Available data suggest that anti-angiogenesis can inhibit tumor growth, but compensatory angiogenesis is involved in anti-angiogenic therapeutic resistance. Important vascular stimulators have been studied in HCC, including vascular endothelial growth factor (VEGF)/vascular endothelial growth factor receptor (VEGFR), fibroblast growth factor (FGF)/fibroblast growth factor receptor (FGFR), platelet-derived growth factor (PDGF)/platelet-derived growth factor receptor (PDGFR), endoglin (CD105), and angioportin/tie (17). Among these factors, VEGF is the most important cytokine that facilitates angiogenesis and is composed of five different isoforms: VEGF-A, -B, -C, -D, and -E. The VEGF receptors VEGFR-1, -2, and -3 bind VEGF with different affinities (18). The combination of VEGF and VEGFR can not only trigger the proliferation of vascular endothelial cells but also activate lymphatic metastasis by forming new lymphatics. In 1971, Folkman suggested potential strategies against tumor progression by blocking tumor angiogenesis (19). Angiogenesis inhibitors are administered to patients with HCC, including bevacizumab, cabozantinib, lenvatinib, ramucirumab, regorafenib, and sorafenib, based on the National Comprehensive Cancer Network guidelines (20). Placental growth factor (PLGF) is a pro-angiogenic factor belonging to the VEGF family, which is usually secreted under pathological conditions. The overexpression of PLGF has been noted in several tumors resistant to anti-angiogenesis therapy, suggesting that PLGF is a prospective target in HCC treatment (21–24).

Although current anti-angiogenic drugs, together with immunotherapy, have prolonged the survival of patients to a certain extent, acquired resistance limits the therapeutic efficacy of HCC. To some extent, antiangiogenic drugs prevent the transport of chemotherapy drugs and limit the therapeutic efficacy. Furthermore, vasculogenic mimicry (VM) and vessel co-option (VC) are two emerging theories that explain resistance to anti-angiogenic drugs.

VM is a novel tumor microcirculation model independent of endothelial cells and was first developed by Maniotis et al. in 1999 (25). This model indicates that pluripotent stem cell-like tumor cells (PSCLTCs) acquire endothelium-like properties and form stable tubular structures without endothelial cells by secreting heparin sulfate, proteoglycans, collagen IV and VI, and laminin to transport nutrients for tumor growth (26). The VM process is unsurprisingly complex and is usually induced by a hypoxic TME and favored by many molecular mechanisms, mainly involving epithelial–mesenchymal transition (EMT), cancer stemness, response to hypoxia, and extracellular matrix remodeling (27, 28). VEGF/VEGFR inhibitors disrupt the formation of tumor blood vessels, leading to an acidic environment that compensates for the production of VM structures. Therefore, VM is an important cause of drug resistance after the application of anti-VEGF drugs to solid tumors. Sorafenib is a small-molecule tyrosine kinase inhibitor with anti-angiogenic activity. Clinical trials in breast cancer have been limited and have stopped at phase III. Mao et al. found that compared to the less invasive MCF7 cell line, only breast cancer stem-like cells (BCSLCs) and ALDH1+ MDA-MB-231 cells showed angiogenic ability and were directly involved in the development of VM. Sorafenib only inhibits endothelial angiogenesis and has no effect on VM. Mechanistically, they found that HIF-1α contributes to the proportion of BCSLCs (29). Furthermore, VM usually occurs in tumors with higher grades, shorter survival, and more aggressive disease (30, 31). Sorafenib resistance has been studied most frequently in HCC. Shi et al. found that the high expression of ITGA5 and ITGB1 in sorafenib-resistant liver cancer tissues promoted the degree of hypoxia and the generation of VM structures (32). The most important molecular mechanisms favoring VM in HCC include the following: (a) HIF1-α promotes VM production by regulating LOXL2 and is positively correlated with poor prognosis of HCC patients (33); (b) m6A methyltransferase METTL3 promotes VM generation and HCC metastasis by enhancing the translation efficiency of YAP1 mRNA and is related to Hippo pathway (34); (c) exosomes derived from hepatocytes, circRNA-100338, can affect the angiogenesis of liver cancer, including VM, thereby promoting the metastasis of liver cancer (35); (d) high expression of CD276 protein can activate the PI3K/AKT/MMPs pathway to promote VM formation, which is related with poor prognosis in HCC (36); (e) BMP4, migration-inducing gene 7 (MIG7), long noncoding RNA n339260, RhoC/ROCK2, Slug, and osteopontin were also found to be factors inducing VM formation in HCC (37–42); (f) EMT pathway can be induced by Hsp90B, Notch 1, ZEB2 and Twist1 to promote VM formation (43–46). These molecules are thought to be potential therapeutic targets for tumors resistant to anti-angiogenic therapy in patients with HCC.

Vessel co-option is a non-angiogenic mechanism in which tumor cells adhere to the existing normal vessels of host organs and grow infiltratively by migrating along them without angiogenesis (26). Kuczynski et al. performed an experiment to illustrate the co-option process in sorafenib-resistant HCC using Hep3B-hCG orthotopic HCC xenografts. Under sorafenib treatment, angiogenic vessels of the tumor are depleted but co-opted pre-existing vessels are preserved. Meanwhile, EMT-like molecules are increased, leading to both tumor invasion and incorporation into the liver parenchyma, co-opting the normal liver vascular system. These changes can be reversed by discontinuing sorafenib treatment (47, 48). Interestingly, following the onset of sorafenib resistance, tumor cells begin to surround the hepatic sinuses and major blood vessels and invade rapidly. The tumor invasion signal weakens and becomes angiogenesis-dependent only when sorafenib treatment is discontinued (49). In cases of colorectal cancer with liver metastases, vascular co-option is associated with resistance to anti-angiogenic therapy. Cancer cells replace hepatocytes by inducing apoptosis proteins, motility, and EMT, and enter sinusoidal vessels to establish vascular co-option (50). However, these findings lack specific markers of co-opted vessels and need to be verified in patients with HCC in future studies. If possible, anti-angiogenic drugs together with anti-co-opted vessels will be an optimal choice for patients with metastatic HCC.

Recently, some VM-targeted drugs have been developed to inhibit VM in HCC and have shown satisfactory results in reversing drug resistance. The promising drugs and functional mechanisms that inhibit VM formation and overcome anti-angiogenic therapy resistance are summarized in Table 1 .

The imbalance between the recruitment of immunosuppressive cells (MDSCs, TAMs, and Tregs) and the reduction of anti-tumor effector cells (cytotoxic T lymphocytes [CTLs], NKs, and dendritic cells [DCs]) results in an immunosuppressive microenvironment for HCC, leading to immunotherapy resistance (61). A high-level overview of these mechanisms is shown in Figure 1 . The mechanisms and potential targets of drug resistance in these cells are described.

Tregs are a subset of CD4+CD25+ immunosuppressive T cells, characterized by Forkhead box protein P3 (Foxp3) expression, which determines the development and differentiation of Tregs. In HCC tissues, a high concentration of Tregs combined with a low concentration of CD8+T cells is an independent factor that affects the survival and recurrence of patients (62). However, the mechanisms by which Tregs induce resistance to ICIs and sorafenib in HCC have not been fully elucidated. However, a number of molecules and signaling pathways have been implicated in Treg-induced immune tolerance: (a) the emergence of other immune checkpoints from Tregs, such as TIM-3, TIGIT, LAG3, and VISTA, blocks effector T cell activation, and their high expression is often associated with poor prognosis in HCC patients after anti-PD-1/PD-L1 treatment (63, 64); (b) TGF-B secreted by Tregs induces EMT and promotes the invasion and migration of HEPA1-6 cells. The combination of TGF-B inhibitors significantly enhances the sensitivity of HCC cells to regorafenib and sorafenib (65, 66); (c) CD25 expressed by Tregs competitively consumes IL-2, reduces effector T cell activation, and induces metabolic disorders (67); (d) CD73, CD39, and cyclic AMP-regulated adenosine A2 receptor (A2AR) can reduce T-cell toxicity and shift the microenvironment into a tolerant state (68); (e) Tregs secrete granzyme B/perforin that directly lyses NK and CD8+ T cells (69). Moreover, CCR4+ Tregs have been found to be an important type of Tregs in HBV⁺ HCC, correlated with sorafenib resistance and increased IL-10 and IL-35 levels, and treatment with a CCR4⁺ antagonist has been shown to successfully reverse sorafenib resistance and sensitize liver cancers to PD-L1 checkpoint blockade (70).

MDSCs are heterogeneous and immature cells derived from the bone marrow, which can be induced in almost all types of tumors and in pathological conditions such as infection, autoimmune disease, and trauma. The mechanism of drug resistance in MDSCs is mainly through damaging T cell function: (a) MDSCs overconsume essential amino acids such as arginine-1 and cysteine in the body, leading to T cell proliferation and activation dysfunction, as well as tryptophan by overexpression of indoleamine-pyrrole 2,3-dioxygenase (IDO) (71–73); (b) Cystine deprivation downregulates the T-cell receptor (TCR) ζ chain of hepatic CD8(+) T cells and helps tumor cells to escape; and (c) MDSCs also express cytotoxic reactive oxygen species (ROS), inducible nitric oxide synthase (iNOS), and nitric oxide (NO) to influence T cell viability (74). Furthermore, many tumor-derived cytokines such as FGF1, HIF-1, VEGF, IL-6, and G-CSF have been shown to promote the accumulation of MDSCs and are related to therapy resistance (75–80). For example, Xu et al. argued that chemotherapy-resistant HCC cell-released IL-6 boosts the silencing and activity of MDSCs and, when anti-IL6 neutralizing antibody is combined with 5-FU chemotherapy, tumor growth is significantly decreased (80). TGF-B secreted by MDSCs participates in a variety of drug resistance mechanisms, such as EMT induction, CSC promotion, and immunosuppression (81). Downregulation of TGF-B expression can overcome sorafenib resistance in liver cancer (82). Furthermore, hepatoma-intrinsic CCRK inhibition reduces MDSCs immunosuppression and enhances the blocking effect of the immune checkpoint (PD-L1) (83). SB265610, a CXCR2 antagonist, may reverse immunosuppression by inhibiting the chemotaxis of MDSC to the tumor, thereby promoting the anti-tumor immunity of CD8+ T cells and inhibiting tumor immune escape (84). MDSC-targeted therapy has achieved great progress in solid cancers, such as melanoma, breast cancer, NSCLC, and glioblastoma, and has even entered the stage of clinical trials, but research on liver cancer is still lacking.

TAMs are another type of marrow-derived cells that are divided into M1 (tumor suppressor macrophages) and M2 (tumor-promoting macrophages) subtypes, with type M2 predominating in liver cancer (85). In multiple cancers, the infiltration of M2 TAMs is highly related to poor prognosis and treatment resistance (86, 87). The main mechanism by which TAMs exert immune tolerance in HCC is through the secretion of molecules. For example, hepatocyte growth factor (HGF), derived from polarized M2 TAMs, confers HCC resistance to sorafenib in a feed-forward manner. Accumulated HGF can activate the HGF/c-Met, ERK1/2/MAPK, and PI3K/AKT pathways in tumor cells, recruit more M2 TAMs, and promote tumor growth. Therefore, the combination of the HGF inhibitor cabozantinib and sorafenib is reasonable for improving the efficacy of first-line systemic therapies (88). TAMs can also induce oxaliplatin resistance through autophagy in HCC cells (89). Therefore, blocking the recruitment of TAMs or reprogramming the polarization of TAMs is a promising measure to enhance the sensitivity of HCC treatment. Antagonists targeting the chemokine C-C motif ligand 2 receptor (CCR2) and stromal cell-derived factor 1 α (SDF-1α/CXCL12) have shown good efficacy in blocking TAM recruitment (90–92). Targeted colony stimulating factor (CSF-1), autophagy, NF-KB, Mir-214, IL-6, and toll-like receptors (TLRs) can transform TAMs from a pro-tumor phenotype into an antitumor phenotype (M1) (93–99).

NK cells are abundant in liver tissue, exerting their ‘killing’ function by regulating the activation receptors (NKG2C, NKG2D, CD244, CD266, and NCR) and inhibitory receptors (CD94/NKG2A and KIR) expressed on the surface of NK cells (100). The balance between the signals elicited by these receptors determines whether NK cells are activated and perform their effector functions. NK cell abnormalities induced by the TME are the main reason why tumor cells escape the immune response. For example: (a) immunosuppressive factors such as TGF-B, IL-10, IL-6, and IL-23 secreted by Tregs, MDSCs, or TAMs suppress NK cell function and induce tumor evasion and progression (101); (b) NK cells are excessively activated by mononuclear macrophages through CD48/2B4 interactions, thus inducing NK cell exhaustion and death (102); and (c) increased lactic acid content in tumor cells leads to metabolic reprogramming, resulting in NK cell dysfunction (103). In addition, the influence of HCC cells and their microenvironment can limit the sensitivity to NK cell cytotoxicity. NK cells targeting CSCs are known to inhibit tumor progression; however, the resulting overexpression of CEACAM1 renders EpCAM⁺ HCC cells resistant to NK toxicity. Anti-CEACAM1 antibody has been shown to restore the cytotoxicity of NK cells against EpCAM⁺ Huh-7 cells (104). Similarly, the inhibition of CNOT7 and the enhancement of zeste homolog 2 (EZH2), miR-889, granulin-epithelin precursor (GEP), and MICA/B can successfully reverse NK cell resistance, suggesting that these molecules are promising targets for immunotherapy in HCC patients (105–109).

As mentioned above, the immune system greatly influences tumor response to various treatments in the microenvironment. Therefore, based on the above immune tolerance mechanisms, significant efforts have been made to attempt to reverse tumor drug resistance, including: (a) combining immune checkpoint inhibitors; (b) inhibiting the recruitment of Tregs, MDSCs, and M2 TAMs in tumor tissue; (c) reprogramming the polarization of TAMs; and (d) reinforcing the anti-tumor capabilities of the immune system.

Multiple studies have shown that PD-1/PD-L1 inhibitors in combination with multi-tyrosine kinase inhibitors, VEGF inhibitors, or CTLA-4 inhibitors are superior to monotherapy (110). Two CTLA-4 inhibitors, ipilimumab and tremelimumab, have recently gained increasing attention, with FDA approval in 2011 for patients diagnosed with advanced or unresectable melanoma and in 2015 for patients with malignant mesothelioma (111, 112). In contrast to the immunonormalizing effect of PD-1/PD-L1 inhibitors, CTLA-4 suppression tends to largely enhance the toxic function of T cells, thus increasing the sensitivity to other treatments. Currently, the combination of ipilimumab and nivolumab/pembrolizumab has been associated with encouraging survival outcomes in patients with advanced HCC with primary resistance to prior immune checkpoint inhibitors (113). Moreover, this combination as a first-line therapy for patients with advanced HCC is currently undergoing evaluation (NCT04039607). Tremelimumab, another CTLA-4 inhibitor, in combination with durvalumab for patients with unresectable hepatocellular carcinoma, has been associated with a reasonable outcome (median overall survival of 18.7 months) and a satisfactory benefit-risk profile (114). With the emergence of more immune checkpoints and anti-PD-1/L1 resistance, the combination of immune checkpoint inhibitors to overcome drug resistance has become a popular trend. For example, the inhibition of TIGIT together with anti-PD-1 has been shown to significantly decrease tumor growth and increase the proportion of cytotoxic T cells in tumors (64). Fibrinogen-like protein 1 (FGL1) was recently identified as a major immune inhibitory ligand of LAG-3 and its overexpression was positively associated with poor prognosis, anti-PD-1 therapy resistance, and sorafenib resistance (115, 116). Although the exact mechanism by which FGL1/LAG3 regulates the immune environment remains unclear, synergistic inhibitory effects of anti-FGL1 and anti-PD-1 have been identified in animal studies. Similarly, a combined inhibitory effect of anti-LAG-3 and anti-PD-1 has been identified in several cancers such as melanoma, NSCLC, and breast cancer (117–119). Therefore, FGL1 and LAG3 can be used as next-generation immunotherapy targets independent of PD-1/PD-L1.

Anti-tumor efficacy can be enhanced by inhibiting Tregs, MDSCs, and TAMs, or by targeting important molecules to enhance NK cell activity. Related targets and potential reversal of drug resistance are described in Section 2.2.1. Adoptive cell transfer is an emerging therapeutic strategy with wide potential value in the treatment of HCC, and is deemed to be a highly individualized cancer therapy because of the adhibition of the patients’ own effect factors (120). What’s more, antiviral treatment for HBV-HCC was proved effective in increasing the postoperative survival (121). In the clinical study of PD-1 antibody, when ICIs repair immune function, HBV infection can be significantly reduced. In mouse models, it can also be seen that the virus infection rate of mice treated with PD-1 antibody is significantly reduced. Thus, the immune escape mechanisms of viruses and tumors actually work in the same pathway (122). Therefore, it is worth exploring whether there is a relationship between viral load and treatment resistance and whether the combination of antiviral drugs and immunotherapy or anti-angiogenesis is more effective. At present, adoptive cell therapy has been developed into a phase III clinical treatment, including chimeric antigen receptor (CAR) T cell therapy, tumor-infiltrating lymphocyte (TIL) therapy, engineered TCR therapy, cytokine-induced killer (CIK) cell therapy, and NK cell therapy. However, more effort is needed to achieve the desired clinical effect. Thus, due to the high heterogeneity of HCC tissues, targeting the immune system requires a combination of other molecular targeting inhibitors or multiple types of therapies.

Abnormal vascular morphology in HCC tumors results in a preliminary anoxic and acidic microenvironment after the treatment of anti-angiogenic therapy (147). The lack of oxygen and nutrients often causes cancer cells to undergo glycolysis rather than metabolism, producing large amounts of lactic acid. The degree of hypoxia in the HCC microenvironment is heterogeneous. As the degree of hypoxia increases along with depth from the blood vessels into the tumor, HCC cells are exposed to a smaller concentration of oxygen and the pH decreases (148, 149). Tyrosine kinase inhibitors, anti-angiogenic drugs, or other treatments that restrict the blood supply within the tumor can exacerbate the development of an anoxic and acidic environment, leading to tumor progression and drug resistance (150). Hypoxia-regulated transcription factor HIF-1 is an extensively authoritative marker to modify cancer sensitivity to therapeutic agents, and comprises a heterodimeric DNA-binding complex composed of α and B subunits (151–154). Hypoxia-induced HIF-1α elevation and multiple mechanisms contributing to chemotherapy resistance are related to: (a) alterations in metabolic pathways. Glucose uptake and glycolysis are strongly activated in hypoxic environments to meet the demands of tumor cell growth. HIF-1α induces many glycolytic genes including glucose transporter type 1 (GLUT1) and hexokinase 2 (HK2) (155, 156). Genistein, as a natural isoflavone, can directly downregulate HIF-1α, thus inactivating GLUT1 and HK2 to inhibit aerobic glycolysis, and therefore induce apoptosis of aerobic glycolysis HCC cells. At the same time, Genistein enhanced the sensitivity of sorafenib resistant HCC cells in vivo and in vitro (157); (b) crosstalk between autophagy and mitophagy. In the context of increased metabolic demands, mitophagy and autophagy may play key roles in inducing chemoresistance by regulating the adaptation of tumor cells to hypoxia, increasing oxidative stress and DNA damage, and providing nutrition and energy to tumor cells (158). B-cell lymphoma-2/adenovirus E1B 19 kDa-interacting protein 3 (BNIP3) and BNIP3-like protein X (NIX) are hypoxia-induced HIF-1α target genes that can bind LC3 and trigger the mitophagy response (159). Upregulation of BNIP3 by HIF-1α can promote autophagy and oxidative resistance in HCC cells (160). However, treatment with sorafenib did not inhibit its cytoprotective action (161); (c) drug efflux. ATP-binding cassette (ABC) transporters, such as MDR1, can pump chemotherapeutic drugs from the intracellular to extracellular regions of HCC cells to produce drug resistance. The reporter gene assay and electrophoretic mobility shift assay proved that HIF-1α is a critical factor for MDR1 gene overexpression (162); and (d) apoptosis inhibition. HIF-1α is the promoter activator of CBR1, and CBR1 overexpression induces apoptosis resistance by reducing oxidative stress associated with hypoxia, cisplatin, and doxorubicin treatment (163). Furthermore, several important signaling pathways and other gene responsible for chemotherapy resistance under hypoxia are summarized in Table 3 ; however, whether they are regulated by HIFs has not been verified.

Hypoxia is a typical factor that leads to radiotherapy resistance in solid tumors. Radiation can directly damage the DNA strands or damage the DNA strands through the production of radicals, resulting in the death of tumor cells (182). In an aerobic environment, oxygen reacts with free radicals on broken DNA strands, forming a more stable pattern of DNA damage. It’s also a stable hydrogen peroxide structure, which can promote more DNA breaks. In anoxic environments, tumor cells repair DNA damage by removing hydrogen from free sulfhydryl groups, thereby developing resistance to radiotherapy (183). Similarly, HIF-1α overexpression during hypoxia is an important factor in radioresistance. Bai Bing et al. found that inhibition of the EZH2/Mir-138-5p/HIF-1α pathway could enhance the sensitivity of radiotherapy (184). In addition, the activation of PI3K/AKT signaling is known to induce radioresistance in various tumors by increasing HIF-1α translation efficiency, and elevated PDK1 is a driver of PI3K/AKT/mTOR signaling in HCC, suggesting that this pathway is a potential therapeutic target to reverse radiotherapy resistance (185, 186). Furthermore, HIF-1α can also be upregulated by mTORC1 to confer radiotherapy resistance to HCC through the anabolic integration of glucose and cardiolipin (187).

Hypoxia is a characteristic of HCC and contributes to chemotherapy and radiotherapy resistance. Therefore, reprogramming the hypoxic environment using immunosuppressants, nanoparticles, natural products, and HIF-1α inhibitors is an attractive direction for future research. Rapamycin, an mTOR inhibitor, reverses resistance to adriamycin during hypoxia by decreasing hypoxia-induced HIF-1α accumulation (188). Moreover, rapamycin can increase the sensitivity of HCC cells to cabozantinib (a c-Met inhibitor), which has a synergistic inhibitory effect on HCC cells (189). Nanoparticles are an emerging tool to improve the hypoxic environment of HCC. Nanoparticle delivery of oxygen-generating MnO₂ and anti-angiogenic drugs can normalize the TME by inhibiting hypoxia-induced invasion, EMT, and metastasis, and increasing the expression of M1-type macrophage genes (Nos2, IL-6, CCL5, CXCL9, and CXCL11) (190). Bruceine D and silymarin are two newly discovered natural products that reverse drug resistance by decreasing HIF-1α expression and can directly block the inhibitor of B-catenin and T-cell factor/B-catenin interaction to regulate HCC cell metabolism, and silymarin reduces the expression of MDR1 and P-glycoprotein (191, 192). Furthermore, sorafenib inhibits HIF-1α synthesis and shifts the hypoxia response from the HIF-1α- to HIF-2α-dependent pathway. HIF-2α overexpression activates the TGF-α/EGFR pathway, thereby inducing sorafenib drug resistance. This means that to inhibit cancer growth, it may be necessary to target both HIF-1α and HIF-2α. The HIF-1α mRNA antagonist RO7070179 has shown primary success in a phase Ib study (NCT02564614).

Exosomes function as vectors to transport a large number of molecules between cells in the TME, including messenger RNA (mRNA), long noncoding RNA (lncRNAs), microRNAs (miRNAs), circular RNA (circRNAs), lipids, proteins, and nucleic acids. These molecules are secreted by cancer or stromal cells in the TME and are involved in TME remodeling, tumor progression, invasion, angiogenesis, metastasis, and drug resistance (193). On one hand, exosome-associated therapeutic resistance is achieved through transport from drug-resistant cancer cells to drug-sensitive cells. For example, Fu et al. found that exosome miR-32-5p from a multidrug-resistant cell line (Bel/5-FU) activated the PI3K/Akt signaling pathway by inhibiting PTEN and induced drug resistance in sensitive cells by facilitating EMT and angiogenesis (194). MiR-221/222 from chemotherapy-resistant MCF-7 cells can also render sensitive MCF-7 cells resistant, but this has not been verified in HCC (195). In contrast, exosomes may swallow drug molecules and pump them out with the help of the ATP-binding cassette (ABC) transporter family (such as ABCB1) (196, 197). Recently, several researchers have made progress in the study of exosome drug resistance in HCC. Li et al. found that miR-27a-3p derived from M2 macrophage exosomes is responsible for HCC stemness and directly negatively regulates TXNIP, which has been reported as a tumor repressor gene. Upregulation of TXNIP can reverse drug resistance induced by elevated miR-27a-3p levels (198). Zhang et al. found that overexpressed plasma exosome circUHRF1 could increase TIM-3 expression caused by miR-449c-5p decline to inhibit NK cell function and induce anti-PD-1 therapy resistance in patients with HCC (199).

Communication between cells is a double-edged sword that can not only transmit information to each other to promote the occurrence and development of tumors and drug resistance but also provide treatment targets. In fact, some exosome-mediated RNAs have shown satisfactory potential in relieving therapy-resistant stress. MiR-199a-3p, miR-744, and miR-122 are downregulated in HCC tissues, and chemotherapy resistance in HCC has been successfully reversed through exosome transport (200–202). Furthermore, exosomes as drug carriers have great advantages and a good precedent for the treatment of HCC patients. First, compared to artificial liposomes, exosomes have higher packaging efficiency and stability. Second, exosomes can reduce immune responses in vivo. Third, because of the specific molecules on the exosome surface, interaction with antibodies or coagulation factors is limited, thereby reducing the occurrence of immune responses in vivo (203). Notably, several major applications in exosome packaging include small-molecule chemical drugs, proteins, peptides, and nucleic acid drugs. Zhang et al. designed a neutrophil-derived exosome-like vesicle loaded with doxorubicin and decorated it with superparamagnetic iron oxide nanoparticles, which showed precise targeting in gastric cancer, hepatoma, and colon cancer cell lines (204). Another exosomal miR-155 inhibitor markedly improved cisplatin susceptibility in a cisplatin-resistant oral squamous cell carcinoma 3D model by decreasing MET and drug efflux transporter proteins (205). Although great progress has been made over the last few decades and multiple preclinical trials show promising results, there are still obstacles to be overcome (206). For example, there is a lack of standardized exosome isolation and purification techniques. Traditional separation techniques are often contaminated by other types of exosomes, which can significantly affect therapeutic efficiency. Second, thorough and accurate research on exosome characterization is required because exosomes from different cell sources may have opposite therapeutic effects. These technical barriers must be addressed so that exosomes can be used for the diagnosis or prognostic monitoring of cancer and innovative, personalized exosome-based therapies.