatocellular injury triggers the expression of pro‑inflammatory factors, leading to macrophage activation, ROS release, and HSC activation, which collectively promote fibrogenesis. In MAFLD, increased oxidation of FFAs and mitochondrial dysfunction in MASH result in incomplete FAO and ROS production. Additionally, FFAs can elicit ER stress and impaired protein folding, further augmenting ROS levels. The accumulation of ROS, in turn, triggers oxidation of ER‑resident proteins, establishing a vicious cycle. Given that peritumoral tissues in HCC patients with underlying liver injury, fibrosis, or cirrhosis are inherently predisposed to ROS accumulation, the targeted activation of single ROS‑responsive prodrugs—in the absence of complementary strategies—remains a concern. Hypoxic Microenvironment Due to a disrupted blood supply, HCC often develops highly heterogeneous hypoxic regions where the expression of hypoxia-inducible factor-1α (HIF-1α) is significantly increased. This is closely associated with tumour invasion, drug resistance and a poor prognosis.67 Hypoxic regions express high levels of reductases, such as azoreductases, nitroreductases (NTRs) and quinone oxidoreductases. This provides an environment in which hypoxia-activated prodrugs (HAPs) can be reduced by enzymes.66,98 HAPs are chemically modified prodrugs that, under specific hypoxic and/or highly reducing conditions in the TME, can be reduced by specific enzymes or reducing agents to release cytotoxic drugs or generate cytotoxic substances. Design strategies can be classified into two categories. One category encompasses the incorporation of hypoxia-sensitive groups into chemotherapeutic drugs through chemical modification. The hypoxia-activated portion is integrated within the molecular backbone to preserve structural integrity. The other category involves the conjugation of hypoxia-sensitive groups to drug molecules as leaving groups. Hypoxia triggers the cleavage of the conjugate, resulting in the release of the cytotoxic drug. At present, five classifications of hypoxia-sensitive groups have been identified.66 The first class encompasses nitroaromatic compounds. The reduction of the nitro group to form hydroxylamine or aniline is a multi-step, single-electron or two-electron process catalyzed by cytochrome P450 reductase (P-450R) or nitroreductase. The active drug is released via 1,6-elimination. The second class is that of nitro compounds, which are further classified into two distinct categories: aromatic and aliphatic. Mono-electron reductases (for example NADPH cytochrome P450 oxidoreductase) reduce nitro compounds to nitroxyl radicals, which are subsequently converted into active metabolites under hypoxic conditions. In the instance of two-electron reduction, enzymes such as NQO1 or AKR1C3 directly transfer two electrons, thereby reducing nitro compounds into active amine metabolites. Thirdly, quinones, including benzoquinone, naphthoquinone, and anthraquinone, undergo single-electron reduction under hypoxic conditions. This process is catalyzed by enzymes such as CYP450, NADH: CyB5, or mitochondrial complex I. The result of this reaction is the generation of semiquinone radicals which can undergo a subsequent reduction to hydroquinone, resulting in DNA cross-linking. The process of two-electron reduction is predominantly facilitated by NQO1, leading to the generation of hydroquinone. The fourth class is comprised of azo compounds. The core skeleton of the active drug is composed of azobenzene, to which the drug is linked to the azo bond either directly or via a linking arm. Under hypoxic conditions, azoreductase catalyzes the reduction and cleavage of the azo bond, generating an amine moiety and releasing the conjugated drug. The fifth class encompasses metal complexes. These mechanisms employ the redox activity of transition metals (platinum, cobalt, and ruthenium) and hypoxia-responsive coordination chemistry to achieve controlled drug activation. Platinum(IV) is reduced to platinum(II), a toxic form of the element, by endogenous reductase. This process releases the axially coordinated drug. The reduction of Cobalt(III) to Cobalt(II) initiates a series of reactions, including ligand dissociation and the release of the cytotoxic drug. Ruthenium complexes achieve activation through transitions between different oxidation states. TACE therapy has been shown to induce tumor hypoxia and a hypoxic response in residual tumor cells following embolization. The hypoxic and immunosuppressive microenvironment limits the application of conventional chemotherapeutic agents, yet provides favorable conditions for the activation of HAPs. Tirapazamine (TPZ) is a prodrug whose anticancer effects are activated exclusively under hypoxic conditions. As a bioreductive agent, TPZ is converted into a toxic free radical under hypoxia, inducing cell necrosis, and reverts to its inactive form upon re‑exposure to oxygen. In addition, TPZ blocks the production of HIF‑1α and VEGF. However, the therapeutic efficacy of TPZ is compromised by unsustainable drug release and insufficient hypoxia induction.68 A first‑in‑human trial of TPZ chemoembolization in patients with intermediate unresectable treatment‑naïve HCC reported a complete response (CR) rate of 60% within the targeted tumor field, with a duration of response (DOR) exceeding 12 months and a median OS of 52 months. In a separate Phase I study with an exclusively Asian population and broader inclusion criteria (eg, Child‑Pugh scores of 5–7, and prior surgery, radiofrequency ablation, or embolization), with hepatitis B as the predominant etiology, TPZ chemoembolization achieved a CR rate of 47.1%, an objective response rate (CR + PR) of 64.7%, a median time to progression of 12.6 months, and a median OS of 29.3 months.82 Building on these clinical findings, the combination of TPZ with TACE addresses both the lack of effective agents for TACE recurrence and the clinical shortcomings of TPZ monotherapy. Evofosfamide (TH302), a nitroimidazole‑based hypoxia‑activated prodrug, is activated under hypoxic conditions to release bromoisophosphoramide mustard (Br‑IPM), a potent DNA cross‑linking agent with hypoxia‑dependent cytotoxic effects. A phase IB study (NCCTG N1135) evaluating sorafenib in combination with TH302 demonstrated survival outcomes comparable to other second‑line HCC therapies and established the recommended Phase II dose.99 However, insufficient hypoxia induction similarly limits the therapeutic efficacy of TH302. Drug‑eluting microspheres (DEMs) incorporating a metal‑phenolic network (MPN) and integrating TH302 (TH302‑DS@FMT) achieve dual pH‑ and hypoxia‑responsive drug release, effectively targeting residual tumor cells in hypoxic regions and reversing the immunosuppressive tumor microenvironment through a ferroptosis‑immunity cascade, demonstrating efficacy in treating embolized tumors.83 Specifically, the formulation employs DiaSpheres (DS), a clinically used DEM formulation, which were loaded with TH302 for selective targeting of hypoxic tumor cells following TAE. The MPN (FMT), composed of Fe3⁺, Mn2⁺, and tannic acid (TA), enhanced the drug loading capacity of DS, ensured controlled pH‑responsive release, induced ROS burst and lipid peroxidation (LPO), and triggered ferroptosis. Mn2⁺ further activated the cGAS‑STING pathway, promoting dendritic cell maturation and antitumor immunity. Additionally, combined with an anti-PD-L1 inhibitor (αPD-L1), distant non-embolized liver tumors were reduced. Hypoxia and HIFs are considered important drivers of hepatic fibrosis. In MAFLD, hypoxia may mediate hepatic steatosis and abnormal lipid metabolism, and the aberrant immune and metabolic microenvironment it induces is also a critical factor in the development and progression of fibrosis and cirrhosis. HIFs can stimulate Cxcl12 expression via TGF‑β, thereby activating HSCs and promoting fibrogenesis, positioning them as key regulators in HSC‑targeted anti‑fibrotic strategies.100 The hypoxic environment in HCC originates during inflammation and progressively intensifies with the advancement of liver injury, fibrosis, cirrhosis, and cancer, resulting in O2 tension drops to 6 mmHg in tumor tissues, whereas O2 tension in the portal tract and perivenous regions of normal liver is approximately 60–65 mmHg and 30–35 mmHg, respectively.101 Even though hypoxia levels are elevated in HCC, it should not be assumed that peritumoral tissues are normoxic; this consideration warrants attention in prodrug design. Enzymes Specifically Overexpressed The upregulation of various enzymes in the HCC microenvironment can serve as a “molecular switch” to activate prodrugs. Cathepsin B (Cat B) Its expression is particularly pronounced in HCC cells and tumor-associated macrophages, where it is confined to their plasma membranes. This enzyme plays a critical role in the process of cancer invasion, facilitating the penetration of basement membranes and other extracellular barriers. It specifically hydrolyzes peptide substrates such as Val-Cit and Phe-Lys. Peptide-drug conjugates (PDCs) designed based on this mechanism exhibit stability within the circulatory system; upon entering tumor tissue, they undergo cleavage by the enzyme to release the active drug. For instance, a smart prodrug of doxorubicin (DOX), Ac-Phe-Lys-PABC-DOX (PDOX), in which a Cat-B specific dipeptide is introduced, along with a spacer PABC (para-aminobenzyloxycarbonyl) to increase the distance between dipeptide and DOX, so that the dipeptide can enter the active site of Cat B, has been shown to exhibit enhanced anti-metastatic properties and a reduced incidence of adverse effects in the highly metastatic HCC model.84 Cat B influences the development and progression of hepatic fibrosis/cirrhosis by participating in apoptosis, promoting HSC activation, autophagy, and various inflammatory responses.102 At the fibrotic/cirrhotic stage, Cat B is upregulated in hepatocytes, HSCs, and NK/NKT cells. Moreover, early studies have reported that plasma Cat B levels in patients with chronic liver disease are positively correlated with the progression of hepatic fibrosis/cirrhosis, suggesting that for HCC arising in a fibrotic/cirrhotic background, sole reliance on Cat B‑responsive prodrugs may pose a risk of off‑target activation in adjacent peritumoral tissues. Matrix Metalloproteinase 2/9 (MMP2/9) It is secreted by tumor cells and activated HSCs, and it plays a role in tumor cell proliferation, migration, apoptosis, angiogenesis, tissue regeneration, and immune processes. The employment of MMP-sensitive peptide chains, such as PLGLAG and GPLGIAGQ, as linking arms facilitates the activation of prodrugs within the tumor stroma. This, in turn, enables passive targeting and enhances permeability and retention effects.85,103 MMP2‑responsive cleaved cell‑penetrating peptide (PF) and folate (FA) co‑modified liposome‑coated calcium arsenate nanoparticles (FA/PF‑LP‑CaAs) leverage overexpressed MMP2 and folate receptors on the cell membrane to enhance drug accumulation and accelerate endocytosis. The encapsulated pentavalent arsenic is subsequently reduced by GSH to the more toxic trivalent arsenic, thereby achieving effective tumor accumulation with minimized side effects.85 The MMP2‑responsive mechanism is as follows: upon arrival at the tumor site, overexpressed MMP2 cleaves the PLGLAG substrate sequence within the PF peptide, leading to PEG2 detachment and exposure of the cell‑penetrating peptide, which enables charge reversal and enhanced drug penetration. Although MMP2 response does not directly participate in the activation of the pentavalent arsenic prodrug, this system exemplifies a multi‑responsive prodrug platform with MMP2‑mediated targeting enhancement. Immunohistochemical staining of paired HCC tumors and adjacent tissues from 30 patients revealed that MMP2 and MMP9 expressions were significantly higher in tumor tissues than in adjacent tissues. Moreover, significant differences in expression were observed between T3/T4 and T1/T2 stages, suggesting that MMP2 and MMP9 may serve as predictive markers for local tumor invasion and staging.104 However, this study did not perform specific grading of adjacent tissues. Additionally, the patient cohort predominantly comprises individuals with HBV infection or alcoholic hepatitis, and the methodological approach warrants further optimization. γ-Glutamyltransferase (GGT) GGT is a significant tumor marker with a strong correlation to the progression of liver cancer, with elevated GGT levels being indicative of advanced disease. The incorporation of a γ-glutamyl group as a masking group for chemotherapeutic drugs enables GGT to cleave and remove it from the cell membrane surface, consequently facilitating the release of the active drug and impeding nonspecific release prior to cellular internalization. This approach can also be used for GGT imaging and on-demand treatment of liver cancer. Based on this, Li et al utilized hyaluronic acid (HA)-modified liposomal nanoparticles to deliver a GGT-responsive prodrug (Pro-Dox) in combination with an immune checkpoint inhibitor. In the Pro-Dox probe, the fluorescence and pharmaceutical effect of the DOX molecule are suppressed, as its amino group is occupied by the γ‑glutamyl group. In the presence of γ‑GGT, the γ‑glutamyl group is cleaved, releasing DOX and restoring both fluorescence and pharmacological activity. This approach has been shown to prolong the circulation time and augment the tumor-targeting capabilities of the treatment, thereby effectively inhibiting the growth of both proximal and distant tumors. This, in turn, facilitates the diagnosis and precision treatment of liver cancer.86 GGT levels have been widely used as a marker of liver injury. It influences oxidative stress by promoting the extracellular metabolism of GSH. Elevated GGT levels reflect prooxidant activity and cellular damage. In the context of liver disease, serum GGT levels are positively correlated with disease severity and serve as a non‑invasive criterion for assessing liver fibrosis in patients with chronic HBV infection.105 However, GGT elevation can also be observed in chronic patients without HCC. Furthermore, serum GGT cannot substitute for spatial enzymatic activity measurements in tumor and adjacent tissues, and no studies to date have examined GGT content specifically in tumor versus adjacent diseased tissues. Consequently, the selectivity of GGT‑responsive prodrugs remains to be confirmed. β-Glucuronidase (GUS/GUSB) In tumors, β-glucuronidase (GUS/GUSB) is secreted by inflammatory cells (monocytes/granulocytes) in necrotic areas outside the cells, while in normal tissues, its activity is confined to lysosomes. Consequently, higher GUS activity is observed in most cancers compared to normal tissues.106 GUS has been identified as being associated with tumor invasion, apoptosis, metastasis, and proliferation. It has been demonstrated to hydrolyze glucuronide bonds. In the design of GUS-responsive prodrugs, the active drug is covalently linked to glucuronide. By targeting enzymes that are specific to the TME, these prodrugs can selectively deliver potent anticancer drugs to the vicinity of malignant cells. Moreover, due to the hydrophilicity imparted by the glucuronide moiety, GUS-responsive prodrugs facilitate enhanced drug deposition in tumor tissue while maintaining low drug concentrations in normal tissue.106 Early in vitro studies showed that the glucuronide prodrug BHAMG, constructed by coupling the phenolic hydroxyl group of the alkylating agent HAM with glucuronic acid via a β‑glycosidic bond, exhibited significantly lower cytotoxicity to HepG2 and AS‑30D hepatoma cells than its parent drug. Upon accumulation of the antibody‑glucuronidase conjugate at the tumor site, BHAMG could be activated by glucuronidase to release HAM and exert antitumor effects.107 However, this antibody‑directed enzyme prodrug therapy (ADEPT) system relies on exogenous enzyme delivery and has not demonstrated that endogenous GUS alone is sufficient to achieve prodrug activation; it thus serves only as a reference for the feasibility of our design. In recent years, GUS‑responsive moieties in HCC research have been primarily employed for fluorescence‑guided imaging rather than therapeutic drug release. Wang et al observed that the GUS activity in HCC cells is markedly more pronounced than in normal liver cells. This heightened activity has been shown to result in the accelerated breakdown of glucuronides, thereby diminishing their capacity to bind to noxious substances. Consequently, this results in the protracted accumulation of toxic substances within the body. GUS has been identified as a potential biomarker, with the potential to facilitate real-time monitoring and early diagnosis of liver cancer. A GUS-responsive fluorescent probe containing an electron-withdrawing group (-NO2) was designed to facilitate highly sensitive and rapid imaging of liver cancer tissue, thereby assisting in surgical resection.108 To date, studies on endogenous GUS‑responsive prodrugs for HCC are scarce. GUS may therefore be considered an exploratory trigger mechanism that remains to be validated in HCC prodrug design. As a housekeeping enzyme, GUS is widely expressed in various tissues and plays important roles in the remodeling of extracellular matrix composition under physiological and inflammatory conditions.109 Currently, there is insufficient evidence that GUS activity in HCC is consistently higher than that in paired diseased liver tissues. Lactate-Coupled Prodrug Activation Enhanced aerobic glycolysis and lactate accumulation are prominent metabolic features of HCC. Reprogramming of glycolysis and enhanced glutamine catabolism increase lactate production.73 Lactate contributes to extracellular acidification, immune suppression and lactate-dependent oncogenic signaling, while MCT1 and MCT4 participate in lactate/proton transport and metabolic adaptation in HCC cells.73,110 However, elevated lactate does not necessarily constitute a direct prodrug-activation signal. In this subsection, bona fide prodrugs are restricted to systems in which a therapeutic entity is chemically masked and subsequently converted to, or releases, its active form. Lactate may initiate prodrug activation indirectly through an enzymatic cascade. Lactate oxidase (LOX) oxidizes lactate to pyruvate and H2O2, thereby converting elevated tumor lactate into a secondary chemical signal that can cleave oxidation-responsive linkers.111,112 LOX-mediated oxygen consumption may also intensify local hypoxia and thereby facilitate the activation of HAPs.111 In a recent reported polyprodrug nanoreactor, LOX-generated H2O2 cleaved oxalate ester linkers and released the indoleamine 2,3-dioxygenase 1 (IDO1) inhibitor NLG919, directly coupling lactate consumption to prodrug activation.112 Although this system was not evaluated in HCC, it established a chemically defined strategy in which lactate functions as the upstream substrate of a prodrug-activation cascade. Comparable bona fide prodrugs employing this mechanism remain insufficiently documented in HCC. Lactate dysregulation may also be targeted downstream through prodrugs that inhibit lactylation-associated signaling. ABCF1-K430 lactylation is increased in HCC and promotes tumor progression through activation of the KDM3A-H3K9me2-HIF1A pathway. Tubuloside A (TubA) was identified as an inhibitor of this modification.113 Based on this finding, an mPEG-NH2/2-FPBA/TubA polymeric micelle was constructed through a three-component dynamic covalent assembly.114 The micelle disassembled under acidic conditions and released TubA, thereby suppressing ABCF1-K430la-associated signaling in HCC models.114 In this system, acidity is the proximal activation signal, whereas lactylation is the downstream therapeutic target rather than the trigger for drug release. Prodrug activation can additionally be combined with MCT-dependent lactate trapping. A self-activating prodrug-lactate-modulated multifunctional prodrug nanocomposite (GD-A NPs), integrates a pentavalent arsenical prodrug, the H2S donor GYY4137, and the lactate-efflux inhibitor diclofenac (DCF).75 In the acidic TME, H2S released from GYY4137 reduced the less toxic As(V) prodrug to cytotoxic As(III), constituting the principal prodrug-activation step. Concurrently, H2S promoted glycolytic lactate production, whereas DCF inhibited MCT1/4-mediated lactate efflux. It resulted in the obstruction of lactate transport, leading to the accumulation of lactate within the intracellular environment. This accumulation induced acidosis, disrupted protein folding, and triggered severe endoplasmic reticulum stress. Consequently, these events ultimately activated immunogenic cell death (ICD) in cancer cells, thereby reshaping the microenvironment. Collectively, lactate can participate at different positions within a prodrug-activation cascade. It may serve as an upstream substrate for the generation of a secondary activating signal, as the metabolic source of a lactylation pathway targeted by the released drug, or as an intracellular metabolite whose accumulation amplifies prodrug-induced cytotoxic stress.75,112–114 However, only the first mechanism represents direct lactate-initiated activation. Current HCC studies predominantly combine prodrug activation by acidity, oxidation, or H2S with downstream modulation of lactate transport or lactylation. Direct lactate-initiated bona fide prodrugs therefore remain an underexplored strategy in HCC. A global lactylome analysis of tumor and adjacent liver tissues from 52 untreated patients with HBV-related HCC revealed widespread upregulation of lysine lactylation (Kla) sites in tumors, with significantly more Kla sites per protein than in adjacent tissues. The tumor-associated lactylome was extensively involved in metabolic reprogramming. Kla was also detected in adjacent liver tissues, where its abundance was inversely associated with the histological inflammation grade, suggesting an association with the inflammatory microenvironment that differed from the glycolysis-associated lactylation observed in HCC.115 Consistently, CCl4-induced liver fibrosis increased serum lactate levels in mice, while activated HSCs exhibited enhanced glycolysis and produced lactate that supported histone lactylation.116 Collectively, these findings indicate that lactate accumulation and lactylation are not restricted to HCC but may also occur in adjacent inflamed or fibrotic liver tissues. From a prodrug-design perspective, this biological overlap suggests that lactate abundance alone may provide insufficient tumor selectivity and may permit off-target activation in diseased liver. Combining lactate-associated activation with an additional tumor-enriched signal may therefore improve selectivity. TME-Responsive HCC Prodrugs: Potential Advantages and Key Challenges Potential Advantages and Current Evidence Conditional Drug Activation By masking pharmacological activity until exposure to a defined trigger, a bona fide prodrug can reduce active-drug exposure outside the intended site. In HCC, this design objective is supported mainly by preclinical systems, including acid-responsive His–CA, the GSH-cleavable FU–SS–IND conjugate, and the hypoxia-activated agents TPZ and TH302.74,79,82,83,99 Tumor selectivity, however, cannot be inferred from stimulus responsiveness or tumor accumulation alone. It requires direct measurement of intact prodrug and released active drug in HCC, paired diseased liver, and dose-limiting organs.117,118 5.1.2 Mechanistic Routes to Address Drug Resistance. These systems may address specific mechanisms that limit conventional therapy, but “overcoming resistance” should be reserved for studies that directly demonstrate reversal of a defined resistant phenotype. In a 5-FU-resistant HCC model, FU–SS–IND combined GSH-responsive release with IDO inhibition and improved antitumor activity, including in combination with PD-L1 blockade.74 The Cat B-cleavable doxorubicin prodrug PDOX retained tumor-growth inhibition and reduced several toxicity measures relative to Dox in an orthotopic HCC model, although the study did not establish reversal of chemoresistance in a multidrug-resistant setting.84 TH302 targets hypoxic cells that may be poorly served by conventional therapy; nevertheless, its activity remains dependent on sufficient hypoxia, adequate expression of activating reductases, and the intrinsic sensitivity of tumor cells to the released DNA‑crosslinking mustard.119,120 Potential to Widen the Therapeutic Window Reducing exposure to released drug in non-target tissues could widen the therapeutic window, but this remains a design objective rather than an established property of the class. Preclinical reductions in systemic or organ toxicity should not be interpreted as evidence that higher clinical doses can be administered. A credible assessment requires exposure–response data for both the prodrug and active moiety, comparison with the parent drug at pharmacologically equivalent doses, and liver-specific safety evaluation in models with impaired hepatic reserve.117,118,121 Rationale for Combination Therapy Combination strategies are most persuasive when the trigger is created or intensified by the partner treatment, or when the released payload addresses a defined resistance mechanism. TACE-induced hypoxia provides a mechanistic rationale for regional delivery of TPZ or TH302. Early phase I studies of TPZ with embolization reported feasibility and efficacy signals, whereas the TH302-DS@FMT and BSA–TPZ platforms remain preclinical.68,82,83 FU–SS–IND plus PD-L1 blockade and GD-A-mediated lactate trapping and immunogenic cell death also improved outcomes in HCC-bearing mice.74,75 These findings support further testing, but they do not establish clinical synergy, improved survival, or reduced toxicity in patients. Key Challenges Interpatient and Intratumoral Heterogeneity HCC heterogeneity occurs across etiologies, patients, lesions, and intratumoral regions.60 This variation affects the abundance and spatial distribution of candidate activation signals, as well as perfusion, immune composition, and drug penetration. The conflicting GSH measurements in paired HCC specimens and the stage-dependent expression of MMP2/9 illustrate why a uniform activation threshold cannot be assumed across patients.95,96,104 A prodrug may therefore remain insufficiently activated in trigger-low regions while releasing more active drug in trigger-high regions, producing heterogeneous intratumoral exposure and allowing poorly exposed clones to persist. Development should define quantitative activation thresholds, sample multiple tumor regions or lesions where feasible, and establish biomarkers that identify patients likely to achieve adequate intratumoral conversion. Biological Overlap and off-Target Activation in Adjacent Diseased Liver The preceding sections show that the principal endogenous triggers considered for HCC are not tumor exclusive. Acidification can occur in fibrotic liver, oxidative stress accompanies chronic liver injury, Cat B is involved in fibrosis and cirrhosis, GGT increases with liver injury, and lactate accumulation or lactylation can occur in activated hepatic stellate cells and fibrotic liver.92,97,102,105,115,116 Paired-tissue studies have also produced inconsistent results for GSH abundance.95,96 This biological overlap creates a plausible risk that a single-trigger prodrug could be activated within adjacent inflamed, fibrotic, or cirrhotic liver, thereby increasing local exposure to the active drug outside the tumor. Evidence of trigger overlap alone does not prove off-target conversion; the risk must be tested by measuring intact prodrug, released active drug, and pharmacodynamic effects in paired tumor and diseased liver.117 Multi-input activation may improve discrimination, but its selectivity should likewise be demonstrated rather than inferred from in vitro responsiveness. Impaired Hepatic Function and a Restricted Therapeutic Window Most pivotal systemic-therapy trials reviewed in this article enrolled patients with preserved liver function, and evidence remains limited for patients with Child–Pugh class B or C disease.23,53,58 In patients with impaired hepatic reserve, baseline liver dysfunction and treatment-related hepatic injury reduce the capacity to tolerate unintended active-drug exposure. The therapeutic window of a TME-responsive prodrug may therefore remain narrow even when systemic exposure is reduced, particularly if activation also occurs in adjacent diseased liver. A clinical development program should characterize the pharmacokinetics of both the intact prodrug and active moiety across relevant levels of hepatic impairment, define liver-specific stopping criteria, and demonstrate a benefit–risk advantage over the parent drug or standard treatment.23,117 Systemic Stability versus Activation Kinetics A TME-responsive prodrug must remain sufficiently stable during manufacture, storage, and systemic circulation, yet convert efficiently after reaching the intended site. Conventional carriers may add low drug loading, burst release, immunogenicity, or carrier-related toxicity, whereas reduction-sensitive linkers may undergo premature cleavage in biological fluids.121,122 Steric engineering can slow off-target conversion, but excessive shielding may also reduce nanostructure flexibility or slow release kinetics at the tumor site.118 Activation kinetics should therefore be optimized together with systemic exposure and tissue-level conversion, rather than assessed from in vitro release profiles alone. Limitations of Preclinical Evidence No single current model reproduces chronic liver disease, hepatic immune environment, spatial heterogeneity, and human drug metabolism that shape HCC treatment response. Subcutaneous xenografts are useful for initial screening but do not reproduce the diseased liver background. Orthotopic, immunocompetent, genetically engineered, and patient-derived models provide complementary information, although each has practical and biological limitations.60 A translational evidence package should include at least one model with relevant liver injury or fibrosis, spatial measurement of trigger abundance and prodrug conversion, and comparisons with the parent drug and trigger-insensitive controls. Clinical Development, Manufacturing, and Regulatory Requirements Biological translation requires prospective identification of patients whose tumors contain sufficient trigger activity. For HAPs, hypoxia alone is inadequate: efficacy also depends on activating reductases and the susceptibility of malignant clones to the released effector.120 Clinical development therefore requires pharmacodynamic assays that link trigger status to intratumoral conversion and response. It must also account for hepatic impairment, because altered metabolism and reduced hepatic reserve may narrow the margin between tumor exposure and liver toxicity.23 Chemical complexity can also compromise developability. The phosphoramidate prodrug ABBV-167 illustrates how solid-state instability can affect formulation stability and activity over time.123 Multi-component nanoprodrugs may introduce additional challenges in batch consistency, scale-up, sterilization, storage, and release testing.124 These risks should be addressed early through predefined critical quality attributes and validated assays for the intact prodrug, active moiety, and major conversion products. The applicable regulatory pathway depends on whether the product is treated as a new molecular entity, a modified drug, or a complex formulation. Existing pathways, including 505(b)(2) in the United States and the modified-new-drug framework in China, do not remove the need to demonstrate added clinical value.125–127 For TME-responsive prodrugs, the development plan should connect reproducible chemistry, manufacturing, and controls with evidence that conditional activation improves the benefit–risk profile relative to the parent drug or an appropriate standard treatment. Future Perspectives Three priorities emerge from the limitations identified above: quantitative and biomarker-guided activation, disease-relevant preclinical validation, and mechanism-based development of combinations that remain chemically and operationally feasible. Quantitative, Multi-Input Activation and Biomarker-Guided Design Multi-input designs may improve discrimination when no single endogenous signal is tumor exclusive. Existing AND-gated systems in non-HCC models provide proof of concept for sequential or coincident activation but added responsiveness does not guarantee better HCC selectivity and may increase chemical and manufacturing complexity.61,118,128,129 The immediate priority is to quantify trigger combinations in paired HCC and diseased liver, define activation thresholds prospectively, and link them to a practical biomarker assay. AI and generative models may assist candidate enumeration or multi-parameter optimization, but current evidence such as DrugGEN validates target-directed small-molecule generation rather than HCC prodrug activation or clinical translation.130 Experimental confirmation of stability, conversion kinetics, and tissue-level selectivity therefore remains indispensable. Disease-Relevant Models and Spatial Pharmacology Models should reproduce underlying liver injury, fibrosis or cirrhosis, hepatic immunity, and HCC heterogeneity rather than tumor growth alone.60 Orthotopic, immunocompetent, genetically engineered, and patient-derived systems provide complementary information; no single model is sufficient. Development should quantify trigger abundance, intact prodrug, active moiety, and pharmacodynamic effects across tumor regions, adjacent diseased liver, plasma, and dose-limiting organs.117,120 The spatial resolution of such data is critical because the same trigger signal (for example, hypoxia) can exist in normal tissues, limiting the tumor selectivity achievable by single-trigger activation alone.117 These data are needed to distinguish tumor accumulation from tumor-selective activation and to define liver-specific safety margins before clinical testing. Mechanism-Based Combinations and Developability Combination development should begin with a testable mechanistic link. Examples include coupling embolization-induced hypoxia with a HAP, or pairing an immunomodulatory prodrug with checkpoint blockade when the prodrug produces a measurable immune change.59,74,75,82,83 Each regimen should be compared with its individual components and standard therapy, and should demonstrate incremental benefit without disproportionate hepatic, vascular, or immune toxicity. Chemistry, manufacturing, and controls; analytical measurement of prodrug conversion; and the regulatory strategy should be addressed before late preclinical optimization, because added components and activation steps can erode the practical advantage of conditional delivery.123–127 Conclusion This review identifies three issues as decisive for the future of TME-responsive prodrugs in HCC. First, tumor selectivity must be demonstrated against adjacent fibrotic, inflamed, or cirrhotic liver, because several proposed triggers are shared with the disease background.92,95–97,102,115,116 Second, heterogeneous trigger distribution requires quantitative biomarkers and spatial pharmacokinetic/pharmacodynamic assessment rather than a binary “responsive” classification.60,117,120 Third, greater chemical complexity is justified only when multi-input activation or combination therapy produces a measurable improvement in benefit–risk and remains manufacturable.118,123,124 Most HCC-directed systems remain at the cellular or animal stage, and the early clinical experience with TPZ embolization does not establish a class effect.74,75,82–84 Progress therefore depends less on adding new nominally responsive formulations than on rigorous activation mapping, disease-relevant models, liver-specific safety assessment, and prospective patient selection. If these requirements are met, TME-responsive prodrugs may provide a useful route for developing HCC therapies with more controlled intratumoral drug exposure; their clinical value, however, remains to be demonstrated.